linux-stable/fs/nilfs2/segment.c

2828 lines
74 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0+
/*
* segment.c - NILFS segment constructor.
*
* Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
*
* Written by Ryusuke Konishi.
*
*/
#include <linux/pagemap.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
#include <linux/bitops.h>
#include <linux/bio.h>
#include <linux/completion.h>
#include <linux/blkdev.h>
#include <linux/backing-dev.h>
#include <linux/freezer.h>
#include <linux/kthread.h>
#include <linux/crc32.h>
#include <linux/pagevec.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/slab.h>
#include <linux/sched/signal.h>
#include "nilfs.h"
#include "btnode.h"
#include "page.h"
#include "segment.h"
#include "sufile.h"
#include "cpfile.h"
#include "ifile.h"
#include "segbuf.h"
/*
* Segment constructor
*/
#define SC_N_INODEVEC 16 /* Size of locally allocated inode vector */
#define SC_MAX_SEGDELTA 64 /*
* Upper limit of the number of segments
* appended in collection retry loop
*/
/* Construction mode */
enum {
SC_LSEG_SR = 1, /* Make a logical segment having a super root */
SC_LSEG_DSYNC, /*
* Flush data blocks of a given file and make
* a logical segment without a super root.
*/
SC_FLUSH_FILE, /*
* Flush data files, leads to segment writes without
* creating a checkpoint.
*/
SC_FLUSH_DAT, /*
* Flush DAT file. This also creates segments
* without a checkpoint.
*/
};
/* Stage numbers of dirty block collection */
enum {
NILFS_ST_INIT = 0,
NILFS_ST_GC, /* Collecting dirty blocks for GC */
NILFS_ST_FILE,
NILFS_ST_IFILE,
NILFS_ST_CPFILE,
NILFS_ST_SUFILE,
NILFS_ST_DAT,
NILFS_ST_SR, /* Super root */
NILFS_ST_DSYNC, /* Data sync blocks */
NILFS_ST_DONE,
};
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
#define CREATE_TRACE_POINTS
#include <trace/events/nilfs2.h>
/*
* nilfs_sc_cstage_inc(), nilfs_sc_cstage_set(), nilfs_sc_cstage_get() are
* wrapper functions of stage count (nilfs_sc_info->sc_stage.scnt). Users of
* the variable must use them because transition of stage count must involve
* trace events (trace_nilfs2_collection_stage_transition).
*
* nilfs_sc_cstage_get() isn't required for the above purpose because it doesn't
* produce tracepoint events. It is provided just for making the intention
* clear.
*/
static inline void nilfs_sc_cstage_inc(struct nilfs_sc_info *sci)
{
sci->sc_stage.scnt++;
trace_nilfs2_collection_stage_transition(sci);
}
static inline void nilfs_sc_cstage_set(struct nilfs_sc_info *sci, int next_scnt)
{
sci->sc_stage.scnt = next_scnt;
trace_nilfs2_collection_stage_transition(sci);
}
static inline int nilfs_sc_cstage_get(struct nilfs_sc_info *sci)
{
return sci->sc_stage.scnt;
}
/* State flags of collection */
#define NILFS_CF_NODE 0x0001 /* Collecting node blocks */
#define NILFS_CF_IFILE_STARTED 0x0002 /* IFILE stage has started */
#define NILFS_CF_SUFREED 0x0004 /* segment usages has been freed */
#define NILFS_CF_HISTORY_MASK (NILFS_CF_IFILE_STARTED | NILFS_CF_SUFREED)
/* Operations depending on the construction mode and file type */
struct nilfs_sc_operations {
int (*collect_data)(struct nilfs_sc_info *, struct buffer_head *,
struct inode *);
int (*collect_node)(struct nilfs_sc_info *, struct buffer_head *,
struct inode *);
int (*collect_bmap)(struct nilfs_sc_info *, struct buffer_head *,
struct inode *);
void (*write_data_binfo)(struct nilfs_sc_info *,
struct nilfs_segsum_pointer *,
union nilfs_binfo *);
void (*write_node_binfo)(struct nilfs_sc_info *,
struct nilfs_segsum_pointer *,
union nilfs_binfo *);
};
/*
* Other definitions
*/
static void nilfs_segctor_start_timer(struct nilfs_sc_info *);
static void nilfs_segctor_do_flush(struct nilfs_sc_info *, int);
static void nilfs_segctor_do_immediate_flush(struct nilfs_sc_info *);
static void nilfs_dispose_list(struct the_nilfs *, struct list_head *, int);
#define nilfs_cnt32_gt(a, b) \
(typecheck(__u32, a) && typecheck(__u32, b) && \
((__s32)(b) - (__s32)(a) < 0))
#define nilfs_cnt32_ge(a, b) \
(typecheck(__u32, a) && typecheck(__u32, b) && \
((__s32)(a) - (__s32)(b) >= 0))
#define nilfs_cnt32_lt(a, b) nilfs_cnt32_gt(b, a)
#define nilfs_cnt32_le(a, b) nilfs_cnt32_ge(b, a)
static int nilfs_prepare_segment_lock(struct super_block *sb,
struct nilfs_transaction_info *ti)
{
struct nilfs_transaction_info *cur_ti = current->journal_info;
void *save = NULL;
if (cur_ti) {
if (cur_ti->ti_magic == NILFS_TI_MAGIC)
return ++cur_ti->ti_count;
/*
* If journal_info field is occupied by other FS,
* it is saved and will be restored on
* nilfs_transaction_commit().
*/
nilfs_warn(sb, "journal info from a different FS");
save = current->journal_info;
}
if (!ti) {
ti = kmem_cache_alloc(nilfs_transaction_cachep, GFP_NOFS);
if (!ti)
return -ENOMEM;
ti->ti_flags = NILFS_TI_DYNAMIC_ALLOC;
} else {
ti->ti_flags = 0;
}
ti->ti_count = 0;
ti->ti_save = save;
ti->ti_magic = NILFS_TI_MAGIC;
current->journal_info = ti;
return 0;
}
/**
* nilfs_transaction_begin - start indivisible file operations.
* @sb: super block
* @ti: nilfs_transaction_info
* @vacancy_check: flags for vacancy rate checks
*
* nilfs_transaction_begin() acquires a reader/writer semaphore, called
* the segment semaphore, to make a segment construction and write tasks
* exclusive. The function is used with nilfs_transaction_commit() in pairs.
* The region enclosed by these two functions can be nested. To avoid a
* deadlock, the semaphore is only acquired or released in the outermost call.
*
* This function allocates a nilfs_transaction_info struct to keep context
* information on it. It is initialized and hooked onto the current task in
* the outermost call. If a pre-allocated struct is given to @ti, it is used
* instead; otherwise a new struct is assigned from a slab.
*
* When @vacancy_check flag is set, this function will check the amount of
* free space, and will wait for the GC to reclaim disk space if low capacity.
*
* Return Value: On success, 0 is returned. On error, one of the following
* negative error code is returned.
*
* %-ENOMEM - Insufficient memory available.
*
* %-ENOSPC - No space left on device
*/
int nilfs_transaction_begin(struct super_block *sb,
struct nilfs_transaction_info *ti,
int vacancy_check)
{
struct the_nilfs *nilfs;
int ret = nilfs_prepare_segment_lock(sb, ti);
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
struct nilfs_transaction_info *trace_ti;
if (unlikely(ret < 0))
return ret;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
if (ret > 0) {
trace_ti = current->journal_info;
trace_nilfs2_transaction_transition(sb, trace_ti,
trace_ti->ti_count, trace_ti->ti_flags,
TRACE_NILFS2_TRANSACTION_BEGIN);
return 0;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
}
sb_start_intwrite(sb);
nilfs = sb->s_fs_info;
down_read(&nilfs->ns_segctor_sem);
if (vacancy_check && nilfs_near_disk_full(nilfs)) {
up_read(&nilfs->ns_segctor_sem);
ret = -ENOSPC;
goto failed;
}
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_ti = current->journal_info;
trace_nilfs2_transaction_transition(sb, trace_ti, trace_ti->ti_count,
trace_ti->ti_flags,
TRACE_NILFS2_TRANSACTION_BEGIN);
return 0;
failed:
ti = current->journal_info;
current->journal_info = ti->ti_save;
if (ti->ti_flags & NILFS_TI_DYNAMIC_ALLOC)
kmem_cache_free(nilfs_transaction_cachep, ti);
sb_end_intwrite(sb);
return ret;
}
/**
* nilfs_transaction_commit - commit indivisible file operations.
* @sb: super block
*
* nilfs_transaction_commit() releases the read semaphore which is
* acquired by nilfs_transaction_begin(). This is only performed
* in outermost call of this function. If a commit flag is set,
* nilfs_transaction_commit() sets a timer to start the segment
* constructor. If a sync flag is set, it starts construction
* directly.
*/
int nilfs_transaction_commit(struct super_block *sb)
{
struct nilfs_transaction_info *ti = current->journal_info;
struct the_nilfs *nilfs = sb->s_fs_info;
int err = 0;
BUG_ON(ti == NULL || ti->ti_magic != NILFS_TI_MAGIC);
ti->ti_flags |= NILFS_TI_COMMIT;
if (ti->ti_count > 0) {
ti->ti_count--;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_COMMIT);
return 0;
}
if (nilfs->ns_writer) {
struct nilfs_sc_info *sci = nilfs->ns_writer;
if (ti->ti_flags & NILFS_TI_COMMIT)
nilfs_segctor_start_timer(sci);
if (atomic_read(&nilfs->ns_ndirtyblks) > sci->sc_watermark)
nilfs_segctor_do_flush(sci, 0);
}
up_read(&nilfs->ns_segctor_sem);
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_COMMIT);
current->journal_info = ti->ti_save;
if (ti->ti_flags & NILFS_TI_SYNC)
err = nilfs_construct_segment(sb);
if (ti->ti_flags & NILFS_TI_DYNAMIC_ALLOC)
kmem_cache_free(nilfs_transaction_cachep, ti);
sb_end_intwrite(sb);
return err;
}
void nilfs_transaction_abort(struct super_block *sb)
{
struct nilfs_transaction_info *ti = current->journal_info;
struct the_nilfs *nilfs = sb->s_fs_info;
BUG_ON(ti == NULL || ti->ti_magic != NILFS_TI_MAGIC);
if (ti->ti_count > 0) {
ti->ti_count--;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_ABORT);
return;
}
up_read(&nilfs->ns_segctor_sem);
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_ABORT);
current->journal_info = ti->ti_save;
if (ti->ti_flags & NILFS_TI_DYNAMIC_ALLOC)
kmem_cache_free(nilfs_transaction_cachep, ti);
sb_end_intwrite(sb);
}
void nilfs_relax_pressure_in_lock(struct super_block *sb)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
if (!sci || !sci->sc_flush_request)
return;
set_bit(NILFS_SC_PRIOR_FLUSH, &sci->sc_flags);
up_read(&nilfs->ns_segctor_sem);
down_write(&nilfs->ns_segctor_sem);
if (sci->sc_flush_request &&
test_bit(NILFS_SC_PRIOR_FLUSH, &sci->sc_flags)) {
struct nilfs_transaction_info *ti = current->journal_info;
ti->ti_flags |= NILFS_TI_WRITER;
nilfs_segctor_do_immediate_flush(sci);
ti->ti_flags &= ~NILFS_TI_WRITER;
}
downgrade_write(&nilfs->ns_segctor_sem);
}
static void nilfs_transaction_lock(struct super_block *sb,
struct nilfs_transaction_info *ti,
int gcflag)
{
struct nilfs_transaction_info *cur_ti = current->journal_info;
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
WARN_ON(cur_ti);
ti->ti_flags = NILFS_TI_WRITER;
ti->ti_count = 0;
ti->ti_save = cur_ti;
ti->ti_magic = NILFS_TI_MAGIC;
current->journal_info = ti;
for (;;) {
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_TRYLOCK);
down_write(&nilfs->ns_segctor_sem);
if (!test_bit(NILFS_SC_PRIOR_FLUSH, &sci->sc_flags))
break;
nilfs_segctor_do_immediate_flush(sci);
up_write(&nilfs->ns_segctor_sem);
cond_resched();
}
if (gcflag)
ti->ti_flags |= NILFS_TI_GC;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_LOCK);
}
static void nilfs_transaction_unlock(struct super_block *sb)
{
struct nilfs_transaction_info *ti = current->journal_info;
struct the_nilfs *nilfs = sb->s_fs_info;
BUG_ON(ti == NULL || ti->ti_magic != NILFS_TI_MAGIC);
BUG_ON(ti->ti_count > 0);
up_write(&nilfs->ns_segctor_sem);
current->journal_info = ti->ti_save;
nilfs2: add a tracepoint for transaction events This patch adds a tracepoint for transaction events of nilfs. With the tracepoint, these events can be tracked: begin, abort, commit, trylock, lock, and unlock. Basically, these events have corresponding functions e.g. begin event corresponds nilfs_transaction_begin(). The unlock event is an exception. It corresponds to the iteration in nilfs_transaction_lock(). Only one tracepoint is introcued: nilfs2_transaction_transition. The above events are distinguished with newly introduced enum. With this tracepoint, we can analyse a critical section of segment constructoin. Sample output by tpoint of perf-tools: cp-4457 [000] ...1 63.266220: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 1 flags = 9 state = BEGIN cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT cp-4457 [000] ...1 63.266221: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800bf5ccc58 count = 0 flags = 9 state = COMMIT segctord-4371 [001] ...1 68.261196: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK segctord-4371 [001] ...1 68.261280: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = LOCK segctord-4371 [001] ...1 68.261877: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 1 flags = 10 state = BEGIN segctord-4371 [001] ...1 68.262116: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = COMMIT segctord-4371 [001] ...1 68.265032: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 18 state = UNLOCK segctord-4371 [001] ...1 132.376847: nilfs2_transaction_transition: sb = ffff8802112b8800 ti = ffff8800b889bdf8 count = 0 flags = 10 state = TRYLOCK This patch also does trivial cleaning of comma usage in collection stage transition event for consistent coding style. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:32:02 +00:00
trace_nilfs2_transaction_transition(sb, ti, ti->ti_count,
ti->ti_flags, TRACE_NILFS2_TRANSACTION_UNLOCK);
}
static void *nilfs_segctor_map_segsum_entry(struct nilfs_sc_info *sci,
struct nilfs_segsum_pointer *ssp,
unsigned int bytes)
{
struct nilfs_segment_buffer *segbuf = sci->sc_curseg;
unsigned int blocksize = sci->sc_super->s_blocksize;
void *p;
if (unlikely(ssp->offset + bytes > blocksize)) {
ssp->offset = 0;
BUG_ON(NILFS_SEGBUF_BH_IS_LAST(ssp->bh,
&segbuf->sb_segsum_buffers));
ssp->bh = NILFS_SEGBUF_NEXT_BH(ssp->bh);
}
p = ssp->bh->b_data + ssp->offset;
ssp->offset += bytes;
return p;
}
/**
* nilfs_segctor_reset_segment_buffer - reset the current segment buffer
* @sci: nilfs_sc_info
*/
static int nilfs_segctor_reset_segment_buffer(struct nilfs_sc_info *sci)
{
struct nilfs_segment_buffer *segbuf = sci->sc_curseg;
struct buffer_head *sumbh;
unsigned int sumbytes;
unsigned int flags = 0;
int err;
if (nilfs_doing_gc())
flags = NILFS_SS_GC;
err = nilfs_segbuf_reset(segbuf, flags, sci->sc_seg_ctime, sci->sc_cno);
if (unlikely(err))
return err;
sumbh = NILFS_SEGBUF_FIRST_BH(&segbuf->sb_segsum_buffers);
sumbytes = segbuf->sb_sum.sumbytes;
sci->sc_finfo_ptr.bh = sumbh; sci->sc_finfo_ptr.offset = sumbytes;
sci->sc_binfo_ptr.bh = sumbh; sci->sc_binfo_ptr.offset = sumbytes;
sci->sc_blk_cnt = sci->sc_datablk_cnt = 0;
return 0;
}
static int nilfs_segctor_feed_segment(struct nilfs_sc_info *sci)
{
sci->sc_nblk_this_inc += sci->sc_curseg->sb_sum.nblocks;
if (NILFS_SEGBUF_IS_LAST(sci->sc_curseg, &sci->sc_segbufs))
return -E2BIG; /*
* The current segment is filled up
* (internal code)
*/
sci->sc_curseg = NILFS_NEXT_SEGBUF(sci->sc_curseg);
return nilfs_segctor_reset_segment_buffer(sci);
}
static int nilfs_segctor_add_super_root(struct nilfs_sc_info *sci)
{
struct nilfs_segment_buffer *segbuf = sci->sc_curseg;
int err;
if (segbuf->sb_sum.nblocks >= segbuf->sb_rest_blocks) {
err = nilfs_segctor_feed_segment(sci);
if (err)
return err;
segbuf = sci->sc_curseg;
}
err = nilfs_segbuf_extend_payload(segbuf, &segbuf->sb_super_root);
if (likely(!err))
segbuf->sb_sum.flags |= NILFS_SS_SR;
return err;
}
/*
* Functions for making segment summary and payloads
*/
static int nilfs_segctor_segsum_block_required(
struct nilfs_sc_info *sci, const struct nilfs_segsum_pointer *ssp,
unsigned int binfo_size)
{
unsigned int blocksize = sci->sc_super->s_blocksize;
/* Size of finfo and binfo is enough small against blocksize */
return ssp->offset + binfo_size +
(!sci->sc_blk_cnt ? sizeof(struct nilfs_finfo) : 0) >
blocksize;
}
static void nilfs_segctor_begin_finfo(struct nilfs_sc_info *sci,
struct inode *inode)
{
sci->sc_curseg->sb_sum.nfinfo++;
sci->sc_binfo_ptr = sci->sc_finfo_ptr;
