2011-03-08 13:14:00 +00:00
|
|
|
/*
|
2012-11-02 15:44:58 +00:00
|
|
|
* Copyright (C) 2011, 2012 STRATO. All rights reserved.
|
2011-03-08 13:14:00 +00:00
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public
|
|
|
|
* License v2 as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public
|
|
|
|
* License along with this program; if not, write to the
|
|
|
|
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
|
|
* Boston, MA 021110-1307, USA.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/blkdev.h>
|
2011-06-13 17:59:12 +00:00
|
|
|
#include <linux/ratelimit.h>
|
2011-03-08 13:14:00 +00:00
|
|
|
#include "ctree.h"
|
|
|
|
#include "volumes.h"
|
|
|
|
#include "disk-io.h"
|
|
|
|
#include "ordered-data.h"
|
2011-06-13 18:04:15 +00:00
|
|
|
#include "transaction.h"
|
2011-06-13 17:59:12 +00:00
|
|
|
#include "backref.h"
|
2011-08-04 16:11:04 +00:00
|
|
|
#include "extent_io.h"
|
2011-11-09 12:44:05 +00:00
|
|
|
#include "check-integrity.h"
|
2012-06-04 18:03:51 +00:00
|
|
|
#include "rcu-string.h"
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* This is only the first step towards a full-features scrub. It reads all
|
|
|
|
* extent and super block and verifies the checksums. In case a bad checksum
|
|
|
|
* is found or the extent cannot be read, good data will be written back if
|
|
|
|
* any can be found.
|
|
|
|
*
|
|
|
|
* Future enhancements:
|
|
|
|
* - In case an unrepairable extent is encountered, track which files are
|
|
|
|
* affected and report them
|
|
|
|
* - track and record media errors, throw out bad devices
|
|
|
|
* - add a mode to also read unallocated space
|
|
|
|
*/
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_block;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
#define SCRUB_PAGES_PER_BIO 16 /* 64k per bio */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
#define SCRUB_BIOS_PER_CTX 16 /* 1 MB per device in flight */
|
2012-11-02 13:58:04 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* the following value times PAGE_SIZE needs to be large enough to match the
|
|
|
|
* largest node/leaf/sector size that shall be supported.
|
|
|
|
* Values larger than BTRFS_STRIPE_LEN are not supported.
|
|
|
|
*/
|
2012-03-27 18:21:27 +00:00
|
|
|
#define SCRUB_MAX_PAGES_PER_BLOCK 16 /* 64k per node/leaf/sector */
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
struct scrub_page {
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_block *sblock;
|
|
|
|
struct page *page;
|
2012-05-25 14:06:08 +00:00
|
|
|
struct btrfs_device *dev;
|
2011-03-08 13:14:00 +00:00
|
|
|
u64 flags; /* extent flags */
|
|
|
|
u64 generation;
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 logical;
|
|
|
|
u64 physical;
|
2012-11-02 13:58:04 +00:00
|
|
|
atomic_t ref_count;
|
2012-03-27 18:21:27 +00:00
|
|
|
struct {
|
|
|
|
unsigned int mirror_num:8;
|
|
|
|
unsigned int have_csum:1;
|
|
|
|
unsigned int io_error:1;
|
|
|
|
};
|
2011-03-08 13:14:00 +00:00
|
|
|
u8 csum[BTRFS_CSUM_SIZE];
|
|
|
|
};
|
|
|
|
|
|
|
|
struct scrub_bio {
|
|
|
|
int index;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_device *dev;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct bio *bio;
|
|
|
|
int err;
|
|
|
|
u64 logical;
|
|
|
|
u64 physical;
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_page *pagev[SCRUB_PAGES_PER_BIO];
|
|
|
|
int page_count;
|
2011-03-08 13:14:00 +00:00
|
|
|
int next_free;
|
|
|
|
struct btrfs_work work;
|
|
|
|
};
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_block {
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *pagev[SCRUB_MAX_PAGES_PER_BLOCK];
|
2012-03-27 18:21:27 +00:00
|
|
|
int page_count;
|
|
|
|
atomic_t outstanding_pages;
|
|
|
|
atomic_t ref_count; /* free mem on transition to zero */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2012-03-27 18:21:27 +00:00
|
|
|
struct {
|
|
|
|
unsigned int header_error:1;
|
|
|
|
unsigned int checksum_error:1;
|
|
|
|
unsigned int no_io_error_seen:1;
|
2012-05-25 14:06:08 +00:00
|
|
|
unsigned int generation_error:1; /* also sets header_error */
|
2012-03-27 18:21:27 +00:00
|
|
|
};
|
|
|
|
};
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx {
|
|
|
|
struct scrub_bio *bios[SCRUB_BIOS_PER_CTX];
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *dev_root;
|
2011-03-08 13:14:00 +00:00
|
|
|
int first_free;
|
|
|
|
int curr;
|
2012-11-02 15:44:58 +00:00
|
|
|
atomic_t bios_in_flight;
|
|
|
|
atomic_t workers_pending;
|
2011-03-08 13:14:00 +00:00
|
|
|
spinlock_t list_lock;
|
|
|
|
wait_queue_head_t list_wait;
|
|
|
|
u16 csum_size;
|
|
|
|
struct list_head csum_list;
|
|
|
|
atomic_t cancel_req;
|
2011-03-23 15:34:19 +00:00
|
|
|
int readonly;
|
2012-03-27 18:21:27 +00:00
|
|
|
int pages_per_bio; /* <= SCRUB_PAGES_PER_BIO */
|
|
|
|
u32 sectorsize;
|
|
|
|
u32 nodesize;
|
|
|
|
u32 leafsize;
|
2012-11-05 17:29:28 +00:00
|
|
|
|
|
|
|
int is_dev_replace;
|
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
/*
|
|
|
|
* statistics
|
|
|
|
*/
|
|
|
|
struct btrfs_scrub_progress stat;
|
|
|
|
spinlock_t stat_lock;
|
|
|
|
};
|
|
|
|
|
2011-06-13 18:04:15 +00:00
|
|
|
struct scrub_fixup_nodatasum {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_device *dev;
|
2011-06-13 18:04:15 +00:00
|
|
|
u64 logical;
|
|
|
|
struct btrfs_root *root;
|
|
|
|
struct btrfs_work work;
|
|
|
|
int mirror_num;
|
|
|
|
};
|
|
|
|
|
2011-06-13 17:59:12 +00:00
|
|
|
struct scrub_warning {
|
|
|
|
struct btrfs_path *path;
|
|
|
|
u64 extent_item_size;
|
|
|
|
char *scratch_buf;
|
|
|
|
char *msg_buf;
|
|
|
|
const char *errstr;
|
|
|
|
sector_t sector;
|
|
|
|
u64 logical;
|
|
|
|
struct btrfs_device *dev;
|
|
|
|
int msg_bufsize;
|
|
|
|
int scratch_bufsize;
|
|
|
|
};
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
|
2012-11-02 15:44:58 +00:00
|
|
|
static void scrub_pending_bio_inc(struct scrub_ctx *sctx);
|
|
|
|
static void scrub_pending_bio_dec(struct scrub_ctx *sctx);
|
|
|
|
static void scrub_pending_trans_workers_inc(struct scrub_ctx *sctx);
|
|
|
|
static void scrub_pending_trans_workers_dec(struct scrub_ctx *sctx);
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_handle_errored_block(struct scrub_block *sblock_to_check);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_setup_recheck_block(struct scrub_ctx *sctx,
|
2012-11-05 14:46:42 +00:00
|
|
|
struct btrfs_fs_info *fs_info,
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 length, u64 logical,
|
|
|
|
struct scrub_block *sblock);
|
2012-11-02 15:16:26 +00:00
|
|
|
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
|
|
|
|
struct scrub_block *sblock, int is_metadata,
|
|
|
|
int have_csum, u8 *csum, u64 generation,
|
|
|
|
u16 csum_size);
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_recheck_block_checksum(struct btrfs_fs_info *fs_info,
|
|
|
|
struct scrub_block *sblock,
|
|
|
|
int is_metadata, int have_csum,
|
|
|
|
const u8 *csum, u64 generation,
|
|
|
|
u16 csum_size);
|
|
|
|
static void scrub_complete_bio_end_io(struct bio *bio, int err);
|
|
|
|
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
|
|
|
|
struct scrub_block *sblock_good,
|
|
|
|
int force_write);
|
|
|
|
static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
|
|
|
|
struct scrub_block *sblock_good,
|
|
|
|
int page_num, int force_write);
|
|
|
|
static int scrub_checksum_data(struct scrub_block *sblock);
|
|
|
|
static int scrub_checksum_tree_block(struct scrub_block *sblock);
|
|
|
|
static int scrub_checksum_super(struct scrub_block *sblock);
|
|
|
|
static void scrub_block_get(struct scrub_block *sblock);
|
|
|
|
static void scrub_block_put(struct scrub_block *sblock);
|
2012-11-02 13:58:04 +00:00
|
|
|
static void scrub_page_get(struct scrub_page *spage);
|
|
|
|
static void scrub_page_put(struct scrub_page *spage);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_add_page_to_bio(struct scrub_ctx *sctx,
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_page *spage);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
|
2012-11-02 12:26:57 +00:00
|
|
|
u64 physical, struct btrfs_device *dev, u64 flags,
|
|
|
|
u64 gen, int mirror_num, u8 *csum, int force);
|
2012-03-27 18:21:26 +00:00
|
|
|
static void scrub_bio_end_io(struct bio *bio, int err);
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_bio_end_io_worker(struct btrfs_work *work);
|
|
|
|
static void scrub_block_complete(struct scrub_block *sblock);
|
2012-03-27 18:21:26 +00:00
|
|
|
|
|
|
|
|
2012-11-02 15:44:58 +00:00
|
|
|
static void scrub_pending_bio_inc(struct scrub_ctx *sctx)
|
|
|
|
{
|
|
|
|
atomic_inc(&sctx->bios_in_flight);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_pending_bio_dec(struct scrub_ctx *sctx)
|
|
|
|
{
|
|
|
|
atomic_dec(&sctx->bios_in_flight);
|
|
|
|
wake_up(&sctx->list_wait);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* used for workers that require transaction commits (i.e., for the
|
|
|
|
* NOCOW case)
|
|
|
|
*/
|
|
|
|
static void scrub_pending_trans_workers_inc(struct scrub_ctx *sctx)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* increment scrubs_running to prevent cancel requests from
|
|
|
|
* completing as long as a worker is running. we must also
|
|
|
|
* increment scrubs_paused to prevent deadlocking on pause
|
|
|
|
* requests used for transactions commits (as the worker uses a
|
|
|
|
* transaction context). it is safe to regard the worker
|
|
|
|
* as paused for all matters practical. effectively, we only
|
|
|
|
* avoid cancellation requests from completing.
