Yu Kuai [Sat, 27 May 2023 01:06:44 +0000 (09:06 +0800)]
blk-sysfs: add a new attr_group for blk_mq
Currently wbt sysfs entry is created for bio based device, and wbt can
be enabled for such device through sysfs while it doesn't make sense
because wbt can only work for rq based device. In the meantime, there
are other similar sysfs entries.
Fix this by adding a new attr_group for blk_mq, and sysfs entries will
only be created when the device is rq based.
wbt_init() will be called from wbt_enable_default() if wbt is not
initialized, currently this is only possible in blk_register_queue(), hence
wbt_init() will never be called from iocost and this warning is false
positive.
However, we might support rq_qos destruction dynamically in the future,
and it's better to prevent that, hence move wbt_enable_default() outside
'ioc->lock'. This is safe because queue is still freezed.
Yu Kuai [Sat, 27 May 2023 01:06:42 +0000 (09:06 +0800)]
blk-wbt: cleanup rwb_enabled() and wbt_disabled()
'wb_normal' will set to 0 if 'min_lat_nsec' is 0, and 'min_lat_nsec' can
only be set to 0 through sysfs configuration where 'WBT_STATE_OFF_MANUAL'
is set together, in the meantime, they can only be cleared together
through sysfs afterwards. Hence 'wb_normal != 0' is the same as
'rwb->enable_state != WBT_STATE_OFF_MANUAL'.
The code is redundan, hence replace the checking of 'wb_normal' to
'enable_state' in rwb_enabled() and reuse rwb_enabled() for
wbt_disabled().
Yu Kuai [Sat, 27 May 2023 01:06:41 +0000 (09:06 +0800)]
blk-wbt: remove dead code to handle wbt enable/disable with io inflight
enable or disable wbt is always called with queue freezed, so that wbt
can never be enabled or disabled while io is still inflight, and this
behaviour should always hold to avoid io hang(There have been reported
several times).
Therefor, the code to handle wbt enable/diskble with io inflight is not
and never will be used, hence remove such dead code.
Yu Kuai [Sat, 27 May 2023 01:06:40 +0000 (09:06 +0800)]
blk-wbt: don't create wbt sysfs entry if CONFIG_BLK_WBT is disabled
sysfs entry /sys/block/[device]/queue/wbt_lat_usec will be created even
if CONFIG_BLK_WBT is disabled, while read and write will always fail.
It doesn't make sense to create a sysfs entry that can't be accessed,
so don't create such entry.
Ming Lei [Sat, 24 Jun 2023 13:01:05 +0000 (21:01 +0800)]
blk-mq: fix two misuses on RQF_USE_SCHED
Request allocated from sched tags can't be issued via ->queue_rqs()
directly, since driver tag isn't allocated yet. This is the 1st misuse
of RQF_USE_SCHED for figuring out plug->has_elevator.
Request allocated from sched tags can't be ended by
blk_mq_end_request_batch() too, fix the 2nd RQF_USE_SCHED misuse
in blk_mq_add_to_batch().
Without this patch, NVMe uring cmd passthrough IO workload can run into
hang easily with real io scheduler.
Jinke Han [Sun, 7 May 2023 17:06:31 +0000 (01:06 +0800)]
blk-throttle: Fix io statistics for cgroup v1
After commit f382fb0bcef4 ("block: remove legacy IO schedulers"),
blkio.throttle.io_serviced and blkio.throttle.io_service_bytes become
the only stable io stats interface of cgroup v1, and these statistics
are done in the blk-throttle code. But the current code only counts the
bios that are actually throttled. When the user does not add the throttle
limit, the io stats for cgroup v1 has nothing. I fix it according to the
statistical method of v2, and made it count all ios accurately.
Jan Kara [Thu, 22 Jun 2023 16:46:55 +0000 (18:46 +0200)]
bcache: Fix bcache device claiming
Commit 2736e8eeb0cc ("block: use the holder as indication for exclusive
opens") introduced a change that blkdev_put() has to get exclusive
holder of the bdev as an argument. However it overlooked that
register_bdev() and register_cache() overwrite the bdev->bd_holder field
in the block device to point to the real owning object which was not
available at the time we called blkdev_get_by_path(). Messing with bdev
internals like this is a layering violation and it also causes
blkdev_put() to issue warning about mismatching holders.
Fix bcache to reopen the block device with appropriate holder once it is
available which also restores the behavior that multiple bcache caches
cannot claim the same device which was broken by commit 29499ab060fe
("bcache: don't pass a stack address to blkdev_get_by_path").
Jan Kara [Thu, 22 Jun 2023 16:46:54 +0000 (18:46 +0200)]
bcache: Alloc holder object before async registration
Allocate holder object (cache or cached_dev) before offloading the
rest of the startup to async work. This will allow us to open the block
block device with proper holder.
Jens Axboe [Fri, 23 Jun 2023 17:59:05 +0000 (11:59 -0600)]
Merge tag 'md-next-20230623' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md into for-6.5/block-late
Pull MD fixes from Song.