nilfs_segctor_map_segsum_entry(
sci, &sci->sc_binfo_ptr, sizeof(struct nilfs_finfo));
if (NILFS_I(inode)->i_root &&
!test_bit(NILFS_SC_HAVE_DELTA, &sci->sc_flags))
set_bit(NILFS_SC_HAVE_DELTA, &sci->sc_flags);
/* skip finfo */
}
static void nilfs_segctor_end_finfo(struct nilfs_sc_info *sci,
struct inode *inode)
{
struct nilfs_finfo *finfo;
struct nilfs_inode_info *ii;
struct nilfs_segment_buffer *segbuf;
__u64 cno;
if (sci->sc_blk_cnt == 0)
return;
ii = NILFS_I(inode);
if (test_bit(NILFS_I_GCINODE, &ii->i_state))
cno = ii->i_cno;
else if (NILFS_ROOT_METADATA_FILE(inode->i_ino))
cno = 0;
else
cno = sci->sc_cno;
finfo = nilfs_segctor_map_segsum_entry(sci, &sci->sc_finfo_ptr,
sizeof(*finfo));
finfo->fi_ino = cpu_to_le64(inode->i_ino);
finfo->fi_nblocks = cpu_to_le32(sci->sc_blk_cnt);
finfo->fi_ndatablk = cpu_to_le32(sci->sc_datablk_cnt);
finfo->fi_cno = cpu_to_le64(cno);
segbuf = sci->sc_curseg;
segbuf->sb_sum.sumbytes = sci->sc_binfo_ptr.offset +
sci->sc_super->s_blocksize * (segbuf->sb_sum.nsumblk - 1);
sci->sc_finfo_ptr = sci->sc_binfo_ptr;
sci->sc_blk_cnt = sci->sc_datablk_cnt = 0;
}
static int nilfs_segctor_add_file_block(struct nilfs_sc_info *sci,
struct buffer_head *bh,
struct inode *inode,
unsigned int binfo_size)
{
struct nilfs_segment_buffer *segbuf;
int required, err = 0;
retry:
segbuf = sci->sc_curseg;
required = nilfs_segctor_segsum_block_required(
sci, &sci->sc_binfo_ptr, binfo_size);
if (segbuf->sb_sum.nblocks + required + 1 > segbuf->sb_rest_blocks) {
nilfs_segctor_end_finfo(sci, inode);
err = nilfs_segctor_feed_segment(sci);
if (err)
return err;
goto retry;
}
if (unlikely(required)) {
err = nilfs_segbuf_extend_segsum(segbuf);
if (unlikely(err))
goto failed;
}
if (sci->sc_blk_cnt == 0)
nilfs_segctor_begin_finfo(sci, inode);
nilfs_segctor_map_segsum_entry(sci, &sci->sc_binfo_ptr, binfo_size);
/* Substitution to vblocknr is delayed until update_blocknr() */
nilfs_segbuf_add_file_buffer(segbuf, bh);
sci->sc_blk_cnt++;
failed:
return err;
}
/*
* Callback functions that enumerate, mark, and collect dirty blocks
*/
static int nilfs_collect_file_data(struct nilfs_sc_info *sci,
struct buffer_head *bh, struct inode *inode)
{
int err;
err = nilfs_bmap_propagate(NILFS_I(inode)->i_bmap, bh);
if (err < 0)
return err;
err = nilfs_segctor_add_file_block(sci, bh, inode,
sizeof(struct nilfs_binfo_v));
if (!err)
sci->sc_datablk_cnt++;
return err;
}
static int nilfs_collect_file_node(struct nilfs_sc_info *sci,
struct buffer_head *bh,
struct inode *inode)
{
return nilfs_bmap_propagate(NILFS_I(inode)->i_bmap, bh);
}
static int nilfs_collect_file_bmap(struct nilfs_sc_info *sci,
struct buffer_head *bh,
struct inode *inode)
{
WARN_ON(!buffer_dirty(bh));
return nilfs_segctor_add_file_block(sci, bh, inode, sizeof(__le64));
}
static void nilfs_write_file_data_binfo(struct nilfs_sc_info *sci,
struct nilfs_segsum_pointer *ssp,
union nilfs_binfo *binfo)
{
struct nilfs_binfo_v *binfo_v = nilfs_segctor_map_segsum_entry(
sci, ssp, sizeof(*binfo_v));
*binfo_v = binfo->bi_v;
}
static void nilfs_write_file_node_binfo(struct nilfs_sc_info *sci,
struct nilfs_segsum_pointer *ssp,
union nilfs_binfo *binfo)
{
__le64 *vblocknr = nilfs_segctor_map_segsum_entry(
sci, ssp, sizeof(*vblocknr));
*vblocknr = binfo->bi_v.bi_vblocknr;
}
static const struct nilfs_sc_operations nilfs_sc_file_ops = {
.collect_data = nilfs_collect_file_data,
.collect_node = nilfs_collect_file_node,
.collect_bmap = nilfs_collect_file_bmap,
.write_data_binfo = nilfs_write_file_data_binfo,
.write_node_binfo = nilfs_write_file_node_binfo,
};
static int nilfs_collect_dat_data(struct nilfs_sc_info *sci,
struct buffer_head *bh, struct inode *inode)
{
int err;
err = nilfs_bmap_propagate(NILFS_I(inode)->i_bmap, bh);
if (err < 0)
return err;
err = nilfs_segctor_add_file_block(sci, bh, inode, sizeof(__le64));
if (!err)
sci->sc_datablk_cnt++;
return err;
}
static int nilfs_collect_dat_bmap(struct nilfs_sc_info *sci,
struct buffer_head *bh, struct inode *inode)
{
WARN_ON(!buffer_dirty(bh));
return nilfs_segctor_add_file_block(sci, bh, inode,
sizeof(struct nilfs_binfo_dat));
}
static void nilfs_write_dat_data_binfo(struct nilfs_sc_info *sci,
struct nilfs_segsum_pointer *ssp,
union nilfs_binfo *binfo)
{
__le64 *blkoff = nilfs_segctor_map_segsum_entry(sci, ssp,
sizeof(*blkoff));
*blkoff = binfo->bi_dat.bi_blkoff;
}
static void nilfs_write_dat_node_binfo(struct nilfs_sc_info *sci,
struct nilfs_segsum_pointer *ssp,
union nilfs_binfo *binfo)
{
struct nilfs_binfo_dat *binfo_dat =
nilfs_segctor_map_segsum_entry(sci, ssp, sizeof(*binfo_dat));
*binfo_dat = binfo->bi_dat;
}
static const struct nilfs_sc_operations nilfs_sc_dat_ops = {
.collect_data = nilfs_collect_dat_data,
.collect_node = nilfs_collect_file_node,
.collect_bmap = nilfs_collect_dat_bmap,
.write_data_binfo = nilfs_write_dat_data_binfo,
.write_node_binfo = nilfs_write_dat_node_binfo,
};
static const struct nilfs_sc_operations nilfs_sc_dsync_ops = {
.collect_data = nilfs_collect_file_data,
.collect_node = NULL,
.collect_bmap = NULL,
.write_data_binfo = nilfs_write_file_data_binfo,
.write_node_binfo = NULL,
};
static size_t nilfs_lookup_dirty_data_buffers(struct inode *inode,
struct list_head *listp,
size_t nlimit,
loff_t start, loff_t end)
{
struct address_space *mapping = inode->i_mapping;
struct pagevec pvec;
pgoff_t index = 0, last = ULONG_MAX;
size_t ndirties = 0;
int i;
if (unlikely(start != 0 || end != LLONG_MAX)) {
/*
* A valid range is given for sync-ing data pages. The
* range is rounded to per-page; extra dirty buffers
* may be included if blocksize < pagesize.
*/
index = start >> PAGE_SHIFT;
last = end >> PAGE_SHIFT;
}
pagevec_init(&pvec);
repeat:
if (unlikely(index > last) ||
!pagevec_lookup_range_tag(&pvec, mapping, &index, last,
PAGECACHE_TAG_DIRTY))
return ndirties;
for (i = 0; i < pagevec_count(&pvec); i++) {
struct buffer_head *bh, *head;
struct page *page = pvec.pages[i];
lock_page(page);
if (!page_has_buffers(page))
create_empty_buffers(page, i_blocksize(inode), 0);
unlock_page(page);
bh = head = page_buffers(page);
do {
nilfs2: fix issue with race condition of competition between segments for dirty blocks Many NILFS2 users were reported about strange file system corruption (for example): NILFS: bad btree node (blocknr=185027): level = 0, flags = 0x0, nchildren = 768 NILFS error (device sda4): nilfs_bmap_last_key: broken bmap (inode number=11540) But such error messages are consequence of file system's issue that takes place more earlier. Fortunately, Jerome Poulin <jeromepoulin@gmail.com> and Anton Eliasson <devel@antoneliasson.se> were reported about another issue not so recently. These reports describe the issue with segctor thread's crash: BUG: unable to handle kernel paging request at 0000000000004c83 IP: nilfs_end_page_io+0x12/0xd0 [nilfs2] Call Trace: nilfs_segctor_do_construct+0xf25/0x1b20 [nilfs2] nilfs_segctor_construct+0x17b/0x290 [nilfs2] nilfs_segctor_thread+0x122/0x3b0 [nilfs2] kthread+0xc0/0xd0 ret_from_fork+0x7c/0xb0 These two issues have one reason. This reason can raise third issue too. Third issue results in hanging of segctor thread with eating of 100% CPU. REPRODUCING PATH: One of the possible way or the issue reproducing was described by Jermoe me Poulin <jeromepoulin@gmail.com>: 1. init S to get to single user mode. 2. sysrq+E to make sure only my shell is running 3. start network-manager to get my wifi connection up 4. login as root and launch "screen" 5. cd /boot/log/nilfs which is a ext3 mount point and can log when NILFS dies. 6. lscp | xz -9e > lscp.txt.xz 7. mount my snapshot using mount -o cp=3360839,ro /dev/vgUbuntu/root /mnt/nilfs 8. start a screen to dump /proc/kmsg to text file since rsyslog is killed 9. start a screen and launch strace -f -o find-cat.log -t find /mnt/nilfs -type f -exec cat {} > /dev/null \; 10. start a screen and launch strace -f -o apt-get.log -t apt-get update 11. launch the last command again as it did not crash the first time 12. apt-get crashes 13. ps aux > ps-aux-crashed.log 13. sysrq+W 14. sysrq+E wait for everything to terminate 15. sysrq+SUSB Simplified way of the issue reproducing is starting kernel compilation task and "apt-get update" in parallel. REPRODUCIBILITY: The issue is reproduced not stable [60% - 80%]. It is very important to have proper environment for the issue reproducing. The critical conditions for successful reproducing: (1) It should have big modified file by mmap() way. (2) This file should have the count of dirty blocks are greater that several segments in size (for example, two or three) from time to time during processing. (3) It should be intensive background activity of files modification in another thread. INVESTIGATION: First of all, it is possible to see that the reason of crash is not valid page address: NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 NILFS [nilfs_segctor_complete_write]:2101 segbuf->sb_segnum 6783 Moreover, value of b_page (0x1a82) is 6786. This value looks like segment number. And b_blocknr with b_size values look like block numbers. So, buffer_head's pointer points on not proper address value. Detailed investigation of the issue is discovered such picture: [-----------------------------SEGMENT 6783-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111149024, segbuf->sb_segnum 6783 [-----------------------------SEGMENT 6784-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff8802174a6798, bh->b_assoc_buffers.prev ffff880221cffee8 NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111150080, segbuf->sb_segnum 6784, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111164416, segbuf->sb_segnum 6784, segbuf->sb_nbio 15 [-----------------------------SEGMENT 6785-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff880219277e80, bh->b_assoc_buffers.prev ffff880221cffc88 NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6785 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111165440, segbuf->sb_segnum 6785, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111177728, segbuf->sb_segnum 6785, segbuf->sb_nbio 12 NILFS [nilfs_segctor_do_construct]:2399 nilfs_segctor_wait NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6783 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6785 NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 BUG: unable to handle kernel paging request at 0000000000001a82 IP: [<ffffffffa024d0f2>] nilfs_end_page_io+0x12/0xd0 [nilfs2] Usually, for every segment we collect dirty files in list. Then, dirty blocks are gathered for every dirty file, prepared for write and submitted by means of nilfs_segbuf_submit_bh() call. Finally, it takes place complete write phase after calling nilfs_end_bio_write() on the block layer. Buffers/pages are marked as not dirty on final phase and processed files removed from the list of dirty files. It is possible to see that we had three prepare_write and submit_bio phases before segbuf_wait and complete_write phase. Moreover, segments compete between each other for dirty blocks because on every iteration of segments processing dirty buffer_heads are added in several lists of payload_buffers: [SEGMENT 6784]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 [SEGMENT 6785]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 The next pointer is the same but prev pointer has changed. It means that buffer_head has next pointer from one list but prev pointer from another. Such modification can be made several times. And, finally, it can be resulted in various issues: (1) segctor hanging, (2) segctor crashing, (3) file system metadata corruption. FIX: This patch adds: (1) setting of BH_Async_Write flag in nilfs_segctor_prepare_write() for every proccessed dirty block; (2) checking of BH_Async_Write flag in nilfs_lookup_dirty_data_buffers() and nilfs_lookup_dirty_node_buffers(); (3) clearing of BH_Async_Write flag in nilfs_segctor_complete_write(), nilfs_abort_logs(), nilfs_forget_buffer(), nilfs_clear_dirty_page(). Reported-by: Jerome Poulin <jeromepoulin@gmail.com> Reported-by: Anton Eliasson <devel@antoneliasson.se> Cc: Paul Fertser <fercerpav@gmail.com> Cc: ARAI Shun-ichi <hermes@ceres.dti.ne.jp> Cc: Piotr Szymaniak <szarpaj@grubelek.pl> Cc: Juan Barry Manuel Canham <Linux@riotingpacifist.net> Cc: Zahid Chowdhury <zahid.chowdhury@starsolutions.com> Cc: Elmer Zhang <freeboy6716@gmail.com> Cc: Kenneth Langga <klangga@gmail.com> Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-30 20:45:12 +00:00
if (!buffer_dirty(bh) || buffer_async_write(bh))
continue;
get_bh(bh);
list_add_tail(&bh->b_assoc_buffers, listp);
ndirties++;
if (unlikely(ndirties >= nlimit)) {
pagevec_release(&pvec);
cond_resched();
return ndirties;
}
} while (bh = bh->b_this_page, bh != head);
}
pagevec_release(&pvec);
cond_resched();
goto repeat;
}
static void nilfs_lookup_dirty_node_buffers(struct inode *inode,
struct list_head *listp)
{
struct nilfs_inode_info *ii = NILFS_I(inode);
struct address_space *mapping = &ii->i_btnode_cache;
struct pagevec pvec;
struct buffer_head *bh, *head;
unsigned int i;
pgoff_t index = 0;
pagevec_init(&pvec);
while (pagevec_lookup_tag(&pvec, mapping, &index,
PAGECACHE_TAG_DIRTY)) {
for (i = 0; i < pagevec_count(&pvec); i++) {
bh = head = page_buffers(pvec.pages[i]);
do {
