generic_make_request has always been very confusingly misnamed, so rename
it to submit_bio_noacct to make it clear that it is submit_bio minus
accounting and a few checks.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
With commit fe64369163c5 ("dm thin: don't allow changing data device
during thin-pool load") it is now possible to re-parent the data
device's flush_bio from the pool_c to pool structure. Doing so offers
improved lifetime guarantees for the flush_bio so that the call to
dm_pool_register_pre_commit_callback can now be done safely from
pool_ctr().
Depends-on: fe64369163c5 ("dm thin: don't allow changing data device during thin-pool load")
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
The existing code allows changing the data device when the thin-pool
target is reloaded.
This capability is not required and only complicates device lifetime
guarantees. This can cause crashes like the one reported here:
https://bugzilla.redhat.com/show_bug.cgi?id=1788596
where the kernel tries to issue a flush bio located in a structure that
was already freed.
Take the first step to simplifying the thin-pool's data device lifetime
by disallowing changing it. Like the thin-pool's metadata device, the
data device is now set in pool_create() and it cannot be changed for a
given thin-pool.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
dm-thin uses struct pool to hold the state of the pool. There may be
multiple pool_c's pointing to a given pool, each pool_c represents a
loaded target. pool_c's may be created and destroyed arbitrarily and the
pool contains a reference count of pool_c's pointing to it.
Since commit 694cfe7f31 ("dm thin: Flush data device before
committing metadata") a pointer to pool_c is passed to
dm_pool_register_pre_commit_callback and this function stores it in
pmd->pre_commit_context. If this pool_c is freed, but pool is not
(because there is another pool_c referencing it), we end up in a
situation where pmd->pre_commit_context structure points to freed
pool_c. It causes a crash in metadata_pre_commit_callback.
Fix this by moving the dm_pool_register_pre_commit_callback() from
pool_ctr() to pool_preresume(). This way the in-core thin-pool metadata
is only ever armed with callback data whose lifetime matches the
active thin-pool target.
In should be noted that this fix preserves the ability to load a
thin-pool table that uses a different data block device (that contains
the same data) -- though it is unclear if that capability is still
useful and/or needed.
Fixes: 694cfe7f31 ("dm thin: Flush data device before committing metadata")
Cc: stable@vger.kernel.org
Reported-by: Zdenek Kabelac <zkabelac@redhat.com>
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
The thin provisioning target maintains per thin device mappings that map
virtual blocks to data blocks in the data device.
When we write to a shared block, in case of internal snapshots, or
provision a new block, in case of external snapshots, we copy the shared
block to a new data block (COW), update the mapping for the relevant
virtual block and then issue the write to the new data block.
Suppose the data device has a volatile write-back cache and the
following sequence of events occur:
1. We write to a shared block
2. A new data block is allocated
3. We copy the shared block to the new data block using kcopyd (COW)
4. We insert the new mapping for the virtual block in the btree for that
thin device.
5. The commit timeout expires and we commit the metadata, that now
includes the new mapping from step (4).
6. The system crashes and the data device's cache has not been flushed,
meaning that the COWed data are lost.
The next time we read that virtual block of the thin device we read it
from the data block allocated in step (2), since the metadata have been
successfully committed. The data are lost due to the crash, so we read
garbage instead of the old, shared data.
This has the following implications:
1. In case of writes to shared blocks, with size smaller than the pool's
block size (which means we first copy the whole block and then issue
the smaller write), we corrupt data that the user never touched.
2. In case of writes to shared blocks, with size equal to the device's
logical block size, we fail to provide atomic sector writes. When the
system recovers the user will read garbage from that sector instead
of the old data or the new data.
3. Even for writes to shared blocks, with size equal to the pool's block
size (overwrites), after the system recovers, the written sectors
will contain garbage instead of a random mix of sectors containing
either old data or new data, thus we fail again to provide atomic
sectors writes.
4. Even when the user flushes the thin device, because we first commit
the metadata and then pass down the flush, the same risk for
corruption exists (if the system crashes after the metadata have been
committed but before the flush is passed down to the data device.)