* tag 'md-next-20230623' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md:
raid10: avoid spin_lock from fastpath from raid10_unplug()
md: fix 'delete_mutex' deadlock
md: use mddev->external to select holder in export_rdev()
md/raid1-10: fix casting from randomized structure in raid1_submit_write()
md/raid10: fix the condition to call bio_end_io_acct()
Yu Kuai [Wed, 21 Jun 2023 14:29:33 +0000 (22:29 +0800)]
md: fix 'delete_mutex' deadlock
Commit 3ce94ce5d05a ("md: fix duplicate filename for rdev") introduce a
new lock 'delete_mutex', and trigger a new deadlock:
t1: remove rdev t2: sysfs writer
rdev_attr_store rdev_attr_store
mddev_lock
state_store
md_kick_rdev_from_array
lock delete_mutex
list_add mddev->deleting
unlock delete_mutex
mddev_unlock
mddev_lock
...
lock delete_mutex
kobject_del
// wait for sysfs writers to be done
mddev_unlock
lock delete_mutex
// wait for delete_mutex, deadlock
'delete_mutex' is used to protect the list 'mddev->deleting', turns out
that this list can be protected by 'reconfig_mutex' directly, and this
lock can be removed.
Fix this problem by removing the lock, and use 'reconfig_mutex' to
protect the list. mddev_unlock() will move this list to a local list to
be handled after 'reconfig_mutex' is dropped.
This means export_rdev() calls blkdev_put with a different holder than the
one used by blkdev_get_by_dev(). This is because mddev->major_version == -2
is not a good check for external metadata. Fix this by using
mddev->external instead.
Also, do not clear mddev->external in md_clean(), as the flag might be used
later in export_rdev().
Li Nan [Fri, 9 Jun 2023 09:43:20 +0000 (17:43 +0800)]
md/raid10: fix the condition to call bio_end_io_acct()
/sys/block/[device]/queue/iostats is used to control whether to count io
stat. Write 0 to it will clear queue_flags QUEUE_FLAG_IO_STAT which means
iostats is disabled. If we disable iostats and later endable it, the io
issued during this period will be counted incorrectly, inflight will be
decreased to -1.
//T1 set iostats
echo 0 > /sys/block/md0/queue/iostats
clear QUEUE_FLAG_IO_STAT
//T2 issue io
if (QUEUE_FLAG_IO_STAT) -> false
bio_start_io_acct
inflight++
echo 1 > /sys/block/md0/queue/iostats
set QUEUE_FLAG_IO_STAT
//T3 io end
if (QUEUE_FLAG_IO_STAT) -> true
bio_end_io_acct
inflight-- -> -1
Also, if iostats is enabled while issuing io but disabled while io end,
inflight will never be decreased.
Fix it by checking start_time when io end. If start_time is not 0, call
bio_end_io_acct().
Yu Kuai [Wed, 21 Jun 2023 16:01:11 +0000 (00:01 +0800)]
scsi/sg: don't grab scsi host module reference
In order to prevent request_queue to be freed before cleaning up
blktrace debugfs entries, commit db59133e9279 ("scsi: sg: fix blktrace
debugfs entries leakage") use scsi_device_get(), however,
scsi_device_get() will also grab scsi module reference and scsi module
can't be removed.
It's reported that blktests can't unload scsi_debug after block/001:
Fix this problem by grabbing request_queue reference directly, so that
scsi host module can still be unloaded while request_queue will be
pinged by sg device.
block: don't return -EINVAL for not found names in devt_from_devname
When we didn't find a device and didn't guess it might be a partition,
it might still show up later, so don't disable rootwait for it by
returning -EINVAL.
Bart Van Assche [Wed, 21 Jun 2023 16:50:54 +0000 (09:50 -0700)]
block: Improve kernel-doc headers
Fix the documentation of the devt_from_partuuid() return value.
Fix the following two recently introduced kernel-doc warnings:
block/bdev.c:570: warning: Function parameter or member 'hops' not described in 'bd_finish_claiming'
block/early-lookup.c:46: warning: Function parameter or member 'devt' not described in 'devt_from_partuuid'
Ming Lei [Wed, 21 Jun 2023 13:22:08 +0000 (21:22 +0800)]
blk-mq: don't insert passthrough request into sw queue
In case of real io scheduler, q->elevator is set, so blk_mq_run_hw_queue()
may just check if scheduler queue has request to dispatch, see
__blk_mq_sched_dispatch_requests(). Then IO hang may be caused because
all passthorugh requests may stay in sw queue.
And any passthrough request should have been inserted to hctx->dispatch
always.
Ivan Orlov [Tue, 20 Jun 2023 18:01:33 +0000 (20:01 +0200)]
bsg: make bsg_class a static const structure
Now that the driver core allows for struct class to be in read-only
memory, move the bsg_class structure to be declared at build time
placing it into read-only memory, instead of having to be dynamically
allocated at boot time.
Ivan Orlov [Tue, 20 Jun 2023 18:01:32 +0000 (20:01 +0200)]
ublk: make ublk_chr_class a static const structure
Now that the driver core allows for struct class to be in read-only
memory, move the ublk_chr_class structure to be declared at build time
placing it into read-only memory, instead of having to be dynamically
allocated at boot time.
Ivan Orlov [Tue, 20 Jun 2023 18:01:31 +0000 (20:01 +0200)]
aoe: make aoe_class a static const structure
Now that the driver core allows for struct class to be in read-only
memory, move the aoe_class structure to be declared at build time
placing it into read-only memory, instead of having to be dynamically
allocated at boot time.
Ivan Orlov [Tue, 20 Jun 2023 18:01:30 +0000 (20:01 +0200)]
block/rnbd: make all 'class' structures const
Now that the driver core allows for struct class to be in read-only
memory, making all 'class' structures to be declared at build time
placing them into read-only memory, instead of having to be dynamically
allocated at load time.
Michael Schmitz [Tue, 20 Jun 2023 20:17:25 +0000 (08:17 +1200)]
block: add overflow checks for Amiga partition support
The Amiga partition parser module uses signed int for partition sector
address and count, which will overflow for disks larger than 1 TB.