nilfs2: fix issue with race condition of competition between segments for dirty blocks Many NILFS2 users were reported about strange file system corruption (for example): NILFS: bad btree node (blocknr=185027): level = 0, flags = 0x0, nchildren = 768 NILFS error (device sda4): nilfs_bmap_last_key: broken bmap (inode number=11540) But such error messages are consequence of file system's issue that takes place more earlier. Fortunately, Jerome Poulin <jeromepoulin@gmail.com> and Anton Eliasson <devel@antoneliasson.se> were reported about another issue not so recently. These reports describe the issue with segctor thread's crash: BUG: unable to handle kernel paging request at 0000000000004c83 IP: nilfs_end_page_io+0x12/0xd0 [nilfs2] Call Trace: nilfs_segctor_do_construct+0xf25/0x1b20 [nilfs2] nilfs_segctor_construct+0x17b/0x290 [nilfs2] nilfs_segctor_thread+0x122/0x3b0 [nilfs2] kthread+0xc0/0xd0 ret_from_fork+0x7c/0xb0 These two issues have one reason. This reason can raise third issue too. Third issue results in hanging of segctor thread with eating of 100% CPU. REPRODUCING PATH: One of the possible way or the issue reproducing was described by Jermoe me Poulin <jeromepoulin@gmail.com>: 1. init S to get to single user mode. 2. sysrq+E to make sure only my shell is running 3. start network-manager to get my wifi connection up 4. login as root and launch "screen" 5. cd /boot/log/nilfs which is a ext3 mount point and can log when NILFS dies. 6. lscp | xz -9e > lscp.txt.xz 7. mount my snapshot using mount -o cp=3360839,ro /dev/vgUbuntu/root /mnt/nilfs 8. start a screen to dump /proc/kmsg to text file since rsyslog is killed 9. start a screen and launch strace -f -o find-cat.log -t find /mnt/nilfs -type f -exec cat {} > /dev/null \; 10. start a screen and launch strace -f -o apt-get.log -t apt-get update 11. launch the last command again as it did not crash the first time 12. apt-get crashes 13. ps aux > ps-aux-crashed.log 13. sysrq+W 14. sysrq+E wait for everything to terminate 15. sysrq+SUSB Simplified way of the issue reproducing is starting kernel compilation task and "apt-get update" in parallel. REPRODUCIBILITY: The issue is reproduced not stable [60% - 80%]. It is very important to have proper environment for the issue reproducing. The critical conditions for successful reproducing: (1) It should have big modified file by mmap() way. (2) This file should have the count of dirty blocks are greater that several segments in size (for example, two or three) from time to time during processing. (3) It should be intensive background activity of files modification in another thread. INVESTIGATION: First of all, it is possible to see that the reason of crash is not valid page address: NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 NILFS [nilfs_segctor_complete_write]:2101 segbuf->sb_segnum 6783 Moreover, value of b_page (0x1a82) is 6786. This value looks like segment number. And b_blocknr with b_size values look like block numbers. So, buffer_head's pointer points on not proper address value. Detailed investigation of the issue is discovered such picture: [-----------------------------SEGMENT 6783-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111149024, segbuf->sb_segnum 6783 [-----------------------------SEGMENT 6784-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff8802174a6798, bh->b_assoc_buffers.prev ffff880221cffee8 NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111150080, segbuf->sb_segnum 6784, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111164416, segbuf->sb_segnum 6784, segbuf->sb_nbio 15 [-----------------------------SEGMENT 6785-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff880219277e80, bh->b_assoc_buffers.prev ffff880221cffc88 NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6785 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111165440, segbuf->sb_segnum 6785, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111177728, segbuf->sb_segnum 6785, segbuf->sb_nbio 12 NILFS [nilfs_segctor_do_construct]:2399 nilfs_segctor_wait NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6783 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6785 NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 BUG: unable to handle kernel paging request at 0000000000001a82 IP: [<ffffffffa024d0f2>] nilfs_end_page_io+0x12/0xd0 [nilfs2] Usually, for every segment we collect dirty files in list. Then, dirty blocks are gathered for every dirty file, prepared for write and submitted by means of nilfs_segbuf_submit_bh() call. Finally, it takes place complete write phase after calling nilfs_end_bio_write() on the block layer. Buffers/pages are marked as not dirty on final phase and processed files removed from the list of dirty files. It is possible to see that we had three prepare_write and submit_bio phases before segbuf_wait and complete_write phase. Moreover, segments compete between each other for dirty blocks because on every iteration of segments processing dirty buffer_heads are added in several lists of payload_buffers: [SEGMENT 6784]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 [SEGMENT 6785]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 The next pointer is the same but prev pointer has changed. It means that buffer_head has next pointer from one list but prev pointer from another. Such modification can be made several times. And, finally, it can be resulted in various issues: (1) segctor hanging, (2) segctor crashing, (3) file system metadata corruption. FIX: This patch adds: (1) setting of BH_Async_Write flag in nilfs_segctor_prepare_write() for every proccessed dirty block; (2) checking of BH_Async_Write flag in nilfs_lookup_dirty_data_buffers() and nilfs_lookup_dirty_node_buffers(); (3) clearing of BH_Async_Write flag in nilfs_segctor_complete_write(), nilfs_abort_logs(), nilfs_forget_buffer(), nilfs_clear_dirty_page(). Reported-by: Jerome Poulin <jeromepoulin@gmail.com> Reported-by: Anton Eliasson <devel@antoneliasson.se> Cc: Paul Fertser <fercerpav@gmail.com> Cc: ARAI Shun-ichi <hermes@ceres.dti.ne.jp> Cc: Piotr Szymaniak <szarpaj@grubelek.pl> Cc: Juan Barry Manuel Canham <Linux@riotingpacifist.net> Cc: Zahid Chowdhury <zahid.chowdhury@starsolutions.com> Cc: Elmer Zhang <freeboy6716@gmail.com> Cc: Kenneth Langga <klangga@gmail.com> Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-30 20:45:12 +00:00
if (buffer_dirty(bh) &&
!buffer_async_write(bh)) {
get_bh(bh);
list_add_tail(&bh->b_assoc_buffers,
listp);
}
bh = bh->b_this_page;
} while (bh != head);
}
pagevec_release(&pvec);
cond_resched();
}
}
static void nilfs_dispose_list(struct the_nilfs *nilfs,
struct list_head *head, int force)
{
struct nilfs_inode_info *ii, *n;
struct nilfs_inode_info *ivec[SC_N_INODEVEC], **pii;
unsigned int nv = 0;
while (!list_empty(head)) {
spin_lock(&nilfs->ns_inode_lock);
list_for_each_entry_safe(ii, n, head, i_dirty) {
list_del_init(&ii->i_dirty);
if (force) {
if (unlikely(ii->i_bh)) {
brelse(ii->i_bh);
ii->i_bh = NULL;
}
} else if (test_bit(NILFS_I_DIRTY, &ii->i_state)) {
set_bit(NILFS_I_QUEUED, &ii->i_state);
list_add_tail(&ii->i_dirty,
&nilfs->ns_dirty_files);
continue;
}
ivec[nv++] = ii;
if (nv == SC_N_INODEVEC)
break;
}
spin_unlock(&nilfs->ns_inode_lock);
for (pii = ivec; nv > 0; pii++, nv--)
iput(&(*pii)->vfs_inode);
}
}
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
static void nilfs_iput_work_func(struct work_struct *work)
{
struct nilfs_sc_info *sci = container_of(work, struct nilfs_sc_info,
sc_iput_work);
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
nilfs_dispose_list(nilfs, &sci->sc_iput_queue, 0);
}
static int nilfs_test_metadata_dirty(struct the_nilfs *nilfs,
struct nilfs_root *root)
{
int ret = 0;
if (nilfs_mdt_fetch_dirty(root->ifile))
ret++;
if (nilfs_mdt_fetch_dirty(nilfs->ns_cpfile))
ret++;
if (nilfs_mdt_fetch_dirty(nilfs->ns_sufile))
ret++;
if ((ret || nilfs_doing_gc()) && nilfs_mdt_fetch_dirty(nilfs->ns_dat))
ret++;
return ret;
}
static int nilfs_segctor_clean(struct nilfs_sc_info *sci)
{
return list_empty(&sci->sc_dirty_files) &&
!test_bit(NILFS_SC_DIRTY, &sci->sc_flags) &&
sci->sc_nfreesegs == 0 &&
(!nilfs_doing_gc() || list_empty(&sci->sc_gc_inodes));
}
static int nilfs_segctor_confirm(struct nilfs_sc_info *sci)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
int ret = 0;
if (nilfs_test_metadata_dirty(nilfs, sci->sc_root))
set_bit(NILFS_SC_DIRTY, &sci->sc_flags);
spin_lock(&nilfs->ns_inode_lock);
if (list_empty(&nilfs->ns_dirty_files) && nilfs_segctor_clean(sci))
ret++;
spin_unlock(&nilfs->ns_inode_lock);
return ret;
}
static void nilfs_segctor_clear_metadata_dirty(struct nilfs_sc_info *sci)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
nilfs_mdt_clear_dirty(sci->sc_root->ifile);
nilfs_mdt_clear_dirty(nilfs->ns_cpfile);
nilfs_mdt_clear_dirty(nilfs->ns_sufile);
nilfs_mdt_clear_dirty(nilfs->ns_dat);
}
static int nilfs_segctor_create_checkpoint(struct nilfs_sc_info *sci)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
struct buffer_head *bh_cp;
struct nilfs_checkpoint *raw_cp;
int err;
/* XXX: this interface will be changed */
err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, nilfs->ns_cno, 1,
&raw_cp, &bh_cp);
if (likely(!err)) {
/*
* The following code is duplicated with cpfile. But, it is
* needed to collect the checkpoint even if it was not newly
* created.
*/
mark_buffer_dirty(bh_cp);
nilfs_mdt_mark_dirty(nilfs->ns_cpfile);
nilfs_cpfile_put_checkpoint(
nilfs->ns_cpfile, nilfs->ns_cno, bh_cp);
} else
WARN_ON(err == -EINVAL || err == -ENOENT);
return err;
}
static int nilfs_segctor_fill_in_checkpoint(struct nilfs_sc_info *sci)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
struct buffer_head *bh_cp;
struct nilfs_checkpoint *raw_cp;
int err;
err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, nilfs->ns_cno, 0,
&raw_cp, &bh_cp);
if (unlikely(err)) {
WARN_ON(err == -EINVAL || err == -ENOENT);
goto failed_ibh;
}
raw_cp->cp_snapshot_list.ssl_next = 0;
raw_cp->cp_snapshot_list.ssl_prev = 0;
raw_cp->cp_inodes_count =
cpu_to_le64(atomic64_read(&sci->sc_root->inodes_count));
raw_cp->cp_blocks_count =
cpu_to_le64(atomic64_read(&sci->sc_root->blocks_count));
raw_cp->cp_nblk_inc =
cpu_to_le64(sci->sc_nblk_inc + sci->sc_nblk_this_inc);
raw_cp->cp_create = cpu_to_le64(sci->sc_seg_ctime);
raw_cp->cp_cno = cpu_to_le64(nilfs->ns_cno);
if (test_bit(NILFS_SC_HAVE_DELTA, &sci->sc_flags))
nilfs_checkpoint_clear_minor(raw_cp);
else
nilfs_checkpoint_set_minor(raw_cp);
nilfs_write_inode_common(sci->sc_root->ifile,
&raw_cp->cp_ifile_inode, 1);
nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, nilfs->ns_cno, bh_cp);
return 0;
failed_ibh:
return err;
}
static void nilfs_fill_in_file_bmap(struct inode *ifile,
struct nilfs_inode_info *ii)
{
struct buffer_head *ibh;
struct nilfs_inode *raw_inode;
if (test_bit(NILFS_I_BMAP, &ii->i_state)) {
ibh = ii->i_bh;
BUG_ON(!ibh);
raw_inode = nilfs_ifile_map_inode(ifile, ii->vfs_inode.i_ino,
ibh);
nilfs_bmap_write(ii->i_bmap, raw_inode);
nilfs_ifile_unmap_inode(ifile, ii->vfs_inode.i_ino, ibh);
}
}
static void nilfs_segctor_fill_in_file_bmap(struct nilfs_sc_info *sci)
{
struct nilfs_inode_info *ii;
list_for_each_entry(ii, &sci->sc_dirty_files, i_dirty) {
nilfs_fill_in_file_bmap(sci->sc_root->ifile, ii);
set_bit(NILFS_I_COLLECTED, &ii->i_state);
}
}
static void nilfs_segctor_fill_in_super_root(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs)
{
struct buffer_head *bh_sr;
struct nilfs_super_root *raw_sr;
unsigned int isz, srsz;
bh_sr = NILFS_LAST_SEGBUF(&sci->sc_segbufs)->sb_super_root;
raw_sr = (struct nilfs_super_root *)bh_sr->b_data;
isz = nilfs->ns_inode_size;
srsz = NILFS_SR_BYTES(isz);
raw_sr->sr_bytes = cpu_to_le16(srsz);
raw_sr->sr_nongc_ctime
= cpu_to_le64(nilfs_doing_gc() ?
nilfs->ns_nongc_ctime : sci->sc_seg_ctime);
raw_sr->sr_flags = 0;
nilfs_write_inode_common(nilfs->ns_dat, (void *)raw_sr +
NILFS_SR_DAT_OFFSET(isz), 1);
nilfs_write_inode_common(nilfs->ns_cpfile, (void *)raw_sr +
NILFS_SR_CPFILE_OFFSET(isz), 1);
nilfs_write_inode_common(nilfs->ns_sufile, (void *)raw_sr +
NILFS_SR_SUFILE_OFFSET(isz), 1);
memset((void *)raw_sr + srsz, 0, nilfs->ns_blocksize - srsz);
}
static void nilfs_redirty_inodes(struct list_head *head)
{
struct nilfs_inode_info *ii;
list_for_each_entry(ii, head, i_dirty) {
if (test_bit(NILFS_I_COLLECTED, &ii->i_state))
clear_bit(NILFS_I_COLLECTED, &ii->i_state);
}
}
static void nilfs_drop_collected_inodes(struct list_head *head)
{
struct nilfs_inode_info *ii;
list_for_each_entry(ii, head, i_dirty) {
if (!test_and_clear_bit(NILFS_I_COLLECTED, &ii->i_state))
continue;
clear_bit(NILFS_I_INODE_SYNC, &ii->i_state);
set_bit(NILFS_I_UPDATED, &ii->i_state);
}
}
static int nilfs_segctor_apply_buffers(struct nilfs_sc_info *sci,
struct inode *inode,
struct list_head *listp,
int (*collect)(struct nilfs_sc_info *,
struct buffer_head *,
struct inode *))
{
struct buffer_head *bh, *n;
int err = 0;
if (collect) {
list_for_each_entry_safe(bh, n, listp, b_assoc_buffers) {
list_del_init(&bh->b_assoc_buffers);
err = collect(sci, bh, inode);
brelse(bh);
if (unlikely(err))
goto dispose_buffers;
}
return 0;
}
dispose_buffers:
while (!list_empty(listp)) {
bh = list_first_entry(listp, struct buffer_head,
b_assoc_buffers);
list_del_init(&bh->b_assoc_buffers);
brelse(bh);
}
return err;
}
static size_t nilfs_segctor_buffer_rest(struct nilfs_sc_info *sci)
{
/* Remaining number of blocks within segment buffer */
return sci->sc_segbuf_nblocks -
(sci->sc_nblk_this_inc + sci->sc_curseg->sb_sum.nblocks);
}
static int nilfs_segctor_scan_file(struct nilfs_sc_info *sci,
struct inode *inode,
const struct nilfs_sc_operations *sc_ops)
{
LIST_HEAD(data_buffers);
LIST_HEAD(node_buffers);
int err;
if (!(sci->sc_stage.flags & NILFS_CF_NODE)) {
size_t n, rest = nilfs_segctor_buffer_rest(sci);
n = nilfs_lookup_dirty_data_buffers(
inode, &data_buffers, rest + 1, 0, LLONG_MAX);
if (n > rest) {
err = nilfs_segctor_apply_buffers(
sci, inode, &data_buffers,
sc_ops->collect_data);
BUG_ON(!err); /* always receive -E2BIG or true error */
goto break_or_fail;
}
}
nilfs_lookup_dirty_node_buffers(inode, &node_buffers);
if (!(sci->sc_stage.flags & NILFS_CF_NODE)) {
err = nilfs_segctor_apply_buffers(
sci, inode, &data_buffers, sc_ops->collect_data);
if (unlikely(err)) {
/* dispose node list */
nilfs_segctor_apply_buffers(
sci, inode, &node_buffers, NULL);
goto break_or_fail;
}
sci->sc_stage.flags |= NILFS_CF_NODE;
}
/* Collect node */
err = nilfs_segctor_apply_buffers(
sci, inode, &node_buffers, sc_ops->collect_node);
if (unlikely(err))
goto break_or_fail;
nilfs_bmap_lookup_dirty_buffers(NILFS_I(inode)->i_bmap, &node_buffers);
err = nilfs_segctor_apply_buffers(
sci, inode, &node_buffers, sc_ops->collect_bmap);
if (unlikely(err))
goto break_or_fail;
nilfs_segctor_end_finfo(sci, inode);
sci->sc_stage.flags &= ~NILFS_CF_NODE;
break_or_fail:
return err;
}
static int nilfs_segctor_scan_file_dsync(struct nilfs_sc_info *sci,
struct inode *inode)
{
LIST_HEAD(data_buffers);
size_t n, rest = nilfs_segctor_buffer_rest(sci);
int err;
n = nilfs_lookup_dirty_data_buffers(inode, &data_buffers, rest + 1,
sci->sc_dsync_start,
sci->sc_dsync_end);
err = nilfs_segctor_apply_buffers(sci, inode, &data_buffers,
nilfs_collect_file_data);
if (!err) {
nilfs_segctor_end_finfo(sci, inode);
BUG_ON(n > rest);
/* always receive -E2BIG or true error if n > rest */
}
return err;
}
static int nilfs_segctor_collect_blocks(struct nilfs_sc_info *sci, int mode)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
struct list_head *head;
struct nilfs_inode_info *ii;
size_t ndone;
int err = 0;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
switch (nilfs_sc_cstage_get(sci)) {
case NILFS_ST_INIT:
/* Pre-processes */
sci->sc_stage.flags = 0;
if (!test_bit(NILFS_SC_UNCLOSED, &sci->sc_flags)) {
sci->sc_nblk_inc = 0;
sci->sc_curseg->sb_sum.flags = NILFS_SS_LOGBGN;
if (mode == SC_LSEG_DSYNC) {
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DSYNC);
goto dsync_mode;
}
}
sci->sc_stage.dirty_file_ptr = NULL;
sci->sc_stage.gc_inode_ptr = NULL;
if (mode == SC_FLUSH_DAT) {
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DAT);
goto dat_stage;
}
nilfs_sc_cstage_inc(sci);
fallthrough;
case NILFS_ST_GC:
if (nilfs_doing_gc()) {
head = &sci->sc_gc_inodes;
ii = list_prepare_entry(sci->sc_stage.gc_inode_ptr,
head, i_dirty);
list_for_each_entry_continue(ii, head, i_dirty) {
err = nilfs_segctor_scan_file(
sci, &ii->vfs_inode,
&nilfs_sc_file_ops);
if (unlikely(err)) {
sci->sc_stage.gc_inode_ptr = list_entry(
ii->i_dirty.prev,
struct nilfs_inode_info,
i_dirty);
goto break_or_fail;
}
set_bit(NILFS_I_COLLECTED, &ii->i_state);
}
sci->sc_stage.gc_inode_ptr = NULL;
}
nilfs_sc_cstage_inc(sci);
fallthrough;
case NILFS_ST_FILE:
head = &sci->sc_dirty_files;
ii = list_prepare_entry(sci->sc_stage.dirty_file_ptr, head,
i_dirty);
list_for_each_entry_continue(ii, head, i_dirty) {
clear_bit(NILFS_I_DIRTY, &ii->i_state);
err = nilfs_segctor_scan_file(sci, &ii->vfs_inode,
&nilfs_sc_file_ops);
if (unlikely(err)) {
sci->sc_stage.dirty_file_ptr =
list_entry(ii->i_dirty.prev,
struct nilfs_inode_info,
i_dirty);
goto break_or_fail;
}
/* sci->sc_stage.dirty_file_ptr = NILFS_I(inode); */
/* XXX: required ? */
}
sci->sc_stage.dirty_file_ptr = NULL;
if (mode == SC_FLUSH_FILE) {
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DONE);
