Commit c76f48eb5c ("block: take bd_mutex around delete_partitions in
del_gendisk") adds disk->part0->bd_mutex in del_gendisk(), this way
causes the following AB/BA deadlock between removing loop and opening
loop:
1) loop_control_ioctl(LOOP_CTL_REMOVE)
-> mutex_lock(&loop_ctl_mutex)
-> del_gendisk
-> mutex_lock(&disk->part0->bd_mutex)
2) blkdev_get_by_dev
-> mutex_lock(&disk->part0->bd_mutex)
-> lo_open
-> mutex_lock(&loop_ctl_mutex)
Add a new Lo_deleting state to remove the need for clearing
->private_data and thus holding loop_ctl_mutex in the ioctl
LOOP_CTL_REMOVE path.
Based on an analysis and earlier patch from
Ming Lei <ming.lei@redhat.com>.
Reported-by: Colin Ian King <colin.king@canonical.com>
Fixes: c76f48eb5c ("block: take bd_mutex around delete_partitions in del_gendisk")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20210605140950.5800-1-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In the cache missing code path of cached device, if a proper location
from the internal B+ tree is matched for a cache miss range, function
cached_dev_cache_miss() will be called in cache_lookup_fn() in the
following code block,
[code block 1]
526 unsigned int sectors = KEY_INODE(k) == s->iop.inode
527 ? min_t(uint64_t, INT_MAX,
528 KEY_START(k) - bio->bi_iter.bi_sector)
529 : INT_MAX;
530 int ret = s->d->cache_miss(b, s, bio, sectors);
Here s->d->cache_miss() is the call backfunction pointer initialized as
cached_dev_cache_miss(), the last parameter 'sectors' is an important
hint to calculate the size of read request to backing device of the
missing cache data.
Current calculation in above code block may generate oversized value of
'sectors', which consequently may trigger 2 different potential kernel
panics by BUG() or BUG_ON() as listed below,
1) BUG_ON() inside bch_btree_insert_key(),
[code block 2]
886 BUG_ON(b->ops->is_extents && !KEY_SIZE(k));
2) BUG() inside biovec_slab(),
[code block 3]
51 default:
52 BUG();
53 return NULL;
All the above panics are original from cached_dev_cache_miss() by the
oversized parameter 'sectors'.
Inside cached_dev_cache_miss(), parameter 'sectors' is used to calculate
the size of data read from backing device for the cache missing. This
size is stored in s->insert_bio_sectors by the following lines of code,
[code block 4]
909 s->insert_bio_sectors = min(sectors, bio_sectors(bio) + reada);
Then the actual key inserting to the internal B+ tree is generated and
stored in s->iop.replace_key by the following lines of code,
[code block 5]
911 s->iop.replace_key = KEY(s->iop.inode,
912 bio->bi_iter.bi_sector + s->insert_bio_sectors,
913 s->insert_bio_sectors);
The oversized parameter 'sectors' may trigger panic 1) by BUG_ON() from
the above code block.
And the bio sending to backing device for the missing data is allocated
with hint from s->insert_bio_sectors by the following lines of code,
[code block 6]
926 cache_bio = bio_alloc_bioset(GFP_NOWAIT,
927 DIV_ROUND_UP(s->insert_bio_sectors, PAGE_SECTORS),
928 &dc->disk.bio_split);
The oversized parameter 'sectors' may trigger panic 2) by BUG() from the
agove code block.
Now let me explain how the panics happen with the oversized 'sectors'.
In code block 5, replace_key is generated by macro KEY(). From the
definition of macro KEY(),
[code block 7]
71 #define KEY(inode, offset, size) \
72 ((struct bkey) { \
73 .high = (1ULL << 63) | ((__u64) (size) << 20) | (inode), \
74 .low = (offset) \
75 })
Here 'size' is 16bits width embedded in 64bits member 'high' of struct
bkey. But in code block 1, if "KEY_START(k) - bio->bi_iter.bi_sector" is
very probably to be larger than (1<<16) - 1, which makes the bkey size
calculation in code block 5 is overflowed. In one bug report the value
of parameter 'sectors' is 131072 (= 1 << 17), the overflowed 'sectors'
results the overflowed s->insert_bio_sectors in code block 4, then makes
size field of s->iop.replace_key to be 0 in code block 5. Then the 0-
sized s->iop.replace_key is inserted into the internal B+ tree as cache
missing check key (a special key to detect and avoid a racing between
normal write request and cache missing read request) as,
[code block 8]
915 ret = bch_btree_insert_check_key(b, &s->op, &s->iop.replace_key);
Then the 0-sized s->iop.replace_key as 3rd parameter triggers the bkey
size check BUG_ON() in code block 2, and causes the kernel panic 1).