Use u64 as type for sector address and size to allow using disks up to
2 TB without LBD support, and disks larger than 2 TB with LBD. The RBD
format allows to specify disk sizes up to 2^128 bytes (though native
OS limitations reduce this somewhat, to max 2^68 bytes), so check for
u64 overflow carefully to protect against overflowing sector_t.
Bail out if sector addresses overflow 32 bits on kernels without LBD
support.
This bug was reported originally in 2012, and the fix was created by
the RDB author, Joanne Dow <[email protected]>. A patch had been
discussed and reviewed on linux-m68k at that time but never officially
submitted (now resubmitted as patch 1 in this series).
This patch adds additional error checking and warning messages.
Michael Schmitz [Tue, 20 Jun 2023 20:17:24 +0000 (08:17 +1200)]
block: change all __u32 annotations to __be32 in affs_hardblocks.h
The Amiga partition parser module uses signed int for partition sector
address and count, which will overflow for disks larger than 1 TB.
Use u64 as type for sector address and size to allow using disks up to
2 TB without LBD support, and disks larger than 2 TB with LBD. The RBD
format allows to specify disk sizes up to 2^128 bytes (though native
OS limitations reduce this somewhat, to max 2^68 bytes), so check for
u64 overflow carefully to protect against overflowing sector_t.
This bug was reported originally in 2012, and the fix was created by
the RDB author, Joanne Dow <[email protected]>. A patch had been
discussed and reviewed on linux-m68k at that time but never officially
submitted (now resubmitted as patch 1 of this series).
Patch 3 (this series) adds additional error checking and warning
messages. One of the error checks now makes use of the previously
unused rdb_CylBlocks field, which causes a 'sparse' warning
(cast to restricted __be32).
Annotate all 32 bit fields in affs_hardblocks.h as __be32, as the
on-disk format of RDB and partition blocks is always big endian.
Michael Schmitz [Tue, 20 Jun 2023 20:17:23 +0000 (08:17 +1200)]
block: fix signed int overflow in Amiga partition support
The Amiga partition parser module uses signed int for partition sector
address and count, which will overflow for disks larger than 1 TB.
Use sector_t as type for sector address and size to allow using disks
up to 2 TB without LBD support, and disks larger than 2 TB with LBD.
This bug was reported originally in 2012, and the fix was created by
the RDB author, Joanne Dow <[email protected]>. A patch had been
discussed and reviewed on linux-m68k at that time but never officially
submitted. This patch differs from Joanne's patch only in its use of
sector_t instead of unsigned int. No checking for overflows is done
(see patch 3 of this series for that).
Min Li [Mon, 19 Jun 2023 09:12:14 +0000 (09:12 +0000)]
block: add capacity validation in bdev_add_partition()
In the function bdev_add_partition(),there is no check that the start
and end sectors exceed the size of the disk before calling add_partition.
When we call the block's ioctl interface directly to add a partition,
and the capacity of the disk is set to 0 by driver,the command will
continue to execute.
Yu Kuai [Tue, 20 Jun 2023 11:13:22 +0000 (19:13 +0800)]
reiserfs: fix blkdev_put() warning from release_journal_dev()
In journal_init_dev(), if super bdev is used as 'j_dev_bd', then
blkdev_get_by_dev() is called with NULL holder, otherwise, holder will
be journal. However, later in release_journal_dev(), blkdev_put() is
called with journal unconditionally, cause following warning:
Yu Kuai [Sun, 18 Jun 2023 14:04:02 +0000 (22:04 +0800)]
block: fix wrong mode for blkdev_get_by_dev() from disk_scan_partitions()
After commit 2736e8eeb0cc ("block: use the holder as indication for
exclusive opens"), blkdev_get_by_dev() will warn if holder is NULL and
mode contains 'FMODE_EXCL'.
holder from blkdev_get_by_dev() from disk_scan_partitions() is always NULL,
hence it should not use 'FMODE_EXCL', which is broben by the commit. For
consequence, WARN_ON_ONCE() will be triggered from blkdev_get_by_dev()
if user scan partitions with device opened exclusively.
Fix this problem by removing 'FMODE_EXCL' from disk_scan_partitions(),
as it used to be.
block: increment diskseq on all media change events
Currently, associating a loop device with a different file descriptor
does not increment its diskseq. This allows the following race
condition:
1. Program X opens a loop device
2. Program X gets the diskseq of the loop device.
3. Program X associates a file with the loop device.
4. Program X passes the loop device major, minor, and diskseq to
something.
5. Program X exits.
6. Program Y detaches the file from the loop device.
7. Program Y attaches a different file to the loop device.
8. The opener finally gets around to opening the loop device and checks
that the diskseq is what it expects it to be. Even though the
diskseq is the expected value, the result is that the opener is
accessing the wrong file.
From discussions with Christoph Hellwig, it appears that
disk_force_media_change() was supposed to call inc_diskseq(), but in
fact it does not. Adding a Fixes: tag to indicate this. Christoph's
Reported-by is because he stated that disk_force_media_change()
calls inc_diskseq(), which is what led me to discover that it should but
does not.