return 0;
}
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_inc(sci);
sci->sc_stage.flags |= NILFS_CF_IFILE_STARTED;
fallthrough;
case NILFS_ST_IFILE:
err = nilfs_segctor_scan_file(sci, sci->sc_root->ifile,
&nilfs_sc_file_ops);
if (unlikely(err))
break;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_inc(sci);
/* Creating a checkpoint */
err = nilfs_segctor_create_checkpoint(sci);
if (unlikely(err))
break;
fallthrough;
case NILFS_ST_CPFILE:
err = nilfs_segctor_scan_file(sci, nilfs->ns_cpfile,
&nilfs_sc_file_ops);
if (unlikely(err))
break;
nilfs_sc_cstage_inc(sci);
fallthrough;
case NILFS_ST_SUFILE:
err = nilfs_sufile_freev(nilfs->ns_sufile, sci->sc_freesegs,
sci->sc_nfreesegs, &ndone);
if (unlikely(err)) {
nilfs_sufile_cancel_freev(nilfs->ns_sufile,
sci->sc_freesegs, ndone,
NULL);
break;
}
sci->sc_stage.flags |= NILFS_CF_SUFREED;
err = nilfs_segctor_scan_file(sci, nilfs->ns_sufile,
&nilfs_sc_file_ops);
if (unlikely(err))
break;
nilfs_sc_cstage_inc(sci);
fallthrough;
case NILFS_ST_DAT:
dat_stage:
err = nilfs_segctor_scan_file(sci, nilfs->ns_dat,
&nilfs_sc_dat_ops);
if (unlikely(err))
break;
if (mode == SC_FLUSH_DAT) {
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DONE);
return 0;
}
nilfs_sc_cstage_inc(sci);
fallthrough;
case NILFS_ST_SR:
if (mode == SC_LSEG_SR) {
/* Appending a super root */
err = nilfs_segctor_add_super_root(sci);
if (unlikely(err))
break;
}
/* End of a logical segment */
sci->sc_curseg->sb_sum.flags |= NILFS_SS_LOGEND;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DONE);
return 0;
case NILFS_ST_DSYNC:
dsync_mode:
sci->sc_curseg->sb_sum.flags |= NILFS_SS_SYNDT;
ii = sci->sc_dsync_inode;
if (!test_bit(NILFS_I_BUSY, &ii->i_state))
break;
err = nilfs_segctor_scan_file_dsync(sci, &ii->vfs_inode);
if (unlikely(err))
break;
sci->sc_curseg->sb_sum.flags |= NILFS_SS_LOGEND;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_DONE);
return 0;
case NILFS_ST_DONE:
return 0;
default:
BUG();
}
break_or_fail:
return err;
}
/**
* nilfs_segctor_begin_construction - setup segment buffer to make a new log
* @sci: nilfs_sc_info
* @nilfs: nilfs object
*/
static int nilfs_segctor_begin_construction(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs)
{
struct nilfs_segment_buffer *segbuf, *prev;
__u64 nextnum;
int err, alloc = 0;
segbuf = nilfs_segbuf_new(sci->sc_super);
if (unlikely(!segbuf))
return -ENOMEM;
if (list_empty(&sci->sc_write_logs)) {
nilfs_segbuf_map(segbuf, nilfs->ns_segnum,
nilfs->ns_pseg_offset, nilfs);
if (segbuf->sb_rest_blocks < NILFS_PSEG_MIN_BLOCKS) {
nilfs_shift_to_next_segment(nilfs);
nilfs_segbuf_map(segbuf, nilfs->ns_segnum, 0, nilfs);
}
segbuf->sb_sum.seg_seq = nilfs->ns_seg_seq;
nextnum = nilfs->ns_nextnum;
if (nilfs->ns_segnum == nilfs->ns_nextnum)
/* Start from the head of a new full segment */
alloc++;
} else {
/* Continue logs */
prev = NILFS_LAST_SEGBUF(&sci->sc_write_logs);
nilfs_segbuf_map_cont(segbuf, prev);
segbuf->sb_sum.seg_seq = prev->sb_sum.seg_seq;
nextnum = prev->sb_nextnum;
if (segbuf->sb_rest_blocks < NILFS_PSEG_MIN_BLOCKS) {
nilfs_segbuf_map(segbuf, prev->sb_nextnum, 0, nilfs);
segbuf->sb_sum.seg_seq++;
alloc++;
}
}
err = nilfs_sufile_mark_dirty(nilfs->ns_sufile, segbuf->sb_segnum);
if (err)
goto failed;
if (alloc) {
err = nilfs_sufile_alloc(nilfs->ns_sufile, &nextnum);
if (err)
goto failed;
}
nilfs_segbuf_set_next_segnum(segbuf, nextnum, nilfs);
BUG_ON(!list_empty(&sci->sc_segbufs));
list_add_tail(&segbuf->sb_list, &sci->sc_segbufs);
sci->sc_segbuf_nblocks = segbuf->sb_rest_blocks;
return 0;
failed:
nilfs_segbuf_free(segbuf);
return err;
}
static int nilfs_segctor_extend_segments(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs, int nadd)
{
struct nilfs_segment_buffer *segbuf, *prev;
struct inode *sufile = nilfs->ns_sufile;
__u64 nextnextnum;
LIST_HEAD(list);
int err, ret, i;
prev = NILFS_LAST_SEGBUF(&sci->sc_segbufs);
/*
* Since the segment specified with nextnum might be allocated during
* the previous construction, the buffer including its segusage may
* not be dirty. The following call ensures that the buffer is dirty
* and will pin the buffer on memory until the sufile is written.
*/
err = nilfs_sufile_mark_dirty(sufile, prev->sb_nextnum);
if (unlikely(err))
return err;
for (i = 0; i < nadd; i++) {
/* extend segment info */
err = -ENOMEM;
segbuf = nilfs_segbuf_new(sci->sc_super);
if (unlikely(!segbuf))
goto failed;
/* map this buffer to region of segment on-disk */
nilfs_segbuf_map(segbuf, prev->sb_nextnum, 0, nilfs);
sci->sc_segbuf_nblocks += segbuf->sb_rest_blocks;
/* allocate the next next full segment */
err = nilfs_sufile_alloc(sufile, &nextnextnum);
if (unlikely(err))
goto failed_segbuf;
segbuf->sb_sum.seg_seq = prev->sb_sum.seg_seq + 1;
nilfs_segbuf_set_next_segnum(segbuf, nextnextnum, nilfs);
list_add_tail(&segbuf->sb_list, &list);
prev = segbuf;
}
list_splice_tail(&list, &sci->sc_segbufs);
return 0;
failed_segbuf:
nilfs_segbuf_free(segbuf);
failed:
list_for_each_entry(segbuf, &list, sb_list) {
ret = nilfs_sufile_free(sufile, segbuf->sb_nextnum);
WARN_ON(ret); /* never fails */
}
nilfs_destroy_logs(&list);
return err;
}
static void nilfs_free_incomplete_logs(struct list_head *logs,
struct the_nilfs *nilfs)
{
struct nilfs_segment_buffer *segbuf, *prev;
struct inode *sufile = nilfs->ns_sufile;
int ret;
segbuf = NILFS_FIRST_SEGBUF(logs);
if (nilfs->ns_nextnum != segbuf->sb_nextnum) {
ret = nilfs_sufile_free(sufile, segbuf->sb_nextnum);
WARN_ON(ret); /* never fails */
}
if (atomic_read(&segbuf->sb_err)) {
/* Case 1: The first segment failed */
if (segbuf->sb_pseg_start != segbuf->sb_fseg_start)
/*
* Case 1a: Partial segment appended into an existing
* segment
*/
nilfs_terminate_segment(nilfs, segbuf->sb_fseg_start,
segbuf->sb_fseg_end);
else /* Case 1b: New full segment */
set_nilfs_discontinued(nilfs);
}
prev = segbuf;
list_for_each_entry_continue(segbuf, logs, sb_list) {
if (prev->sb_nextnum != segbuf->sb_nextnum) {
ret = nilfs_sufile_free(sufile, segbuf->sb_nextnum);
WARN_ON(ret); /* never fails */
}
if (atomic_read(&segbuf->sb_err) &&
segbuf->sb_segnum != nilfs->ns_nextnum)
/* Case 2: extended segment (!= next) failed */
nilfs_sufile_set_error(sufile, segbuf->sb_segnum);
prev = segbuf;
}
}
static void nilfs_segctor_update_segusage(struct nilfs_sc_info *sci,
struct inode *sufile)
{
struct nilfs_segment_buffer *segbuf;
unsigned long live_blocks;
int ret;
list_for_each_entry(segbuf, &sci->sc_segbufs, sb_list) {
live_blocks = segbuf->sb_sum.nblocks +
(segbuf->sb_pseg_start - segbuf->sb_fseg_start);
ret = nilfs_sufile_set_segment_usage(sufile, segbuf->sb_segnum,
live_blocks,
sci->sc_seg_ctime);
WARN_ON(ret); /* always succeed because the segusage is dirty */
}
}
static void nilfs_cancel_segusage(struct list_head *logs, struct inode *sufile)
{
struct nilfs_segment_buffer *segbuf;
int ret;
segbuf = NILFS_FIRST_SEGBUF(logs);
ret = nilfs_sufile_set_segment_usage(sufile, segbuf->sb_segnum,
segbuf->sb_pseg_start -
segbuf->sb_fseg_start, 0);
WARN_ON(ret); /* always succeed because the segusage is dirty */
list_for_each_entry_continue(segbuf, logs, sb_list) {
ret = nilfs_sufile_set_segment_usage(sufile, segbuf->sb_segnum,
0, 0);
WARN_ON(ret); /* always succeed */
}
}
static void nilfs_segctor_truncate_segments(struct nilfs_sc_info *sci,
struct nilfs_segment_buffer *last,
struct inode *sufile)
{
struct nilfs_segment_buffer *segbuf = last;
int ret;
list_for_each_entry_continue(segbuf, &sci->sc_segbufs, sb_list) {
sci->sc_segbuf_nblocks -= segbuf->sb_rest_blocks;
ret = nilfs_sufile_free(sufile, segbuf->sb_nextnum);
WARN_ON(ret);
}
nilfs_truncate_logs(&sci->sc_segbufs, last);
}
static int nilfs_segctor_collect(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs, int mode)
{
struct nilfs_cstage prev_stage = sci->sc_stage;
int err, nadd = 1;
/* Collection retry loop */
for (;;) {
sci->sc_nblk_this_inc = 0;
sci->sc_curseg = NILFS_FIRST_SEGBUF(&sci->sc_segbufs);
err = nilfs_segctor_reset_segment_buffer(sci);
if (unlikely(err))
goto failed;
err = nilfs_segctor_collect_blocks(sci, mode);
sci->sc_nblk_this_inc += sci->sc_curseg->sb_sum.nblocks;
if (!err)
break;
if (unlikely(err != -E2BIG))
goto failed;
/* The current segment is filled up */
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
if (mode != SC_LSEG_SR ||
nilfs_sc_cstage_get(sci) < NILFS_ST_CPFILE)
break;
nilfs_clear_logs(&sci->sc_segbufs);
if (sci->sc_stage.flags & NILFS_CF_SUFREED) {
err = nilfs_sufile_cancel_freev(nilfs->ns_sufile,
sci->sc_freesegs,
sci->sc_nfreesegs,
NULL);
WARN_ON(err); /* do not happen */
nilfs2: fix segctor bug that causes file system corruption There is a bug in the function nilfs_segctor_collect, which results in active data being written to a segment, that is marked as clean. It is possible, that this segment is selected for a later segment construction, whereby the old data is overwritten. The problem shows itself with the following kernel log message: nilfs_sufile_do_cancel_free: segment 6533 must be clean Usually a few hours later the file system gets corrupted: NILFS: bad btree node (blocknr=8748107): level = 0, flags = 0x0, nchildren = 0 NILFS error (device sdc1): nilfs_bmap_last_key: broken bmap (inode number=114660) The issue can be reproduced with a file system that is nearly full and with the cleaner running, while some IO intensive task is running. Although it is quite hard to reproduce. This is what happens: 1. The cleaner starts the segment construction 2. nilfs_segctor_collect is called 3. sc_stage is on NILFS_ST_SUFILE and segments are freed 4. sc_stage is on NILFS_ST_DAT current segment is full 5. nilfs_segctor_extend_segments is called, which allocates a new segment 6. The new segment is one of the segments freed in step 3 7. nilfs_sufile_cancel_freev is called and produces an error message 8. Loop around and the collection starts again 9. sc_stage is on NILFS_ST_SUFILE and segments are freed including the newly allocated segment, which will contain active data and can be allocated at a later time 10. A few hours later another segment construction allocates the segment and causes file system corruption This can be prevented by simply reordering the statements. If nilfs_sufile_cancel_freev is called before nilfs_segctor_extend_segments the freed segments are marked as dirty and cannot be allocated any more. Signed-off-by: Andreas Rohner <andreas.rohner@gmx.net> Reviewed-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Tested-by: Andreas Rohner <andreas.rohner@gmx.net> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-15 01:56:36 +00:00
sci->sc_stage.flags &= ~NILFS_CF_SUFREED;
}
nilfs2: fix segctor bug that causes file system corruption There is a bug in the function nilfs_segctor_collect, which results in active data being written to a segment, that is marked as clean. It is possible, that this segment is selected for a later segment construction, whereby the old data is overwritten. The problem shows itself with the following kernel log message: nilfs_sufile_do_cancel_free: segment 6533 must be clean Usually a few hours later the file system gets corrupted: NILFS: bad btree node (blocknr=8748107): level = 0, flags = 0x0, nchildren = 0 NILFS error (device sdc1): nilfs_bmap_last_key: broken bmap (inode number=114660) The issue can be reproduced with a file system that is nearly full and with the cleaner running, while some IO intensive task is running. Although it is quite hard to reproduce. This is what happens: 1. The cleaner starts the segment construction 2. nilfs_segctor_collect is called 3. sc_stage is on NILFS_ST_SUFILE and segments are freed 4. sc_stage is on NILFS_ST_DAT current segment is full 5. nilfs_segctor_extend_segments is called, which allocates a new segment 6. The new segment is one of the segments freed in step 3 7. nilfs_sufile_cancel_freev is called and produces an error message 8. Loop around and the collection starts again 9. sc_stage is on NILFS_ST_SUFILE and segments are freed including the newly allocated segment, which will contain active data and can be allocated at a later time 10. A few hours later another segment construction allocates the segment and causes file system corruption This can be prevented by simply reordering the statements. If nilfs_sufile_cancel_freev is called before nilfs_segctor_extend_segments the freed segments are marked as dirty and cannot be allocated any more. Signed-off-by: Andreas Rohner <andreas.rohner@gmx.net> Reviewed-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Tested-by: Andreas Rohner <andreas.rohner@gmx.net> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-15 01:56:36 +00:00
err = nilfs_segctor_extend_segments(sci, nilfs, nadd);
if (unlikely(err))
return err;
nadd = min_t(int, nadd << 1, SC_MAX_SEGDELTA);
sci->sc_stage = prev_stage;
}
nilfs_segctor_truncate_segments(sci, sci->sc_curseg, nilfs->ns_sufile);
return 0;
failed:
return err;
}
static void nilfs_list_replace_buffer(struct buffer_head *old_bh,
struct buffer_head *new_bh)
{
BUG_ON(!list_empty(&new_bh->b_assoc_buffers));
list_replace_init(&old_bh->b_assoc_buffers, &new_bh->b_assoc_buffers);
/* The caller must release old_bh */
}
static int
nilfs_segctor_update_payload_blocknr(struct nilfs_sc_info *sci,
struct nilfs_segment_buffer *segbuf,
int mode)
{
struct inode *inode = NULL;
sector_t blocknr;
unsigned long nfinfo = segbuf->sb_sum.nfinfo;
unsigned long nblocks = 0, ndatablk = 0;
const struct nilfs_sc_operations *sc_op = NULL;
struct nilfs_segsum_pointer ssp;
struct nilfs_finfo *finfo = NULL;
union nilfs_binfo binfo;
struct buffer_head *bh, *bh_org;
ino_t ino = 0;
int err = 0;
if (!nfinfo)
goto out;
blocknr = segbuf->sb_pseg_start + segbuf->sb_sum.nsumblk;
ssp.bh = NILFS_SEGBUF_FIRST_BH(&segbuf->sb_segsum_buffers);
ssp.offset = sizeof(struct nilfs_segment_summary);
list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
if (bh == segbuf->sb_super_root)
break;
if (!finfo) {
finfo = nilfs_segctor_map_segsum_entry(
sci, &ssp, sizeof(*finfo));
ino = le64_to_cpu(finfo->fi_ino);
nblocks = le32_to_cpu(finfo->fi_nblocks);
ndatablk = le32_to_cpu(finfo->fi_ndatablk);
inode = bh->b_page->mapping->host;
if (mode == SC_LSEG_DSYNC)
sc_op = &nilfs_sc_dsync_ops;
else if (ino == NILFS_DAT_INO)
sc_op = &nilfs_sc_dat_ops;
else /* file blocks */
sc_op = &nilfs_sc_file_ops;
}
bh_org = bh;
get_bh(bh_org);
err = nilfs_bmap_assign(NILFS_I(inode)->i_bmap, &bh, blocknr,
&binfo);
if (bh != bh_org)
nilfs_list_replace_buffer(bh_org, bh);
brelse(bh_org);
if (unlikely(err))
goto failed_bmap;
if (ndatablk > 0)
sc_op->write_data_binfo(sci, &ssp, &binfo);
else
sc_op->write_node_binfo(sci, &ssp, &binfo);
blocknr++;
if (--nblocks == 0) {
finfo = NULL;
if (--nfinfo == 0)
break;
} else if (ndatablk > 0)
ndatablk--;
}
out:
return 0;
failed_bmap:
return err;
}
static int nilfs_segctor_assign(struct nilfs_sc_info *sci, int mode)
{
struct nilfs_segment_buffer *segbuf;
int err;
list_for_each_entry(segbuf, &sci->sc_segbufs, sb_list) {
err = nilfs_segctor_update_payload_blocknr(sci, segbuf, mode);
if (unlikely(err))
return err;
nilfs_segbuf_fill_in_segsum(segbuf);
}
return 0;
}
static void nilfs_begin_page_io(struct page *page)
{
if (!page || PageWriteback(page))
/*
* For split b-tree node pages, this function may be called
* twice. We ignore the 2nd or later calls by this check.