|
|
|
|
*/
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
atomic_inc(&fs_info->scrubs_running);
|
|
|
|
atomic_inc(&fs_info->scrubs_paused);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
atomic_inc(&sctx->workers_pending);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* used for workers that require transaction commits */
|
|
|
|
static void scrub_pending_trans_workers_dec(struct scrub_ctx *sctx)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* see scrub_pending_trans_workers_inc() why we're pretending
|
|
|
|
* to be paused in the scrub counters
|
|
|
|
*/
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
atomic_dec(&fs_info->scrubs_running);
|
|
|
|
atomic_dec(&fs_info->scrubs_paused);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
atomic_dec(&sctx->workers_pending);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
wake_up(&sctx->list_wait);
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static void scrub_free_csums(struct scrub_ctx *sctx)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
while (!list_empty(&sctx->csum_list)) {
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_ordered_sum *sum;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sum = list_first_entry(&sctx->csum_list,
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_ordered_sum, list);
|
|
|
|
list_del(&sum->list);
|
|
|
|
kfree(sum);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static noinline_for_stack void scrub_free_ctx(struct scrub_ctx *sctx)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (!sctx)
|
2011-03-08 13:14:00 +00:00
|
|
|
return;
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* this can happen when scrub is cancelled */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (sctx->curr != -1) {
|
|
|
|
struct scrub_bio *sbio = sctx->bios[sctx->curr];
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
for (i = 0; i < sbio->page_count; i++) {
|
|
|
|
BUG_ON(!sbio->pagev[i]);
|
|
|
|
BUG_ON(!sbio->pagev[i]->page);
|
|
|
|
scrub_block_put(sbio->pagev[i]->sblock);
|
|
|
|
}
|
|
|
|
bio_put(sbio->bio);
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
for (i = 0; i < SCRUB_BIOS_PER_CTX; ++i) {
|
|
|
|
struct scrub_bio *sbio = sctx->bios[i];
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
if (!sbio)
|
|
|
|
break;
|
|
|
|
kfree(sbio);
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_free_csums(sctx);
|
|
|
|
kfree(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static noinline_for_stack
|
2012-11-05 17:29:28 +00:00
|
|
|
struct scrub_ctx *scrub_setup_ctx(struct btrfs_device *dev, int is_dev_replace)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
int i;
|
|
|
|
struct btrfs_fs_info *fs_info = dev->dev_root->fs_info;
|
2012-03-27 18:21:27 +00:00
|
|
|
int pages_per_bio;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
pages_per_bio = min_t(int, SCRUB_PAGES_PER_BIO,
|
|
|
|
bio_get_nr_vecs(dev->bdev));
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx = kzalloc(sizeof(*sctx), GFP_NOFS);
|
|
|
|
if (!sctx)
|
2011-03-08 13:14:00 +00:00
|
|
|
goto nomem;
|
2012-11-05 17:29:28 +00:00
|
|
|
sctx->is_dev_replace = is_dev_replace;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx->pages_per_bio = pages_per_bio;
|
|
|
|
sctx->curr = -1;
|
2012-11-02 12:26:57 +00:00
|
|
|
sctx->dev_root = dev->dev_root;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
for (i = 0; i < SCRUB_BIOS_PER_CTX; ++i) {
|
2011-03-08 13:14:00 +00:00
|
|
|
struct scrub_bio *sbio;
|
|
|
|
|
|
|
|
sbio = kzalloc(sizeof(*sbio), GFP_NOFS);
|
|
|
|
if (!sbio)
|
|
|
|
goto nomem;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx->bios[i] = sbio;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
sbio->index = i;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sbio->sctx = sctx;
|
2012-03-27 18:21:27 +00:00
|
|
|
sbio->page_count = 0;
|
|
|
|
sbio->work.func = scrub_bio_end_io_worker;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (i != SCRUB_BIOS_PER_CTX - 1)
|
|
|
|
sctx->bios[i]->next_free = i + 1;
|
2011-06-13 18:04:15 +00:00
|
|
|
else
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx->bios[i]->next_free = -1;
|
|
|
|
}
|
|
|
|
sctx->first_free = 0;
|
|
|
|
sctx->nodesize = dev->dev_root->nodesize;
|
|
|
|
sctx->leafsize = dev->dev_root->leafsize;
|
|
|
|
sctx->sectorsize = dev->dev_root->sectorsize;
|
2012-11-02 15:44:58 +00:00
|
|
|
atomic_set(&sctx->bios_in_flight, 0);
|
|
|
|
atomic_set(&sctx->workers_pending, 0);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
atomic_set(&sctx->cancel_req, 0);
|
|
|
|
sctx->csum_size = btrfs_super_csum_size(fs_info->super_copy);
|
|
|
|
INIT_LIST_HEAD(&sctx->csum_list);
|
|
|
|
|
|
|
|
spin_lock_init(&sctx->list_lock);
|
|
|
|
spin_lock_init(&sctx->stat_lock);
|
|
|
|
init_waitqueue_head(&sctx->list_wait);
|
|
|
|
return sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
nomem:
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_free_ctx(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
2011-06-13 17:59:12 +00:00
|
|
|
static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root, void *ctx)
|
|
|
|
{
|
|
|
|
u64 isize;
|
|
|
|
u32 nlink;
|
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
struct extent_buffer *eb;
|
|
|
|
struct btrfs_inode_item *inode_item;
|
|
|
|
struct scrub_warning *swarn = ctx;
|
|
|
|
struct btrfs_fs_info *fs_info = swarn->dev->dev_root->fs_info;
|
|
|
|
struct inode_fs_paths *ipath = NULL;
|
|
|
|
struct btrfs_root *local_root;
|
|
|
|
struct btrfs_key root_key;
|
|
|
|
|
|
|
|
root_key.objectid = root;
|
|
|
|
root_key.type = BTRFS_ROOT_ITEM_KEY;
|
|
|
|
root_key.offset = (u64)-1;
|
|
|
|
local_root = btrfs_read_fs_root_no_name(fs_info, &root_key);
|
|
|
|
if (IS_ERR(local_root)) {
|
|
|
|
ret = PTR_ERR(local_root);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = inode_item_info(inum, 0, local_root, swarn->path);
|
|
|
|
if (ret) {
|
|
|
|
btrfs_release_path(swarn->path);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
eb = swarn->path->nodes[0];
|
|
|
|
inode_item = btrfs_item_ptr(eb, swarn->path->slots[0],
|
|
|
|
struct btrfs_inode_item);
|
|
|
|
isize = btrfs_inode_size(eb, inode_item);
|
|
|
|
nlink = btrfs_inode_nlink(eb, inode_item);
|
|
|
|
btrfs_release_path(swarn->path);
|
|
|
|
|
|
|
|
ipath = init_ipath(4096, local_root, swarn->path);
|
2011-11-16 08:28:01 +00:00
|
|
|
if (IS_ERR(ipath)) {
|
|
|
|
ret = PTR_ERR(ipath);
|
|
|
|
ipath = NULL;
|
|
|
|
goto err;
|
|
|
|
}
|
2011-06-13 17:59:12 +00:00
|
|
|
ret = paths_from_inode(inum, ipath);
|
|
|
|
|
|
|
|
if (ret < 0)
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* we deliberately ignore the bit ipath might have been too small to
|
|
|
|
* hold all of the paths here
|
|
|
|
*/
|
|
|
|
for (i = 0; i < ipath->fspath->elem_cnt; ++i)
|
2012-06-04 18:03:51 +00:00
|
|
|
printk_in_rcu(KERN_WARNING "btrfs: %s at logical %llu on dev "
|
2011-06-13 17:59:12 +00:00
|
|
|
"%s, sector %llu, root %llu, inode %llu, offset %llu, "
|
|
|
|
"length %llu, links %u (path: %s)\n", swarn->errstr,
|
2012-06-04 18:03:51 +00:00
|
|
|
swarn->logical, rcu_str_deref(swarn->dev->name),
|
2011-06-13 17:59:12 +00:00
|
|
|
(unsigned long long)swarn->sector, root, inum, offset,
|
|
|
|
min(isize - offset, (u64)PAGE_SIZE), nlink,
|
2011-11-20 12:31:57 +00:00
|
|
|
(char *)(unsigned long)ipath->fspath->val[i]);
|
2011-06-13 17:59:12 +00:00
|
|
|
|
|
|
|
free_ipath(ipath);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err:
|
2012-06-04 18:03:51 +00:00
|
|
|
printk_in_rcu(KERN_WARNING "btrfs: %s at logical %llu on dev "
|
2011-06-13 17:59:12 +00:00
|
|
|
"%s, sector %llu, root %llu, inode %llu, offset %llu: path "
|
|
|
|
"resolving failed with ret=%d\n", swarn->errstr,
|
2012-06-04 18:03:51 +00:00
|
|
|
swarn->logical, rcu_str_deref(swarn->dev->name),
|
2011-06-13 17:59:12 +00:00
|
|
|
(unsigned long long)swarn->sector, root, inum, offset, ret);
|
|
|
|
|
|
|
|
free_ipath(ipath);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_print_warning(const char *errstr, struct scrub_block *sblock)
|
2011-06-13 17:59:12 +00:00
|
|
|
{
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_device *dev;
|
|
|
|
struct btrfs_fs_info *fs_info;
|
2011-06-13 17:59:12 +00:00
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
struct extent_buffer *eb;
|
|
|
|
struct btrfs_extent_item *ei;
|
|
|
|
struct scrub_warning swarn;
|
2012-09-08 02:01:28 +00:00
|
|
|
unsigned long ptr = 0;
|
|
|
|
u64 extent_item_pos;
|
|
|
|
u64 flags = 0;
|
2011-06-13 17:59:12 +00:00
|
|
|
u64 ref_root;
|
2012-09-08 02:01:28 +00:00
|
|
|
u32 item_size;
|
2011-06-13 17:59:12 +00:00
|
|
|
u8 ref_level;
|
|
|
|
const int bufsize = 4096;
|
2012-09-08 02:01:28 +00:00
|
|
|
int ret;
|
2011-06-13 17:59:12 +00:00
|
|
|
|
2012-11-02 12:26:57 +00:00
|
|
|
WARN_ON(sblock->page_count < 1);
|
2012-11-02 13:58:04 +00:00
|
|
|
dev = sblock->pagev[0]->dev;
|
2012-11-02 12:26:57 +00:00
|
|
|
fs_info = sblock->sctx->dev_root->fs_info;
|
|
|
|
|
2011-06-13 17:59:12 +00:00
|
|
|
path = btrfs_alloc_path();
|
|
|
|
|
|
|
|
swarn.scratch_buf = kmalloc(bufsize, GFP_NOFS);
|
|
|
|
swarn.msg_buf = kmalloc(bufsize, GFP_NOFS);
|
2012-11-02 13:58:04 +00:00
|
|
|
swarn.sector = (sblock->pagev[0]->physical) >> 9;
|
|
|
|
swarn.logical = sblock->pagev[0]->logical;
|
2011-06-13 17:59:12 +00:00
|
|
|
swarn.errstr = errstr;
|
2012-11-02 12:26:57 +00:00
|
|
|
swarn.dev = NULL;
|
2011-06-13 17:59:12 +00:00
|
|
|
swarn.msg_bufsize = bufsize;
|
|
|
|
swarn.scratch_bufsize = bufsize;
|
|
|
|
|
|
|
|
if (!path || !swarn.scratch_buf || !swarn.msg_buf)
|
|
|
|
goto out;
|
|
|
|
|
2012-09-08 02:01:28 +00:00
|
|
|
ret = extent_from_logical(fs_info, swarn.logical, path, &found_key,
|
|
|
|
&flags);
|
2011-06-13 17:59:12 +00:00
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
2011-12-02 13:56:41 +00:00
|
|
|
extent_item_pos = swarn.logical - found_key.objectid;
|
2011-06-13 17:59:12 +00:00
|
|
|
swarn.extent_item_size = found_key.offset;
|
|
|
|
|
|
|
|
eb = path->nodes[0];
|
|
|
|
ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
|
|
|
|
item_size = btrfs_item_size_nr(eb, path->slots[0]);
|
2011-12-02 13:56:41 +00:00
|
|
|
btrfs_release_path(path);
|
2011-06-13 17:59:12 +00:00
|
|
|
|
2012-09-08 02:01:28 +00:00
|
|
|
if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
|
2011-06-13 17:59:12 +00:00
|
|
|
do {
|
|
|
|
ret = tree_backref_for_extent(&ptr, eb, ei, item_size,
|
|
|
|
&ref_root, &ref_level);
|
2012-06-04 18:03:51 +00:00
|
|
|
printk_in_rcu(KERN_WARNING
|
2012-03-27 18:21:26 +00:00
|
|
|
"btrfs: %s at logical %llu on dev %s, "
|
2011-06-13 17:59:12 +00:00
|
|
|
"sector %llu: metadata %s (level %d) in tree "
|
2012-06-04 18:03:51 +00:00
|
|
|
"%llu\n", errstr, swarn.logical,
|
|
|
|
rcu_str_deref(dev->name),
|
2011-06-13 17:59:12 +00:00
|
|
|
(unsigned long long)swarn.sector,
|
|
|
|
ref_level ? "node" : "leaf",
|
|
|
|
ret < 0 ? -1 : ref_level,
|
|
|
|
ret < 0 ? -1 : ref_root);
|
|
|
|
} while (ret != 1);
|
|
|
|
} else {
|
|
|
|
swarn.path = path;
|
2012-11-02 12:26:57 +00:00
|
|
|
swarn.dev = dev;
|
2012-03-23 16:32:28 +00:00
|
|
|
iterate_extent_inodes(fs_info, found_key.objectid,
|
|
|
|
extent_item_pos, 1,
|
2011-06-13 17:59:12 +00:00
|
|
|
scrub_print_warning_inode, &swarn);
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
kfree(swarn.scratch_buf);
|
|
|
|
kfree(swarn.msg_buf);
|
|
|
|
}
|
|
|
|
|
2011-06-13 18:04:15 +00:00
|
|
|
static int scrub_fixup_readpage(u64 inum, u64 offset, u64 root, void *ctx)
|
|
|
|
{
|
2011-08-04 16:11:04 +00:00
|
|
|
struct page *page = NULL;
|
2011-06-13 18:04:15 +00:00
|
|
|
unsigned long index;
|
|
|
|
struct scrub_fixup_nodatasum *fixup = ctx;
|
|
|
|
int ret;
|
2011-08-04 16:11:04 +00:00
|
|
|
int corrected = 0;
|
2011-06-13 18:04:15 +00:00
|
|
|
struct btrfs_key key;
|
2011-08-04 16:11:04 +00:00
|
|
|
struct inode *inode = NULL;
|
2011-06-13 18:04:15 +00:00
|
|
|
u64 end = offset + PAGE_SIZE - 1;
|
|
|
|
struct btrfs_root *local_root;
|
|
|
|
|
|
|
|
key.objectid = root;
|
|
|
|
key.type = BTRFS_ROOT_ITEM_KEY;
|
|
|
|
key.offset = (u64)-1;
|
|
|
|
local_root = btrfs_read_fs_root_no_name(fixup->root->fs_info, &key);
|
|
|
|
if (IS_ERR(local_root))
|
|
|
|
return PTR_ERR(local_root);
|
|
|
|
|
|
|
|
key.type = BTRFS_INODE_ITEM_KEY;
|
|
|
|
key.objectid = inum;
|
|
|
|
key.offset = 0;
|
|
|
|
inode = btrfs_iget(fixup->root->fs_info->sb, &key, local_root, NULL);
|
|
|
|
if (IS_ERR(inode))
|
|
|
|
return PTR_ERR(inode);
|
|
|
|
|
|
|
|
index = offset >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
|
2011-08-04 16:11:04 +00:00
|
|
|
if (!page) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (PageUptodate(page)) {
|
2012-11-05 14:46:42 +00:00
|
|
|
struct btrfs_fs_info *fs_info;
|
2011-08-04 16:11:04 +00:00
|
|
|
if (PageDirty(page)) {
|
|
|
|
/*
|
|
|
|
* we need to write the data to the defect sector. the
|
|
|
|
* data that was in that sector is not in memory,
|
|
|
|
* because the page was modified. we must not write the
|
|
|
|
* modified page to that sector.
|
|
|
|
*
|
|
|
|
* TODO: what could be done here: wait for the delalloc
|
|
|
|
* runner to write out that page (might involve
|
|
|
|
* COW) and see whether the sector is still
|
|
|
|
* referenced afterwards.
|
|
|
|
*
|
|
|
|
* For the meantime, we'll treat this error
|
|
|
|
* incorrectable, although there is a chance that a
|
|
|
|
* later scrub will find the bad sector again and that
|
|
|
|
* there's no dirty page in memory, then.
|
|
|
|
*/
|
|
|
|
ret = -EIO;
|
|
|
|
goto out;
|
|
|
|
}
|
2012-11-05 14:46:42 +00:00
|
|
|
fs_info = BTRFS_I(inode)->root->fs_info;
|
|
|
|
ret = repair_io_failure(fs_info, offset, PAGE_SIZE,
|
2011-08-04 16:11:04 +00:00
|
|
|
fixup->logical, page,
|
|
|
|
fixup->mirror_num);
|
|
|
|
unlock_page(page);
|
|
|
|
corrected = !ret;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* we need to get good data first. the general readpage path
|
|
|
|
* will call repair_io_failure for us, we just have to make
|
|
|
|
* sure we read the bad mirror.
|
|
|
|
*/
|
|
|
|
ret = set_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
|
|
|
|
EXTENT_DAMAGED, GFP_NOFS);
|
|
|
|
if (ret) {
|
|
|
|
/* set_extent_bits should give proper error */
|
|
|
|
WARN_ON(ret > 0);
|
|
|
|
if (ret > 0)
|
|
|
|
ret = -EFAULT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = extent_read_full_page(&BTRFS_I(inode)->io_tree, page,
|
|
|
|
btrfs_get_extent,
|
|
|
|
fixup->mirror_num);
|
|
|
|
wait_on_page_locked(page);
|
|
|
|
|
|
|
|
corrected = !test_range_bit(&BTRFS_I(inode)->io_tree, offset,
|
|
|
|
end, EXTENT_DAMAGED, 0, NULL);
|
|
|
|
if (!corrected)
|
|
|
|
clear_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
|
|
|
|
EXTENT_DAMAGED, GFP_NOFS);
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
if (page)
|
|
|
|
put_page(page);
|
|
|
|
if (inode)
|
|
|
|
iput(inode);
|
2011-06-13 18:04:15 +00:00
|
|
|
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
if (ret == 0 && corrected) {
|
|
|
|
/*
|
|
|
|
* we only need to call readpage for one of the inodes belonging
|
|
|
|
* to this extent. so make iterate_extent_inodes stop
|
|
|
|
*/
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_fixup_nodatasum(struct btrfs_work *work)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct scrub_fixup_nodatasum *fixup;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2011-06-13 18:04:15 +00:00
|
|
|
struct btrfs_trans_handle *trans = NULL;
|
|
|
|
struct btrfs_fs_info *fs_info;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
int uncorrectable = 0;
|
|
|
|
|
|
|
|
fixup = container_of(work, struct scrub_fixup_nodatasum, work);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx = fixup->sctx;
|
2011-06-13 18:04:15 +00:00
|
|
|
fs_info = fixup->root->fs_info;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
++sctx->stat.malloc_errors;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2011-06-13 18:04:15 +00:00
|
|
|
uncorrectable = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
trans = btrfs_join_transaction(fixup->root);
|
|
|
|
if (IS_ERR(trans)) {
|
|
|
|
uncorrectable = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the idea is to trigger a regular read through the standard path. we
|
|
|
|
* read a page from the (failed) logical address by specifying the
|
|
|
|
* corresponding copynum of the failed sector. thus, that readpage is
|
|
|
|
* expected to fail.