The only case which is unaffected is that of writes with size equal to
the pool's block size and with the FUA flag set. But, because FUA writes
trigger metadata commits, this case can trigger the corruption
indirectly.
Moreover, apart from internal and external snapshots, the same issue
exists for newly provisioned blocks, when block zeroing is enabled.
After the system recovers the provisioned blocks might contain garbage
instead of zeroes.
To solve this and avoid the potential data corruption we flush the
pool's data device **before** committing its metadata.
This ensures that the data blocks of any newly inserted mappings are
properly written to non-volatile storage and won't be lost in case of a
crash.
Cc: stable@vger.kernel.org
Signed-off-by: Nikos Tsironis <ntsironis@arrikto.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Single thread fio test (read, bs=4k, ioengine=libaio, iodepth=128,
numjobs=1) over dm-thin device has poor performance versus bare nvme
device.
Further investigation with perf indicates that queue_work_on() consumes
over 20% CPU time when doing IO over dm-thin device. The call stack is
as follows.
- 40.57% thin_map
+ 22.07% queue_work_on
+ 9.95% dm_thin_find_block
+ 2.80% cell_defer_no_holder
1.91% inc_all_io_entry.isra.33.part.34
+ 1.78% bio_detain.isra.35
In cell_defer_no_holder(), wakeup_worker() is always called, no matter
whether the tc->deferred_bio_list list is empty or not. In single thread
IO model, this list is most likely empty. So skip waking up worker thread
if tc->deferred_bio_list list is empty.
Single thread IO performance improves from 448 MiB/s to 646 MiB/s (+44%)
once the needless wake_worker() calls are properly skipped.
Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
If we are in a place where it is known that interrupts are enabled,
functions spin_lock_irq/spin_unlock_irq should be used instead of
spin_lock_irqsave/spin_unlock_irqrestore.
spin_lock_irq and spin_unlock_irq are faster because they don't need to
push and pop the flags register.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Invoking dm_get_device() twice on the same device path with different
modes is dangerous. Because in that case, upgrade_mode() will alloc a
new 'dm_dev' and free the old one, which may be referenced by a previous
caller. Dereferencing the dangling pointer will trigger kernel NULL
pointer dereference.
The following two cases can reproduce this issue. Actually, they are
invalid setups that must be disallowed, e.g.:
1. Creating a thin-pool with read_only mode, and the same device as
both metadata and data.
dmsetup create thinp --table \
"0 41943040 thin-pool /dev/vdb /dev/vdb 128 0 1 read_only"
BUG: unable to handle kernel NULL pointer dereference at 0000000000000080
...
Call Trace:
new_read+0xfb/0x110 [dm_bufio]
dm_bm_read_lock+0x43/0x190 [dm_persistent_data]
? kmem_cache_alloc_trace+0x15c/0x1e0
__create_persistent_data_objects+0x65/0x3e0 [dm_thin_pool]
dm_pool_metadata_open+0x8c/0xf0 [dm_thin_pool]
pool_ctr.cold.79+0x213/0x913 [dm_thin_pool]
? realloc_argv+0x50/0x70 [dm_mod]
dm_table_add_target+0x14e/0x330 [dm_mod]
table_load+0x122/0x2e0 [dm_mod]
? dev_status+0x40/0x40 [dm_mod]
ctl_ioctl+0x1aa/0x3e0 [dm_mod]
dm_ctl_ioctl+0xa/0x10 [dm_mod]
do_vfs_ioctl+0xa2/0x600
? handle_mm_fault+0xda/0x200
? __do_page_fault+0x26c/0x4f0
ksys_ioctl+0x60/0x90
__x64_sys_ioctl+0x16/0x20
do_syscall_64+0x55/0x150
entry_SYSCALL_64_after_hwframe+0x44/0xa9
2. Creating a external snapshot using the same thin-pool device.
dmsetup create thinp --table \
"0 41943040 thin-pool /dev/vdc /dev/vdb 128 0 2 ignore_discard"
dmsetup message /dev/mapper/thinp 0 "create_thin 0"
dmsetup create snap --table \
"0 204800 thin /dev/mapper/thinp 0 /dev/mapper/thinp"
BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
...