Another kernel panic is from code block 6, is by the bvecs number
oversized value s->insert_bio_sectors from code block 4,
min(sectors, bio_sectors(bio) + reada)
There are two possibility for oversized reresult,
- bio_sectors(bio) is valid, but bio_sectors(bio) + reada is oversized.
- sectors < bio_sectors(bio) + reada, but sectors is oversized.
From a bug report the result of "DIV_ROUND_UP(s->insert_bio_sectors,
PAGE_SECTORS)" from code block 6 can be 344, 282, 946, 342 and many
other values which larther than BIO_MAX_VECS (a.k.a 256). When calling
bio_alloc_bioset() with such larger-than-256 value as the 2nd parameter,
this value will eventually be sent to biovec_slab() as parameter
'nr_vecs' in following code path,
bio_alloc_bioset() ==> bvec_alloc() ==> biovec_slab()
Because parameter 'nr_vecs' is larger-than-256 value, the panic by BUG()
in code block 3 is triggered inside biovec_slab().
From the above analysis, we know that the 4th parameter 'sector' sent
into cached_dev_cache_miss() may cause overflow in code block 5 and 6,
and finally cause kernel panic in code block 2 and 3. And if result of
bio_sectors(bio) + reada exceeds valid bvecs number, it may also trigger
kernel panic in code block 3 from code block 6.
Now the almost-useless readahead size for cache missing request back to
backing device is removed, this patch can fix the oversized issue with
more simpler method.
- add a local variable size_limit, set it by the minimum value from
the max bkey size and max bio bvecs number.
- set s->insert_bio_sectors by the minimum value from size_limit,
sectors, and the sectors size of bio.
- replace sectors by s->insert_bio_sectors to do bio_next_split.
By the above method with size_limit, s->insert_bio_sectors will never
result oversized replace_key size or bio bvecs number. And split bio
'miss' from bio_next_split() will always match the size of 'cache_bio',
that is the current maximum bio size we can sent to backing device for
fetching the cache missing data.
Current problmatic code can be partially found since Linux v3.13-rc1,
therefore all maintained stable kernels should try to apply this fix.
Reported-by: Alexander Ullrich <ealex1979@gmail.com>
Reported-by: Diego Ercolani <diego.ercolani@gmail.com>
Reported-by: Jan Szubiak <jan.szubiak@linuxpolska.pl>
Reported-by: Marco Rebhan <me@dblsaiko.net>
Reported-by: Matthias Ferdinand <bcache@mfedv.net>
Reported-by: Victor Westerhuis <victor@westerhu.is>
Reported-by: Vojtech Pavlik <vojtech@suse.cz>
Reported-and-tested-by: Rolf Fokkens <rolf@rolffokkens.nl>
Reported-and-tested-by: Thorsten Knabe <linux@thorsten-knabe.de>
Signed-off-by: Coly Li <colyli@suse.de>
Cc: stable@vger.kernel.org
Cc: Christoph Hellwig <hch@lst.de>
Cc: Kent Overstreet <kent.overstreet@gmail.com>
Cc: Nix <nix@esperi.org.uk>
Cc: Takashi Iwai <tiwai@suse.com>
Link: https://lore.kernel.org/r/20210607125052.21277-3-colyli@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For read cache missing, bcache defines a readahead size for the read I/O
request to the backing device for the missing data. This readahead size
is initialized to 0, and almost no one uses it to avoid unnecessary read
amplifying onto backing device and write amplifying onto cache device.
Considering upper layer file system code has readahead logic allready
and works fine with readahead_cache_policy sysfile interface, we don't
have to keep bcache self-defined readahead anymore.
This patch removes the bcache self-defined readahead for cache missing
request for backing device, and the readahead sysfs file interfaces are
removed as well.
This is the preparation for next patch to fix potential kernel panic due
to oversized request in a simpler method.