|
|
|
|
* that is the point where on-the-fly error correction will kick in
|
|
|
|
* (once it's finished) and rewrite the failed sector if a good copy
|
|
|
|
* can be found.
|
|
|
|
*/
|
|
|
|
ret = iterate_inodes_from_logical(fixup->logical, fixup->root->fs_info,
|
|
|
|
path, scrub_fixup_readpage,
|
|
|
|
fixup);
|
|
|
|
if (ret < 0) {
|
|
|
|
uncorrectable = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
WARN_ON(ret != 1);
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
++sctx->stat.corrected_errors;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2011-06-13 18:04:15 +00:00
|
|
|
|
|
|
|
out:
|
|
|
|
if (trans && !IS_ERR(trans))
|
|
|
|
btrfs_end_transaction(trans, fixup->root);
|
|
|
|
if (uncorrectable) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
++sctx->stat.uncorrectable_errors;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-06-04 18:03:51 +00:00
|
|
|
|
|
|
|
printk_ratelimited_in_rcu(KERN_ERR
|
2012-03-27 18:21:27 +00:00
|
|
|
"btrfs: unable to fixup (nodatasum) error at logical %llu on dev %s\n",
|
2012-06-04 18:03:51 +00:00
|
|
|
(unsigned long long)fixup->logical,
|
2012-11-02 12:26:57 +00:00
|
|
|
rcu_str_deref(fixup->dev->name));
|
2011-06-13 18:04:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_free_path(path);
|
|
|
|
kfree(fixup);
|
|
|
|
|
2012-11-02 15:44:58 +00:00
|
|
|
scrub_pending_trans_workers_dec(sctx);
|
2011-06-13 18:04:15 +00:00
|
|
|
}
|
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
/*
|
2012-03-27 18:21:27 +00:00
|
|
|
* scrub_handle_errored_block gets called when either verification of the
|
|
|
|
* pages failed or the bio failed to read, e.g. with EIO. In the latter
|
|
|
|
* case, this function handles all pages in the bio, even though only one
|
|
|
|
* may be bad.
|
|
|
|
* The goal of this function is to repair the errored block by using the
|
|
|
|
* contents of one of the mirrors.
|
2011-03-08 13:14:00 +00:00
|
|
|
*/
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_handle_errored_block(struct scrub_block *sblock_to_check)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = sblock_to_check->sctx;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_device *dev;
|
2012-03-27 18:21:27 +00:00
|
|
|
struct btrfs_fs_info *fs_info;
|
|
|
|
u64 length;
|
|
|
|
u64 logical;
|
|
|
|
u64 generation;
|
|
|
|
unsigned int failed_mirror_index;
|
|
|
|
unsigned int is_metadata;
|
|
|
|
unsigned int have_csum;
|
|
|
|
u8 *csum;
|
|
|
|
struct scrub_block *sblocks_for_recheck; /* holds one for each mirror */
|
|
|
|
struct scrub_block *sblock_bad;
|
|
|
|
int ret;
|
|
|
|
int mirror_index;
|
|
|
|
int page_num;
|
|
|
|
int success;
|
2011-06-13 17:59:12 +00:00
|
|
|
static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
|
2012-03-27 18:21:27 +00:00
|
|
|
DEFAULT_RATELIMIT_BURST);
|
|
|
|
|
|
|
|
BUG_ON(sblock_to_check->page_count < 1);
|
2012-11-02 12:26:57 +00:00
|
|
|
fs_info = sctx->dev_root->fs_info;
|
2012-03-27 18:21:27 +00:00
|
|
|
length = sblock_to_check->page_count * PAGE_SIZE;
|
2012-11-02 13:58:04 +00:00
|
|
|
logical = sblock_to_check->pagev[0]->logical;
|
|
|
|
generation = sblock_to_check->pagev[0]->generation;
|
|
|
|
BUG_ON(sblock_to_check->pagev[0]->mirror_num < 1);
|
|
|
|
failed_mirror_index = sblock_to_check->pagev[0]->mirror_num - 1;
|
|
|
|
is_metadata = !(sblock_to_check->pagev[0]->flags &
|
2012-03-27 18:21:27 +00:00
|
|
|
BTRFS_EXTENT_FLAG_DATA);
|
2012-11-02 13:58:04 +00:00
|
|
|
have_csum = sblock_to_check->pagev[0]->have_csum;
|
|
|
|
csum = sblock_to_check->pagev[0]->csum;
|
|
|
|
dev = sblock_to_check->pagev[0]->dev;
|
2011-06-13 17:56:13 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* read all mirrors one after the other. This includes to
|
|
|
|
* re-read the extent or metadata block that failed (that was
|
|
|
|
* the cause that this fixup code is called) another time,
|
|
|
|
* page by page this time in order to know which pages
|
|
|
|
* caused I/O errors and which ones are good (for all mirrors).
|
|
|
|
* It is the goal to handle the situation when more than one
|
|
|
|
* mirror contains I/O errors, but the errors do not
|
|
|
|
* overlap, i.e. the data can be repaired by selecting the
|
|
|
|
* pages from those mirrors without I/O error on the
|
|
|
|
* particular pages. One example (with blocks >= 2 * PAGE_SIZE)
|
|
|
|
* would be that mirror #1 has an I/O error on the first page,
|
|
|
|
* the second page is good, and mirror #2 has an I/O error on
|
|
|
|
* the second page, but the first page is good.
|
|
|
|
* Then the first page of the first mirror can be repaired by
|
|
|
|
* taking the first page of the second mirror, and the
|
|
|
|
* second page of the second mirror can be repaired by
|
|
|
|
* copying the contents of the 2nd page of the 1st mirror.
|
|
|
|
* One more note: if the pages of one mirror contain I/O
|
|
|
|
* errors, the checksum cannot be verified. In order to get
|
|
|
|
* the best data for repairing, the first attempt is to find
|
|
|
|
* a mirror without I/O errors and with a validated checksum.
|
|
|
|
* Only if this is not possible, the pages are picked from
|
|
|
|
* mirrors with I/O errors without considering the checksum.
|
|
|
|
* If the latter is the case, at the end, the checksum of the
|
|
|
|
* repaired area is verified in order to correctly maintain
|
|
|
|
* the statistics.
|
|
|
|
*/
|
|
|
|
|
|
|
|
sblocks_for_recheck = kzalloc(BTRFS_MAX_MIRRORS *
|
|
|
|
sizeof(*sblocks_for_recheck),
|
|
|
|
GFP_NOFS);
|
|
|
|
if (!sblocks_for_recheck) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.malloc_errors++;
|
|
|
|
sctx->stat.read_errors++;
|
|
|
|
sctx->stat.uncorrectable_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
|
2012-03-27 18:21:27 +00:00
|
|
|
goto out;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* setup the context, map the logical blocks and alloc the pages */
|
2012-11-05 14:46:42 +00:00
|
|
|
ret = scrub_setup_recheck_block(sctx, fs_info, length,
|
2012-03-27 18:21:27 +00:00
|
|
|
logical, sblocks_for_recheck);
|
|
|
|
if (ret) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.read_errors++;
|
|
|
|
sctx->stat.uncorrectable_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
|
2012-03-27 18:21:27 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
BUG_ON(failed_mirror_index >= BTRFS_MAX_MIRRORS);
|
|
|
|
sblock_bad = sblocks_for_recheck + failed_mirror_index;
|
2011-06-13 17:56:13 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* build and submit the bios for the failed mirror, check checksums */
|
2012-11-02 15:16:26 +00:00
|
|
|
scrub_recheck_block(fs_info, sblock_bad, is_metadata, have_csum,
|
|
|
|
csum, generation, sctx->csum_size);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (!sblock_bad->header_error && !sblock_bad->checksum_error &&
|
|
|
|
sblock_bad->no_io_error_seen) {
|
|
|
|
/*
|
|
|
|
* the error disappeared after reading page by page, or
|
|
|
|
* the area was part of a huge bio and other parts of the
|
|
|
|
* bio caused I/O errors, or the block layer merged several
|
|
|
|
* read requests into one and the error is caused by a
|
|
|
|
* different bio (usually one of the two latter cases is
|
|
|
|
* the cause)
|
|
|
|
*/
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.unverified_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
goto out;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (!sblock_bad->no_io_error_seen) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.read_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (__ratelimit(&_rs))
|
|
|
|
scrub_print_warning("i/o error", sblock_to_check);
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
|
2012-03-27 18:21:27 +00:00
|
|
|
} else if (sblock_bad->checksum_error) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.csum_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (__ratelimit(&_rs))
|
|
|
|
scrub_print_warning("checksum error", sblock_to_check);
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_DEV_STAT_CORRUPTION_ERRS);
|
2012-03-27 18:21:27 +00:00
|
|
|
} else if (sblock_bad->header_error) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.verify_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (__ratelimit(&_rs))
|
|
|
|
scrub_print_warning("checksum/header error",
|
|
|
|
sblock_to_check);
|
2012-05-25 14:06:08 +00:00
|
|
|
if (sblock_bad->generation_error)
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_DEV_STAT_GENERATION_ERRS);
|
|
|
|
else
|
2012-11-02 12:26:57 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(dev,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_DEV_STAT_CORRUPTION_ERRS);
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (sctx->readonly)
|
2012-03-27 18:21:27 +00:00
|
|
|
goto did_not_correct_error;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (!is_metadata && !have_csum) {
|
|
|
|
struct scrub_fixup_nodatasum *fixup_nodatasum;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* !is_metadata and !have_csum, this means that the data
|
|
|
|
* might not be COW'ed, that it might be modified
|
|
|
|
* concurrently. The general strategy to work on the
|
|
|
|
* commit root does not help in the case when COW is not
|
|
|
|
* used.
|
|
|
|
*/
|
|
|
|
fixup_nodatasum = kzalloc(sizeof(*fixup_nodatasum), GFP_NOFS);
|
|
|
|
if (!fixup_nodatasum)
|
|
|
|
goto did_not_correct_error;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
fixup_nodatasum->sctx = sctx;
|
2012-11-02 12:26:57 +00:00
|
|
|
fixup_nodatasum->dev = dev;
|
2012-03-27 18:21:27 +00:00
|
|
|
fixup_nodatasum->logical = logical;
|
|
|
|
fixup_nodatasum->root = fs_info->extent_root;
|
|
|
|
fixup_nodatasum->mirror_num = failed_mirror_index + 1;
|
2012-11-02 15:44:58 +00:00
|
|
|
scrub_pending_trans_workers_inc(sctx);
|
2012-03-27 18:21:27 +00:00
|
|
|
fixup_nodatasum->work.func = scrub_fixup_nodatasum;
|
|
|
|
btrfs_queue_worker(&fs_info->scrub_workers,
|
|
|
|
&fixup_nodatasum->work);
|
|
|
|
goto out;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* now build and submit the bios for the other mirrors, check
|
2012-11-02 15:14:21 +00:00
|
|
|
* checksums.
|
|
|
|
* First try to pick the mirror which is completely without I/O
|
2012-03-27 18:21:27 +00:00
|
|
|
* errors and also does not have a checksum error.
|
|
|
|
* If one is found, and if a checksum is present, the full block
|
|
|
|
* that is known to contain an error is rewritten. Afterwards
|
|
|
|
* the block is known to be corrected.
|
|
|
|
* If a mirror is found which is completely correct, and no
|
|
|
|
* checksum is present, only those pages are rewritten that had
|
|
|
|
* an I/O error in the block to be repaired, since it cannot be
|
|
|
|
* determined, which copy of the other pages is better (and it
|
|
|
|
* could happen otherwise that a correct page would be
|
|
|
|
* overwritten by a bad one).
|
|
|
|
*/
|
|
|
|
for (mirror_index = 0;
|
|
|
|
mirror_index < BTRFS_MAX_MIRRORS &&
|
|
|
|
sblocks_for_recheck[mirror_index].page_count > 0;
|
|
|
|
mirror_index++) {
|
2012-11-02 15:14:21 +00:00
|
|
|
struct scrub_block *sblock_other;
|
2012-03-27 18:21:27 +00:00
|
|
|
|
2012-11-02 15:14:21 +00:00
|
|
|
if (mirror_index == failed_mirror_index)
|
|
|
|
continue;
|
|
|
|
sblock_other = sblocks_for_recheck + mirror_index;
|
|
|
|
|
|
|
|
/* build and submit the bios, check checksums */
|
2012-11-02 15:16:26 +00:00
|
|
|
scrub_recheck_block(fs_info, sblock_other, is_metadata,
|
|
|
|
have_csum, csum, generation,
|
|
|
|
sctx->csum_size);
|
|
|
|
|
|
|
|
if (!sblock_other->header_error &&
|
2012-03-27 18:21:27 +00:00
|
|
|
!sblock_other->checksum_error &&
|
|
|
|
sblock_other->no_io_error_seen) {
|
|
|
|
int force_write = is_metadata || have_csum;
|
|
|
|
|
|
|
|
ret = scrub_repair_block_from_good_copy(sblock_bad,
|
|
|
|
sblock_other,
|
|
|
|
force_write);
|
|
|
|
if (0 == ret)
|
|
|
|
goto corrected_error;
|
|
|
|
}
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
/*
|
2012-03-27 18:21:27 +00:00
|
|
|
* in case of I/O errors in the area that is supposed to be
|
|
|
|
* repaired, continue by picking good copies of those pages.
|
|
|
|
* Select the good pages from mirrors to rewrite bad pages from
|
|
|
|
* the area to fix. Afterwards verify the checksum of the block
|
|
|
|
* that is supposed to be repaired. This verification step is
|
|
|
|
* only done for the purpose of statistic counting and for the
|
|
|
|
* final scrub report, whether errors remain.