*/
return;
lock_page(page);
clear_page_dirty_for_io(page);
set_page_writeback(page);
unlock_page(page);
}
static void nilfs_segctor_prepare_write(struct nilfs_sc_info *sci)
{
struct nilfs_segment_buffer *segbuf;
struct page *bd_page = NULL, *fs_page = NULL;
list_for_each_entry(segbuf, &sci->sc_segbufs, sb_list) {
struct buffer_head *bh;
list_for_each_entry(bh, &segbuf->sb_segsum_buffers,
b_assoc_buffers) {
if (bh->b_page != bd_page) {
if (bd_page) {
lock_page(bd_page);
clear_page_dirty_for_io(bd_page);
set_page_writeback(bd_page);
unlock_page(bd_page);
}
bd_page = bh->b_page;
}
}
list_for_each_entry(bh, &segbuf->sb_payload_buffers,
b_assoc_buffers) {
nilfs2: fix issue with race condition of competition between segments for dirty blocks Many NILFS2 users were reported about strange file system corruption (for example): NILFS: bad btree node (blocknr=185027): level = 0, flags = 0x0, nchildren = 768 NILFS error (device sda4): nilfs_bmap_last_key: broken bmap (inode number=11540) But such error messages are consequence of file system's issue that takes place more earlier. Fortunately, Jerome Poulin <jeromepoulin@gmail.com> and Anton Eliasson <devel@antoneliasson.se> were reported about another issue not so recently. These reports describe the issue with segctor thread's crash: BUG: unable to handle kernel paging request at 0000000000004c83 IP: nilfs_end_page_io+0x12/0xd0 [nilfs2] Call Trace: nilfs_segctor_do_construct+0xf25/0x1b20 [nilfs2] nilfs_segctor_construct+0x17b/0x290 [nilfs2] nilfs_segctor_thread+0x122/0x3b0 [nilfs2] kthread+0xc0/0xd0 ret_from_fork+0x7c/0xb0 These two issues have one reason. This reason can raise third issue too. Third issue results in hanging of segctor thread with eating of 100% CPU. REPRODUCING PATH: One of the possible way or the issue reproducing was described by Jermoe me Poulin <jeromepoulin@gmail.com>: 1. init S to get to single user mode. 2. sysrq+E to make sure only my shell is running 3. start network-manager to get my wifi connection up 4. login as root and launch "screen" 5. cd /boot/log/nilfs which is a ext3 mount point and can log when NILFS dies. 6. lscp | xz -9e > lscp.txt.xz 7. mount my snapshot using mount -o cp=3360839,ro /dev/vgUbuntu/root /mnt/nilfs 8. start a screen to dump /proc/kmsg to text file since rsyslog is killed 9. start a screen and launch strace -f -o find-cat.log -t find /mnt/nilfs -type f -exec cat {} > /dev/null \; 10. start a screen and launch strace -f -o apt-get.log -t apt-get update 11. launch the last command again as it did not crash the first time 12. apt-get crashes 13. ps aux > ps-aux-crashed.log 13. sysrq+W 14. sysrq+E wait for everything to terminate 15. sysrq+SUSB Simplified way of the issue reproducing is starting kernel compilation task and "apt-get update" in parallel. REPRODUCIBILITY: The issue is reproduced not stable [60% - 80%]. It is very important to have proper environment for the issue reproducing. The critical conditions for successful reproducing: (1) It should have big modified file by mmap() way. (2) This file should have the count of dirty blocks are greater that several segments in size (for example, two or three) from time to time during processing. (3) It should be intensive background activity of files modification in another thread. INVESTIGATION: First of all, it is possible to see that the reason of crash is not valid page address: NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 NILFS [nilfs_segctor_complete_write]:2101 segbuf->sb_segnum 6783 Moreover, value of b_page (0x1a82) is 6786. This value looks like segment number. And b_blocknr with b_size values look like block numbers. So, buffer_head's pointer points on not proper address value. Detailed investigation of the issue is discovered such picture: [-----------------------------SEGMENT 6783-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111149024, segbuf->sb_segnum 6783 [-----------------------------SEGMENT 6784-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff8802174a6798, bh->b_assoc_buffers.prev ffff880221cffee8 NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111150080, segbuf->sb_segnum 6784, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111164416, segbuf->sb_segnum 6784, segbuf->sb_nbio 15 [-----------------------------SEGMENT 6785-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff880219277e80, bh->b_assoc_buffers.prev ffff880221cffc88 NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6785 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111165440, segbuf->sb_segnum 6785, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111177728, segbuf->sb_segnum 6785, segbuf->sb_nbio 12 NILFS [nilfs_segctor_do_construct]:2399 nilfs_segctor_wait NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6783 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6785 NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 BUG: unable to handle kernel paging request at 0000000000001a82 IP: [<ffffffffa024d0f2>] nilfs_end_page_io+0x12/0xd0 [nilfs2] Usually, for every segment we collect dirty files in list. Then, dirty blocks are gathered for every dirty file, prepared for write and submitted by means of nilfs_segbuf_submit_bh() call. Finally, it takes place complete write phase after calling nilfs_end_bio_write() on the block layer. Buffers/pages are marked as not dirty on final phase and processed files removed from the list of dirty files. It is possible to see that we had three prepare_write and submit_bio phases before segbuf_wait and complete_write phase. Moreover, segments compete between each other for dirty blocks because on every iteration of segments processing dirty buffer_heads are added in several lists of payload_buffers: [SEGMENT 6784]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 [SEGMENT 6785]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 The next pointer is the same but prev pointer has changed. It means that buffer_head has next pointer from one list but prev pointer from another. Such modification can be made several times. And, finally, it can be resulted in various issues: (1) segctor hanging, (2) segctor crashing, (3) file system metadata corruption. FIX: This patch adds: (1) setting of BH_Async_Write flag in nilfs_segctor_prepare_write() for every proccessed dirty block; (2) checking of BH_Async_Write flag in nilfs_lookup_dirty_data_buffers() and nilfs_lookup_dirty_node_buffers(); (3) clearing of BH_Async_Write flag in nilfs_segctor_complete_write(), nilfs_abort_logs(), nilfs_forget_buffer(), nilfs_clear_dirty_page(). Reported-by: Jerome Poulin <jeromepoulin@gmail.com> Reported-by: Anton Eliasson <devel@antoneliasson.se> Cc: Paul Fertser <fercerpav@gmail.com> Cc: ARAI Shun-ichi <hermes@ceres.dti.ne.jp> Cc: Piotr Szymaniak <szarpaj@grubelek.pl> Cc: Juan Barry Manuel Canham <Linux@riotingpacifist.net> Cc: Zahid Chowdhury <zahid.chowdhury@starsolutions.com> Cc: Elmer Zhang <freeboy6716@gmail.com> Cc: Kenneth Langga <klangga@gmail.com> Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-30 20:45:12 +00:00
set_buffer_async_write(bh);
if (bh == segbuf->sb_super_root) {
if (bh->b_page != bd_page) {
lock_page(bd_page);
clear_page_dirty_for_io(bd_page);
set_page_writeback(bd_page);
unlock_page(bd_page);
bd_page = bh->b_page;
}
break;
}
if (bh->b_page != fs_page) {
nilfs_begin_page_io(fs_page);
fs_page = bh->b_page;
}
}
}
if (bd_page) {
lock_page(bd_page);
clear_page_dirty_for_io(bd_page);
set_page_writeback(bd_page);
unlock_page(bd_page);
}
nilfs_begin_page_io(fs_page);
}
static int nilfs_segctor_write(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs)
{
int ret;
ret = nilfs_write_logs(&sci->sc_segbufs, nilfs);
list_splice_tail_init(&sci->sc_segbufs, &sci->sc_write_logs);
return ret;
}
static void nilfs_end_page_io(struct page *page, int err)
{
if (!page)
return;
if (buffer_nilfs_node(page_buffers(page)) && !PageWriteback(page)) {
/*
* For b-tree node pages, this function may be called twice
* or more because they might be split in a segment.
*/
if (PageDirty(page)) {
/*
* For pages holding split b-tree node buffers, dirty
* flag on the buffers may be cleared discretely.
* In that case, the page is once redirtied for
* remaining buffers, and it must be cancelled if
* all the buffers get cleaned later.
*/
lock_page(page);
if (nilfs_page_buffers_clean(page))
__nilfs_clear_page_dirty(page);
unlock_page(page);
}
return;
}
if (!err) {
if (!nilfs_page_buffers_clean(page))
__set_page_dirty_nobuffers(page);
ClearPageError(page);
} else {
__set_page_dirty_nobuffers(page);
SetPageError(page);
}
end_page_writeback(page);
}
static void nilfs_abort_logs(struct list_head *logs, int err)
{
struct nilfs_segment_buffer *segbuf;
struct page *bd_page = NULL, *fs_page = NULL;
struct buffer_head *bh;
if (list_empty(logs))
return;
list_for_each_entry(segbuf, logs, sb_list) {
list_for_each_entry(bh, &segbuf->sb_segsum_buffers,
b_assoc_buffers) {
if (bh->b_page != bd_page) {
if (bd_page)
end_page_writeback(bd_page);
bd_page = bh->b_page;
}
}
list_for_each_entry(bh, &segbuf->sb_payload_buffers,
b_assoc_buffers) {
nilfs2: fix issue with race condition of competition between segments for dirty blocks Many NILFS2 users were reported about strange file system corruption (for example): NILFS: bad btree node (blocknr=185027): level = 0, flags = 0x0, nchildren = 768 NILFS error (device sda4): nilfs_bmap_last_key: broken bmap (inode number=11540) But such error messages are consequence of file system's issue that takes place more earlier. Fortunately, Jerome Poulin <jeromepoulin@gmail.com> and Anton Eliasson <devel@antoneliasson.se> were reported about another issue not so recently. These reports describe the issue with segctor thread's crash: BUG: unable to handle kernel paging request at 0000000000004c83 IP: nilfs_end_page_io+0x12/0xd0 [nilfs2] Call Trace: nilfs_segctor_do_construct+0xf25/0x1b20 [nilfs2] nilfs_segctor_construct+0x17b/0x290 [nilfs2] nilfs_segctor_thread+0x122/0x3b0 [nilfs2] kthread+0xc0/0xd0 ret_from_fork+0x7c/0xb0 These two issues have one reason. This reason can raise third issue too. Third issue results in hanging of segctor thread with eating of 100% CPU. REPRODUCING PATH: One of the possible way or the issue reproducing was described by Jermoe me Poulin <jeromepoulin@gmail.com>: 1. init S to get to single user mode. 2. sysrq+E to make sure only my shell is running 3. start network-manager to get my wifi connection up 4. login as root and launch "screen" 5. cd /boot/log/nilfs which is a ext3 mount point and can log when NILFS dies. 6. lscp | xz -9e > lscp.txt.xz 7. mount my snapshot using mount -o cp=3360839,ro /dev/vgUbuntu/root /mnt/nilfs 8. start a screen to dump /proc/kmsg to text file since rsyslog is killed 9. start a screen and launch strace -f -o find-cat.log -t find /mnt/nilfs -type f -exec cat {} > /dev/null \; 10. start a screen and launch strace -f -o apt-get.log -t apt-get update 11. launch the last command again as it did not crash the first time 12. apt-get crashes 13. ps aux > ps-aux-crashed.log 13. sysrq+W 14. sysrq+E wait for everything to terminate 15. sysrq+SUSB Simplified way of the issue reproducing is starting kernel compilation task and "apt-get update" in parallel. REPRODUCIBILITY: The issue is reproduced not stable [60% - 80%]. It is very important to have proper environment for the issue reproducing. The critical conditions for successful reproducing: (1) It should have big modified file by mmap() way. (2) This file should have the count of dirty blocks are greater that several segments in size (for example, two or three) from time to time during processing. (3) It should be intensive background activity of files modification in another thread. INVESTIGATION: First of all, it is possible to see that the reason of crash is not valid page address: NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 NILFS [nilfs_segctor_complete_write]:2101 segbuf->sb_segnum 6783 Moreover, value of b_page (0x1a82) is 6786. This value looks like segment number. And b_blocknr with b_size values look like block numbers. So, buffer_head's pointer points on not proper address value. Detailed investigation of the issue is discovered such picture: [-----------------------------SEGMENT 6783-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111149024, segbuf->sb_segnum 6783 [-----------------------------SEGMENT 6784-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff8802174a6798, bh->b_assoc_buffers.prev ffff880221cffee8 NILFS [nilfs_segctor_do_construct]:2336 nilfs_segctor_assign NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 1, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111150080, segbuf->sb_segnum 6784, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111164416, segbuf->sb_segnum 6784, segbuf->sb_nbio 15 [-----------------------------SEGMENT 6785-------------------------------] NILFS [nilfs_segctor_do_construct]:2310 nilfs_segctor_begin_construction NILFS [nilfs_segctor_do_construct]:2321 nilfs_segctor_collect NILFS [nilfs_lookup_dirty_data_buffers]:782 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_lookup_dirty_data_buffers]:783 bh->b_assoc_buffers.next ffff880219277e80, bh->b_assoc_buffers.prev ffff880221cffc88 NILFS [nilfs_segctor_do_construct]:2367 nilfs_segctor_update_segusage NILFS [nilfs_segctor_do_construct]:2371 nilfs_segctor_prepare_write NILFS [nilfs_segctor_do_construct]:2376 nilfs_add_checksums_on_logs NILFS [nilfs_segctor_do_construct]:2381 nilfs_segctor_write NILFS [nilfs_segbuf_submit_bh]:575 bh->b_count 2, bh->b_page ffffea000709b000, page->index 0, i_ino 1033103, i_size 25165824 NILFS [nilfs_segbuf_submit_bh]:576 segbuf->sb_segnum 6785 NILFS [nilfs_segbuf_submit_bh]:577 bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111165440, segbuf->sb_segnum 6785, segbuf->sb_nbio 0 [----------] ditto NILFS [nilfs_segbuf_submit_bio]:464 bio->bi_sector 111177728, segbuf->sb_segnum 6785, segbuf->sb_nbio 12 NILFS [nilfs_segctor_do_construct]:2399 nilfs_segctor_wait NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6783 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6784 NILFS [nilfs_segbuf_wait]:676 segbuf->sb_segnum 6785 NILFS [nilfs_segctor_complete_write]:2100 bh->b_count 0, bh->b_blocknr 13895680, bh->b_size 13897727, bh->b_page 0000000000001a82 BUG: unable to handle kernel paging request at 0000000000001a82 IP: [<ffffffffa024d0f2>] nilfs_end_page_io+0x12/0xd0 [nilfs2] Usually, for every segment we collect dirty files in list. Then, dirty blocks are gathered for every dirty file, prepared for write and submitted by means of nilfs_segbuf_submit_bh() call. Finally, it takes place complete write phase after calling nilfs_end_bio_write() on the block layer. Buffers/pages are marked as not dirty on final phase and processed files removed from the list of dirty files. It is possible to see that we had three prepare_write and submit_bio phases before segbuf_wait and complete_write phase. Moreover, segments compete between each other for dirty blocks because on every iteration of segments processing dirty buffer_heads are added in several lists of payload_buffers: [SEGMENT 6784]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880218bcdf50 [SEGMENT 6785]: bh->b_assoc_buffers.next ffff880218a0d5f8, bh->b_assoc_buffers.prev ffff880222cc7ee8 The next pointer is the same but prev pointer has changed. It means that buffer_head has next pointer from one list but prev pointer from another. Such modification can be made several times. And, finally, it can be resulted in various issues: (1) segctor hanging, (2) segctor crashing, (3) file system metadata corruption. FIX: This patch adds: (1) setting of BH_Async_Write flag in nilfs_segctor_prepare_write() for every proccessed dirty block; (2) checking of BH_Async_Write flag in nilfs_lookup_dirty_data_buffers() and nilfs_lookup_dirty_node_buffers(); (3) clearing of BH_Async_Write flag in nilfs_segctor_complete_write(), nilfs_abort_logs(), nilfs_forget_buffer(), nilfs_clear_dirty_page(). Reported-by: Jerome Poulin <jeromepoulin@gmail.com> Reported-by: Anton Eliasson <devel@antoneliasson.se> Cc: Paul Fertser <fercerpav@gmail.com> Cc: ARAI Shun-ichi <hermes@ceres.dti.ne.jp> Cc: Piotr Szymaniak <szarpaj@grubelek.pl> Cc: Juan Barry Manuel Canham <Linux@riotingpacifist.net> Cc: Zahid Chowdhury <zahid.chowdhury@starsolutions.com> Cc: Elmer Zhang <freeboy6716@gmail.com> Cc: Kenneth Langga <klangga@gmail.com> Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-30 20:45:12 +00:00
clear_buffer_async_write(bh);
if (bh == segbuf->sb_super_root) {
if (bh->b_page != bd_page) {
end_page_writeback(bd_page);
bd_page = bh->b_page;
}
break;
}
if (bh->b_page != fs_page) {
nilfs_end_page_io(fs_page, err);
fs_page = bh->b_page;
}
}
}
if (bd_page)
end_page_writeback(bd_page);
nilfs_end_page_io(fs_page, err);
}
static void nilfs_segctor_abort_construction(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs, int err)
{
LIST_HEAD(logs);
int ret;
list_splice_tail_init(&sci->sc_write_logs, &logs);
ret = nilfs_wait_on_logs(&logs);
nilfs_abort_logs(&logs, ret ? : err);
list_splice_tail_init(&sci->sc_segbufs, &logs);
nilfs_cancel_segusage(&logs, nilfs->ns_sufile);
nilfs_free_incomplete_logs(&logs, nilfs);
if (sci->sc_stage.flags & NILFS_CF_SUFREED) {
ret = nilfs_sufile_cancel_freev(nilfs->ns_sufile,
sci->sc_freesegs,
sci->sc_nfreesegs,
NULL);
WARN_ON(ret); /* do not happen */
}
nilfs_destroy_logs(&logs);
}
static void nilfs_set_next_segment(struct the_nilfs *nilfs,
struct nilfs_segment_buffer *segbuf)
{
nilfs->ns_segnum = segbuf->sb_segnum;
nilfs->ns_nextnum = segbuf->sb_nextnum;
nilfs->ns_pseg_offset = segbuf->sb_pseg_start - segbuf->sb_fseg_start
+ segbuf->sb_sum.nblocks;
nilfs->ns_seg_seq = segbuf->sb_sum.seg_seq;
nilfs->ns_ctime = segbuf->sb_sum.ctime;
}
static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
{
struct nilfs_segment_buffer *segbuf;
struct page *bd_page = NULL, *fs_page = NULL;
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
int update_sr = false;
list_for_each_entry(segbuf, &sci->sc_write_logs, sb_list) {
struct buffer_head *bh;
list_for_each_entry(bh, &segbuf->sb_segsum_buffers,
b_assoc_buffers) {
set_buffer_uptodate(bh);
clear_buffer_dirty(bh);
if (bh->b_page != bd_page) {
if (bd_page)
end_page_writeback(bd_page);
bd_page = bh->b_page;
}
}
/*
* We assume that the buffers which belong to the same page
* continue over the buffer list.