Call Trace:
? __alloc_pages_nodemask+0x13c/0x2e0
retrieve_status+0xa5/0x1f0 [dm_mod]
? dm_get_live_or_inactive_table.isra.7+0x20/0x20 [dm_mod]
table_status+0x61/0xa0 [dm_mod]
ctl_ioctl+0x1aa/0x3e0 [dm_mod]
dm_ctl_ioctl+0xa/0x10 [dm_mod]
do_vfs_ioctl+0xa2/0x600
ksys_ioctl+0x60/0x90
? ksys_write+0x4f/0xb0
__x64_sys_ioctl+0x16/0x20
do_syscall_64+0x55/0x150
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Signed-off-by: Jason Cai (Xiang Feng) <jason.cai@linux.alibaba.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
There is no need to have DM core split discards on behalf of a DM target
now that blk_queue_split() handles splitting discards based on the
queue_limits. A DM target just needs to set max_discard_sectors,
discard_granularity, etc, in queue_limits.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
When provisioning a new data block for a virtual block, either because
the block was previously unallocated or because we are breaking sharing,
if the whole block of data is being overwritten the bio that triggered
the provisioning is issued immediately, skipping copying or zeroing of
the data block.
When this bio completes the new mapping is inserted in to the pool's
metadata by process_prepared_mapping(), where the bio completion is
signaled to the upper layers.
This completion is signaled without first committing the metadata. If
the bio in question has the REQ_FUA flag set and the system crashes
right after its completion and before the next metadata commit, then the
write is lost despite the REQ_FUA flag requiring that I/O completion for
this request must only be signaled after the data has been committed to
non-volatile storage.
Fix this by deferring the completion of overwrite bios, with the REQ_FUA
flag set, until after the metadata has been committed.
Cc: stable@vger.kernel.org
Signed-off-by: Nikos Tsironis <ntsironis@arrikto.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Acked-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Commit 00a0ea33b4 ("dm thin: do not queue freed thin mapping for next
stage processing") changed process_prepared_discard_passdown_pt1() to
increment all the blocks being discarded until after the passdown had
completed to avoid them being prematurely reused.
IO issued to a thin device that breaks sharing with a snapshot, followed
by a discard issued to snapshot(s) that previously shared the block(s),
results in passdown_double_checking_shared_status() being called to
iterate through the blocks double checking their reference count is zero
and issuing the passdown if so. So a side effect of commit 00a0ea33b4
is passdown_double_checking_shared_status() was broken.
Fix this by checking if the block reference count is greater than 1.
Also, rename dm_pool_block_is_used() to dm_pool_block_is_shared().
Fixes: 00a0ea33b4 ("dm thin: do not queue freed thin mapping for next stage processing")
Cc: stable@vger.kernel.org # 4.9+
Reported-by: ryan.p.norwood@gmail.com
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Decoupled version bump from commit f6c367585d ("dm thin: send event
about thin-pool state change _after_ making it") because version bumps
just create conflicts when backporting to the stable trees.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Sending a DM event before a thin-pool state change is about to happen is
a bug. It wasn't realized until it became clear that userspace response
to the event raced with the actual state change that the event was
meant to notify about.
Fix this by first updating internal thin-pool state to reflect what the
DM event is being issued about. This fixes a long-standing racey/buggy
userspace device-mapper-test-suite 'resize_io' test that would get an
event but not find the state it was looking for -- so it would just go
on to hang because no other events caused the test to reevaluate the
thin-pool's state.
Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
The API surrounding refcount_t should be used in place of atomic_t
when variables are being used as reference counters. It can
potentially prevent reference counter overflows and use-after-free
conditions. In the dm thin layer, one such example is tc->refcount.
Change this from the atomic_t API to the refcount_t API to prevent
mentioned conditions.
Signed-off-by: John Pittman <jpittman@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Committing a transaction can consume some metadata of it's own, we now
reserve a small amount of metadata to cover this. Free metadata
reported by the kernel will not include this reserve.