|
|
|
|
* A perfect algorithm could make use of the checksum and try
|
|
|
|
* all possible combinations of pages from the different mirrors
|
|
|
|
* until the checksum verification succeeds. For example, when
|
|
|
|
* the 2nd page of mirror #1 faces I/O errors, and the 2nd page
|
|
|
|
* of mirror #2 is readable but the final checksum test fails,
|
|
|
|
* then the 2nd page of mirror #3 could be tried, whether now
|
|
|
|
* the final checksum succeedes. But this would be a rare
|
|
|
|
* exception and is therefore not implemented. At least it is
|
|
|
|
* avoided that the good copy is overwritten.
|
|
|
|
* A more useful improvement would be to pick the sectors
|
|
|
|
* without I/O error based on sector sizes (512 bytes on legacy
|
|
|
|
* disks) instead of on PAGE_SIZE. Then maybe 512 byte of one
|
|
|
|
* mirror could be repaired by taking 512 byte of a different
|
|
|
|
* mirror, even if other 512 byte sectors in the same PAGE_SIZE
|
|
|
|
* area are unreadable.
|
2011-03-08 13:14:00 +00:00
|
|
|
*/
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* can only fix I/O errors from here on */
|
|
|
|
if (sblock_bad->no_io_error_seen)
|
|
|
|
goto did_not_correct_error;
|
|
|
|
|
|
|
|
success = 1;
|
|
|
|
for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *page_bad = sblock_bad->pagev[page_num];
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
if (!page_bad->io_error)
|
2011-03-08 13:14:00 +00:00
|
|
|
continue;
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
for (mirror_index = 0;
|
|
|
|
mirror_index < BTRFS_MAX_MIRRORS &&
|
|
|
|
sblocks_for_recheck[mirror_index].page_count > 0;
|
|
|
|
mirror_index++) {
|
|
|
|
struct scrub_block *sblock_other = sblocks_for_recheck +
|
|
|
|
mirror_index;
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *page_other = sblock_other->pagev[
|
|
|
|
page_num];
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
if (!page_other->io_error) {
|
|
|
|
ret = scrub_repair_page_from_good_copy(
|
|
|
|
sblock_bad, sblock_other, page_num, 0);
|
|
|
|
if (0 == ret) {
|
|
|
|
page_bad->io_error = 0;
|
|
|
|
break; /* succeeded for this page */
|
|
|
|
}
|
|
|
|
}
|
2011-04-09 11:27:01 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (page_bad->io_error) {
|
|
|
|
/* did not find a mirror to copy the page from */
|
|
|
|
success = 0;
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (success) {
|
|
|
|
if (is_metadata || have_csum) {
|
|
|
|
/*
|
|
|
|
* need to verify the checksum now that all
|
|
|
|
* sectors on disk are repaired (the write
|
|
|
|
* request for data to be repaired is on its way).
|
|
|
|
* Just be lazy and use scrub_recheck_block()
|
|
|
|
* which re-reads the data before the checksum
|
|
|
|
* is verified, but most likely the data comes out
|
|
|
|
* of the page cache.
|
|
|
|
*/
|
2012-11-02 15:16:26 +00:00
|
|
|
scrub_recheck_block(fs_info, sblock_bad,
|
|
|
|
is_metadata, have_csum, csum,
|
|
|
|
generation, sctx->csum_size);
|
|
|
|
if (!sblock_bad->header_error &&
|
2012-03-27 18:21:27 +00:00
|
|
|
!sblock_bad->checksum_error &&
|
|
|
|
sblock_bad->no_io_error_seen)
|
|
|
|
goto corrected_error;
|
|
|
|
else
|
|
|
|
goto did_not_correct_error;
|
|
|
|
} else {
|
|
|
|
corrected_error:
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.corrected_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-06-04 18:03:51 +00:00
|
|
|
printk_ratelimited_in_rcu(KERN_ERR
|
2012-03-27 18:21:27 +00:00
|
|
|
"btrfs: fixed up error at logical %llu on dev %s\n",
|
2012-06-04 18:03:51 +00:00
|
|
|
(unsigned long long)logical,
|
2012-11-02 12:26:57 +00:00
|
|
|
rcu_str_deref(dev->name));
|
2011-03-23 15:34:19 +00:00
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
} else {
|
|
|
|
did_not_correct_error:
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.uncorrectable_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-06-04 18:03:51 +00:00
|
|
|
printk_ratelimited_in_rcu(KERN_ERR
|
2012-03-27 18:21:27 +00:00
|
|
|
"btrfs: unable to fixup (regular) error at logical %llu on dev %s\n",
|
2012-06-04 18:03:51 +00:00
|
|
|
(unsigned long long)logical,
|
2012-11-02 12:26:57 +00:00
|
|
|
rcu_str_deref(dev->name));
|
2011-04-09 11:27:01 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
out:
|
|
|
|
if (sblocks_for_recheck) {
|
|
|
|
for (mirror_index = 0; mirror_index < BTRFS_MAX_MIRRORS;
|
|
|
|
mirror_index++) {
|
|
|
|
struct scrub_block *sblock = sblocks_for_recheck +
|
|
|
|
mirror_index;
|
|
|
|
int page_index;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
for (page_index = 0; page_index < sblock->page_count;
|
|
|
|
page_index++) {
|
|
|
|
sblock->pagev[page_index]->sblock = NULL;
|
|
|
|
scrub_page_put(sblock->pagev[page_index]);
|
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
|
|
|
kfree(sblocks_for_recheck);
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_setup_recheck_block(struct scrub_ctx *sctx,
|
2012-11-05 14:46:42 +00:00
|
|
|
struct btrfs_fs_info *fs_info,
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 length, u64 logical,
|
|
|
|
struct scrub_block *sblocks_for_recheck)
|
|
|
|
{
|
|
|
|
int page_index;
|
|
|
|
int mirror_index;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
2012-11-02 13:58:04 +00:00
|
|
|
* note: the two members ref_count and outstanding_pages
|
2012-03-27 18:21:27 +00:00
|
|
|
* are not used (and not set) in the blocks that are used for
|
|
|
|
* the recheck procedure
|
|
|
|
*/
|
|
|
|
|
|
|
|
page_index = 0;
|
|
|
|
while (length > 0) {
|
|
|
|
u64 sublen = min_t(u64, length, PAGE_SIZE);
|
|
|
|
u64 mapped_length = sublen;
|
|
|
|
struct btrfs_bio *bbio = NULL;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* with a length of PAGE_SIZE, each returned stripe
|
|
|
|
* represents one mirror
|
|
|
|
*/
|
2012-11-05 14:46:42 +00:00
|
|
|
ret = btrfs_map_block(fs_info, WRITE, logical, &mapped_length,
|
2012-03-27 18:21:27 +00:00
|
|
|
&bbio, 0);
|
|
|
|
if (ret || !bbio || mapped_length < sublen) {
|
|
|
|
kfree(bbio);
|
|
|
|
return -EIO;
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
BUG_ON(page_index >= SCRUB_PAGES_PER_BIO);
|
|
|
|
for (mirror_index = 0; mirror_index < (int)bbio->num_stripes;
|
|
|
|
mirror_index++) {
|
|
|
|
struct scrub_block *sblock;
|
|
|
|
struct scrub_page *page;
|
|
|
|
|
|
|
|
if (mirror_index >= BTRFS_MAX_MIRRORS)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
sblock = sblocks_for_recheck + mirror_index;
|
2012-11-02 13:58:04 +00:00
|
|
|
sblock->sctx = sctx;
|
|
|
|
page = kzalloc(sizeof(*page), GFP_NOFS);
|
|
|
|
if (!page) {
|
|
|
|
leave_nomem:
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.malloc_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-09-02 13:44:51 +00:00
|
|
|
kfree(bbio);
|
2012-03-27 18:21:27 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2012-11-02 13:58:04 +00:00
|
|
|
scrub_page_get(page);
|
|
|
|
sblock->pagev[page_index] = page;
|
|
|
|
page->logical = logical;
|
|
|
|
page->physical = bbio->stripes[mirror_index].physical;
|
|
|
|
/* for missing devices, dev->bdev is NULL */
|
|
|
|
page->dev = bbio->stripes[mirror_index].dev;
|
|
|
|
page->mirror_num = mirror_index + 1;
|
2012-03-27 18:21:27 +00:00
|
|
|
sblock->page_count++;
|
2012-11-02 13:58:04 +00:00
|
|
|
page->page = alloc_page(GFP_NOFS);
|
|
|
|
if (!page->page)
|
|
|
|
goto leave_nomem;
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
|
|
|
kfree(bbio);
|
|
|
|
length -= sublen;
|
|
|
|
logical += sublen;
|
|
|
|
page_index++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
2011-04-09 11:27:01 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* this function will check the on disk data for checksum errors, header
|
|
|
|
* errors and read I/O errors. If any I/O errors happen, the exact pages
|
|
|
|
* which are errored are marked as being bad. The goal is to enable scrub
|
|
|
|
* to take those pages that are not errored from all the mirrors so that
|
|
|
|
* the pages that are errored in the just handled mirror can be repaired.
|
|
|
|
*/
|
2012-11-02 15:16:26 +00:00
|
|
|
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
|
|
|
|
struct scrub_block *sblock, int is_metadata,
|
|
|
|
int have_csum, u8 *csum, u64 generation,
|
|
|
|
u16 csum_size)
|
2011-04-09 11:27:01 +00:00
|
|
|
{
|
2012-03-27 18:21:27 +00:00
|
|
|
int page_num;
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
sblock->no_io_error_seen = 1;
|
|
|
|
sblock->header_error = 0;
|
|
|
|
sblock->checksum_error = 0;
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
for (page_num = 0; page_num < sblock->page_count; page_num++) {
|
|
|
|
struct bio *bio;
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *page = sblock->pagev[page_num];
|
2012-03-27 18:21:27 +00:00
|
|
|
DECLARE_COMPLETION_ONSTACK(complete);
|
|
|
|
|
2012-05-25 14:06:08 +00:00
|
|
|
if (page->dev->bdev == NULL) {
|
2012-05-04 19:16:07 +00:00
|
|
|
page->io_error = 1;
|
|
|
|
sblock->no_io_error_seen = 0;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
WARN_ON(!page->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
bio = bio_alloc(GFP_NOFS, 1);
|
2012-11-02 15:16:26 +00:00
|
|
|
if (!bio) {
|
|
|
|
page->io_error = 1;
|
|
|
|
sblock->no_io_error_seen = 0;
|
|
|
|
continue;
|
|
|
|
}
|
2012-05-25 14:06:08 +00:00
|
|
|
bio->bi_bdev = page->dev->bdev;
|
2012-03-27 18:21:27 +00:00
|
|
|
bio->bi_sector = page->physical >> 9;
|
|
|
|
bio->bi_end_io = scrub_complete_bio_end_io;
|
|
|
|
bio->bi_private = &complete;
|
|
|
|
|
2012-11-02 15:16:26 +00:00
|
|
|
bio_add_page(bio, page->page, PAGE_SIZE, 0);
|
2012-03-27 18:21:27 +00:00
|
|
|
btrfsic_submit_bio(READ, bio);
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* this will also unplug the queue */
|
|
|
|
wait_for_completion(&complete);
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
page->io_error = !test_bit(BIO_UPTODATE, &bio->bi_flags);
|
|
|
|
if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
|
|
|
|
sblock->no_io_error_seen = 0;
|
|
|
|
bio_put(bio);
|
|
|
|
}
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (sblock->no_io_error_seen)
|
|
|
|
scrub_recheck_block_checksum(fs_info, sblock, is_metadata,
|
|
|
|
have_csum, csum, generation,
|
|
|
|
csum_size);
|
|
|
|
|
2012-11-02 15:16:26 +00:00
|
|
|
return;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_recheck_block_checksum(struct btrfs_fs_info *fs_info,
|
|
|
|
struct scrub_block *sblock,
|
|
|
|
int is_metadata, int have_csum,
|
|
|
|
const u8 *csum, u64 generation,
|
|
|
|
u16 csum_size)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
2012-03-27 18:21:27 +00:00
|
|
|
int page_num;
|
|
|
|
u8 calculated_csum[BTRFS_CSUM_SIZE];
|
|
|
|
u32 crc = ~(u32)0;
|
|
|
|
struct btrfs_root *root = fs_info->extent_root;
|
|
|
|
void *mapped_buffer;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
WARN_ON(!sblock->pagev[0]->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (is_metadata) {
|
|
|
|
struct btrfs_header *h;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
mapped_buffer = kmap_atomic(sblock->pagev[0]->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
h = (struct btrfs_header *)mapped_buffer;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (sblock->pagev[0]->logical != le64_to_cpu(h->bytenr) ||
|
2012-03-27 18:21:27 +00:00
|
|
|
memcmp(h->fsid, fs_info->fsid, BTRFS_UUID_SIZE) ||
|
|
|
|
memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_UUID_SIZE)) {
|
2012-03-27 18:21:27 +00:00
|
|
|
sblock->header_error = 1;
|
2012-05-25 14:06:08 +00:00
|
|
|
} else if (generation != le64_to_cpu(h->generation)) {
|
|
|
|
sblock->header_error = 1;
|
|
|
|
sblock->generation_error = 1;
|
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
csum = h->csum;
|
|
|
|
} else {
|
|
|
|
if (!have_csum)
|
|
|
|
return;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
mapped_buffer = kmap_atomic(sblock->pagev[0]->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
for (page_num = 0;;) {
|
|
|
|
if (page_num == 0 && is_metadata)
|
|
|
|
crc = btrfs_csum_data(root,
|
|
|
|
((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE,
|
|
|
|
crc, PAGE_SIZE - BTRFS_CSUM_SIZE);
|
|
|
|
else
|
|
|
|
crc = btrfs_csum_data(root, mapped_buffer, crc,
|
|
|
|
PAGE_SIZE);
|
|
|
|
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
kunmap_atomic(mapped_buffer);
|
2012-03-27 18:21:27 +00:00
|
|
|
page_num++;
|
|
|
|
if (page_num >= sblock->page_count)
|
|
|
|
break;
|
2012-11-02 13:58:04 +00:00
|
|
|
WARN_ON(!sblock->pagev[page_num]->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
mapped_buffer = kmap_atomic(sblock->pagev[page_num]->page);
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_csum_final(crc, calculated_csum);
|
|
|
|
if (memcmp(calculated_csum, csum, csum_size))
|
|
|
|
sblock->checksum_error = 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_complete_bio_end_io(struct bio *bio, int err)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