* Under this assumption, the last BHs of pages is
* identifiable by the discontinuity of bh->b_page
* (page != fs_page).
*
* For B-tree node blocks, however, this assumption is not
* guaranteed. The cleanup code of B-tree node pages needs
* special care.
*/
list_for_each_entry(bh, &segbuf->sb_payload_buffers,
b_assoc_buffers) {
const unsigned long set_bits = BIT(BH_Uptodate);
const unsigned long clear_bits =
(BIT(BH_Dirty) | BIT(BH_Async_Write) |
BIT(BH_Delay) | BIT(BH_NILFS_Volatile) |
BIT(BH_NILFS_Redirected));
set_mask_bits(&bh->b_state, clear_bits, set_bits);
if (bh == segbuf->sb_super_root) {
if (bh->b_page != bd_page) {
end_page_writeback(bd_page);
bd_page = bh->b_page;
}
update_sr = true;
break;
}
if (bh->b_page != fs_page) {
nilfs_end_page_io(fs_page, 0);
fs_page = bh->b_page;
}
}
if (!nilfs_segbuf_simplex(segbuf)) {
if (segbuf->sb_sum.flags & NILFS_SS_LOGBGN) {
set_bit(NILFS_SC_UNCLOSED, &sci->sc_flags);
sci->sc_lseg_stime = jiffies;
}
if (segbuf->sb_sum.flags & NILFS_SS_LOGEND)
clear_bit(NILFS_SC_UNCLOSED, &sci->sc_flags);
}
}
/*
* Since pages may continue over multiple segment buffers,
* end of the last page must be checked outside of the loop.
*/
if (bd_page)
end_page_writeback(bd_page);
nilfs_end_page_io(fs_page, 0);
nilfs_drop_collected_inodes(&sci->sc_dirty_files);
if (nilfs_doing_gc())
nilfs_drop_collected_inodes(&sci->sc_gc_inodes);
else
nilfs->ns_nongc_ctime = sci->sc_seg_ctime;
sci->sc_nblk_inc += sci->sc_nblk_this_inc;
segbuf = NILFS_LAST_SEGBUF(&sci->sc_write_logs);
nilfs_set_next_segment(nilfs, segbuf);
if (update_sr) {
nilfs->ns_flushed_device = 0;
nilfs_set_last_segment(nilfs, segbuf->sb_pseg_start,
segbuf->sb_sum.seg_seq, nilfs->ns_cno++);
clear_bit(NILFS_SC_HAVE_DELTA, &sci->sc_flags);
clear_bit(NILFS_SC_DIRTY, &sci->sc_flags);
set_bit(NILFS_SC_SUPER_ROOT, &sci->sc_flags);
nilfs_segctor_clear_metadata_dirty(sci);
} else
clear_bit(NILFS_SC_SUPER_ROOT, &sci->sc_flags);
}
static int nilfs_segctor_wait(struct nilfs_sc_info *sci)
{
int ret;
ret = nilfs_wait_on_logs(&sci->sc_write_logs);
if (!ret) {
nilfs_segctor_complete_write(sci);
nilfs_destroy_logs(&sci->sc_write_logs);
}
return ret;
}
static int nilfs_segctor_collect_dirty_files(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs)
{
struct nilfs_inode_info *ii, *n;
struct inode *ifile = sci->sc_root->ifile;
spin_lock(&nilfs->ns_inode_lock);
retry:
list_for_each_entry_safe(ii, n, &nilfs->ns_dirty_files, i_dirty) {
if (!ii->i_bh) {
struct buffer_head *ibh;
int err;
spin_unlock(&nilfs->ns_inode_lock);
err = nilfs_ifile_get_inode_block(
ifile, ii->vfs_inode.i_ino, &ibh);
if (unlikely(err)) {
nilfs_warn(sci->sc_super,
"log writer: error %d getting inode block (ino=%lu)",
err, ii->vfs_inode.i_ino);
return err;
}
spin_lock(&nilfs->ns_inode_lock);
if (likely(!ii->i_bh))
ii->i_bh = ibh;
else
brelse(ibh);
goto retry;
}
nilfs2: fix race condition that causes file system corruption There is a race condition between nilfs_dirty_inode() and nilfs_set_file_dirty(). When a file is opened, nilfs_dirty_inode() is called to update the access timestamp in the inode. It calls __nilfs_mark_inode_dirty() in a separate transaction. __nilfs_mark_inode_dirty() caches the ifile buffer_head in the i_bh field of the inode info structure and marks it as dirty. After some data was written to the file in another transaction, the function nilfs_set_file_dirty() is called, which adds the inode to the ns_dirty_files list. Then the segment construction calls nilfs_segctor_collect_dirty_files(), which goes through the ns_dirty_files list and checks the i_bh field. If there is a cached buffer_head in i_bh it is not marked as dirty again. Since nilfs_dirty_inode() and nilfs_set_file_dirty() use separate transactions, it is possible that a segment construction that writes out the ifile occurs in-between the two. If this happens the inode is not on the ns_dirty_files list, but its ifile block is still marked as dirty and written out. In the next segment construction, the data for the file is written out and nilfs_bmap_propagate() updates the b-tree. Eventually the bmap root is written into the i_bh block, which is not dirty, because it was written out in another segment construction. As a result the bmap update can be lost, which leads to file system corruption. Either the virtual block address points to an unallocated DAT block, or the DAT entry will be reused for something different. The error can remain undetected for a long time. A typical error message would be one of the "bad btree" errors or a warning that a DAT entry could not be found. This bug can be reproduced reliably by a simple benchmark that creates and overwrites millions of 4k files. Link: http://lkml.kernel.org/r/1509367935-3086-2-git-send-email-konishi.ryusuke@lab.ntt.co.jp Signed-off-by: Andreas Rohner <andreas.rohner@gmx.net> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Tested-by: Andreas Rohner <andreas.rohner@gmx.net> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-17 23:29:35 +00:00
// Always redirty the buffer to avoid race condition
mark_buffer_dirty(ii->i_bh);
nilfs_mdt_mark_dirty(ifile);
clear_bit(NILFS_I_QUEUED, &ii->i_state);
set_bit(NILFS_I_BUSY, &ii->i_state);
list_move_tail(&ii->i_dirty, &sci->sc_dirty_files);
}
spin_unlock(&nilfs->ns_inode_lock);
return 0;
}
static void nilfs_segctor_drop_written_files(struct nilfs_sc_info *sci,
struct the_nilfs *nilfs)
{
struct nilfs_inode_info *ii, *n;
Rename superblock flags (MS_xyz -> SB_xyz) This is a pure automated search-and-replace of the internal kernel superblock flags. The s_flags are now called SB_*, with the names and the values for the moment mirroring the MS_* flags that they're equivalent to. Note how the MS_xyz flags are the ones passed to the mount system call, while the SB_xyz flags are what we then use in sb->s_flags. The script to do this was: # places to look in; re security/*: it generally should *not* be # touched (that stuff parses mount(2) arguments directly), but # there are two places where we really deal with superblock flags. FILES="drivers/mtd drivers/staging/lustre fs ipc mm \ include/linux/fs.h include/uapi/linux/bfs_fs.h \ security/apparmor/apparmorfs.c security/apparmor/include/lib.h" # the list of MS_... constants SYMS="RDONLY NOSUID NODEV NOEXEC SYNCHRONOUS REMOUNT MANDLOCK \ DIRSYNC NOATIME NODIRATIME BIND MOVE REC VERBOSE SILENT \ POSIXACL UNBINDABLE PRIVATE SLAVE SHARED RELATIME KERNMOUNT \ I_VERSION STRICTATIME LAZYTIME SUBMOUNT NOREMOTELOCK NOSEC BORN \ ACTIVE NOUSER" SED_PROG= for i in $SYMS; do SED_PROG="$SED_PROG -e s/MS_$i/SB_$i/g"; done # we want files that contain at least one of MS_..., # with fs/namespace.c and fs/pnode.c excluded. L=$(for i in $SYMS; do git grep -w -l MS_$i $FILES; done| sort|uniq|grep -v '^fs/namespace.c'|grep -v '^fs/pnode.c') for f in $L; do sed -i $f $SED_PROG; done Requested-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-27 21:05:09 +00:00
int during_mount = !(sci->sc_super->s_flags & SB_ACTIVE);
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
int defer_iput = false;
spin_lock(&nilfs->ns_inode_lock);
list_for_each_entry_safe(ii, n, &sci->sc_dirty_files, i_dirty) {
if (!test_and_clear_bit(NILFS_I_UPDATED, &ii->i_state) ||
test_bit(NILFS_I_DIRTY, &ii->i_state))
continue;
clear_bit(NILFS_I_BUSY, &ii->i_state);
brelse(ii->i_bh);
ii->i_bh = NULL;
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
list_del_init(&ii->i_dirty);
nilfs2: fix deadlock of segment constructor during recovery According to a report from Yuxuan Shui, nilfs2 in kernel 3.19 got stuck during recovery at mount time. The code path that caused the deadlock was as follows: nilfs_fill_super() load_nilfs() nilfs_salvage_orphan_logs() * Do roll-forwarding, attach segment constructor for recovery, and kick it. nilfs_segctor_thread() nilfs_segctor_thread_construct() * A lock is held with nilfs_transaction_lock() nilfs_segctor_do_construct() nilfs_segctor_drop_written_files() iput() iput_final() write_inode_now() writeback_single_inode() __writeback_single_inode() do_writepages() nilfs_writepage() nilfs_construct_dsync_segment() nilfs_transaction_lock() --> deadlock This can happen if commit 7ef3ff2fea8b ("nilfs2: fix deadlock of segment constructor over I_SYNC flag") is applied and roll-forward recovery was performed at mount time. The roll-forward recovery can happen if datasync write is done and the file system crashes immediately after that. For instance, we can reproduce the issue with the following steps: < nilfs2 is mounted on /nilfs (device: /dev/sdb1) > # dd if=/dev/zero of=/nilfs/test bs=4k count=1 && sync # dd if=/dev/zero of=/nilfs/test conv=notrunc oflag=dsync bs=4k count=1 && reboot -nfh < the system will immediately reboot > # mount -t nilfs2 /dev/sdb1 /nilfs The deadlock occurs because iput() can run segment constructor through writeback_single_inode() if MS_ACTIVE flag is not set on sb->s_flags. The above commit changed segment constructor so that it calls iput() asynchronously for inodes with i_nlink == 0, but that change was imperfect. This fixes the another deadlock by deferring iput() in segment constructor even for the case that mount is not finished, that is, for the case that MS_ACTIVE flag is not set. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Reported-by: Yuxuan Shui <yshuiv7@gmail.com> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-03-12 23:26:00 +00:00
if (!ii->vfs_inode.i_nlink || during_mount) {
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
/*
nilfs2: fix deadlock of segment constructor during recovery According to a report from Yuxuan Shui, nilfs2 in kernel 3.19 got stuck during recovery at mount time. The code path that caused the deadlock was as follows: nilfs_fill_super() load_nilfs() nilfs_salvage_orphan_logs() * Do roll-forwarding, attach segment constructor for recovery, and kick it. nilfs_segctor_thread() nilfs_segctor_thread_construct() * A lock is held with nilfs_transaction_lock() nilfs_segctor_do_construct() nilfs_segctor_drop_written_files() iput() iput_final() write_inode_now() writeback_single_inode() __writeback_single_inode() do_writepages() nilfs_writepage() nilfs_construct_dsync_segment() nilfs_transaction_lock() --> deadlock This can happen if commit 7ef3ff2fea8b ("nilfs2: fix deadlock of segment constructor over I_SYNC flag") is applied and roll-forward recovery was performed at mount time. The roll-forward recovery can happen if datasync write is done and the file system crashes immediately after that. For instance, we can reproduce the issue with the following steps: < nilfs2 is mounted on /nilfs (device: /dev/sdb1) > # dd if=/dev/zero of=/nilfs/test bs=4k count=1 && sync # dd if=/dev/zero of=/nilfs/test conv=notrunc oflag=dsync bs=4k count=1 && reboot -nfh < the system will immediately reboot > # mount -t nilfs2 /dev/sdb1 /nilfs The deadlock occurs because iput() can run segment constructor through writeback_single_inode() if MS_ACTIVE flag is not set on sb->s_flags. The above commit changed segment constructor so that it calls iput() asynchronously for inodes with i_nlink == 0, but that change was imperfect. This fixes the another deadlock by deferring iput() in segment constructor even for the case that mount is not finished, that is, for the case that MS_ACTIVE flag is not set. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Reported-by: Yuxuan Shui <yshuiv7@gmail.com> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-03-12 23:26:00 +00:00
* Defer calling iput() to avoid deadlocks if
* i_nlink == 0 or mount is not yet finished.
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
*/
list_add_tail(&ii->i_dirty, &sci->sc_iput_queue);
defer_iput = true;
} else {
spin_unlock(&nilfs->ns_inode_lock);
iput(&ii->vfs_inode);
spin_lock(&nilfs->ns_inode_lock);
}
}
spin_unlock(&nilfs->ns_inode_lock);
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
if (defer_iput)
schedule_work(&sci->sc_iput_work);
}
/*
* Main procedure of segment constructor
*/
static int nilfs_segctor_do_construct(struct nilfs_sc_info *sci, int mode)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
int err;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_set(sci, NILFS_ST_INIT);
sci->sc_cno = nilfs->ns_cno;
err = nilfs_segctor_collect_dirty_files(sci, nilfs);
if (unlikely(err))
goto out;
if (nilfs_test_metadata_dirty(nilfs, sci->sc_root))
set_bit(NILFS_SC_DIRTY, &sci->sc_flags);
if (nilfs_segctor_clean(sci))
goto out;
do {
sci->sc_stage.flags &= ~NILFS_CF_HISTORY_MASK;
err = nilfs_segctor_begin_construction(sci, nilfs);
if (unlikely(err))
goto out;
/* Update time stamp */
sci->sc_seg_ctime = ktime_get_real_seconds();
err = nilfs_segctor_collect(sci, nilfs, mode);
if (unlikely(err))
goto failed;
/* Avoid empty segment */
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
if (nilfs_sc_cstage_get(sci) == NILFS_ST_DONE &&
nilfs_segbuf_empty(sci->sc_curseg)) {
nilfs_segctor_abort_construction(sci, nilfs, 1);
goto out;
}
err = nilfs_segctor_assign(sci, mode);
if (unlikely(err))
goto failed;
if (sci->sc_stage.flags & NILFS_CF_IFILE_STARTED)
nilfs_segctor_fill_in_file_bmap(sci);
if (mode == SC_LSEG_SR &&
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
nilfs_sc_cstage_get(sci) >= NILFS_ST_CPFILE) {
err = nilfs_segctor_fill_in_checkpoint(sci);
if (unlikely(err))
goto failed_to_write;
nilfs_segctor_fill_in_super_root(sci, nilfs);
}
nilfs_segctor_update_segusage(sci, nilfs->ns_sufile);
/* Write partial segments */
nilfs_segctor_prepare_write(sci);
nilfs_add_checksums_on_logs(&sci->sc_segbufs,
nilfs->ns_crc_seed);
err = nilfs_segctor_write(sci, nilfs);
if (unlikely(err))
goto failed_to_write;
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
if (nilfs_sc_cstage_get(sci) == NILFS_ST_DONE ||
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
nilfs->ns_blocksize_bits != PAGE_SHIFT) {
/*
* At this point, we avoid double buffering
* for blocksize < pagesize because page dirty
* flag is turned off during write and dirty
* buffers are not properly collected for
* pages crossing over segments.