Jens Axboe [Fri, 16 Jun 2023 15:57:40 +0000 (09:57 -0600)]
Merge tag 'nvme-6.5-2023-06-16' of git://git.infradead.org/nvme into for-6.5/block
Pull NVMe updates from Keith:
"nvme updates for Linux 6.5
- Various cleanups all around (Irvin, Chaitanya, Christophe)
- Better struct packing (Christophe JAILLET)
- Reduce controller error logs for optional commands (Keith)
- Support for >=64KiB block sizes (Daniel Gomez)
- Fabrics fixes and code organization (Max, Chaitanya, Daniel Wagner)"
* tag 'nvme-6.5-2023-06-16' of git://git.infradead.org/nvme: (27 commits)
nvme: forward port sysfs delete fix
nvme: skip optional id ctrl csi if it failed
nvme-core: use nvme_ns_head_multipath instead of ns->head->disk
nvmet-fcloop: Do not wait on completion when unregister fails
nvme-fabrics: open code __nvmf_host_find()
nvme-fabrics: error out to unlock the mutex
nvme: Increase block size variable size to 32-bit
nvme-fcloop: no need to return from void function
nvmet-auth: remove unnecessary break after goto
nvmet-auth: remove some dead code
nvme-core: remove redundant check from nvme_init_ns_head
nvme: move sysfs code to a dedicated sysfs.c file
nvme-fabrics: prevent overriding of existing host
nvme-fabrics: check hostid using uuid_equal
nvme-fabrics: unify common code in admin and io queue connect
nvmet: reorder fields in 'struct nvmefc_fcp_req'
nvmet: reorder fields in 'struct nvme_dhchap_queue_context'
nvmet: reorder fields in 'struct nvmf_ctrl_options'
nvme: reorder fields in 'struct nvme_ctrl'
nvmet: reorder fields in 'struct nvmet_sq'
...
Keith Busch [Thu, 15 Jun 2023 15:17:57 +0000 (08:17 -0700)]
nvme: forward port sysfs delete fix
We had a late fix that modified nvme_sysfs_delete() after the staging
branch for the next merge window relocated the function to a new file.
Port commit 2eb94dd56a4a4 ("nvme: do not let the user delete a ctrl
before a complete") to the latest to avoid a potentially confusing merge
conflict.
bch_btree_check_thread() and bch_dirty_init_thread() may call
mca_cannibalize() to cannibalize other cached btree nodes. Only one thread
can do it at a time, so the op of other threads will be added to the
btree_cache_wait list.
We must call finish_wait() to remove op from btree_cache_wait before free
it's memory address. Otherwise, the list will be damaged. Also should call
bch_cannibalize_unlock() to release the btree_cache_alloc_lock and wake_up
other waiters.
Zheng Wang [Thu, 15 Jun 2023 12:12:22 +0000 (20:12 +0800)]
bcache: Fix __bch_btree_node_alloc to make the failure behavior consistent
In some specific situations, the return value of __bch_btree_node_alloc
may be NULL. This may lead to a potential NULL pointer dereference in
caller function like a calling chain :
btree_split->bch_btree_node_alloc->__bch_btree_node_alloc.
Fix it by initializing the return value in __bch_btree_node_alloc.
Zheng Wang [Thu, 15 Jun 2023 12:12:21 +0000 (20:12 +0800)]
bcache: Remove unnecessary NULL point check in node allocations
Due to the previous fix of __bch_btree_node_alloc, the return value will
never be a NULL pointer. So IS_ERR is enough to handle the failure
situation. Fix it by replacing IS_ERR_OR_NULL check by an IS_ERR check.
ye xingchen [Thu, 15 Jun 2023 12:12:18 +0000 (20:12 +0800)]
bcache: Convert to use sysfs_emit()/sysfs_emit_at() APIs
Follow the advice of the Documentation/filesystems/sysfs.rst and show()
should only use sysfs_emit() or sysfs_emit_at() when formatting the
value to be returned to user space.
Yu Kuai [Sat, 10 Jun 2023 02:20:03 +0000 (10:20 +0800)]
block: fix blktrace debugfs entries leakage
Commit 99d055b4fd4b ("block: remove per-disk debugfs files in
blk_unregister_queue") moves blk_trace_shutdown() from
blk_release_queue() to blk_unregister_queue(), this is safe if blktrace
is created through sysfs, however, there is a regression in corner
case.
blktrace can still be enabled after del_gendisk() through ioctl if
the disk is opened before del_gendisk(), and if blktrace is not shutdown
through ioctl before closing the disk, debugfs entries will be leaked.
Fix this problem by shutdown blktrace in disk_release(), this is safe
because blk_trace_remove() is reentrant.
Yu Kuai [Sat, 10 Jun 2023 02:20:02 +0000 (10:20 +0800)]
scsi: sg: fix blktrace debugfs entries leakage
sg_ioctl() support to enable blktrace, which will create debugfs entries
"/sys/kernel/debug/block/sgx/", however, there is no guarantee that user
will remove these entries through ioctl, and deleting sg device doesn't
cleanup these blktrace entries.
This problem can be fixed by cleanup blktrace while releasing
request_queue, however, it's not a good idea to do this special handling
in common layer just for sg device.
Fix this problem by shutdown bltkrace in sg_device_destroy(), where the
device is deleted and all the users close the device, also grab a
scsi_device reference from sg_add_device() to prevent scsi_device to be
freed before sg_device_destroy();
Yu Kuai [Sat, 10 Jun 2023 02:20:01 +0000 (10:20 +0800)]
blktrace: use inline function for blk_trace_remove() while blktrace is disabled
If config is disabled, call blk_trace_remove() directly will trigger
build warning, hence use inline function instead, prepare to fix
blktrace debugfs entries leakage.