2012-03-27 18:21:27 +00:00
|
|
|
complete((struct completion *)bio->bi_private);
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
|
|
|
|
struct scrub_block *sblock_good,
|
|
|
|
int force_write)
|
|
|
|
{
|
|
|
|
int page_num;
|
|
|
|
int ret = 0;
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
|
|
|
|
int ret_sub;
|
2011-04-09 11:27:01 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
ret_sub = scrub_repair_page_from_good_copy(sblock_bad,
|
|
|
|
sblock_good,
|
|
|
|
page_num,
|
|
|
|
force_write);
|
|
|
|
if (ret_sub)
|
|
|
|
ret = ret_sub;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
|
|
|
|
struct scrub_block *sblock_good,
|
|
|
|
int page_num, int force_write)
|
|
|
|
{
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *page_bad = sblock_bad->pagev[page_num];
|
|
|
|
struct scrub_page *page_good = sblock_good->pagev[page_num];
|
2012-03-27 18:21:27 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
BUG_ON(page_bad->page == NULL);
|
|
|
|
BUG_ON(page_good->page == NULL);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (force_write || sblock_bad->header_error ||
|
|
|
|
sblock_bad->checksum_error || page_bad->io_error) {
|
|
|
|
struct bio *bio;
|
|
|
|
int ret;
|
|
|
|
DECLARE_COMPLETION_ONSTACK(complete);
|
|
|
|
|
|
|
|
bio = bio_alloc(GFP_NOFS, 1);
|
2012-04-12 20:03:56 +00:00
|
|
|
if (!bio)
|
|
|
|
return -EIO;
|
2012-05-25 14:06:08 +00:00
|
|
|
bio->bi_bdev = page_bad->dev->bdev;
|
2012-03-27 18:21:27 +00:00
|
|
|
bio->bi_sector = page_bad->physical >> 9;
|
|
|
|
bio->bi_end_io = scrub_complete_bio_end_io;
|
|
|
|
bio->bi_private = &complete;
|
|
|
|
|
|
|
|
ret = bio_add_page(bio, page_good->page, PAGE_SIZE, 0);
|
|
|
|
if (PAGE_SIZE != ret) {
|
|
|
|
bio_put(bio);
|
|
|
|
return -EIO;
|
2011-06-13 17:56:13 +00:00
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
btrfsic_submit_bio(WRITE, bio);
|
|
|
|
|
|
|
|
/* this will also unplug the queue */
|
|
|
|
wait_for_completion(&complete);
|
2012-05-25 14:06:08 +00:00
|
|
|
if (!bio_flagged(bio, BIO_UPTODATE)) {
|
|
|
|
btrfs_dev_stat_inc_and_print(page_bad->dev,
|
|
|
|
BTRFS_DEV_STAT_WRITE_ERRS);
|
|
|
|
bio_put(bio);
|
|
|
|
return -EIO;
|
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
bio_put(bio);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_checksum(struct scrub_block *sblock)
|
|
|
|
{
|
|
|
|
u64 flags;
|
|
|
|
int ret;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
WARN_ON(sblock->page_count < 1);
|
|
|
|
flags = sblock->pagev[0]->flags;
|
2012-03-27 18:21:27 +00:00
|
|
|
ret = 0;
|
|
|
|
if (flags & BTRFS_EXTENT_FLAG_DATA)
|
|
|
|
ret = scrub_checksum_data(sblock);
|
|
|
|
else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
|
|
|
|
ret = scrub_checksum_tree_block(sblock);
|
|
|
|
else if (flags & BTRFS_EXTENT_FLAG_SUPER)
|
|
|
|
(void)scrub_checksum_super(sblock);
|
|
|
|
else
|
|
|
|
WARN_ON(1);
|
|
|
|
if (ret)
|
|
|
|
scrub_handle_errored_block(sblock);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_checksum_data(struct scrub_block *sblock)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = sblock->sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
u8 csum[BTRFS_CSUM_SIZE];
|
2012-03-27 18:21:27 +00:00
|
|
|
u8 *on_disk_csum;
|
|
|
|
struct page *page;
|
|
|
|
void *buffer;
|
2011-03-08 13:14:00 +00:00
|
|
|
u32 crc = ~(u32)0;
|
|
|
|
int fail = 0;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *root = sctx->dev_root;
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 len;
|
|
|
|
int index;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
BUG_ON(sblock->page_count < 1);
|
2012-11-02 13:58:04 +00:00
|
|
|
if (!sblock->pagev[0]->have_csum)
|
2011-03-08 13:14:00 +00:00
|
|
|
return 0;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
on_disk_csum = sblock->pagev[0]->csum;
|
|
|
|
page = sblock->pagev[0]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
len = sctx->sectorsize;
|
2012-03-27 18:21:27 +00:00
|
|
|
index = 0;
|
|
|
|
for (;;) {
|
|
|
|
u64 l = min_t(u64, len, PAGE_SIZE);
|
|
|
|
|
|
|
|
crc = btrfs_csum_data(root, buffer, crc, l);
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
kunmap_atomic(buffer);
|
2012-03-27 18:21:27 +00:00
|
|
|
len -= l;
|
|
|
|
if (len == 0)
|
|
|
|
break;
|
|
|
|
index++;
|
|
|
|
BUG_ON(index >= sblock->page_count);
|
2012-11-02 13:58:04 +00:00
|
|
|
BUG_ON(!sblock->pagev[index]->page);
|
|
|
|
page = sblock->pagev[index]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
btrfs_csum_final(crc, csum);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (memcmp(csum, on_disk_csum, sctx->csum_size))
|
2011-03-08 13:14:00 +00:00
|
|
|
fail = 1;
|
|
|
|
|
|
|
|
return fail;
|
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_checksum_tree_block(struct scrub_block *sblock)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = sblock->sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_header *h;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *root = sctx->dev_root;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
2012-03-27 18:21:27 +00:00
|
|
|
u8 calculated_csum[BTRFS_CSUM_SIZE];
|
|
|
|
u8 on_disk_csum[BTRFS_CSUM_SIZE];
|
|
|
|
struct page *page;
|
|
|
|
void *mapped_buffer;
|
|
|
|
u64 mapped_size;
|
|
|
|
void *p;
|
2011-03-08 13:14:00 +00:00
|
|
|
u32 crc = ~(u32)0;
|
|
|
|
int fail = 0;
|
|
|
|
int crc_fail = 0;
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 len;
|
|
|
|
int index;
|
|
|
|
|
|
|
|
BUG_ON(sblock->page_count < 1);
|
2012-11-02 13:58:04 +00:00
|
|
|
page = sblock->pagev[0]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
mapped_buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
h = (struct btrfs_header *)mapped_buffer;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
memcpy(on_disk_csum, h->csum, sctx->csum_size);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we don't use the getter functions here, as we
|
|
|
|
* a) don't have an extent buffer and
|
|
|
|
* b) the page is already kmapped
|
|
|
|
*/
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (sblock->pagev[0]->logical != le64_to_cpu(h->bytenr))
|
2011-03-08 13:14:00 +00:00
|
|
|
++fail;
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (sblock->pagev[0]->generation != le64_to_cpu(h->generation))
|
2011-03-08 13:14:00 +00:00
|
|
|
++fail;
|
|
|
|
|
|
|
|
if (memcmp(h->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
|
|
|
|
++fail;
|
|
|
|
|
|
|
|
if (memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
|
|
|
|
BTRFS_UUID_SIZE))
|
|
|
|
++fail;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
BUG_ON(sctx->nodesize != sctx->leafsize);
|
|
|
|
len = sctx->nodesize - BTRFS_CSUM_SIZE;
|
2012-03-27 18:21:27 +00:00
|
|
|
mapped_size = PAGE_SIZE - BTRFS_CSUM_SIZE;
|
|
|
|
p = ((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE;
|
|
|
|
index = 0;
|
|
|
|
for (;;) {
|
|
|
|
u64 l = min_t(u64, len, mapped_size);
|
|
|
|
|
|
|
|
crc = btrfs_csum_data(root, p, crc, l);
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
kunmap_atomic(mapped_buffer);
|
2012-03-27 18:21:27 +00:00
|
|
|
len -= l;
|
|
|
|
if (len == 0)
|
|
|
|
break;
|
|
|
|
index++;
|
|
|
|
BUG_ON(index >= sblock->page_count);
|
2012-11-02 13:58:04 +00:00
|
|
|
BUG_ON(!sblock->pagev[index]->page);
|
|
|
|
page = sblock->pagev[index]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
mapped_buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
mapped_size = PAGE_SIZE;
|
|
|
|
p = mapped_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_csum_final(crc, calculated_csum);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (memcmp(calculated_csum, on_disk_csum, sctx->csum_size))
|
2011-03-08 13:14:00 +00:00
|
|
|
++crc_fail;
|
|
|
|
|
|
|
|
return fail || crc_fail;
|
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static int scrub_checksum_super(struct scrub_block *sblock)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_super_block *s;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = sblock->sctx;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *root = sctx->dev_root;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
2012-03-27 18:21:27 +00:00
|
|
|
u8 calculated_csum[BTRFS_CSUM_SIZE];
|
|
|
|
u8 on_disk_csum[BTRFS_CSUM_SIZE];
|
|
|
|
struct page *page;
|
|
|
|
void *mapped_buffer;
|
|
|
|
u64 mapped_size;
|
|
|
|
void *p;
|
2011-03-08 13:14:00 +00:00
|
|
|
u32 crc = ~(u32)0;
|
2012-05-25 14:06:08 +00:00
|
|
|
int fail_gen = 0;
|
|
|
|
int fail_cor = 0;
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 len;
|
|
|
|
int index;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
BUG_ON(sblock->page_count < 1);
|
2012-11-02 13:58:04 +00:00
|
|
|
page = sblock->pagev[0]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
mapped_buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
s = (struct btrfs_super_block *)mapped_buffer;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
memcpy(on_disk_csum, s->csum, sctx->csum_size);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (sblock->pagev[0]->logical != le64_to_cpu(s->bytenr))
|
2012-05-25 14:06:08 +00:00
|
|
|
++fail_cor;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (sblock->pagev[0]->generation != le64_to_cpu(s->generation))
|
2012-05-25 14:06:08 +00:00
|
|
|
++fail_gen;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
if (memcmp(s->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
|
2012-05-25 14:06:08 +00:00
|
|
|
++fail_cor;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
len = BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE;
|
|
|
|
mapped_size = PAGE_SIZE - BTRFS_CSUM_SIZE;
|
|
|
|
p = ((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE;
|
|
|
|
index = 0;
|
|
|
|
for (;;) {
|
|
|
|
u64 l = min_t(u64, len, mapped_size);
|
|
|
|
|
|
|
|
crc = btrfs_csum_data(root, p, crc, l);
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
kunmap_atomic(mapped_buffer);
|
2012-03-27 18:21:27 +00:00
|
|
|
len -= l;
|
|
|
|
if (len == 0)
|
|
|
|
break;
|
|
|
|
index++;
|
|
|
|
BUG_ON(index >= sblock->page_count);
|
2012-11-02 13:58:04 +00:00
|
|
|
BUG_ON(!sblock->pagev[index]->page);
|
|
|
|
page = sblock->pagev[index]->page;
|
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs
Pull btrfs fixes and features from Chris Mason:
"We've merged in the error handling patches from SuSE. These are
already shipping in the sles kernel, and they give btrfs the ability
to abort transactions and go readonly on errors. It involves a lot of
churn as they clarify BUG_ONs, and remove the ones we now properly
deal with.
Josef reworked the way our metadata interacts with the page cache.
page->private now points to the btrfs extent_buffer object, which
makes everything faster. He changed it so we write an whole extent
buffer at a time instead of allowing individual pages to go down,,
which will be important for the raid5/6 code (for the 3.5 merge
window ;)
Josef also made us more aggressive about dropping pages for metadata
blocks that were freed due to COW. Overall, our metadata caching is
much faster now.
We've integrated my patch for metadata bigger than the page size.
This allows metadata blocks up to 64KB in size. In practice 16K and
32K seem to work best. For workloads with lots of metadata, this cuts
down the size of the extent allocation tree dramatically and fragments
much less.
Scrub was updated to support the larger block sizes, which ended up
being a fairly large change (thanks Stefan Behrens).
We also have an assortment of fixes and updates, especially to the
balancing code (Ilya Dryomov), the back ref walker (Jan Schmidt) and
the defragging code (Liu Bo)."
Fixed up trivial conflicts in fs/btrfs/scrub.c that were just due to
removal of the second argument to k[un]map_atomic() in commit
7ac687d9e047.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (75 commits)
Btrfs: update the checks for mixed block groups with big metadata blocks
Btrfs: update to the right index of defragment
Btrfs: do not bother to defrag an extent if it is a big real extent
Btrfs: add a check to decide if we should defrag the range
Btrfs: fix recursive defragment with autodefrag option
Btrfs: fix the mismatch of page->mapping
Btrfs: fix race between direct io and autodefrag
Btrfs: fix deadlock during allocating chunks
Btrfs: show useful info in space reservation tracepoint
Btrfs: don't use crc items bigger than 4KB
Btrfs: flush out and clean up any block device pages during mount
btrfs: disallow unequal data/metadata blocksize for mixed block groups
Btrfs: enhance superblock sanity checks
Btrfs: change scrub to support big blocks
Btrfs: minor cleanup in scrub
Btrfs: introduce common define for max number of mirrors
Btrfs: fix infinite loop in btrfs_shrink_device()
Btrfs: fix memory leak in resolver code
Btrfs: allow dup for data chunks in mixed mode
Btrfs: validate target profiles only if we are going to use them
...