*/
err = nilfs_segctor_wait(sci);
if (err)
goto failed_to_write;
}
nilfs2: add a tracepoint for tracking stage transition of segment construction This patch adds a tracepoint for tracking stage transition of block collection in segment construction. With the tracepoint, we can analysis the behavior of segment construction in depth. It would be useful for bottleneck detection and debugging, etc. The tracepoint is created with the standard trace API of linux (like ext3, ext4, f2fs and btrfs). So we can analysis with existing tools easily. Of course, more detailed analysis will be possible if we can create nilfs specific analysis tools. Below is an example of event dump with Brendan Gregg's perf-tools (https://github.com/brendangregg/perf-tools). Time consumption between each stage can be obtained. $ sudo bin/tpoint nilfs2:nilfs2_collection_stage_transition Tracing nilfs2:nilfs2_collection_stage_transition. Ctrl-C to end. segctord-14875 [003] ...1 28311.067794: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_INIT segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_GC segctord-14875 [003] ...1 28311.068139: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_FILE segctord-14875 [003] ...1 28311.068486: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_IFILE segctord-14875 [003] ...1 28311.068540: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_CPFILE segctord-14875 [003] ...1 28311.068561: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SUFILE segctord-14875 [003] ...1 28311.068565: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DAT segctord-14875 [003] ...1 28311.068573: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_SR segctord-14875 [003] ...1 28311.068574: nilfs2_collection_stage_transition: sci = ffff8800ce6de000 stage = ST_DONE For capturing transition correctly, this patch adds wrappers for the member scnt of nilfs_cstage. With this change, every transition of the stage can produce trace event in a correct manner. Signed-off-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:31:59 +00:00
} while (nilfs_sc_cstage_get(sci) != NILFS_ST_DONE);
out:
nilfs_segctor_drop_written_files(sci, nilfs);
return err;
failed_to_write:
if (sci->sc_stage.flags & NILFS_CF_IFILE_STARTED)
nilfs_redirty_inodes(&sci->sc_dirty_files);
failed:
if (nilfs_doing_gc())
nilfs_redirty_inodes(&sci->sc_gc_inodes);
nilfs_segctor_abort_construction(sci, nilfs, err);
goto out;
}
/**
* nilfs_segctor_start_timer - set timer of background write
* @sci: nilfs_sc_info
*
* If the timer has already been set, it ignores the new request.
* This function MUST be called within a section locking the segment
* semaphore.
*/
static void nilfs_segctor_start_timer(struct nilfs_sc_info *sci)
{
spin_lock(&sci->sc_state_lock);
if (!(sci->sc_state & NILFS_SEGCTOR_COMMIT)) {
sci->sc_timer.expires = jiffies + sci->sc_interval;
add_timer(&sci->sc_timer);
sci->sc_state |= NILFS_SEGCTOR_COMMIT;
}
spin_unlock(&sci->sc_state_lock);
}
static void nilfs_segctor_do_flush(struct nilfs_sc_info *sci, int bn)
{
spin_lock(&sci->sc_state_lock);
if (!(sci->sc_flush_request & BIT(bn))) {
unsigned long prev_req = sci->sc_flush_request;
sci->sc_flush_request |= BIT(bn);
if (!prev_req)
wake_up(&sci->sc_wait_daemon);
}
spin_unlock(&sci->sc_state_lock);
}
/**
* nilfs_flush_segment - trigger a segment construction for resource control
* @sb: super block
* @ino: inode number of the file to be flushed out.
*/
void nilfs_flush_segment(struct super_block *sb, ino_t ino)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
if (!sci || nilfs_doing_construction())
return;
nilfs_segctor_do_flush(sci, NILFS_MDT_INODE(sb, ino) ? ino : 0);
/* assign bit 0 to data files */
}
struct nilfs_segctor_wait_request {
wait_queue_entry_t wq;
__u32 seq;
int err;
atomic_t done;
};
static int nilfs_segctor_sync(struct nilfs_sc_info *sci)
{
struct nilfs_segctor_wait_request wait_req;
int err = 0;
spin_lock(&sci->sc_state_lock);
init_wait(&wait_req.wq);
wait_req.err = 0;
atomic_set(&wait_req.done, 0);
wait_req.seq = ++sci->sc_seq_request;
spin_unlock(&sci->sc_state_lock);
init_waitqueue_entry(&wait_req.wq, current);
add_wait_queue(&sci->sc_wait_request, &wait_req.wq);
set_current_state(TASK_INTERRUPTIBLE);
wake_up(&sci->sc_wait_daemon);
for (;;) {
if (atomic_read(&wait_req.done)) {
err = wait_req.err;
break;
}
if (!signal_pending(current)) {
schedule();
continue;
}
err = -ERESTARTSYS;
break;
}
finish_wait(&sci->sc_wait_request, &wait_req.wq);
return err;
}
static void nilfs_segctor_wakeup(struct nilfs_sc_info *sci, int err)
{
struct nilfs_segctor_wait_request *wrq, *n;
unsigned long flags;
spin_lock_irqsave(&sci->sc_wait_request.lock, flags);
sched/wait: Disambiguate wq_entry->task_list and wq_head->task_list naming So I've noticed a number of instances where it was not obvious from the code whether ->task_list was for a wait-queue head or a wait-queue entry. Furthermore, there's a number of wait-queue users where the lists are not for 'tasks' but other entities (poll tables, etc.), in which case the 'task_list' name is actively confusing. To clear this all up, name the wait-queue head and entry list structure fields unambiguously: struct wait_queue_head::task_list => ::head struct wait_queue_entry::task_list => ::entry For example, this code: rqw->wait.task_list.next != &wait->task_list ... is was pretty unclear (to me) what it's doing, while now it's written this way: rqw->wait.head.next != &wait->entry ... which makes it pretty clear that we are iterating a list until we see the head. Other examples are: list_for_each_entry_safe(pos, next, &x->task_list, task_list) { list_for_each_entry(wq, &fence->wait.task_list, task_list) { ... where it's unclear (to me) what we are iterating, and during review it's hard to tell whether it's trying to walk a wait-queue entry (which would be a bug), while now it's written as: list_for_each_entry_safe(pos, next, &x->head, entry) { list_for_each_entry(wq, &fence->wait.head, entry) { Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-06-20 10:06:46 +00:00
list_for_each_entry_safe(wrq, n, &sci->sc_wait_request.head, wq.entry) {
if (!atomic_read(&wrq->done) &&
nilfs_cnt32_ge(sci->sc_seq_done, wrq->seq)) {
wrq->err = err;
atomic_set(&wrq->done, 1);
}
if (atomic_read(&wrq->done)) {
wrq->wq.func(&wrq->wq,
TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE,
0, NULL);
}
}
spin_unlock_irqrestore(&sci->sc_wait_request.lock, flags);
}
/**
* nilfs_construct_segment - construct a logical segment
* @sb: super block
*
* Return Value: On success, 0 is retured. On errors, one of the following
* negative error code is returned.
*
* %-EROFS - Read only filesystem.
*
* %-EIO - I/O error
*
* %-ENOSPC - No space left on device (only in a panic state).
*
* %-ERESTARTSYS - Interrupted.
*
* %-ENOMEM - Insufficient memory available.
*/
int nilfs_construct_segment(struct super_block *sb)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
struct nilfs_transaction_info *ti;
int err;
if (!sci)
return -EROFS;
/* A call inside transactions causes a deadlock. */
BUG_ON((ti = current->journal_info) && ti->ti_magic == NILFS_TI_MAGIC);
err = nilfs_segctor_sync(sci);
return err;
}
/**
* nilfs_construct_dsync_segment - construct a data-only logical segment
* @sb: super block
* @inode: inode whose data blocks should be written out
* @start: start byte offset
* @end: end byte offset (inclusive)
*
* Return Value: On success, 0 is retured. On errors, one of the following
* negative error code is returned.
*
* %-EROFS - Read only filesystem.
*
* %-EIO - I/O error
*
* %-ENOSPC - No space left on device (only in a panic state).
*
* %-ERESTARTSYS - Interrupted.
*
* %-ENOMEM - Insufficient memory available.
*/
int nilfs_construct_dsync_segment(struct super_block *sb, struct inode *inode,
loff_t start, loff_t end)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
struct nilfs_inode_info *ii;
struct nilfs_transaction_info ti;
int err = 0;
if (!sci)
return -EROFS;
nilfs_transaction_lock(sb, &ti, 0);
ii = NILFS_I(inode);
if (test_bit(NILFS_I_INODE_SYNC, &ii->i_state) ||
nilfs_test_opt(nilfs, STRICT_ORDER) ||
test_bit(NILFS_SC_UNCLOSED, &sci->sc_flags) ||
nilfs_discontinued(nilfs)) {
nilfs_transaction_unlock(sb);
err = nilfs_segctor_sync(sci);
return err;
}
spin_lock(&nilfs->ns_inode_lock);
if (!test_bit(NILFS_I_QUEUED, &ii->i_state) &&
!test_bit(NILFS_I_BUSY, &ii->i_state)) {
spin_unlock(&nilfs->ns_inode_lock);
nilfs_transaction_unlock(sb);
return 0;
}
spin_unlock(&nilfs->ns_inode_lock);
sci->sc_dsync_inode = ii;
sci->sc_dsync_start = start;
sci->sc_dsync_end = end;
err = nilfs_segctor_do_construct(sci, SC_LSEG_DSYNC);
if (!err)
nilfs->ns_flushed_device = 0;
nilfs_transaction_unlock(sb);
return err;
}
#define FLUSH_FILE_BIT (0x1) /* data file only */
#define FLUSH_DAT_BIT BIT(NILFS_DAT_INO) /* DAT only */
/**
* nilfs_segctor_accept - record accepted sequence count of log-write requests
* @sci: segment constructor object
*/
static void nilfs_segctor_accept(struct nilfs_sc_info *sci)
{
spin_lock(&sci->sc_state_lock);
sci->sc_seq_accepted = sci->sc_seq_request;
spin_unlock(&sci->sc_state_lock);
del_timer_sync(&sci->sc_timer);
}
/**
* nilfs_segctor_notify - notify the result of request to caller threads
* @sci: segment constructor object
* @mode: mode of log forming
* @err: error code to be notified
*/
static void nilfs_segctor_notify(struct nilfs_sc_info *sci, int mode, int err)
{
/* Clear requests (even when the construction failed) */
spin_lock(&sci->sc_state_lock);
if (mode == SC_LSEG_SR) {
sci->sc_state &= ~NILFS_SEGCTOR_COMMIT;
sci->sc_seq_done = sci->sc_seq_accepted;
nilfs_segctor_wakeup(sci, err);
sci->sc_flush_request = 0;
} else {
if (mode == SC_FLUSH_FILE)
sci->sc_flush_request &= ~FLUSH_FILE_BIT;
else if (mode == SC_FLUSH_DAT)
sci->sc_flush_request &= ~FLUSH_DAT_BIT;
/* re-enable timer if checkpoint creation was not done */
if ((sci->sc_state & NILFS_SEGCTOR_COMMIT) &&
time_before(jiffies, sci->sc_timer.expires))
add_timer(&sci->sc_timer);
}
spin_unlock(&sci->sc_state_lock);
}
/**
* nilfs_segctor_construct - form logs and write them to disk
* @sci: segment constructor object
* @mode: mode of log forming
*/
static int nilfs_segctor_construct(struct nilfs_sc_info *sci, int mode)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
struct nilfs_super_block **sbp;
int err = 0;
nilfs_segctor_accept(sci);
if (nilfs_discontinued(nilfs))
mode = SC_LSEG_SR;
if (!nilfs_segctor_confirm(sci))
err = nilfs_segctor_do_construct(sci, mode);
if (likely(!err)) {
if (mode != SC_FLUSH_DAT)
atomic_set(&nilfs->ns_ndirtyblks, 0);
if (test_bit(NILFS_SC_SUPER_ROOT, &sci->sc_flags) &&
nilfs_discontinued(nilfs)) {
down_write(&nilfs->ns_sem);
err = -EIO;
sbp = nilfs_prepare_super(sci->sc_super,
nilfs_sb_will_flip(nilfs));
if (likely(sbp)) {
nilfs_set_log_cursor(sbp[0], nilfs);
err = nilfs_commit_super(sci->sc_super,
NILFS_SB_COMMIT);
}
up_write(&nilfs->ns_sem);
}
}
nilfs_segctor_notify(sci, mode, err);
return err;
}
static void nilfs_construction_timeout(struct timer_list *t)
{
struct nilfs_sc_info *sci = from_timer(sci, t, sc_timer);
wake_up_process(sci->sc_timer_task);
}
static void
nilfs_remove_written_gcinodes(struct the_nilfs *nilfs, struct list_head *head)
{
struct nilfs_inode_info *ii, *n;
list_for_each_entry_safe(ii, n, head, i_dirty) {
if (!test_bit(NILFS_I_UPDATED, &ii->i_state))
continue;
list_del_init(&ii->i_dirty);
truncate_inode_pages(&ii->vfs_inode.i_data, 0);
nilfs_btnode_cache_clear(&ii->i_btnode_cache);
iput(&ii->vfs_inode);
}
}
nilfs2: fix lock order reversal in nilfs_clean_segments ioctl This is a companion patch to ("nilfs2: fix possible circular locking for get information ioctls"). This corrects lock order reversal between mm->mmap_sem and nilfs->ns_segctor_sem in nilfs_clean_segments() which was detected by lockdep check: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.30-rc3-nilfs-00003-g360bdc1 #7 ------------------------------------------------------- mmap/5294 is trying to acquire lock: (&nilfs->ns_segctor_sem){++++.+}, at: [<d0d0e846>] nilfs_transaction_begin+0xb6/0x10c [nilfs2] but task is already holding lock: (&mm->mmap_sem){++++++}, at: [<c043700a>] do_page_fault+0x1d8/0x30a which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&mm->mmap_sem){++++++}: [<c01470a5>] __lock_acquire+0x1066/0x13b0 [<c01474a9>] lock_acquire+0xba/0xdd [<c01836bc>] might_fault+0x68/0x88 [<c023c61d>] copy_from_user+0x2a/0x111 [<d0d120d0>] nilfs_ioctl_prepare_clean_segments+0x1d/0xf1 [nilfs2] [<d0d0e2aa>] nilfs_clean_segments+0x6d/0x1b9 [nilfs2] [<d0d11f68>] nilfs_ioctl+0x2ad/0x318 [nilfs2] [<c01a3be7>] vfs_ioctl+0x22/0x69 [<c01a408e>] do_vfs_ioctl+0x460/0x499 [<c01a4107>] sys_ioctl+0x40/0x5a [<c01031a4>] sysenter_do_call+0x12/0x38 [<ffffffff>] 0xffffffff -> #0 (&nilfs->ns_segctor_sem){++++.+}: [<c0146e0b>] __lock_acquire+0xdcc/0x13b0 [<c01474a9>] lock_acquire+0xba/0xdd [<c0433f1d>] down_read+0x2a/0x3e [<d0d0e846>] nilfs_transaction_begin+0xb6/0x10c [nilfs2] [<d0cfe0e5>] nilfs_page_mkwrite+0xe7/0x154 [nilfs2] [<c0183b0b>] __do_fault+0x165/0x376 [<c01855cd>] handle_mm_fault+0x287/0x5d1 [<c043712d>] do_page_fault+0x2fb/0x30a [<c0435462>] error_code+0x72/0x78 [<ffffffff>] 0xffffffff where nilfs_clean_segments() holds: nilfs->ns_segctor_sem -> copy_from_user() --> page fault -> mm->mmap_sem And, page fault path may hold: page fault -> mm->mmap_sem --> nilfs_page_mkwrite() -> nilfs->ns_segctor_sem Even though nilfs_clean_segments() does not perform write access on given user pages, it may cause deadlock because nilfs->ns_segctor_sem is shared per device and mm->mmap_sem can be shared with other tasks. To avoid this problem, this patch moves all calls of copy_from_user() outside the nilfs->ns_segctor_sem lock in the ioctl. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
2009-05-10 13:41:43 +00:00
int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
void **kbufs)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci = nilfs->ns_writer;
struct nilfs_transaction_info ti;
int err;
if (unlikely(!sci))
return -EROFS;
nilfs_transaction_lock(sb, &ti, 1);
err = nilfs_mdt_save_to_shadow_map(nilfs->ns_dat);
if (unlikely(err))
goto out_unlock;
nilfs2: fix lock order reversal in nilfs_clean_segments ioctl This is a companion patch to ("nilfs2: fix possible circular locking for get information ioctls"). This corrects lock order reversal between mm->mmap_sem and nilfs->ns_segctor_sem in nilfs_clean_segments() which was detected by lockdep check: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.30-rc3-nilfs-00003-g360bdc1 #7 ------------------------------------------------------- mmap/5294 is trying to acquire lock: (&nilfs->ns_segctor_sem){++++.+}, at: [<d0d0e846>] nilfs_transaction_begin+0xb6/0x10c [nilfs2] but task is already holding lock: (&mm->mmap_sem){++++++}, at: [<c043700a>] do_page_fault+0x1d8/0x30a which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&mm->mmap_sem){++++++}: [<c01470a5>] __lock_acquire+0x1066/0x13b0 [<c01474a9>] lock_acquire+0xba/0xdd [<c01836bc>] might_fault+0x68/0x88 [<c023c61d>] copy_from_user+0x2a/0x111 [<d0d120d0>] nilfs_ioctl_prepare_clean_segments+0x1d/0xf1 [nilfs2] [<d0d0e2aa>] nilfs_clean_segments+0x6d/0x1b9 [nilfs2] [<d0d11f68>] nilfs_ioctl+0x2ad/0x318 [nilfs2] [<c01a3be7>] vfs_ioctl+0x22/0x69 [<c01a408e>] do_vfs_ioctl+0x460/0x499 [<c01a4107>] sys_ioctl+0x40/0x5a [<c01031a4>] sysenter_do_call+0x12/0x38 [<ffffffff>] 0xffffffff -> #0 (&nilfs->ns_segctor_sem){++++.+}: [<c0146e0b>] __lock_acquire+0xdcc/0x13b0 [<c01474a9>] lock_acquire+0xba/0xdd [<c0433f1d>] down_read+0x2a/0x3e [<d0d0e846>] nilfs_transaction_begin+0xb6/0x10c [nilfs2] [<d0cfe0e5>] nilfs_page_mkwrite+0xe7/0x154 [nilfs2] [<c0183b0b>] __do_fault+0x165/0x376 [<c01855cd>] handle_mm_fault+0x287/0x5d1 [<c043712d>] do_page_fault+0x2fb/0x30a [<c0435462>] error_code+0x72/0x78 [<ffffffff>] 0xffffffff where nilfs_clean_segments() holds: nilfs->ns_segctor_sem -> copy_from_user() --> page fault -> mm->mmap_sem And, page fault path may hold: page fault -> mm->mmap_sem --> nilfs_page_mkwrite() -> nilfs->ns_segctor_sem Even though nilfs_clean_segments() does not perform write access on given user pages, it may cause deadlock because nilfs->ns_segctor_sem is shared per device and mm->mmap_sem can be shared with other tasks. To avoid this problem, this patch moves all calls of copy_from_user() outside the nilfs->ns_segctor_sem lock in the ioctl. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
2009-05-10 13:41:43 +00:00
err = nilfs_ioctl_prepare_clean_segments(nilfs, argv, kbufs);
if (unlikely(err)) {
nilfs_mdt_restore_from_shadow_map(nilfs->ns_dat);
goto out_unlock;
}
sci->sc_freesegs = kbufs[4];
sci->sc_nfreesegs = argv[4].v_nmembs;
list_splice_tail_init(&nilfs->ns_gc_inodes, &sci->sc_gc_inodes);
for (;;) {
err = nilfs_segctor_construct(sci, SC_LSEG_SR);
nilfs_remove_written_gcinodes(nilfs, &sci->sc_gc_inodes);
if (likely(!err))
break;
nilfs_warn(sb, "error %d cleaning segments", err);
set_current_state(TASK_INTERRUPTIBLE);
schedule_timeout(sci->sc_interval);
}
if (nilfs_test_opt(nilfs, DISCARD)) {
int ret = nilfs_discard_segments(nilfs, sci->sc_freesegs,
sci->sc_nfreesegs);
if (ret) {
nilfs_warn(sb,
"error %d on discard request, turning discards off for the device",
ret);
nilfs_clear_opt(nilfs, DISCARD);
}
}
out_unlock:
sci->sc_freesegs = NULL;
sci->sc_nfreesegs = 0;
nilfs_mdt_clear_shadow_map(nilfs->ns_dat);
nilfs_transaction_unlock(sb);
return err;
}
static void nilfs_segctor_thread_construct(struct nilfs_sc_info *sci, int mode)
{
struct nilfs_transaction_info ti;
nilfs_transaction_lock(sci->sc_super, &ti, 0);
nilfs_segctor_construct(sci, mode);
/*
* Unclosed segment should be retried. We do this using sc_timer.