If any of the reserve has been used after a commit we enter a new
internal state PM_OUT_OF_METADATA_SPACE. This is reported as
PM_READ_ONLY, so no userland changes are needed. If the metadata
device is resized the pool will move back to PM_WRITE.
These changes mean we never need to abort and rollback a transaction due
to running out of metadata space. This is particularly important
because there have been a handful of reports of data corruption against
DM thin-provisioning that can all be attributed to the thin-pool having
ran out of metadata space.
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Now both check_for_space() and do_no_space_timeout() will read & write
pool->pf.error_if_no_space. If these functions run concurrently, as
shown in the following case, the default setting of "queue_if_no_space"
can get lost.
precondition:
* error_if_no_space = false (aka "queue_if_no_space")
* pool is in Out-of-Data-Space (OODS) mode
* no_space_timeout worker has been queued
CPU 0: CPU 1:
// delete a thin device
process_delete_mesg()
// check_for_space() invoked by commit()
set_pool_mode(pool, PM_WRITE)
pool->pf.error_if_no_space = \
pt->requested_pf.error_if_no_space
// timeout, pool is still in OODS mode
do_no_space_timeout
// "queue_if_no_space" config is lost
pool->pf.error_if_no_space = true
pool->pf.mode = new_mode
Fix it by stopping no_space_timeout worker when switching to write mode.
Fixes: bcc696fac1 ("dm thin: stay in out-of-data-space mode once no_space_timeout expires")
Cc: stable@vger.kernel.org
Signed-off-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
dm_kcopyd_copy() only ever returns 0 so there is no need for callers to
account for possible failure. Same goes for dm_kcopyd_zero().
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
The metadata low watermark threshold is set by the kernel. But the
kernel depends on userspace to extend the thinpool metadata device when
the threshold is crossed.
Since the metadata low watermark threshold is not visible to userspace,
upon receiving an event, userspace cannot tell that the kernel wants the
metadata device extended, instead of some other eventing condition.
Making it visible (but not settable) enables userspace to affirmatively
know the kernel is asking for a metadata device extension, by comparing
metadata_low_watermark against nr_free_blocks_metadata, also reported in
status.
Current solutions like dmeventd have their own thresholds for extending
the data and metadata devices, and both devices are checked against
their thresholds on each event. This lessens the value of the kernel-set
threshold, since userspace will either extend the metadata device sooner,
when receiving another event; or will receive the metadata lowater event
and do nothing, if dmeventd's threshold is less than the kernel's.
(This second case is dangerous. The metadata lowater event will not be
re-sent, so no further event will be generated before the metadata
device is out if space, unless some other event causes userspace to
recheck its thresholds.)
Signed-off-by: Andy Grover <agrover@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Discards issued to a DM thin device can complete to userspace (via
fstrim) _before_ the metadata changes associated with the discards is
reflected in the thinp superblock (e.g. free blocks). As such, if a
user constructs a test that loops repeatedly over these steps, block
allocation can fail due to discards not having completed yet:
1) fill thin device via filesystem file
2) remove file
3) fstrim
From initial report, here:
https://www.redhat.com/archives/dm-devel/2018-April/msg00022.html
"The root cause of this issue is that dm-thin will first remove
mapping and increase corresponding blocks' reference count to prevent
them from being reused before DISCARD bios get processed by the
underlying layers. However. increasing blocks' reference count could
also increase the nr_allocated_this_transaction in struct sm_disk
which makes smd->old_ll.nr_allocated +
smd->nr_allocated_this_transaction bigger than smd->old_ll.nr_blocks.
In this case, alloc_data_block() will never commit metadata to reset
the begin pointer of struct sm_disk, because sm_disk_get_nr_free()
always return an underflow value."
While there is room for improvement to the space-map accounting that
thinp is making use of: the reality is this test is inherently racey and
will result in the previous iteration's fstrim's discard(s) completing
vs concurrent block allocation, via dd, in the next iteration of the
loop.
No amount of space map accounting improvements will be able to allow
user's to use a block before a discard of that block has completed.