2012-03-30 19:44:29 +00:00
|
|
|
mapped_buffer = kmap_atomic(page);
|
2012-03-27 18:21:27 +00:00
|
|
|
mapped_size = PAGE_SIZE;
|
|
|
|
p = mapped_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_csum_final(crc, calculated_csum);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (memcmp(calculated_csum, on_disk_csum, sctx->csum_size))
|
2012-05-25 14:06:08 +00:00
|
|
|
++fail_cor;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-05-25 14:06:08 +00:00
|
|
|
if (fail_cor + fail_gen) {
|
2011-03-08 13:14:00 +00:00
|
|
|
/*
|
|
|
|
* if we find an error in a super block, we just report it.
|
|
|
|
* They will get written with the next transaction commit
|
|
|
|
* anyway
|
|
|
|
*/
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
++sctx->stat.super_errors;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-05-25 14:06:08 +00:00
|
|
|
if (fail_cor)
|
2012-11-02 13:58:04 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(sblock->pagev[0]->dev,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_DEV_STAT_CORRUPTION_ERRS);
|
|
|
|
else
|
2012-11-02 13:58:04 +00:00
|
|
|
btrfs_dev_stat_inc_and_print(sblock->pagev[0]->dev,
|
2012-05-25 14:06:08 +00:00
|
|
|
BTRFS_DEV_STAT_GENERATION_ERRS);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-05-25 14:06:08 +00:00
|
|
|
return fail_cor + fail_gen;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_block_get(struct scrub_block *sblock)
|
|
|
|
{
|
|
|
|
atomic_inc(&sblock->ref_count);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_block_put(struct scrub_block *sblock)
|
|
|
|
{
|
|
|
|
if (atomic_dec_and_test(&sblock->ref_count)) {
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < sblock->page_count; i++)
|
2012-11-02 13:58:04 +00:00
|
|
|
scrub_page_put(sblock->pagev[i]);
|
2012-03-27 18:21:27 +00:00
|
|
|
kfree(sblock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
static void scrub_page_get(struct scrub_page *spage)
|
|
|
|
{
|
|
|
|
atomic_inc(&spage->ref_count);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_page_put(struct scrub_page *spage)
|
|
|
|
{
|
|
|
|
if (atomic_dec_and_test(&spage->ref_count)) {
|
|
|
|
if (spage->page)
|
|
|
|
__free_page(spage->page);
|
|
|
|
kfree(spage);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static void scrub_submit(struct scrub_ctx *sctx)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct scrub_bio *sbio;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (sctx->curr == -1)
|
2012-03-27 18:21:26 +00:00
|
|
|
return;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sbio = sctx->bios[sctx->curr];
|
|
|
|
sctx->curr = -1;
|
2012-11-02 15:44:58 +00:00
|
|
|
scrub_pending_bio_inc(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2011-11-09 12:44:05 +00:00
|
|
|
btrfsic_submit_bio(READ, sbio->bio);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_add_page_to_bio(struct scrub_ctx *sctx,
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_page *spage)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
2012-03-27 18:21:27 +00:00
|
|
|
struct scrub_block *sblock = spage->sblock;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct scrub_bio *sbio;
|
2011-11-11 13:17:10 +00:00
|
|
|
int ret;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
again:
|
|
|
|
/*
|
|
|
|
* grab a fresh bio or wait for one to become available
|
|
|
|
*/
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
while (sctx->curr == -1) {
|
|
|
|
spin_lock(&sctx->list_lock);
|
|
|
|
sctx->curr = sctx->first_free;
|
|
|
|
if (sctx->curr != -1) {
|
|
|
|
sctx->first_free = sctx->bios[sctx->curr]->next_free;
|
|
|
|
sctx->bios[sctx->curr]->next_free = -1;
|
|
|
|
sctx->bios[sctx->curr]->page_count = 0;
|
|
|
|
spin_unlock(&sctx->list_lock);
|
2011-03-08 13:14:00 +00:00
|
|
|
} else {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_unlock(&sctx->list_lock);
|
|
|
|
wait_event(sctx->list_wait, sctx->first_free != -1);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
}
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sbio = sctx->bios[sctx->curr];
|
2012-03-27 18:21:27 +00:00
|
|
|
if (sbio->page_count == 0) {
|
2011-11-11 13:17:10 +00:00
|
|
|
struct bio *bio;
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
sbio->physical = spage->physical;
|
|
|
|
sbio->logical = spage->logical;
|
2012-11-02 12:26:57 +00:00
|
|
|
sbio->dev = spage->dev;
|
2012-03-27 18:21:27 +00:00
|
|
|
bio = sbio->bio;
|
|
|
|
if (!bio) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
bio = bio_alloc(GFP_NOFS, sctx->pages_per_bio);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (!bio)
|
|
|
|
return -ENOMEM;
|
|
|
|
sbio->bio = bio;
|
|
|
|
}
|
2011-11-11 13:17:10 +00:00
|
|
|
|
|
|
|
bio->bi_private = sbio;
|
|
|
|
bio->bi_end_io = scrub_bio_end_io;
|
2012-11-02 12:26:57 +00:00
|
|
|
bio->bi_bdev = sbio->dev->bdev;
|
|
|
|
bio->bi_sector = sbio->physical >> 9;
|
2011-11-11 13:17:10 +00:00
|
|
|
sbio->err = 0;
|
2012-03-27 18:21:27 +00:00
|
|
|
} else if (sbio->physical + sbio->page_count * PAGE_SIZE !=
|
|
|
|
spage->physical ||
|
|
|
|
sbio->logical + sbio->page_count * PAGE_SIZE !=
|
2012-11-02 12:26:57 +00:00
|
|
|
spage->logical ||
|
|
|
|
sbio->dev != spage->dev) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_submit(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
goto again;
|
|
|
|
}
|
2011-11-11 13:17:10 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
sbio->pagev[sbio->page_count] = spage;
|
|
|
|
ret = bio_add_page(sbio->bio, spage->page, PAGE_SIZE, 0);
|
|
|
|
if (ret != PAGE_SIZE) {
|
|
|
|
if (sbio->page_count < 1) {
|
|
|
|
bio_put(sbio->bio);
|
|
|
|
sbio->bio = NULL;
|
|
|
|
return -EIO;
|
|
|
|
}
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_submit(sctx);
|
2011-11-11 13:17:10 +00:00
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
scrub_block_get(sblock); /* one for the added page */
|
|
|
|
atomic_inc(&sblock->outstanding_pages);
|
|
|
|
sbio->page_count++;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (sbio->page_count == sctx->pages_per_bio)
|
|
|
|
scrub_submit(sctx);
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
|
2012-11-02 12:26:57 +00:00
|
|
|
u64 physical, struct btrfs_device *dev, u64 flags,
|
|
|
|
u64 gen, int mirror_num, u8 *csum, int force)
|
2012-03-27 18:21:27 +00:00
|
|
|
{
|
|
|
|
struct scrub_block *sblock;
|
|
|
|
int index;
|
|
|
|
|
|
|
|
sblock = kzalloc(sizeof(*sblock), GFP_NOFS);
|
|
|
|
if (!sblock) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.malloc_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
return -ENOMEM;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
2012-03-27 18:21:27 +00:00
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
/* one ref inside this function, plus one for each page added to
|
|
|
|
* a bio later on */
|
2012-03-27 18:21:27 +00:00
|
|
|
atomic_set(&sblock->ref_count, 1);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sblock->sctx = sctx;
|
2012-03-27 18:21:27 +00:00
|
|
|
sblock->no_io_error_seen = 1;
|
|
|
|
|
|
|
|
for (index = 0; len > 0; index++) {
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *spage;
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 l = min_t(u64, len, PAGE_SIZE);
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
spage = kzalloc(sizeof(*spage), GFP_NOFS);
|
|
|
|
if (!spage) {
|
|
|
|
leave_nomem:
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.malloc_errors++;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-11-02 13:58:04 +00:00
|
|
|
scrub_block_put(sblock);
|
2012-03-27 18:21:27 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2012-11-02 13:58:04 +00:00
|
|
|
BUG_ON(index >= SCRUB_MAX_PAGES_PER_BLOCK);
|
|
|
|
scrub_page_get(spage);
|
|
|
|
sblock->pagev[index] = spage;
|
2012-03-27 18:21:27 +00:00
|
|
|
spage->sblock = sblock;
|
2012-11-02 12:26:57 +00:00
|
|
|
spage->dev = dev;
|
2012-03-27 18:21:27 +00:00
|
|
|
spage->flags = flags;
|
|
|
|
spage->generation = gen;
|
|
|
|
spage->logical = logical;
|
|
|
|
spage->physical = physical;
|
|
|
|
spage->mirror_num = mirror_num;
|
|
|
|
if (csum) {
|
|
|
|
spage->have_csum = 1;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
memcpy(spage->csum, csum, sctx->csum_size);
|
2012-03-27 18:21:27 +00:00
|
|
|
} else {
|
|
|
|
spage->have_csum = 0;
|
|
|
|
}
|
|
|
|
sblock->page_count++;
|
2012-11-02 13:58:04 +00:00
|
|
|
spage->page = alloc_page(GFP_NOFS);
|
|
|
|
if (!spage->page)
|
|
|
|
goto leave_nomem;
|
2012-03-27 18:21:27 +00:00
|
|
|
len -= l;
|
|
|
|
logical += l;
|
|
|
|
physical += l;
|
|
|
|
}
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
WARN_ON(sblock->page_count == 0);
|
2012-03-27 18:21:27 +00:00
|
|
|
for (index = 0; index < sblock->page_count; index++) {
|
2012-11-02 13:58:04 +00:00
|
|
|
struct scrub_page *spage = sblock->pagev[index];
|
2011-05-28 19:57:55 +00:00
|
|
|
int ret;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
ret = scrub_add_page_to_bio(sctx, spage);
|
2012-03-27 18:21:27 +00:00
|
|
|
if (ret) {
|
|
|
|
scrub_block_put(sblock);
|
2011-05-28 19:57:55 +00:00
|
|
|
return ret;
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
2011-05-28 19:57:55 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
if (force)
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_submit(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/* last one frees, either here or in bio completion for last page */
|
|
|
|
scrub_block_put(sblock);
|
2011-03-08 13:14:00 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
static void scrub_bio_end_io(struct bio *bio, int err)
|
|
|
|
{
|
|
|
|
struct scrub_bio *sbio = bio->bi_private;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_fs_info *fs_info = sbio->dev->dev_root->fs_info;
|
2012-03-27 18:21:27 +00:00
|
|
|
|
|
|
|
sbio->err = err;
|
|
|
|
sbio->bio = bio;
|
|
|
|
|
|
|
|
btrfs_queue_worker(&fs_info->scrub_workers, &sbio->work);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_bio_end_io_worker(struct btrfs_work *work)
|
|
|
|
{
|
|
|
|
struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = sbio->sctx;
|
2012-03-27 18:21:27 +00:00
|
|
|
int i;
|
|
|
|
|
|
|
|
BUG_ON(sbio->page_count > SCRUB_PAGES_PER_BIO);
|
|
|
|
if (sbio->err) {
|
|
|
|
for (i = 0; i < sbio->page_count; i++) {
|
|
|
|
struct scrub_page *spage = sbio->pagev[i];
|
|
|
|
|
|
|
|
spage->io_error = 1;
|
|
|
|
spage->sblock->no_io_error_seen = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* now complete the scrub_block items that have all pages completed */
|
|
|
|
for (i = 0; i < sbio->page_count; i++) {
|
|
|
|
struct scrub_page *spage = sbio->pagev[i];
|
|
|
|
struct scrub_block *sblock = spage->sblock;
|
|
|
|
|
|
|
|
if (atomic_dec_and_test(&sblock->outstanding_pages))
|
|
|
|
scrub_block_complete(sblock);
|
|
|
|
scrub_block_put(sblock);
|
|
|
|
}
|
|
|
|
|
|
|
|
bio_put(sbio->bio);
|
|
|
|
sbio->bio = NULL;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->list_lock);
|
|
|
|
sbio->next_free = sctx->first_free;
|
|
|
|
sctx->first_free = sbio->index;
|
|
|
|
spin_unlock(&sctx->list_lock);
|
2012-11-02 15:44:58 +00:00
|
|
|
scrub_pending_bio_dec(sctx);
|
2012-03-27 18:21:27 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void scrub_block_complete(struct scrub_block *sblock)
|
|
|
|
{
|
|
|
|
if (!sblock->no_io_error_seen)
|
|
|
|
scrub_handle_errored_block(sblock);
|
|
|
|
else
|
|
|
|
scrub_checksum(sblock);
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u64 len,
|
2011-03-08 13:14:00 +00:00
|
|
|
u8 *csum)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_sum *sum = NULL;
|
|
|
|
int ret = 0;
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long num_sectors;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
while (!list_empty(&sctx->csum_list)) {
|
|
|
|
sum = list_first_entry(&sctx->csum_list,
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_ordered_sum, list);
|
|
|
|
if (sum->bytenr > logical)
|
|
|
|
return 0;
|
|
|
|
if (sum->bytenr + sum->len > logical)
|
|
|
|
break;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
++sctx->stat.csum_discards;
|
2011-03-08 13:14:00 +00:00
|
|
|
list_del(&sum->list);
|
|
|
|
kfree(sum);
|
|
|
|
sum = NULL;
|
|
|
|
}
|
|
|
|
if (!sum)
|
|
|
|
return 0;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
num_sectors = sum->len / sctx->sectorsize;
|
2011-03-08 13:14:00 +00:00
|
|
|
for (i = 0; i < num_sectors; ++i) {
|
|
|
|
if (sum->sums[i].bytenr == logical) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
memcpy(csum, &sum->sums[i].sum, sctx->csum_size);
|
2011-03-08 13:14:00 +00:00
|
|
|
ret = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (ret && i == num_sectors - 1) {
|
|
|
|
list_del(&sum->list);
|
|
|
|
kfree(sum);
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* scrub extent tries to collect up to 64 kB for each bio */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static int scrub_extent(struct scrub_ctx *sctx, u64 logical, u64 len,
|
2012-11-02 12:26:57 +00:00
|
|
|
u64 physical, struct btrfs_device *dev, u64 flags,
|
|
|
|
u64 gen, int mirror_num)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
u8 csum[BTRFS_CSUM_SIZE];
|
2012-03-27 18:21:27 +00:00
|
|
|
u32 blocksize;
|
|
|
|
|
|
|
|
if (flags & BTRFS_EXTENT_FLAG_DATA) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
blocksize = sctx->sectorsize;
|
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.data_extents_scrubbed++;
|
|
|
|
sctx->stat.data_bytes_scrubbed += len;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
BUG_ON(sctx->nodesize != sctx->leafsize);
|
|
|
|
blocksize = sctx->nodesize;
|
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.tree_extents_scrubbed++;
|
|
|
|
sctx->stat.tree_bytes_scrubbed += len;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2012-03-27 18:21:27 +00:00
|
|
|
} else {
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
blocksize = sctx->sectorsize;
|
2012-03-27 18:21:27 +00:00
|
|
|
BUG_ON(1);
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
while (len) {
|
2012-03-27 18:21:27 +00:00
|
|
|
u64 l = min_t(u64, len, blocksize);
|
2011-03-08 13:14:00 +00:00
|
|
|
int have_csum = 0;
|
|
|
|
|
|
|
|
if (flags & BTRFS_EXTENT_FLAG_DATA) {
|
|
|
|
/* push csums to sbio */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