* Timeout of sc_timer will invoke complete construction which leads
* to close the current logical segment.
*/
if (test_bit(NILFS_SC_UNCLOSED, &sci->sc_flags))
nilfs_segctor_start_timer(sci);
nilfs_transaction_unlock(sci->sc_super);
}
static void nilfs_segctor_do_immediate_flush(struct nilfs_sc_info *sci)
{
int mode = 0;
spin_lock(&sci->sc_state_lock);
mode = (sci->sc_flush_request & FLUSH_DAT_BIT) ?
SC_FLUSH_DAT : SC_FLUSH_FILE;
spin_unlock(&sci->sc_state_lock);
if (mode) {
2015-11-07 00:32:14 +00:00
nilfs_segctor_do_construct(sci, mode);
spin_lock(&sci->sc_state_lock);
sci->sc_flush_request &= (mode == SC_FLUSH_FILE) ?
~FLUSH_FILE_BIT : ~FLUSH_DAT_BIT;
spin_unlock(&sci->sc_state_lock);
}
clear_bit(NILFS_SC_PRIOR_FLUSH, &sci->sc_flags);
}
static int nilfs_segctor_flush_mode(struct nilfs_sc_info *sci)
{
if (!test_bit(NILFS_SC_UNCLOSED, &sci->sc_flags) ||
time_before(jiffies, sci->sc_lseg_stime + sci->sc_mjcp_freq)) {
if (!(sci->sc_flush_request & ~FLUSH_FILE_BIT))
return SC_FLUSH_FILE;
else if (!(sci->sc_flush_request & ~FLUSH_DAT_BIT))
return SC_FLUSH_DAT;
}
return SC_LSEG_SR;
}
/**
* nilfs_segctor_thread - main loop of the segment constructor thread.
* @arg: pointer to a struct nilfs_sc_info.
*
* nilfs_segctor_thread() initializes a timer and serves as a daemon
* to execute segment constructions.
*/
static int nilfs_segctor_thread(void *arg)
{
struct nilfs_sc_info *sci = (struct nilfs_sc_info *)arg;
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
int timeout = 0;
sci->sc_timer_task = current;
/* start sync. */
sci->sc_task = current;
wake_up(&sci->sc_wait_task); /* for nilfs_segctor_start_thread() */
nilfs_info(sci->sc_super,
"segctord starting. Construction interval = %lu seconds, CP frequency < %lu seconds",
sci->sc_interval / HZ, sci->sc_mjcp_freq / HZ);
spin_lock(&sci->sc_state_lock);
loop:
for (;;) {
int mode;
if (sci->sc_state & NILFS_SEGCTOR_QUIT)
goto end_thread;
if (timeout || sci->sc_seq_request != sci->sc_seq_done)
mode = SC_LSEG_SR;
else if (sci->sc_flush_request)
mode = nilfs_segctor_flush_mode(sci);
else
break;
spin_unlock(&sci->sc_state_lock);
nilfs_segctor_thread_construct(sci, mode);
spin_lock(&sci->sc_state_lock);
timeout = 0;
}
if (freezing(current)) {
spin_unlock(&sci->sc_state_lock);
try_to_freeze();
spin_lock(&sci->sc_state_lock);
} else {
DEFINE_WAIT(wait);
int should_sleep = 1;
prepare_to_wait(&sci->sc_wait_daemon, &wait,
TASK_INTERRUPTIBLE);
if (sci->sc_seq_request != sci->sc_seq_done)
should_sleep = 0;
else if (sci->sc_flush_request)
should_sleep = 0;
else if (sci->sc_state & NILFS_SEGCTOR_COMMIT)
should_sleep = time_before(jiffies,
sci->sc_timer.expires);
if (should_sleep) {
spin_unlock(&sci->sc_state_lock);
schedule();
spin_lock(&sci->sc_state_lock);
}
finish_wait(&sci->sc_wait_daemon, &wait);
timeout = ((sci->sc_state & NILFS_SEGCTOR_COMMIT) &&
time_after_eq(jiffies, sci->sc_timer.expires));
if (nilfs_sb_dirty(nilfs) && nilfs_sb_need_update(nilfs))
set_nilfs_discontinued(nilfs);
}
goto loop;
end_thread:
spin_unlock(&sci->sc_state_lock);
/* end sync. */
sci->sc_task = NULL;
wake_up(&sci->sc_wait_task); /* for nilfs_segctor_kill_thread() */
return 0;
}
static int nilfs_segctor_start_thread(struct nilfs_sc_info *sci)
{
struct task_struct *t;
t = kthread_run(nilfs_segctor_thread, sci, "segctord");
if (IS_ERR(t)) {
int err = PTR_ERR(t);
nilfs_err(sci->sc_super, "error %d creating segctord thread",
err);
return err;
}
wait_event(sci->sc_wait_task, sci->sc_task != NULL);
return 0;
}
static void nilfs_segctor_kill_thread(struct nilfs_sc_info *sci)
__acquires(&sci->sc_state_lock)
__releases(&sci->sc_state_lock)
{
sci->sc_state |= NILFS_SEGCTOR_QUIT;
while (sci->sc_task) {
wake_up(&sci->sc_wait_daemon);
spin_unlock(&sci->sc_state_lock);
wait_event(sci->sc_wait_task, sci->sc_task == NULL);
spin_lock(&sci->sc_state_lock);
}
}
/*
* Setup & clean-up functions
*/
static struct nilfs_sc_info *nilfs_segctor_new(struct super_block *sb,
struct nilfs_root *root)
{
struct the_nilfs *nilfs = sb->s_fs_info;
struct nilfs_sc_info *sci;
sci = kzalloc(sizeof(*sci), GFP_KERNEL);
if (!sci)
return NULL;
sci->sc_super = sb;
nilfs_get_root(root);
sci->sc_root = root;
init_waitqueue_head(&sci->sc_wait_request);
init_waitqueue_head(&sci->sc_wait_daemon);
init_waitqueue_head(&sci->sc_wait_task);
spin_lock_init(&sci->sc_state_lock);
INIT_LIST_HEAD(&sci->sc_dirty_files);
INIT_LIST_HEAD(&sci->sc_segbufs);
INIT_LIST_HEAD(&sci->sc_write_logs);
INIT_LIST_HEAD(&sci->sc_gc_inodes);
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
INIT_LIST_HEAD(&sci->sc_iput_queue);
INIT_WORK(&sci->sc_iput_work, nilfs_iput_work_func);
timer_setup(&sci->sc_timer, nilfs_construction_timeout, 0);
sci->sc_interval = HZ * NILFS_SC_DEFAULT_TIMEOUT;
sci->sc_mjcp_freq = HZ * NILFS_SC_DEFAULT_SR_FREQ;
sci->sc_watermark = NILFS_SC_DEFAULT_WATERMARK;
if (nilfs->ns_interval)
sci->sc_interval = HZ * nilfs->ns_interval;
if (nilfs->ns_watermark)
sci->sc_watermark = nilfs->ns_watermark;
return sci;
}
static void nilfs_segctor_write_out(struct nilfs_sc_info *sci)
{
int ret, retrycount = NILFS_SC_CLEANUP_RETRY;
/*
* The segctord thread was stopped and its timer was removed.
* But some tasks remain.
*/
do {
struct nilfs_transaction_info ti;
nilfs_transaction_lock(sci->sc_super, &ti, 0);
ret = nilfs_segctor_construct(sci, SC_LSEG_SR);
nilfs_transaction_unlock(sci->sc_super);
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
flush_work(&sci->sc_iput_work);
} while (ret && retrycount-- > 0);
}
/**
* nilfs_segctor_destroy - destroy the segment constructor.
* @sci: nilfs_sc_info
*
* nilfs_segctor_destroy() kills the segctord thread and frees
* the nilfs_sc_info struct.
* Caller must hold the segment semaphore.
*/
static void nilfs_segctor_destroy(struct nilfs_sc_info *sci)
{
struct the_nilfs *nilfs = sci->sc_super->s_fs_info;
int flag;
up_write(&nilfs->ns_segctor_sem);
spin_lock(&sci->sc_state_lock);
nilfs_segctor_kill_thread(sci);
flag = ((sci->sc_state & NILFS_SEGCTOR_COMMIT) || sci->sc_flush_request
|| sci->sc_seq_request != sci->sc_seq_done);
spin_unlock(&sci->sc_state_lock);
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
if (flush_work(&sci->sc_iput_work))
flag = true;
if (flag || !nilfs_segctor_confirm(sci))
nilfs_segctor_write_out(sci);
if (!list_empty(&sci->sc_dirty_files)) {
nilfs_warn(sci->sc_super,
"disposed unprocessed dirty file(s) when stopping log writer");
nilfs_dispose_list(nilfs, &sci->sc_dirty_files, 1);
}
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
if (!list_empty(&sci->sc_iput_queue)) {
nilfs_warn(sci->sc_super,
"disposed unprocessed inode(s) in iput queue when stopping log writer");
nilfs2: fix deadlock of segment constructor over I_SYNC flag Nilfs2 eventually hangs in a stress test with fsstress program. This issue was caused by the following deadlock over I_SYNC flag between nilfs_segctor_thread() and writeback_sb_inodes(): nilfs_segctor_thread() nilfs_segctor_thread_construct() nilfs_segctor_unlock() nilfs_dispose_list() iput() iput_final() evict() inode_wait_for_writeback() * wait for I_SYNC flag writeback_sb_inodes() * set I_SYNC flag on inode->i_state __writeback_single_inode() do_writepages() nilfs_writepages() nilfs_construct_dsync_segment() nilfs_segctor_sync() * wait for completion of segment constructor inode_sync_complete() * clear I_SYNC flag after __writeback_single_inode() completed writeback_sb_inodes() calls do_writepages() for dirty inodes after setting I_SYNC flag on inode->i_state. do_writepages() in turn calls nilfs_writepages(), which can run segment constructor and wait for its completion. On the other hand, segment constructor calls iput(), which can call evict() and wait for the I_SYNC flag on inode_wait_for_writeback(). Since segment constructor doesn't know when I_SYNC will be set, it cannot know whether iput() will block or not unless inode->i_nlink has a non-zero count. We can prevent evict() from being called in iput() by implementing sop->drop_inode(), but it's not preferable to leave inodes with i_nlink == 0 for long periods because it even defers file truncation and inode deallocation. So, this instead resolves the deadlock by calling iput() asynchronously with a workqueue for inodes with i_nlink == 0. Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-05 20:25:20 +00:00
nilfs_dispose_list(nilfs, &sci->sc_iput_queue, 1);
}
WARN_ON(!list_empty(&sci->sc_segbufs));
WARN_ON(!list_empty(&sci->sc_write_logs));
nilfs_put_root(sci->sc_root);
down_write(&nilfs->ns_segctor_sem);
del_timer_sync(&sci->sc_timer);
kfree(sci);
}
/**
* nilfs_attach_log_writer - attach log writer
* @sb: super block instance
* @root: root object of the current filesystem tree
*
* This allocates a log writer object, initializes it, and starts the
* log writer.
*
* Return Value: On success, 0 is returned. On error, one of the following
* negative error code is returned.
*
* %-ENOMEM - Insufficient memory available.
*/
int nilfs_attach_log_writer(struct super_block *sb, struct nilfs_root *root)
{
struct the_nilfs *nilfs = sb->s_fs_info;
int err;
if (nilfs->ns_writer) {
/*
* This happens if the filesystem was remounted
* read/write after nilfs_error degenerated it into a
* read-only mount.
*/
nilfs_detach_log_writer(sb);
}
nilfs->ns_writer = nilfs_segctor_new(sb, root);
if (!nilfs->ns_writer)
return -ENOMEM;
nilfs2: fix null pointer dereference at nilfs_segctor_do_construct() After commit c3aab9a0bd91 ("mm/filemap.c: don't initiate writeback if mapping has no dirty pages"), the following null pointer dereference has been reported on nilfs2: BUG: kernel NULL pointer dereference, address: 00000000000000a8 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 0 P4D 0 Oops: 0000 [#1] SMP PTI ... RIP: 0010:percpu_counter_add_batch+0xa/0x60 ... Call Trace: __test_set_page_writeback+0x2d3/0x330 nilfs_segctor_do_construct+0x10d3/0x2110 [nilfs2] nilfs_segctor_construct+0x168/0x260 [nilfs2] nilfs_segctor_thread+0x127/0x3b0 [nilfs2] kthread+0xf8/0x130 ... This crash turned out to be caused by set_page_writeback() call for segment summary buffers at nilfs_segctor_prepare_write(). set_page_writeback() can call inc_wb_stat(inode_to_wb(inode), WB_WRITEBACK) where inode_to_wb(inode) is NULL if the inode of underlying block device does not have an associated wb. This fixes the issue by calling inode_attach_wb() in advance to ensure to associate the bdev inode with its wb. Fixes: c3aab9a0bd91 ("mm/filemap.c: don't initiate writeback if mapping has no dirty pages") Reported-by: Walton Hoops <me@waltonhoops.com> Reported-by: Tomas Hlavaty <tom@logand.com> Reported-by: ARAI Shun-ichi <hermes@ceres.dti.ne.jp> Reported-by: Hideki EIRAKU <hdk1983@gmail.com> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com> Cc: <stable@vger.kernel.org> [5.4+] Link: http://lkml.kernel.org/r/20200608.011819.1399059588922299158.konishi.ryusuke@gmail.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-11 01:41:35 +00:00
inode_attach_wb(nilfs->ns_bdev->bd_inode, NULL);
err = nilfs_segctor_start_thread(nilfs->ns_writer);
if (err) {
kfree(nilfs->ns_writer);
nilfs->ns_writer = NULL;
}
return err;
}
/**
* nilfs_detach_log_writer - destroy log writer
* @sb: super block instance
*
* This kills log writer daemon, frees the log writer object, and
* destroys list of dirty files.
*/
void nilfs_detach_log_writer(struct super_block *sb)
{
struct the_nilfs *nilfs = sb->s_fs_info;
LIST_HEAD(garbage_list);
down_write(&nilfs->ns_segctor_sem);
if (nilfs->ns_writer) {
nilfs_segctor_destroy(nilfs->ns_writer);
nilfs->ns_writer = NULL;
}
/* Force to free the list of dirty files */
spin_lock(&nilfs->ns_inode_lock);
if (!list_empty(&nilfs->ns_dirty_files)) {
list_splice_init(&nilfs->ns_dirty_files, &garbage_list);
nilfs_warn(sb,
"disposed unprocessed dirty file(s) when detaching log writer");
}
spin_unlock(&nilfs->ns_inode_lock);
up_write(&nilfs->ns_segctor_sem);
nilfs_dispose_list(nilfs, &garbage_list, 1);
}