Reported-by: Alexander Ullrich <ealex1979@gmail.com>
Reported-by: Diego Ercolani <diego.ercolani@gmail.com>
Reported-by: Jan Szubiak <jan.szubiak@linuxpolska.pl>
Reported-by: Marco Rebhan <me@dblsaiko.net>
Reported-by: Matthias Ferdinand <bcache@mfedv.net>
Reported-by: Victor Westerhuis <victor@westerhu.is>
Reported-by: Vojtech Pavlik <vojtech@suse.cz>
Reported-and-tested-by: Rolf Fokkens <rolf@rolffokkens.nl>
Reported-and-tested-by: Thorsten Knabe <linux@thorsten-knabe.de>
Signed-off-by: Coly Li <colyli@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: stable@vger.kernel.org
Cc: Kent Overstreet <kent.overstreet@gmail.com>
Cc: Nix <nix@esperi.org.uk>
Cc: Takashi Iwai <tiwai@suse.com>
Link: https://lore.kernel.org/r/20210607125052.21277-2-colyli@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In case p2p device was found but the p2p pool is empty, the nvme target
is still trying to free the sgl from the p2p pool instead of the
regular sgl pool and causing a crash (BUG() is called). Instead, assign
the p2p_dev for the request only if it was allocated from p2p pool.
This is the crash that was caused:
[Sun May 30 19:13:53 2021] ------------[ cut here ]------------
[Sun May 30 19:13:53 2021] kernel BUG at lib/genalloc.c:518!
[Sun May 30 19:13:53 2021] invalid opcode: 0000 [#1] SMP PTI
...
[Sun May 30 19:13:53 2021] kernel BUG at lib/genalloc.c:518!
...
[Sun May 30 19:13:53 2021] RIP: 0010:gen_pool_free_owner+0xa8/0xb0
...
[Sun May 30 19:13:53 2021] Call Trace:
[Sun May 30 19:13:53 2021] ------------[ cut here ]------------
[Sun May 30 19:13:53 2021] pci_free_p2pmem+0x2b/0x70
[Sun May 30 19:13:53 2021] pci_p2pmem_free_sgl+0x4f/0x80
[Sun May 30 19:13:53 2021] nvmet_req_free_sgls+0x1e/0x80 [nvmet]
[Sun May 30 19:13:53 2021] kernel BUG at lib/genalloc.c:518!
[Sun May 30 19:13:53 2021] nvmet_rdma_release_rsp+0x4e/0x1f0 [nvmet_rdma]
[Sun May 30 19:13:53 2021] nvmet_rdma_send_done+0x1c/0x60 [nvmet_rdma]
Fixes: c6e3f13398 ("nvmet: add metadata support for block devices")
Reviewed-by: Israel Rukshin <israelr@nvidia.com>
Signed-off-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
During concurrent reset and delete calls the reset workqueue is
flushed, causing nvme_loop_reset_ctrl_work() to be executed when
the controller is in state DELETING or DELETING_NOIO.
But this is expected, so we shouldn't issue a WARN_ON here.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
We need to check the NVME_LOOP_Q_LIVE flag in
nvme_loop_destroy_admin_queue() to protect against duplicate
invocations eg during concurrent reset and remove calls.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When the call to nvme_enable_ctrl() in nvme_loop_configure_admin_queue()
fails the NVME_LOOP_Q_LIVE flag is not cleared.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
The queue count is increased in nvme_loop_init_io_queues(), so we
need to reset it to 1 at the end of nvme_loop_destroy_io_queues().
Otherwise the function is not re-entrant safe, and crash will happen
during concurrent reset and remove calls.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
We have only 2 inline sg entries and we allow 4 sg entries for the send
wr sge. Larger sgls entries will be chained. However when we build
in-capsule send wr sge, we iterate without taking into account that the
sgl may be chained and still fit in-capsule (which can happen if the sgl
is bigger than 2, but lower-equal to 4).
Fix in-capsule data mapping to correctly iterate chained sgls.