have_csum = scrub_find_csum(sctx, logical, l, csum);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (have_csum == 0)
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
++sctx->stat.no_csum;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
2012-11-02 12:26:57 +00:00
|
|
|
ret = scrub_pages(sctx, logical, l, physical, dev, flags, gen,
|
2012-03-27 18:21:27 +00:00
|
|
|
mirror_num, have_csum ? csum : NULL, 0);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
len -= l;
|
|
|
|
logical += l;
|
|
|
|
physical += l;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
|
2012-11-02 12:26:57 +00:00
|
|
|
struct map_lookup *map,
|
|
|
|
struct btrfs_device *scrub_dev,
|
|
|
|
int num, u64 base, u64 length)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_path *path;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_root *root = fs_info->extent_root;
|
|
|
|
struct btrfs_root *csum_root = fs_info->csum_root;
|
|
|
|
struct btrfs_extent_item *extent;
|
2011-05-28 20:58:38 +00:00
|
|
|
struct blk_plug plug;
|
2011-03-08 13:14:00 +00:00
|
|
|
u64 flags;
|
|
|
|
int ret;
|
|
|
|
int slot;
|
|
|
|
int i;
|
|
|
|
u64 nstripes;
|
|
|
|
struct extent_buffer *l;
|
|
|
|
struct btrfs_key key;
|
|
|
|
u64 physical;
|
|
|
|
u64 logical;
|
|
|
|
u64 generation;
|
2011-06-17 13:55:21 +00:00
|
|
|
int mirror_num;
|
2011-06-10 10:39:23 +00:00
|
|
|
struct reada_control *reada1;
|
|
|
|
struct reada_control *reada2;
|
|
|
|
struct btrfs_key key_start;
|
|
|
|
struct btrfs_key key_end;
|
2011-03-08 13:14:00 +00:00
|
|
|
u64 increment = map->stripe_len;
|
|
|
|
u64 offset;
|
|
|
|
|
|
|
|
nstripes = length;
|
|
|
|
offset = 0;
|
|
|
|
do_div(nstripes, map->stripe_len);
|
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
|
|
|
|
offset = map->stripe_len * num;
|
|
|
|
increment = map->stripe_len * map->num_stripes;
|
2011-06-13 17:56:54 +00:00
|
|
|
mirror_num = 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
|
|
|
|
int factor = map->num_stripes / map->sub_stripes;
|
|
|
|
offset = map->stripe_len * (num / map->sub_stripes);
|
|
|
|
increment = map->stripe_len * factor;
|
2011-06-13 17:56:54 +00:00
|
|
|
mirror_num = num % map->sub_stripes + 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
|
|
|
|
increment = map->stripe_len;
|
2011-06-13 17:56:54 +00:00
|
|
|
mirror_num = num % map->num_stripes + 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
|
|
|
|
increment = map->stripe_len;
|
2011-06-13 17:56:54 +00:00
|
|
|
mirror_num = num % map->num_stripes + 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
} else {
|
|
|
|
increment = map->stripe_len;
|
2011-06-13 17:56:54 +00:00
|
|
|
mirror_num = 1;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* work on commit root. The related disk blocks are static as
|
|
|
|
* long as COW is applied. This means, it is save to rewrite
|
|
|
|
* them to repair disk errors without any race conditions
|
|
|
|
*/
|
2011-03-08 13:14:00 +00:00
|
|
|
path->search_commit_root = 1;
|
|
|
|
path->skip_locking = 1;
|
|
|
|
|
|
|
|
/*
|
2011-06-10 10:39:23 +00:00
|
|
|
* trigger the readahead for extent tree csum tree and wait for
|
|
|
|
* completion. During readahead, the scrub is officially paused
|
|
|
|
* to not hold off transaction commits
|
2011-03-08 13:14:00 +00:00
|
|
|
*/
|
|
|
|
logical = base + offset;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
wait_event(sctx->list_wait,
|
2012-11-02 15:44:58 +00:00
|
|
|
atomic_read(&sctx->bios_in_flight) == 0);
|
2011-06-10 10:39:23 +00:00
|
|
|
atomic_inc(&fs_info->scrubs_paused);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
|
|
|
|
/* FIXME it might be better to start readahead at commit root */
|
|
|
|
key_start.objectid = logical;
|
|
|
|
key_start.type = BTRFS_EXTENT_ITEM_KEY;
|
|
|
|
key_start.offset = (u64)0;
|
|
|
|
key_end.objectid = base + offset + nstripes * increment;
|
|
|
|
key_end.type = BTRFS_EXTENT_ITEM_KEY;
|
|
|
|
key_end.offset = (u64)0;
|
|
|
|
reada1 = btrfs_reada_add(root, &key_start, &key_end);
|
|
|
|
|
|
|
|
key_start.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
|
|
|
|
key_start.type = BTRFS_EXTENT_CSUM_KEY;
|
|
|
|
key_start.offset = logical;
|
|
|
|
key_end.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
|
|
|
|
key_end.type = BTRFS_EXTENT_CSUM_KEY;
|
|
|
|
key_end.offset = base + offset + nstripes * increment;
|
|
|
|
reada2 = btrfs_reada_add(csum_root, &key_start, &key_end);
|
|
|
|
|
|
|
|
if (!IS_ERR(reada1))
|
|
|
|
btrfs_reada_wait(reada1);
|
|
|
|
if (!IS_ERR(reada2))
|
|
|
|
btrfs_reada_wait(reada2);
|
|
|
|
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
while (atomic_read(&fs_info->scrub_pause_req)) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wait_event(fs_info->scrub_pause_wait,
|
|
|
|
atomic_read(&fs_info->scrub_pause_req) == 0);
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
2011-06-10 10:39:23 +00:00
|
|
|
atomic_dec(&fs_info->scrubs_paused);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* collect all data csums for the stripe to avoid seeking during
|
|
|
|
* the scrub. This might currently (crc32) end up to be about 1MB
|
|
|
|
*/
|
2011-05-28 20:58:38 +00:00
|
|
|
blk_start_plug(&plug);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* now find all extents for each stripe and scrub them
|
|
|
|
*/
|
2011-06-10 10:39:23 +00:00
|
|
|
logical = base + offset;
|
|
|
|
physical = map->stripes[num].physical;
|
2011-03-08 13:14:00 +00:00
|
|
|
ret = 0;
|
2011-06-10 10:39:23 +00:00
|
|
|
for (i = 0; i < nstripes; ++i) {
|
2011-03-08 13:14:00 +00:00
|
|
|
/*
|
|
|
|
* canceled?
|
|
|
|
*/
|
|
|
|
if (atomic_read(&fs_info->scrub_cancel_req) ||
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
atomic_read(&sctx->cancel_req)) {
|
2011-03-08 13:14:00 +00:00
|
|
|
ret = -ECANCELED;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* check to see if we have to pause
|
|
|
|
*/
|
|
|
|
if (atomic_read(&fs_info->scrub_pause_req)) {
|
|
|
|
/* push queued extents */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_submit(sctx);
|
|
|
|
wait_event(sctx->list_wait,
|
2012-11-02 15:44:58 +00:00
|
|
|
atomic_read(&sctx->bios_in_flight) == 0);
|
2011-03-08 13:14:00 +00:00
|
|
|
atomic_inc(&fs_info->scrubs_paused);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
while (atomic_read(&fs_info->scrub_pause_req)) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wait_event(fs_info->scrub_pause_wait,
|
|
|
|
atomic_read(&fs_info->scrub_pause_req) == 0);
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
atomic_dec(&fs_info->scrubs_paused);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
}
|
|
|
|
|
2011-06-10 10:39:23 +00:00
|
|
|
ret = btrfs_lookup_csums_range(csum_root, logical,
|
|
|
|
logical + map->stripe_len - 1,
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
&sctx->csum_list, 1);
|
2011-06-10 10:39:23 +00:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
key.objectid = logical;
|
|
|
|
key.type = BTRFS_EXTENT_ITEM_KEY;
|
|
|
|
key.offset = (u64)0;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
2011-06-03 08:09:26 +00:00
|
|
|
if (ret > 0) {
|
2011-03-08 13:14:00 +00:00
|
|
|
ret = btrfs_previous_item(root, path, 0,
|
|
|
|
BTRFS_EXTENT_ITEM_KEY);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
2011-06-03 08:09:26 +00:00
|
|
|
if (ret > 0) {
|
|
|
|
/* there's no smaller item, so stick with the
|
|
|
|
* larger one */
|
|
|
|
btrfs_release_path(path);
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key,
|
|
|
|
path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
l = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
if (slot >= btrfs_header_nritems(l)) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret == 0)
|
|
|
|
continue;
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
btrfs_item_key_to_cpu(l, &key, slot);
|
|
|
|
|
|
|
|
if (key.objectid + key.offset <= logical)
|
|
|
|
goto next;
|
|
|
|
|
|
|
|
if (key.objectid >= logical + map->stripe_len)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
|
|
|
|
goto next;
|
|
|
|
|
|
|
|
extent = btrfs_item_ptr(l, slot,
|
|
|
|
struct btrfs_extent_item);
|
|
|
|
flags = btrfs_extent_flags(l, extent);
|
|
|
|
generation = btrfs_extent_generation(l, extent);
|
|
|
|
|
|
|
|
if (key.objectid < logical &&
|
|
|
|
(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)) {
|
|
|
|
printk(KERN_ERR
|
|
|
|
"btrfs scrub: tree block %llu spanning "
|
|
|
|
"stripes, ignored. logical=%llu\n",
|
|
|
|
(unsigned long long)key.objectid,
|
|
|
|
(unsigned long long)logical);
|
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* trim extent to this stripe
|
|
|
|
*/
|
|
|
|
if (key.objectid < logical) {
|
|
|
|
key.offset -= logical - key.objectid;
|
|
|
|
key.objectid = logical;
|
|
|
|
}
|
|
|
|
if (key.objectid + key.offset >
|
|
|
|
logical + map->stripe_len) {
|
|
|
|
key.offset = logical + map->stripe_len -
|
|
|
|
key.objectid;
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
ret = scrub_extent(sctx, key.objectid, key.offset,
|
2011-03-08 13:14:00 +00:00
|
|
|
key.objectid - logical + physical,
|
2012-11-02 12:26:57 +00:00
|
|
|
scrub_dev, flags, generation,
|
|
|
|
mirror_num);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
next:
|
|
|
|
path->slots[0]++;
|
|
|
|
}
|
2011-05-23 10:30:52 +00:00
|
|
|
btrfs_release_path(path);
|
2011-03-08 13:14:00 +00:00
|
|
|
logical += increment;
|
|
|
|
physical += map->stripe_len;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
spin_lock(&sctx->stat_lock);
|
|
|
|
sctx->stat.last_physical = physical;
|
|
|
|
spin_unlock(&sctx->stat_lock);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
/* push queued extents */
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_submit(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
out:
|
2011-05-28 20:58:38 +00:00
|
|
|
blk_finish_plug(&plug);
|
2011-03-08 13:14:00 +00:00
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret < 0 ? ret : 0;
|
|
|
|
}
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_device *scrub_dev,
|
|
|
|
u64 chunk_tree, u64 chunk_objectid,
|
|
|
|
u64 chunk_offset, u64 length,
|
|
|
|
u64 dev_offset)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_mapping_tree *map_tree =
|
2012-11-02 12:26:57 +00:00
|
|
|
&sctx->dev_root->fs_info->mapping_tree;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct map_lookup *map;
|
|
|
|
struct extent_map *em;
|
|
|
|
int i;
|
|
|
|
int ret = -EINVAL;
|
|
|
|
|
|
|
|
read_lock(&map_tree->map_tree.lock);
|
|
|
|
em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1);
|
|
|
|
read_unlock(&map_tree->map_tree.lock);
|
|
|
|
|
|
|
|
if (!em)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
if (em->start != chunk_offset)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (em->len < length)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
for (i = 0; i < map->num_stripes; ++i) {
|
2012-11-02 12:26:57 +00:00
|
|
|
if (map->stripes[i].dev->bdev == scrub_dev->bdev &&
|
2012-02-09 14:09:02 +00:00
|
|
|
map->stripes[i].physical == dev_offset) {
|
2012-11-02 12:26:57 +00:00
|
|
|
ret = scrub_stripe(sctx, map, scrub_dev, i,
|
|
|
|
chunk_offset, length);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
free_extent_map(em);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static noinline_for_stack
|
2012-11-02 12:26:57 +00:00
|
|
|
int scrub_enumerate_chunks(struct scrub_ctx *sctx,
|
|
|
|
struct btrfs_device *scrub_dev, u64 start, u64 end)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_dev_extent *dev_extent = NULL;
|
|
|
|
struct btrfs_path *path;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *root = sctx->dev_root;
|
2011-03-08 13:14:00 +00:00
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
|
|
u64 length;
|
|
|
|
u64 chunk_tree;
|
|
|
|
u64 chunk_objectid;
|
|
|
|
u64 chunk_offset;
|
|
|
|
int ret;
|
|
|
|
int slot;
|
|
|
|
struct extent_buffer *l;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
struct btrfs_block_group_cache *cache;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
path->reada = 2;
|
|
|
|
path->search_commit_root = 1;
|
|
|
|
path->skip_locking = 1;
|
|
|
|
|
2012-11-02 12:26:57 +00:00
|
|
|
key.objectid = scrub_dev->devid;
|
2011-03-08 13:14:00 +00:00
|
|
|
key.offset = 0ull;
|
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
2011-06-03 08:09:26 +00:00
|
|
|
break;
|
|
|
|
if (ret > 0) {
|
|
|
|
if (path->slots[0] >=
|
|
|
|
btrfs_header_nritems(path->nodes[0])) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
l = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
|
|
|
|
btrfs_item_key_to_cpu(l, &found_key, slot);
|
|
|
|
|
2012-11-02 12:26:57 +00:00
|
|
|
if (found_key.objectid != scrub_dev->devid)
|
2011-03-08 13:14:00 +00:00
|
|
|
break;
|
|
|
|
|
2011-06-03 08:09:26 +00:00
|
|
|
if (btrfs_key_type(&found_key) != BTRFS_DEV_EXTENT_KEY)
|
2011-03-08 13:14:00 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
if (found_key.offset >= end)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (found_key.offset < key.offset)
|
|
|
|
break;
|
|
|
|
|
|
|
|
dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
|
|
|
|
length = btrfs_dev_extent_length(l, dev_extent);
|
|
|
|
|
|
|
|
if (found_key.offset + length <= start) {
|
|
|
|
key.offset = found_key.offset + length;
|
2011-05-23 10:30:52 +00:00
|
|
|
btrfs_release_path(path);
|
2011-03-08 13:14:00 +00:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
|
|
|
|
chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
|
|
|
|
chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get a reference on the corresponding block group to prevent
|
|
|
|
* the chunk from going away while we scrub it
|
|
|
|
*/
|
|
|
|
cache = btrfs_lookup_block_group(fs_info, chunk_offset);
|
|
|
|
if (!cache) {
|
|
|
|
ret = -ENOENT;
|
2011-06-03 08:09:26 +00:00
|
|
|
break;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
2012-11-02 12:26:57 +00:00
|
|
|
ret = scrub_chunk(sctx, scrub_dev, chunk_tree, chunk_objectid,
|
2012-02-09 14:09:02 +00:00
|
|
|
chunk_offset, length, found_key.offset);
|
2011-03-08 13:14:00 +00:00
|
|
|
btrfs_put_block_group(cache);
|
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
|
|
|
|
key.offset = found_key.offset + length;
|
2011-05-23 10:30:52 +00:00