Ed Tsai [Wed, 14 Jun 2023 00:25:29 +0000 (08:25 +0800)]
blk-mq: check on cpu id when there is only one ctx mapping
commit f168420c62e7 ("blk-mq: don't redirect completion for hctx withs
only one ctx mapping") When nvme applies a 1:1 mapping of hctx and ctx,
there will be no remote request.
But for ufs, the submission and completion queues could be asymmetric.
(e.g. Multiple SQs share one CQ) Therefore, 1:1 mapping of hctx and
ctx won't complete request on the submission cpu. In this situation,
this nr_ctx check could violate the QUEUE_FLAG_SAME_FORCE, as a result,
check on cpu id when there is only one ctx mapping.
Jens Axboe [Wed, 14 Jun 2023 12:58:43 +0000 (06:58 -0600)]
Merge tag 'md-next-20230613' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md into for-6.5/block
Pull MD updates from Song:
"The major changes are:
1. Protect md_thread with rcu, by Yu Kuai;
2. Various non-urgent raid5 and raid1/10 fixes, by Yu Kuai;
3. Non-urgent raid10 fixes, by Li Nan."
* tag 'md-next-20230613' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md: (29 commits)
md/raid1-10: limit the number of plugged bio
md/raid1-10: don't handle pluged bio by daemon thread
md/md-bitmap: add a new helper to unplug bitmap asynchrously
md/raid1-10: submit write io directly if bitmap is not enabled
md/raid1-10: factor out a helper to submit normal write
md/raid1-10: factor out a helper to add bio to plug
md/raid10: prevent soft lockup while flush writes
md/raid10: fix io loss while replacement replace rdev
md/raid10: Do not add spare disk when recovery fails
md/raid10: clean up md_add_new_disk()
md/raid10: prioritize adding disk to 'removed' mirror
md/raid10: improve code of mrdev in raid10_sync_request
md/raid10: fix null-ptr-deref of mreplace in raid10_sync_request
md/raid5: don't start reshape when recovery or replace is in progress
md: protect md_thread with rcu
md/bitmap: factor out a helper to set timeout
md/bitmap: always wake up md_thread in timeout_store
dm-raid: remove useless checking in raid_message()
md: factor out a helper to wake up md_thread directly
md: fix duplicate filename for rdev
...
David Howells [Tue, 13 Jun 2023 21:54:39 +0000 (22:54 +0100)]
block: Fix dio_cleanup() to advance the head index
Fix dio_bio_cleanup() to advance the head index into the list of pages past
the pages it has released, as __blockdev_direct_IO() will call it twice if
do_direct_IO() fails.
The issue was causing:
WARNING: CPU: 6 PID: 2220 at mm/gup.c:76 try_get_folio
This can be triggered by setting up a clean pair of UDF filesystems on
loopback devices and running the generic/451 xfstest with them as the
scratch and test partitions. Something like the following:
Yu Kuai [Mon, 29 May 2023 13:11:05 +0000 (21:11 +0800)]
md/raid1-10: don't handle pluged bio by daemon thread
current->bio_list will be set under submit_bio() context, in this case
bitmap io will be added to the list and wait for current io submission to
finish, while current io submission must wait for bitmap io to be done.
commit 874807a83139 ("md/raid1{,0}: fix deadlock in bitmap_unplug.") fix
the deadlock by handling plugged bio by daemon thread.
On the one hand, the deadlock won't exist after commit a214b949d8e3
("blk-mq: only flush requests from the plug in blk_mq_submit_bio"). On
the other hand, current solution makes it impossible to flush plugged bio
in raid1/10_make_request(), because this will cause that all the writes
will goto daemon thread.
In order to limit the number of plugged bio, commit 874807a83139
("md/raid1{,0}: fix deadlock in bitmap_unplug.") is reverted, and the
deadlock is fixed by handling bitmap io asynchronously.
Yu Kuai [Mon, 29 May 2023 13:11:04 +0000 (21:11 +0800)]
md/md-bitmap: add a new helper to unplug bitmap asynchrously
If bitmap is enabled, bitmap must update before submitting write io, this
is why unplug callback must move these io to 'conf->pending_io_list' if
'current->bio_list' is not empty, which will suffer performance
degradation.
A new helper md_bitmap_unplug_async() is introduced to submit bitmap io
in a kworker, so that submit bitmap io in raid10_unplug() doesn't require
that 'current->bio_list' is empty.
This patch prepare to limit the number of plugged bio.
Yu Kuai [Mon, 29 May 2023 13:11:03 +0000 (21:11 +0800)]
md/raid1-10: submit write io directly if bitmap is not enabled
Commit 6cce3b23f6f8 ("[PATCH] md: write intent bitmap support for raid10")
add bitmap support, and it changed that write io is submitted through
daemon thread because bitmap need to be updated before write io. And
later, plug is used to fix performance regression because all the write io
will go to demon thread, which means io can't be issued concurrently.
However, if bitmap is not enabled, the write io should not go to daemon
thread in the first place, and plug is not needed as well.
Yu Kuai [Mon, 29 May 2023 13:11:00 +0000 (21:11 +0800)]
md/raid10: prevent soft lockup while flush writes
Currently, there is no limit for raid1/raid10 plugged bio. While flushing
writes, raid1 has cond_resched() while raid10 doesn't, and too many
writes can cause soft lockup.
Follow up soft lockup can be triggered easily with writeback test for
raid10 with ramdisks:
Fix the problem by adding cond_resched() to raid10 like what raid1 did.