So the best we can really do is allow DM thinp to gracefully handle such
aggressive use of all the pool's data by degrading the pool into
out-of-data-space (OODS) mode. We _should_ get that behaviour already
(if space map accounting didn't falsely cause alloc_data_block() to
believe free space was available).. but short of that we handle the
current reality that dm_pool_alloc_data_block() can return -ENOSPC.
Reported-by: Dennis Yang <dennisyang@qnap.com>
Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
- Additional struct_size() conversions (Matthew, Kees)
- Explicitly reported overflow fixes (Silvio, Kees)
- Add missing kvcalloc() function (Kees)
- Treewide conversions of allocators to use either 2-factor argument
variant when available, or array_size() and array3_size() as needed (Kees)
-----BEGIN PGP SIGNATURE-----
Comment: Kees Cook <kees@outflux.net>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=zUUp
-----END PGP SIGNATURE-----
Merge tag 'overflow-v4.18-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux
Pull more overflow updates from Kees Cook:
"The rest of the overflow changes for v4.18-rc1.
This includes the explicit overflow fixes from Silvio, further
struct_size() conversions from Matthew, and a bug fix from Dan.
But the bulk of it is the treewide conversions to use either the
2-factor argument allocators (e.g. kmalloc(a * b, ...) into
kmalloc_array(a, b, ...) or the array_size() macros (e.g. vmalloc(a *
b) into vmalloc(array_size(a, b)).
Coccinelle was fighting me on several fronts, so I've done a bunch of
manual whitespace updates in the patches as well.
Summary:
- Error path bug fix for overflow tests (Dan)
- Additional struct_size() conversions (Matthew, Kees)
- Explicitly reported overflow fixes (Silvio, Kees)
- Add missing kvcalloc() function (Kees)
- Treewide conversions of allocators to use either 2-factor argument
variant when available, or array_size() and array3_size() as needed
(Kees)"
* tag 'overflow-v4.18-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux: (26 commits)
treewide: Use array_size in f2fs_kvzalloc()
treewide: Use array_size() in f2fs_kzalloc()
treewide: Use array_size() in f2fs_kmalloc()
treewide: Use array_size() in sock_kmalloc()
treewide: Use array_size() in kvzalloc_node()
treewide: Use array_size() in vzalloc_node()
treewide: Use array_size() in vzalloc()
treewide: Use array_size() in vmalloc()
treewide: devm_kzalloc() -> devm_kcalloc()
treewide: devm_kmalloc() -> devm_kmalloc_array()
treewide: kvzalloc() -> kvcalloc()
treewide: kvmalloc() -> kvmalloc_array()
treewide: kzalloc_node() -> kcalloc_node()
treewide: kzalloc() -> kcalloc()
treewide: kmalloc() -> kmalloc_array()
mm: Introduce kvcalloc()
video: uvesafb: Fix integer overflow in allocation
UBIFS: Fix potential integer overflow in allocation
leds: Use struct_size() in allocation
Convert intel uncore to struct_size
...
Eliminate most holes in DM data structures that were modified by
commit 6f1c819c21 ("dm: convert to bioset_init()/mempool_init()").
Also prevent structure members from unnecessarily spanning cache
lines.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
mempool_init()/bioset_init() require that the mempools/biosets be zeroed
first; they probably should not _require_ this, but not allocating those
structs with kzalloc is a fairly nonsensical thing to do (calling
mempool_exit()/bioset_exit() on an uninitialized mempool/bioset is legal
and safe, but only works if said memory was zeroed.)
Acked-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Convert dm to embedded bio sets.
Acked-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Could be useful for a target to return stats or other information.
If a target does DMEMIT() anything to @result from its .message method
then it must return 1 to the caller.
Signed-off-By: Mike Snitzer <snitzer@redhat.com>
The trailing semicolon is an empty statement that does no operation.
Removing it since it doesn't do anything.