Fixes: 38e1800275 ("nvme-rdma: Avoid preallocating big SGL for data")
Reported-by: Walker, Benjamin <benjamin.walker@intel.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
- fix a memory leak in nvme_cdev_add (Guoqing Jiang)
- fix inline data size comparison in nvmet_tcp_queue_response (Hou Pu)
- fix false keep-alive timeout when a controller is torn down
(Sagi Grimberg)
- fix a nvme-tcp Kconfig dependency (Sagi Grimberg)
- short-circuit reconnect retries for FC (Hannes Reinecke)
- decode host pathing error for connect (Hannes Reinecke)
-----BEGIN PGP SIGNATURE-----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=JT0/
-----END PGP SIGNATURE-----
Merge tag 'nvme-5.13-2021-05-27' of git://git.infradead.org/nvme into block-5.13
Pull NVMe fixes from Christoph:
"nvme fixes for Linux 5.13
- fix a memory leak in nvme_cdev_add (Guoqing Jiang)
- fix inline data size comparison in nvmet_tcp_queue_response (Hou Pu)
- fix false keep-alive timeout when a controller is torn down
(Sagi Grimberg)
- fix a nvme-tcp Kconfig dependency (Sagi Grimberg)
- short-circuit reconnect retries for FC (Hannes Reinecke)
- decode host pathing error for connect (Hannes Reinecke)"
* tag 'nvme-5.13-2021-05-27' of git://git.infradead.org/nvme:
nvmet: fix false keep-alive timeout when a controller is torn down
nvmet-tcp: fix inline data size comparison in nvmet_tcp_queue_response
nvme-tcp: remove incorrect Kconfig dep in BLK_DEV_NVME
nvme-fabrics: decode host pathing error for connect
nvme-fc: short-circuit reconnect retries
nvme: fix potential memory leaks in nvme_cdev_add
Controller teardown flow may take some time in case it has many I/O
queues, and the host may not send us keep-alive during this period.
Hence reset the traffic based keep-alive timer so we don't trigger
a controller teardown as a result of a keep-alive expiration.
Reported-by: Yi Zhang <yi.zhang@redhat.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Tested-by: Yi Zhang <yi.zhang@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Using "<=" instead "<" to compare inline data size.
Fixes: bdaf132791 ("nvmet-tcp: fix a segmentation fault during io parsing error")
Signed-off-by: Hou Pu <houpu.main@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
We need to select NVME_CORE.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Now that the original bdev is stored in the bio this assert is incorrect
and will trigger for any partitioned raid5 device.
Reported-by: Florian Dazinger <spam02@dazinger.net>
Tested-by: Florian Dazinger <spam02@dazinger.net>
Cc: stable@vger.kernel.org # 5.12
Fixes: 309dca309f ("block: store a block_device pointer in struct bio"),
Reviewed-by: Guoqing Jiang <jiangguoqing@kylinos.cn>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Song Liu <song@kernel.org>
Fix crash with illegal operation exception in dasd_device_tasklet.
Commit b729493288 ("s390/dasd: Prepare for additional path event handling")
renamed the verify_path function for ECKD but not for FBA and DIAG.
This leads to a panic when the path verification function is called for a
FBA or DIAG device.
Fix by defining a wrapper function for dasd_generic_verify_path().
Fixes: b729493288 ("s390/dasd: Prepare for additional path event handling")
Cc: <stable@vger.kernel.org> #5.11
Reviewed-by: Jan Hoeppner <hoeppner@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Link: https://lore.kernel.org/r/20210525125006.157531-2-sth@linux.ibm.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Returning an nvme status from nvme_fc_create_association() indicates
that the association is established, and we should honour the DNR bit.
If it's set a reconnect attempt will just return the same error, so
we can short-circuit the reconnect attempts and fail the connection
directly.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Reviewed-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When BLKRRPART is called concurrently with del_gendisk, the partitions
rescan can create a stale partition that will never be be cleaned up.
Fix this by checking the the disk is up before rescanning partitions
while under bd_mutex.
Signed-off-by: Gulam Mohamed <gulam.mohamed@oracle.com>
[hch: split from a larger patch]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20210514131842.1600568-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
As an artifact of how gendisk lookup used to work in earlier kernels,
GENHD_FL_UP is only cleared very late in del_gendisk, and a global lock
is used to prevent opens from succeeding while del_gendisk is tearing
down the gendisk. Switch to clearing the flag early and under bd_mutex
so that callers can use bd_mutex to stabilize the flag, which removes
the need for the global mutex.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20210514131842.1600568-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The __nvmf_check_ready() routine used to bounce all filesystem io if the
controller state isn't LIVE. However, a later patch changed the logic so
that it rejection ends up being based on the Q live check. The FC
transport has a slightly different sequence from rdma and tcp for
shutting down queues/marking them non-live. FC marks its queue non-live
after aborting all ios and waiting for their termination, leaving a
rather large window for filesystem io to continue to hit the transport.
Unfortunately this resulted in filesystem I/O or applications seeing I/O
errors.
Change the FC transport to mark the queues non-live at the first sign of
teardown for the association (when I/O is initially terminated).