Note that unlimited plugged bio still need to be optimized, for example,
in the case of lots of dirty pages writeback, this will take lots of
memory and io will spend a long time in plug, hence io latency is bad.
Li Nan [Fri, 2 Jun 2023 09:18:39 +0000 (17:18 +0800)]
md/raid10: fix io loss while replacement replace rdev
When removing a disk with replacement, the replacement will be used to
replace rdev. During this process, there is a brief window in which both
rdev and replacement are read as NULL in raid10_write_request(). This
will result in io not being submitted but it should be.
Li Nan [Fri, 2 Jun 2023 09:18:38 +0000 (17:18 +0800)]
md/raid10: Do not add spare disk when recovery fails
In raid10_sync_request(), if data cannot be read from any disk for
recovery, it will go to 'giveup' and let 'chunks_skipped' + 1. After
multiple 'giveup', when 'chunks_skipped >= geo.raid_disks', it will
return 'max_sector', indicating that the recovery has been completed.
However, the recovery is just aborted and the data remains inconsistent.
Fix it by setting mirror->recovery_disabled, which will prevent the spare
disk from being added to this mirror. The same issue also exists during
resync, it will be fixed afterwards.
Li Nan [Sat, 27 May 2023 10:18:51 +0000 (18:18 +0800)]
md/raid10: clean up md_add_new_disk()
Commit 1a855a060665 ("md: fix bug with re-adding of partially recovered
device.") only add device which is set to In_sync. But it let devices
without metadata cannot be added when they should be.
Commit bf572541ab44 ("md: fix regression with re-adding devices to arrays
with no metadata") fix the above issue, it set device without metadata to
In_sync when add new disk.
However, after commit f466722ca614 ("md: Change handling of save_raid_disk
and metadata update during recovery.") deletes changes of the first patch,
setting In_sync for devcie without metadata is meanless because the flag
will be cleared soon and will not be used during this period. Clean it up.
Li Nan [Sat, 27 May 2023 09:20:07 +0000 (17:20 +0800)]
md/raid10: prioritize adding disk to 'removed' mirror
When add a new disk to raid10, it will traverse conf->mirror from start
and find one of the following mirror to add:
1. mirror->rdev is set to WantReplacement and it have no replacement,
set new disk to mirror->replacement.
2. no mirror->rdev, set new disk to mirror->rdev.
There is a array as below (sda is set to WantReplacement):
Number Major Minor RaidDevice State
0 8 0 0 active sync set-A /dev/sda
- 0 0 1 removed
2 8 32 2 active sync set-A /dev/sdc
3 8 48 3 active sync set-B /dev/sdd
Use 'mdadm --add' to add a new disk to this array, the new disk will
become sda's replacement instead of add to removed position, which is
confusing for users. Meanwhile, after new disk recovery success, sda
will be set to Faulty.
Prioritize adding disk to 'removed' mirror is a better choice. In the
above scenario, the behavior is the same as before, except sda will not
be deleted. Before other disks are added, continued use sda is more
reliable.
Li Nan [Sat, 27 May 2023 07:22:16 +0000 (15:22 +0800)]
md/raid10: improve code of mrdev in raid10_sync_request
'need_recover' and 'mrdev' are equivalent in raid10_sync_request(), and
inc mrdev->nr_pending is unreasonable if don't need recovery. Replace
'need_recover' with 'mrdev', and only inc nr_pending when needed.
Li Nan [Sat, 27 May 2023 07:22:15 +0000 (15:22 +0800)]
md/raid10: fix null-ptr-deref of mreplace in raid10_sync_request
There are two check of 'mreplace' in raid10_sync_request(). In the first
check, 'need_replace' will be set and 'mreplace' will be used later if
no-Faulty 'mreplace' exists, In the second check, 'mreplace' will be
set to NULL if it is Faulty, but 'need_replace' will not be changed
accordingly. null-ptr-deref occurs if Faulty is set between two check.
Fix it by merging two checks into one. And replace 'need_replace' with
'mreplace' because their values are always the same.
Yu Kuai [Mon, 29 May 2023 13:34:10 +0000 (21:34 +0800)]
md/raid5: don't start reshape when recovery or replace is in progress
When recovery is interrupted (reboot, etc.) check for MD_RECOVERY_RUNNING
is not enough to tell recovery is in progress. Also check recovery_cp
before starting reshape.
Yu Kuai [Tue, 23 May 2023 02:10:17 +0000 (10:10 +0800)]
md: protect md_thread with rcu
Currently, there are many places that md_thread can be accessed without
protection, following are known scenarios that can cause
null-ptr-dereference or uaf:
1) sync_thread that is allocated and started from md_start_sync()
2) mddev->thread can be accessed directly from timeout_store() and
md_bitmap_daemon_work()
3) md_unregister_thread() from action_store().
Currently, a global spinlock 'pers_lock' is borrowed to protect
'mddev->thread' in some places, this problem can be fixed likewise,
however, use a global lock for all the cases is not good.
Fix this problem by protecting all md_thread with rcu.
Yu Kuai [Tue, 23 May 2023 02:10:16 +0000 (10:10 +0800)]
md/bitmap: factor out a helper to set timeout
Register/unregister 'mddev->thread' are both under 'reconfig_mutex',
however, some context didn't hold the mutex to access mddev->thread,
which can cause null-ptr-deference:
1) md_bitmap_daemon_work() can be called from md_check_recovery() where
'reconfig_mutex' is not held, deference 'mddev->thread' might cause
null-ptr-deference, because md_unregister_thread() reset the pointer
before stopping the thread.