|
|
|
btrfs_release_path(path);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_free_path(path);
|
2011-06-03 08:09:26 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* ret can still be 1 from search_slot or next_leaf,
|
|
|
|
* that's not an error
|
|
|
|
*/
|
|
|
|
return ret < 0 ? ret : 0;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-11-02 12:26:57 +00:00
|
|
|
static noinline_for_stack int scrub_supers(struct scrub_ctx *sctx,
|
|
|
|
struct btrfs_device *scrub_dev)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
u64 bytenr;
|
|
|
|
u64 gen;
|
|
|
|
int ret;
|
2012-11-02 12:26:57 +00:00
|
|
|
struct btrfs_root *root = sctx->dev_root;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-03-12 15:03:00 +00:00
|
|
|
if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
|
|
|
|
return -EIO;
|
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
gen = root->fs_info->last_trans_committed;
|
|
|
|
|
|
|
|
for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
|
|
|
|
bytenr = btrfs_sb_offset(i);
|
2012-11-02 12:26:57 +00:00
|
|
|
if (bytenr + BTRFS_SUPER_INFO_SIZE > scrub_dev->total_bytes)
|
2011-03-08 13:14:00 +00:00
|
|
|
break;
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
ret = scrub_pages(sctx, bytenr, BTRFS_SUPER_INFO_SIZE, bytenr,
|
2012-11-02 12:26:57 +00:00
|
|
|
scrub_dev, BTRFS_EXTENT_FLAG_SUPER, gen, i,
|
|
|
|
NULL, 1);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
2012-11-02 15:44:58 +00:00
|
|
|
wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get a reference count on fs_info->scrub_workers. start worker if necessary
|
|
|
|
*/
|
2012-11-05 16:03:39 +00:00
|
|
|
static noinline_for_stack int scrub_workers_get(struct btrfs_fs_info *fs_info)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
2011-11-18 19:37:27 +00:00
|
|
|
int ret = 0;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
2011-06-10 10:07:07 +00:00
|
|
|
if (fs_info->scrub_workers_refcnt == 0) {
|
|
|
|
btrfs_init_workers(&fs_info->scrub_workers, "scrub",
|
|
|
|
fs_info->thread_pool_size, &fs_info->generic_worker);
|
|
|
|
fs_info->scrub_workers.idle_thresh = 4;
|
2011-11-18 19:37:27 +00:00
|
|
|
ret = btrfs_start_workers(&fs_info->scrub_workers);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
2011-06-10 10:07:07 +00:00
|
|
|
}
|
2011-03-08 13:14:00 +00:00
|
|
|
++fs_info->scrub_workers_refcnt;
|
2011-11-18 19:37:27 +00:00
|
|
|
out:
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
|
2011-11-18 19:37:27 +00:00
|
|
|
return ret;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
static noinline_for_stack void scrub_workers_put(struct btrfs_fs_info *fs_info)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
if (--fs_info->scrub_workers_refcnt == 0)
|
|
|
|
btrfs_stop_workers(&fs_info->scrub_workers);
|
|
|
|
WARN_ON(fs_info->scrub_workers_refcnt < 0);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
|
|
|
|
u64 end, struct btrfs_scrub_progress *progress,
|
2012-11-05 17:29:28 +00:00
|
|
|
int readonly, int is_dev_replace)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
int ret;
|
|
|
|
struct btrfs_device *dev;
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
if (btrfs_fs_closing(fs_info))
|
2011-03-08 13:14:00 +00:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* check some assumptions
|
|
|
|
*/
|
2012-11-05 16:03:39 +00:00
|
|
|
if (fs_info->chunk_root->nodesize != fs_info->chunk_root->leafsize) {
|
2012-03-27 18:21:27 +00:00
|
|
|
printk(KERN_ERR
|
|
|
|
"btrfs_scrub: size assumption nodesize == leafsize (%d == %d) fails\n",
|
2012-11-05 16:03:39 +00:00
|
|
|
fs_info->chunk_root->nodesize,
|
|
|
|
fs_info->chunk_root->leafsize);
|
2012-03-27 18:21:27 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
if (fs_info->chunk_root->nodesize > BTRFS_STRIPE_LEN) {
|
2012-03-27 18:21:27 +00:00
|
|
|
/*
|
|
|
|
* in this case scrub is unable to calculate the checksum
|
|
|
|
* the way scrub is implemented. Do not handle this
|
|
|
|
* situation at all because it won't ever happen.
|
|
|
|
*/
|
|
|
|
printk(KERN_ERR
|
|
|
|
"btrfs_scrub: size assumption nodesize <= BTRFS_STRIPE_LEN (%d <= %d) fails\n",
|
2012-11-05 16:03:39 +00:00
|
|
|
fs_info->chunk_root->nodesize, BTRFS_STRIPE_LEN);
|
2012-03-27 18:21:27 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
if (fs_info->chunk_root->sectorsize != PAGE_SIZE) {
|
2012-03-27 18:21:27 +00:00
|
|
|
/* not supported for data w/o checksums */
|
|
|
|
printk(KERN_ERR
|
|
|
|
"btrfs_scrub: size assumption sectorsize != PAGE_SIZE (%d != %lld) fails\n",
|
2012-11-05 16:03:39 +00:00
|
|
|
fs_info->chunk_root->sectorsize,
|
|
|
|
(unsigned long long)PAGE_SIZE);
|
2011-03-08 13:14:00 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2012-11-02 13:58:04 +00:00
|
|
|
if (fs_info->chunk_root->nodesize >
|
|
|
|
PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK ||
|
|
|
|
fs_info->chunk_root->sectorsize >
|
|
|
|
PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK) {
|
|
|
|
/*
|
|
|
|
* would exhaust the array bounds of pagev member in
|
|
|
|
* struct scrub_block
|
|
|
|
*/
|
|
|
|
pr_err("btrfs_scrub: size assumption nodesize and sectorsize <= SCRUB_MAX_PAGES_PER_BLOCK (%d <= %d && %d <= %d) fails\n",
|
|
|
|
fs_info->chunk_root->nodesize,
|
|
|
|
SCRUB_MAX_PAGES_PER_BLOCK,
|
|
|
|
fs_info->chunk_root->sectorsize,
|
|
|
|
SCRUB_MAX_PAGES_PER_BLOCK);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
ret = scrub_workers_get(fs_info);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_lock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
dev = btrfs_find_device(fs_info, devid, NULL, NULL);
|
2012-11-05 17:29:28 +00:00
|
|
|
if (!dev || (dev->missing && !is_dev_replace)) {
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
scrub_workers_put(fs_info);
|
2011-03-08 13:14:00 +00:00
|
|
|
return -ENODEV;
|
|
|
|
}
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
|
2012-11-05 17:29:28 +00:00
|
|
|
if (!dev->in_fs_metadata || dev->is_tgtdev_for_dev_replace) {
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
scrub_workers_put(fs_info);
|
|
|
|
return -EIO;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (dev->scrub_device) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
scrub_workers_put(fs_info);
|
2011-03-08 13:14:00 +00:00
|
|
|
return -EINPROGRESS;
|
|
|
|
}
|
2012-11-05 17:29:28 +00:00
|
|
|
sctx = scrub_setup_ctx(dev, is_dev_replace);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
if (IS_ERR(sctx)) {
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
scrub_workers_put(fs_info);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
return PTR_ERR(sctx);
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx->readonly = readonly;
|
|
|
|
dev->scrub_device = sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
atomic_inc(&fs_info->scrubs_running);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
2012-11-05 16:03:39 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
down_read(&fs_info->scrub_super_lock);
|
2012-11-02 12:26:57 +00:00
|
|
|
ret = scrub_supers(sctx, dev);
|
2011-03-08 13:14:00 +00:00
|
|
|
up_read(&fs_info->scrub_super_lock);
|
|
|
|
|
|
|
|
if (!ret)
|
2012-11-02 12:26:57 +00:00
|
|
|
ret = scrub_enumerate_chunks(sctx, dev, start, end);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
2012-11-02 15:44:58 +00:00
|
|
|
wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
|
2011-03-08 13:14:00 +00:00
|
|
|
atomic_dec(&fs_info->scrubs_running);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
|
2012-11-02 15:44:58 +00:00
|
|
|
wait_event(sctx->list_wait, atomic_read(&sctx->workers_pending) == 0);
|
2011-06-13 18:04:15 +00:00
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
if (progress)
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
memcpy(progress, &sctx->stat, sizeof(*progress));
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
dev->scrub_device = NULL;
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
scrub_free_ctx(sctx);
|
2012-11-05 16:03:39 +00:00
|
|
|
scrub_workers_put(fs_info);
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2012-03-01 13:56:26 +00:00
|
|
|
void btrfs_scrub_pause(struct btrfs_root *root)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
|
|
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
atomic_inc(&fs_info->scrub_pause_req);
|
|
|
|
while (atomic_read(&fs_info->scrubs_paused) !=
|
|
|
|
atomic_read(&fs_info->scrubs_running)) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wait_event(fs_info->scrub_pause_wait,
|
|
|
|
atomic_read(&fs_info->scrubs_paused) ==
|
|
|
|
atomic_read(&fs_info->scrubs_running));
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
|
2012-03-01 13:56:26 +00:00
|
|
|
void btrfs_scrub_continue(struct btrfs_root *root)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
|
|
|
|
|
|
atomic_dec(&fs_info->scrub_pause_req);
|
|
|
|
wake_up(&fs_info->scrub_pause_wait);
|
|
|
|
}
|
|
|
|
|
2012-03-01 13:56:26 +00:00
|
|
|
void btrfs_scrub_pause_super(struct btrfs_root *root)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
down_write(&root->fs_info->scrub_super_lock);
|
|
|
|
}
|
|
|
|
|
2012-03-01 13:56:26 +00:00
|
|
|
void btrfs_scrub_continue_super(struct btrfs_root *root)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
up_write(&root->fs_info->scrub_super_lock);
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
int btrfs_scrub_cancel(struct btrfs_fs_info *fs_info)
|
2011-03-08 13:14:00 +00:00
|
|
|
{
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
if (!atomic_read(&fs_info->scrubs_running)) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
return -ENOTCONN;
|
|
|
|
}
|
|
|
|
|
|
|
|
atomic_inc(&fs_info->scrub_cancel_req);
|
|
|
|
while (atomic_read(&fs_info->scrubs_running)) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wait_event(fs_info->scrub_pause_wait,
|
|
|
|
atomic_read(&fs_info->scrubs_running) == 0);
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
atomic_dec(&fs_info->scrub_cancel_req);
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-11-05 16:03:39 +00:00
|
|
|
int btrfs_scrub_cancel_dev(struct btrfs_fs_info *fs_info,
|
|
|
|
struct btrfs_device *dev)
|
2012-03-01 16:24:58 +00:00
|
|
|
{
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx = dev->scrub_device;
|
|
|
|
if (!sctx) {
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
return -ENOTCONN;
|
|
|
|
}
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
atomic_inc(&sctx->cancel_req);
|
2011-03-08 13:14:00 +00:00
|
|
|
while (dev->scrub_device) {
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
wait_event(fs_info->scrub_pause_wait,
|
|
|
|
dev->scrub_device == NULL);
|
|
|
|
mutex_lock(&fs_info->scrub_lock);
|
|
|
|
}
|
|
|
|
mutex_unlock(&fs_info->scrub_lock);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2012-03-27 18:21:26 +00:00
|
|
|
|
2011-03-08 13:14:00 +00:00
|
|
|
int btrfs_scrub_cancel_devid(struct btrfs_root *root, u64 devid)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
|
|
struct btrfs_device *dev;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* we have to hold the device_list_mutex here so the device
|
|
|
|
* does not go away in cancel_dev. FIXME: find a better solution
|
|
|
|
*/
|
|
|
|
mutex_lock(&fs_info->fs_devices->device_list_mutex);
|
2012-11-05 16:03:39 +00:00
|
|
|
dev = btrfs_find_device(fs_info, devid, NULL, NULL);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (!dev) {
|
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
return -ENODEV;
|
|
|
|
}
|
2012-11-05 16:03:39 +00:00
|
|
|
ret = btrfs_scrub_cancel_dev(fs_info, dev);
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&fs_info->fs_devices->device_list_mutex);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_scrub_progress(struct btrfs_root *root, u64 devid,
|
|
|
|
struct btrfs_scrub_progress *progress)
|
|
|
|
{
|
|
|
|
struct btrfs_device *dev;
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
struct scrub_ctx *sctx = NULL;
|
2011-03-08 13:14:00 +00:00
|
|
|
|
|
|
|
mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
|
2012-11-05 16:03:39 +00:00
|
|
|
dev = btrfs_find_device(root->fs_info, devid, NULL, NULL);
|
2011-03-08 13:14:00 +00:00
|
|
|
if (dev)
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
sctx = dev->scrub_device;
|
|
|
|
if (sctx)
|
|
|
|
memcpy(progress, &sctx->stat, sizeof(*progress));
|
2011-03-08 13:14:00 +00:00
|
|
|
mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
|
|
|
|
|
Btrfs: rename the scrub context structure
The device replace procedure makes use of the scrub code. The scrub
code is the most efficient code to read the allocated data of a disk,
i.e. it reads sequentially in order to avoid disk head movements, it
skips unallocated blocks, it uses read ahead mechanisms, and it
contains all the code to detect and repair defects.
This commit is a first preparation step to adapt the scrub code to
be shareable for the device replace procedure.
The block device will be removed from the scrub context state
structure in a later step. It used to be the source block device.
The scrub code as it is used for the device replace procedure reads
the source data from whereever it is optimal. The source device might
even be gone (disconnected, for instance due to a hardware failure).
Or the drive can be so faulty so that the device replace procedure
tries to avoid access to the faulty source drive as much as possible,
and only if all other mirrors are damaged, as a last resort, the
source disk is accessed.
The modified scrub code operates as if it would handle the source
drive and thereby generates an exact copy of the source disk on the
target disk, even if the source disk is not present at all. Therefore
the block device pointer to the source disk is removed in a later
patch, and therefore the context structure is renamed (this is the
goal of the current patch) to reflect that no source block device
scope is there anymore.
Summary:
This first preparation step consists of a textual substitution of the
term "dev" to the term "ctx" whereever the scrub context is used.
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2012-11-02 08:58:09 +00:00
|
|
|
return dev ? (sctx ? 0 : -ENOTCONN) : -ENODEV;
|
2011-03-08 13:14:00 +00:00
|
|
|
}
|