Fixes: 73a5379937 ("nvme-fabrics: allow to queue requests for live queues")
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
A possible race condition exists where the request to send data is
enqueued from nvme_tcp_handle_r2t()'s will not be observed by
nvme_tcp_send_all() if it happens to be running. The driver relies on
io_work to send the enqueued request when it is runs again, but the
concurrently running nvme_tcp_send_all() may not have released the
send_mutex at that time. If no future commands are enqueued to re-kick
the io_work, the request will timeout in the SEND_H2C state, resulting
in a timeout error like:
nvme nvme0: queue 1: timeout request 0x3 type 6
Ensure the io_work continues to run as long as the req_list is not empty.
Fixes: db5ad6b7f8 ("nvme-tcp: try to send request in queue_rq context")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Commit db5ad6b7f8 ("nvme-tcp: try to send request in queue_rq
context") added a second context that may perform a network send.
This means that now RX and TX are not serialized in nvme_tcp_io_work
and can run concurrently.
While there is correct mutual exclusion in the TX path (where
the send_mutex protect the queue socket send activity) RX activity,
and more specifically request completion may run concurrently.
This means we must guarantee that any mutation of the request state
related to its lifetime, bytes sent must not be accessed when a completion
may have possibly arrived back (and processed).
The race may trigger when a request completion arrives, processed
_and_ reused as a fresh new request, exactly in the (relatively short)
window between the last data payload sent and before the request iov_iter
is advanced.
Consider the following race:
1. 16K write request is queued
2. The nvme command and the data is sent to the controller (in-capsule
or solicited by r2t)
3. After the last payload is sent but before the req.iter is advanced,
the controller sends back a completion.
4. The completion is processed, the request is completed, and reused
to transfer a new request (write or read)
5. The new request is queued, and the driver reset the request parameters
(nvme_tcp_setup_cmd_pdu).
6. Now context in (2) resumes execution and advances the req.iter
==> use-after-completion as this is already a new request.
Fix this by making sure the request is not advanced after the last
data payload send, knowing that a completion may have arrived already.
An alternative solution would have been to delay the request completion
or state change waiting for reference counting on the TX path, but besides
adding atomic operations to the hot-path, it may present challenges in
multi-stage R2T scenarios where a r2t handler needs to be deferred to
an async execution.
Reported-by: Narayan Ayalasomayajula <narayan.ayalasomayajula@wdc.com>
Tested-by: Anil Mishra <anil.mishra@wdc.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Cc: stable@vger.kernel.org # v5.8+
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When creating loop ctrl in nvme_loop_create_ctrl(), if nvme_init_ctrl()
fails, the loop ctrl should be freed before jumping to the "out" label.
Fixes: 3a85a5de29 ("nvme-loop: add a NVMe loopback host driver")
Signed-off-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When creating ctrl in nvmet_alloc_ctrl(), if the cntlid_min is larger
than cntlid_max of the subsystem, and jumps to the
"out_free_changed_ns_list" label, but the ctrl->sqs lack of be freed.
Fix this by jumping to the "out_free_sqs" label.
Fixes: 94a39d61f8 ("nvmet: make ctrl-id configurable")
Signed-off-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
If a tag set is shared across request queues (e.g. SCSI LUNs) then the
block layer core keeps track of the number of active request queues in
tags->active_queues. blk_mq_tag_busy() and blk_mq_tag_idle() update that
atomic counter if the hctx flag BLK_MQ_F_TAG_QUEUE_SHARED is set. Make
sure that blk_mq_exit_queue() calls blk_mq_tag_idle() before that flag is
cleared by blk_mq_del_queue_tag_set().
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Hannes Reinecke <hare@suse.com>
Fixes: 0d2602ca30 ("blk-mq: improve support for shared tags maps")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20210513171529.7977-1-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In case of shared sbitmap, request won't be held in plug list any more
sine commit 32bc15afed ("blk-mq: Facilitate a shared sbitmap per
tagset"), this way makes request merge from flush plug list & batching
submission not possible, so cause performance regression.
Yanhui reports performance regression when running sequential IO
test(libaio, 16 jobs, 8 depth for each job) in VM, and the VM disk
is emulated with image stored on xfs/megaraid_sas.
Fix the issue by recovering original behavior to allow to hold request
in plug list.