2) timeout_store() access 'mddev->thread' multiple times,
null-ptr-deference can be triggered if 'mddev->thread' is reset in the
middle.
This patch factor out a helper to set timeout, the new helper always
check if 'mddev->thread' is null first, so that problem 1 can be fixed.
Now that this helper only access 'mddev->thread' once, but it's possible
that 'mddev->thread' can be freed while this helper is still in progress,
hence the problem is not fixed yet. Follow up patches will fix this by
protecting md_thread with rcu.
Yu Kuai [Tue, 23 May 2023 02:10:15 +0000 (10:10 +0800)]
md/bitmap: always wake up md_thread in timeout_store
md_wakeup_thread() can handle the case that pass in md_thread is NULL,
the only difference is that md_wakeup_thread() will be called when
current timeout is 'MAX_SCHEDULE_TIMEOUT', this should not matter
because timeout_store() is not hot path, and the daemon process is
woke up more than demand from other context already.
Yu Kuai [Tue, 23 May 2023 02:10:13 +0000 (10:10 +0800)]
md: factor out a helper to wake up md_thread directly
md_wakeup_thread() can't wakeup md_thread->tsk if md_thread->run is
still in progress, and in some cases md_thread->tsk need to be woke up
directly, like md_set_readonly() and do_md_stop().
Commit 9dfbdafda3b3 ("md: unlock mddev before reap sync_thread in
action_store") introduce a new scenario where unregister sync_thread is
not protected by 'reconfig_mutex', this can cause null-ptr-deference in
theroy:
t1: md_set_readonly t2: action_store
md_unregister_thread
// 'reconfig_mutex' is not held
// 'reconfig_mutex' is held by caller
if (mddev->sync_thread)
thread = *threadp
*threadp = NULL
wake_up_process(mddev->sync_thread->tsk)
// null-ptr-deference
Fix this problem by factoring out a helper to wake up md_thread directly,
so that 'sync_thread' won't be accessed multiple times from the reader
side. This helper also prepare to protect md_thread with rcu.
Noted that later patches is going to fix that unregister sync_thread is
not protected by 'reconfig_mutex' from action_store().
Yu Kuai [Tue, 23 May 2023 01:27:27 +0000 (09:27 +0800)]
md: fix duplicate filename for rdev
Commit 5792a2856a63 ("[PATCH] md: avoid a deadlock when removing a device
from an md array via sysfs") delays the deletion of rdev, however, this
introduces a window that rdev can be added again while the deletion is
not done yet, and sysfs will complain about duplicate filename.
Follow up patches try to fix this problem by flushing workqueue, however,
flush_rdev_wq() is just dead code, the progress in
md_kick_rdev_from_array():
So in flush_rdev_wq(), if rdev is found in the list, work_pending() can
never pass, in the meantime, if work is queued, then rdev can never be
found in the list.
flush_rdev_wq() can be replaced by flush_workqueue() directly, however,
this approach is not good:
- the workqueue is global, this synchronization for all raid disks is
not necessary.
- flush_workqueue can't be called under 'reconfig_mutex', there is still
a small window between flush_workqueue() and mddev_lock() that other
contexts can queue new work, hence the problem is not solved completely.
sysfs already has apis to support delete itself through writer, and
these apis, specifically sysfs_break/unbreak_active_protection(), is used
to support deleting rdev synchronously. Therefore, the above commit can be
reverted, and sysfs duplicate filename can be avoided.
A new mdadm regression test is proposed as well([1]).
Li Nan [Mon, 22 May 2023 07:25:33 +0000 (15:25 +0800)]
md/raid10: fix overflow of md/safe_mode_delay
There is no input check when echo md/safe_mode_delay in safe_delay_store().
And msec might also overflow when HZ < 1000 in safe_delay_show(), Fix it by
checking overflow in safe_delay_store() and use unsigned long conversion in
safe_delay_show().
Yu Kuai [Fri, 12 May 2023 01:56:10 +0000 (09:56 +0800)]
md/raid5: fix a deadlock in the case that reshape is interrupted
If reshape is in progress and io across reshape_position is issued, such
io will wait for reshape to make progress(see details in the case that
make_stripe_request() return STRIPE_SCHEDULE_AND_RETRY).
It has been reported several times that if system reboot while growing
raid5 to raid6, array assemble will hang infinitely([1, 2]). This is
because following deadlock is triggered:
1) a normal io is waiting for reshape to progress, this io can be from
system-udevd or mdadm.
2) while assemble, mdadm tries to suspend the array, hence
'reconfig_mutex' is held and mddev_suspend() must wait for normal io
to be done.
3) daemon thread can't start reshape because 'reconfig_mutex' can't be
held.
1) and 3) is unbreakable because they're foundation design. In order to
break 2), following is possible solutions that I can think of:
a) Let mddev_suspend() fail is not a good option, because this will
break many scenarios since mddev_suspend() doesn't fail before.
b) Fail the io that is waiting for reshape to make progress from
mddev_suspend().
c) Return false for the io that is waiting for reshape to make
progress from raid5_make_request(), and these io will wait for
suspend to be done in md_handle_request(), where 'active_io' is
not grabbed.
c) sounds better than b), however, b) is used because it's easy and
straightforward, and it's verified that mdadm can assemble in this case.
On the other hand, c) breaks the logic that mddev_suspend() will wait
for submitted io to be completely handled.
Fix the problem by checking reshape in mddev_suspend(), if reshape can't
make progress and there are still some io waiting for reshape, fail
those io.