Signed-off-by: Luis de Bethencourt <luisbg@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
A NULL pointer is seen if two concurrent "vgchange -ay -K <vg name>"
processes race to load the dm-thin-pool module:
PID: 25992 TASK: ffff883cd7d23500 CPU: 4 COMMAND: "vgchange"
#0 [ffff883cd743d600] machine_kexec at ffffffff81038fa9
0000001 [ffff883cd743d660] crash_kexec at ffffffff810c5992
0000002 [ffff883cd743d730] oops_end at ffffffff81515c90
0000003 [ffff883cd743d760] no_context at ffffffff81049f1b
0000004 [ffff883cd743d7b0] __bad_area_nosemaphore at ffffffff8104a1a5
0000005 [ffff883cd743d800] bad_area at ffffffff8104a2ce
0000006 [ffff883cd743d830] __do_page_fault at ffffffff8104aa6f
0000007 [ffff883cd743d950] do_page_fault at ffffffff81517bae
0000008 [ffff883cd743d980] page_fault at ffffffff81514f95
[exception RIP: kmem_cache_alloc+108]
RIP: ffffffff8116ef3c RSP: ffff883cd743da38 RFLAGS: 00010046
RAX: 0000000000000004 RBX: ffffffff81121b90 RCX: ffff881bf1e78cc0
RDX: 0000000000000000 RSI: 00000000000000d0 RDI: 0000000000000000
RBP: ffff883cd743da68 R8: ffff881bf1a4eb00 R9: 0000000080042000
R10: 0000000000002000 R11: 0000000000000000 R12: 00000000000000d0
R13: 0000000000000000 R14: 00000000000000d0 R15: 0000000000000246
ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
0000009 [ffff883cd743da70] mempool_alloc_slab at ffffffff81121ba5
0000010 [ffff883cd743da80] mempool_create_node at ffffffff81122083
0000011 [ffff883cd743dad0] mempool_create at ffffffff811220f4
0000012 [ffff883cd743dae0] pool_ctr at ffffffffa08de049 [dm_thin_pool]
0000013 [ffff883cd743dbd0] dm_table_add_target at ffffffffa0005f2f [dm_mod]
0000014 [ffff883cd743dc30] table_load at ffffffffa0008ba9 [dm_mod]
0000015 [ffff883cd743dc90] ctl_ioctl at ffffffffa0009dc4 [dm_mod]
The race results in a NULL pointer because:
Process A (vgchange -ay -K):
a. send DM_LIST_VERSIONS_CMD ioctl;
b. pool_target not registered;
c. modprobe dm_thin_pool and wait until end.
Process B (vgchange -ay -K):
a. send DM_LIST_VERSIONS_CMD ioctl;
b. pool_target registered;
c. table_load->dm_table_add_target->pool_ctr;
d. _new_mapping_cache is NULL and panic.
Note:
1. process A and process B are two concurrent processes.
2. pool_target can be detected by process B but
_new_mapping_cache initialization has not ended.
To fix dm-thin-pool, and other targets (cache, multipath, and snapshot)
with the same problem, simply dm_register_target() after all resources
created during module init (as labelled with __init) are finished.
Cc: stable@vger.kernel.org
Signed-off-by: monty <monty_pavel@sina.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Please do not apply this to mainline directly, instead please re-run the
coccinelle script shown below and apply its output.
For several reasons, it is desirable to use {READ,WRITE}_ONCE() in
preference to ACCESS_ONCE(), and new code is expected to use one of the
former. So far, there's been no reason to change most existing uses of
ACCESS_ONCE(), as these aren't harmful, and changing them results in
churn.