Cc: Yanhui Ma <yama@redhat.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Bart Van Assche <bvanassche@acm.org>
Cc: kashyap.desai@broadcom.com
Fixes: 32bc15afed ("blk-mq: Facilitate a shared sbitmap per tagset")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20210514022052.1047665-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- correct the check for using the inline bio in nvmet
(Chaitanya Kulkarni)
- demote unsupported command warnings (Chaitanya Kulkarni)
- fix corruption due to double initializing ANA state (me, Hou Pu)
- reset ns->file when open fails (Daniel Wagner)
- fix a NULL deref when SEND is completed with error in nvmet-rdma
(Michal Kalderon)
-----BEGIN PGP SIGNATURE-----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=ZwLd
-----END PGP SIGNATURE-----
Merge tag 'nvme-5.13-2021-05-13' of git://git.infradead.org/nvme into block-5.13
Pull NVMe fixes from Christoph:
"nvme fix for Linux 5.13
- correct the check for using the inline bio in nvmet
(Chaitanya Kulkarni)
- demote unsupported command warnings (Chaitanya Kulkarni)
- fix corruption due to double initializing ANA state (me, Hou Pu)
- reset ns->file when open fails (Daniel Wagner)
- fix a NULL deref when SEND is completed with error in nvmet-rdma
(Michal Kalderon)"
* tag 'nvme-5.13-2021-05-13' of git://git.infradead.org/nvme:
nvmet: use new ana_log_size instead the old one
nvmet: seset ns->file when open fails
nvmet: demote fabrics cmd parse err msg to debug
nvmet: use helper to remove the duplicate code
nvmet: demote discovery cmd parse err msg to debug
nvmet-rdma: Fix NULL deref when SEND is completed with error
nvmet: fix inline bio check for passthru
nvmet: fix inline bio check for bdev-ns
nvme-multipath: fix double initialization of ANA state
Reset the ns->file value to NULL also in the error case in
nvmet_file_ns_enable().
The ns->file variable points either to file object or contains the
error code after the filp_open() call. This can lead to following
problem:
When the user first setups an invalid file backend and tries to enable
the ns, it will fail. Then the user switches over to a bdev backend
and enables successfully the ns. The first received I/O will crash the
system because the IO backend is chosen based on the ns->file value:
static u16 nvmet_parse_io_cmd(struct nvmet_req *req)
{
[...]
if (req->ns->file)
return nvmet_file_parse_io_cmd(req);
return nvmet_bdev_parse_io_cmd(req);
}
Reported-by: Enzo Matsumiya <ematsumiya@suse.com>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Replace the following two statements by the statement “goto put_nbd;”
nbd_put(nbd);
return 0;
Signed-off-by: Sun Ke <sunke32@huawei.com>
Suggested-by: Markus Elfring <Markus.Elfring@web.de>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Link: https://lore.kernel.org/r/20210512114331.1233964-3-sunke32@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Last users of blk_account_rq gone with patch commit a1ce35fa49
("block: remove dead elevator code") and now it gets no caller, it can
be safely removed.
Signed-off-by: Lin Feng <linf@wangsu.com>
Link: https://lore.kernel.org/r/20210512100124.173769-1-linf@wangsu.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
BFQ may merge a new bfq_queue, stably, with the last bfq_queue
created. In particular, BFQ first waits a little bit for some I/O to
flow inside the new queue, say Q2, if this is needed to understand
whether it is better or worse to merge Q2 with the last queue created,
say Q1. This delayed stable merge is performed by assigning
bic->stable_merge_bfqq = Q1, for the bic associated with Q1.
Yet, while waiting for some I/O to flow in Q2, a non-stable queue
merge of Q2 with Q1 may happen, causing the bic previously associated
with Q2 to be associated with exactly Q1 (bic->bfqq = Q1). After that,
Q2 and Q1 may happen to be split, and, in the split, Q1 may happen to
be recycled as a non-shared bfq_queue. In that case, Q1 may then
happen to undergo a stable merge with the bfq_queue pointed by
bic->stable_merge_bfqq. Yet bic->stable_merge_bfqq still points to
Q1. So Q1 would be merged with itself.
This commit fixes this error by intercepting this situation, and
canceling the schedule of the stable merge.
Fixes: 430a67f9d6 ("block, bfq: merge bursts of newly-created queues")
Signed-off-by: Pietro Pedroni <pedroni.pietro.96@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Link: https://lore.kernel.org/r/20210512094352.85545-2-paolo.valente@linaro.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When the weight of an active iocg is updated, weight_updated() is called
which in turn calls __propagate_weights() to update the active and inuse
weights so that the effective hierarchical weights are update accordingly.
The current implementation is incorrect for inner active nodes. For an
active leaf iocg, inuse can be any value between 1 and active and the
difference represents how much the iocg is donating. When weight is updated,
as long as inuse is clamped between 1 and the new weight, we're alright and
this is what __propagate_weights() currently implements.