Yu Kuai [Fri, 12 May 2023 01:56:06 +0000 (09:56 +0800)]
md/raid5: don't allow replacement while reshape is in progress
If reshape is interrupted(for example, echo frozen to sync_action), then
rdev replacement can be set. It's safe because reshape is always prior to
resync in md_check_recovery(). However, if system reboots, then kernel will
complain cannot handle concurrent replacement and reshape and this array
is not able to assemble anymore.
Fix this problem by don't allow replacement until reshape is done.
Li Nan [Mon, 15 May 2023 13:48:05 +0000 (21:48 +0800)]
md/raid10: check slab-out-of-bounds in md_bitmap_get_counter
If we write a large number to md/bitmap_set_bits, md_bitmap_checkpage()
will return -EINVAL because 'page >= bitmap->pages', but the return value
was not checked immediately in md_bitmap_get_counter() in order to set
*blocks value and slab-out-of-bounds occurs.
Move check of 'page >= bitmap->pages' to md_bitmap_get_counter() and
return directly if true.
Arnd Bergmann [Wed, 17 May 2023 13:22:12 +0000 (15:22 +0200)]
raid6: neon: add missing prototypes
The raid6 syndrome functions are generated for different sizes and have
no generic prototype, while in the inner functions have a prototype
in a header that cannot be included from the correct file. In both
cases, the compiler warns about missing prototypes:
lib/raid6/recov_neon_inner.c:27:6: warning: no previous prototype for '__raid6_2data_recov_neon' [-Wmissing-prototypes]
lib/raid6/recov_neon_inner.c:77:6: warning: no previous prototype for '__raid6_datap_recov_neon' [-Wmissing-prototypes]
lib/raid6/neon1.c:56:6: warning: no previous prototype for 'raid6_neon1_gen_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon1.c:86:6: warning: no previous prototype for 'raid6_neon1_xor_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon2.c:56:6: warning: no previous prototype for 'raid6_neon2_gen_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon2.c:97:6: warning: no previous prototype for 'raid6_neon2_xor_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon4.c:56:6: warning: no previous prototype for 'raid6_neon4_gen_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon4.c:119:6: warning: no previous prototype for 'raid6_neon4_xor_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon8.c:56:6: warning: no previous prototype for 'raid6_neon8_gen_syndrome_real' [-Wmissing-prototypes]
lib/raid6/neon8.c:163:6: warning: no previous prototype for 'raid6_neon8_xor_syndrome_real' [-Wmissing-prototypes]
Add a new header file that contains the prototypes for both to avoid
the warnings.
Keith Busch [Tue, 30 May 2023 15:18:20 +0000 (08:18 -0700)]
nvme: skip optional id ctrl csi if it failed
A frequently recieved report is the driver requests the optional Command
Set Specific Identify Controller structure. Some controllers report this
in their error log, which tiggers other warnings to user space
monitoring the devices.
These error entries are harmless and of questionable value to save in
the log, but let's reduce their occurance by not resending the command
if it previously failed. This will not prevent the errors on the initial
module load, but will greatly reduce their occurance on any rescans and
resumes from suspend.
Daniel Wagner [Tue, 6 Jun 2023 12:24:11 +0000 (14:24 +0200)]
nvmet-fcloop: Do not wait on completion when unregister fails
The nvme_fc_unregister_localport() returns an error code in case that
the locaport pointer is NULL or has already been unegisterd. localport is
is either in the ONLINE state (all resources allocated) or has already
been put into DELETED state.
In this case we will never receive an wakeup call and thus any caller
will hang, e.g. module unload.
There is no point in maintaining a separate funciton __nvmf_host_find()
that has only one caller nvmf_host_add() especially when caller and
callee both are small enough to merge.
Due to this we are actually repeating the error handling code in both
callee and caller for no reason that can be avoided, but instead we have
to read both function to establish the correctness along with additional
lockdep warning check due to involved locking.
Just open code __nvmf_host_find() in nvme_host_alloc() with appropriate
comment that removes repeated error checks in the callee/caller and
lockdep check that is needed for the nvmf_hosts_mutex involvement,
diffstats :-
Currently, in the nvmf_host_add() function, if the nvmf_host_alloc()
call failed to allocate memory for the host, the code would directly
return -ENOMEM without unlocking the nvmf_hosts_mutex. This could
lead to potential issues with mutex synchronization.
Fix that error handling mechanism by jumping to the out_unlock label
when nvmf_host_alloc() fails. This ensures that the mutex is unlocked
before returning the error code. The updated code enhances avoids
possible deadlocks.
Max Gurtovoy [Mon, 24 Apr 2023 21:12:42 +0000 (00:12 +0300)]
nvme: move sysfs code to a dedicated sysfs.c file
The core.c file became long and hard to maintain. Create a dedicated
file to centralize the sysfs functionality. This is a common practice to
separate sysfs/configfs related logic from the main driver logic .c file.
For example, in the nvmet module the configfs interface has its own
dedicated file.
This patch does not include any functional changes.
Max Gurtovoy [Fri, 12 May 2023 15:41:55 +0000 (18:41 +0300)]
nvme-fabrics: prevent overriding of existing host
When first connecting a target using the "default" host parameters,
setting the hostid from the command line during a subsequent connection
establishment would override the "default" hostid parameter. This would
cause an existing connection that is already using the host definitions
to lose its hostid.
To address this issue, the code has been modified to allow only 1:1
mapping between hostnqn and hostid. This will maintain unambiguous host
identification. Any non 1:1 mapping will be rejected during connection
establishment.
projects / linux.git / log