However, that's not how an active inner node's inuse is set. An inner node's
inuse is solely determined by the ratio between the sums of inuse's and
active's of its children - ie. they're results of propagating the leaves'
active and inuse weights upwards. __propagate_weights() incorrectly applies
the same clamping as for a leaf when an active inner node's weight is
updated. Consider a hierarchy which looks like the following with saturating
workloads in AA and BB.
R
/ \
A B
| |
AA BB
1. For both A and B, active=100, inuse=100, hwa=0.5, hwi=0.5.
2. echo 200 > A/io.weight
3. __propagate_weights() update A's active to 200 and leave inuse at 100 as
it's already between 1 and the new active, making A:active=200,
A:inuse=100. As R's active_sum is updated along with A's active,
A:hwa=2/3, B:hwa=1/3. However, because the inuses didn't change, the
hwi's remain unchanged at 0.5.
4. The weight of A is now twice that of B but AA and BB still have the same
hwi of 0.5 and thus are doing the same amount of IOs.
Fix it by making __propgate_weights() always calculate the inuse of an
active inner iocg based on the ratio of child_inuse_sum to child_active_sum.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Dan Schatzberg <dschatzberg@fb.com>
Fixes: 7caa47151a ("blkcg: implement blk-iocost")
Cc: stable@vger.kernel.org # v5.4+
Link: https://lore.kernel.org/r/YJsxnLZV1MnBcqjj@slm.duckdns.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Host can send invalid commands and flood the target with error messages.
Demote the error message from pr_err() to pr_debug() in
nvmet_parse_fabrics_cmd() and nvmet_parse_connect_cmd().
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Use the helper nvmet_report_invalid_opcode() to report invalid opcode
so we can remove the duplicate code.
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Host can send invalid commands and flood the target with error messages
for the discovery controller. Demote the error message from pr_err() to
pr_debug( in nvmet_parse_discovery_cmd().
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When handling passthru commands, for inline bio allocation we only
consider the transfer size. This works well when req->sg_cnt fits into
the req->inline_bvec, but it will result in the early return from
bio_add_hw_page() when req->sg_cnt > NVMET_MAX_INLINE_BVEC.
Consider an I/O of size 32768 and first buffer is not aligned to the
page boundary, then I/O is split in following manner :-
[ 2206.256140] nvmet: sg->length 3440 sg->offset 656
[ 2206.256144] nvmet: sg->length 4096 sg->offset 0
[ 2206.256148] nvmet: sg->length 4096 sg->offset 0
[ 2206.256152] nvmet: sg->length 4096 sg->offset 0
[ 2206.256155] nvmet: sg->length 4096 sg->offset 0
[ 2206.256159] nvmet: sg->length 4096 sg->offset 0
[ 2206.256163] nvmet: sg->length 4096 sg->offset 0
[ 2206.256166] nvmet: sg->length 4096 sg->offset 0
[ 2206.256170] nvmet: sg->length 656 sg->offset 0
Now the req->transfer_size == NVMET_MAX_INLINE_DATA_LEN i.e. 32768, but
the req->sg_cnt is (9) > NVMET_MAX_INLINE_BIOVEC which is (8).
This will result in early return in the following code path :-
nvmet_bdev_execute_rw()
bio_add_pc_page()
bio_add_hw_page()
if (bio_full(bio, len))
return 0;
Use previously introduced helper nvmet_use_inline_bvec() to consider
req->sg_cnt when using inline bio. This only affects nvme-loop
transport.
Fixes: dab3902b19 ("nvmet: use inline bio for passthru fast path")
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When handling rw commands, for inline bio case we only consider
transfer size. This works well when req->sg_cnt fits into the
req->inline_bvec, but it will result in the warning in
__bio_add_page() when req->sg_cnt > NVMET_MAX_INLINE_BVEC.
Consider an I/O size 32768 and first page is not aligned to the page
boundary, then I/O is split in following manner :-
[ 2206.256140] nvmet: sg->length 3440 sg->offset 656
[ 2206.256144] nvmet: sg->length 4096 sg->offset 0
[ 2206.256148] nvmet: sg->length 4096 sg->offset 0
[ 2206.256152] nvmet: sg->length 4096 sg->offset 0
[ 2206.256155] nvmet: sg->length 4096 sg->offset 0
[ 2206.256159] nvmet: sg->length 4096 sg->offset 0
[ 2206.256163] nvmet: sg->length 4096 sg->offset 0
[ 2206.256166] nvmet: sg->length 4096 sg->offset 0
[ 2206.256170] nvmet: sg->length 656 sg->offset 0
Now the req->transfer_size == NVMET_MAX_INLINE_DATA_LEN i.e. 32768, but
the req->sg_cnt is (9) > NVMET_MAX_INLINE_BIOVEC which is (8).
This will result in the following warning message :-
nvmet_bdev_execute_rw()
bio_add_page()
__bio_add_page()
WARN_ON_ONCE(bio_full(bio, len));
This scenario is very hard to reproduce on the nvme-loop transport only
with rw commands issued with the passthru IOCTL interface from the host
application and the data buffer is allocated with the malloc() and not
the posix_memalign().
Fixes: 73383adfad ("nvmet: don't split large I/Os unconditionally")
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
nvme_init_identify and thus nvme_mpath_init can be called multiple
times and thus must not overwrite potentially initialized or in-use
fields. Split out a helper for the basic initialization when the
controller is initialized and make sure the init_identify path does
not blindly change in-use data structures.
Fixes: 0d0b660f21 ("nvme: add ANA support")
Reported-by: Martin Wilck <mwilck@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Hannes Reinecke <hare@suse.de>
__blk_mq_sched_bio_merge() gets the ctx and hctx for the current CPU and
passes the hctx to ->bio_merge(). kyber_bio_merge() then gets the ctx
for the current CPU again and uses that to get the corresponding Kyber
context in the passed hctx. However, the thread may be preempted between
the two calls to blk_mq_get_ctx(), and the ctx returned the second time
may no longer correspond to the passed hctx. This "works" accidentally
most of the time, but it can cause us to read garbage if the second ctx
came from an hctx with more ctx's than the first one (i.e., if
ctx->index_hw[hctx->type] > hctx->nr_ctx).
This manifested as this UBSAN array index out of bounds error reported
by Jakub:
UBSAN: array-index-out-of-bounds in ../kernel/locking/qspinlock.c:130:9
index 13106 is out of range for type 'long unsigned int [128]'
Call Trace:
dump_stack+0xa4/0xe5
ubsan_epilogue+0x5/0x40
__ubsan_handle_out_of_bounds.cold.13+0x2a/0x34
queued_spin_lock_slowpath+0x476/0x480
do_raw_spin_lock+0x1c2/0x1d0
kyber_bio_merge+0x112/0x180
blk_mq_submit_bio+0x1f5/0x1100
submit_bio_noacct+0x7b0/0x870
submit_bio+0xc2/0x3a0
btrfs_map_bio+0x4f0/0x9d0
btrfs_submit_data_bio+0x24e/0x310
submit_one_bio+0x7f/0xb0
submit_extent_page+0xc4/0x440
__extent_writepage_io+0x2b8/0x5e0
__extent_writepage+0x28d/0x6e0
extent_write_cache_pages+0x4d7/0x7a0
extent_writepages+0xa2/0x110
do_writepages+0x8f/0x180
__writeback_single_inode+0x99/0x7f0
writeback_sb_inodes+0x34e/0x790
__writeback_inodes_wb+0x9e/0x120
wb_writeback+0x4d2/0x660
wb_workfn+0x64d/0xa10
process_one_work+0x53a/0xa80
worker_thread+0x69/0x5b0
kthread+0x20b/0x240
ret_from_fork+0x1f/0x30
Only Kyber uses the hctx, so fix it by passing the request_queue to
->bio_merge() instead. BFQ and mq-deadline just use that, and Kyber can
map the queues itself to avoid the mismatch.
Fixes: a6088845c2 ("block: kyber: make kyber more friendly with merging")
Reported-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Link: https://lore.kernel.org/r/c7598605401a48d5cfeadebb678abd10af22b83f.1620691329.git.osandov@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Fix the comment mentioning ioctl command range used for zoned block
devices to reflect the range of commands actually implemented.
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Link: https://lore.kernel.org/r/20210509234806.3000-1-damien.lemoal@wdc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This reverts commit cd2c7545ae.
Alex reports that the commit causes corruption with LUKS on ext4. Revert
it for now so that this can be investigated properly.
Link: https://lore.kernel.org/linux-block/1620493841.bxdq8r5haw.none@localhost/
Reported-by: Alex Xu (Hello71) <alex_y_xu@yahoo.ca>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