However, for some features, the read/write distinction is critical to
correct operation. To distinguish these cases, separate read/write
accessors must be used. This patch migrates (most) remaining
ACCESS_ONCE() instances to {READ,WRITE}_ONCE(), using the following
coccinelle script:
----
// Convert trivial ACCESS_ONCE() uses to equivalent READ_ONCE() and
// WRITE_ONCE()
// $ make coccicheck COCCI=/home/mark/once.cocci SPFLAGS="--include-headers" MODE=patch
virtual patch
@ depends on patch @
expression E1, E2;
@@
- ACCESS_ONCE(E1) = E2
+ WRITE_ONCE(E1, E2)
@ depends on patch @
expression E;
@@
- ACCESS_ONCE(E)
+ READ_ONCE(E)
----
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: davem@davemloft.net
Cc: linux-arch@vger.kernel.org
Cc: mpe@ellerman.id.au
Cc: shuah@kernel.org
Cc: snitzer@redhat.com
Cc: thor.thayer@linux.intel.com
Cc: tj@kernel.org
Cc: viro@zeniv.linux.org.uk
Cc: will.deacon@arm.com
Link: http://lkml.kernel.org/r/1508792849-3115-19-git-send-email-paulmck@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
- Constify a few variables in DM core and DM integrity
- Add bufio optimization and checksum failure accounting to DM integrity
- Fix DM integrity to avoid checking integrity of failed reads
- Fix DM integrity to use init_completion
- A couple DM log-writes target fixes
- Simplify DAX flushing by eliminating the unnecessary flush abstraction
that was stood up for DM's use.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJZuo8UAAoJEMUj8QotnQNa5BEIANO4mHh1nrzEbH72a4RCLgxV
H1Pk1zZx/W1bhOOmcRRhxCSM85dPgsCegc5EmpwLZEMavQrP9UZblHcYOUsyIx7W
S/lWa+soOq/5N2OveROc4WdoWVs50UFmc1+BcClc4YrEe+15XC3R0VMkjX2b/hUL
o2eYhPjpMlgaorMtRRU6MAooo2fBRQ9m05aPeVgd35fxibrE7PZm+EYW09wa0STi
9ufuDXJf8+TtFP/38BD41LbUEskuHUZTSDeAJ+3DBaTtfEZcZYxsst4P9JangsHx
jqqqI9aYzFD2a27fl9WLhCvm40YFiKp5nwzED0RZjzWxVa/jTShX7a49BdzTTfw=
=rkSB
-----END PGP SIGNATURE-----
Merge tag 'for-4.14/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer:
- Some request-based DM core and DM multipath fixes and cleanups
- Constify a few variables in DM core and DM integrity
- Add bufio optimization and checksum failure accounting to DM
integrity
- Fix DM integrity to avoid checking integrity of failed reads
- Fix DM integrity to use init_completion
- A couple DM log-writes target fixes
- Simplify DAX flushing by eliminating the unnecessary flush
abstraction that was stood up for DM's use.
* tag 'for-4.14/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
dax: remove the pmem_dax_ops->flush abstraction
dm integrity: use init_completion instead of COMPLETION_INITIALIZER_ONSTACK
dm integrity: make blk_integrity_profile structure const
dm integrity: do not check integrity for failed read operations
dm log writes: fix >512b sectorsize support
dm log writes: don't use all the cpu while waiting to log blocks
dm ioctl: constify ioctl lookup table
dm: constify argument arrays
dm integrity: count and display checksum failures
dm integrity: optimize writing dm-bufio buffers that are partially changed
dm rq: do not update rq partially in each ending bio
dm rq: make dm-sq requeuing behavior consistent with dm-mq behavior
dm mpath: complain about unsupported __multipath_map_bio() return values
dm mpath: avoid that building with W=1 causes gcc 7 to complain about fall-through
The arrays of 'struct dm_arg' are never modified by the device-mapper
core, so constify them so that they are placed in .rodata.
(Exception: the args array in dm-raid cannot be constified because it is
allocated on the stack and modified.)
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
This way we don't need a block_device structure to submit I/O. The
block_device has different life time rules from the gendisk and
request_queue and is usually only available when the block device node
is open. Other callers need to explicitly create one (e.g. the lightnvm
passthrough code, or the new nvme multipathing code).
For the actual I/O path all that we need is the gendisk, which exists
once per block device. But given that the block layer also does
partition remapping we additionally need a partition index, which is
used for said remapping in generic_make_request.
Note that all the block drivers generally want request_queue or
sometimes the gendisk, so this removes a layer of indirection all
over the stack.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull core block/IO updates from Jens Axboe:
"This is the main pull request for the block layer for 4.13. Not a huge
round in terms of features, but there's a lot of churn related to some
core cleanups.
Note this depends on the UUID tree pull request, that Christoph
already sent out.
This pull request contains:
- A series from Christoph, unifying the error/stats codes in the
block layer. We now use blk_status_t everywhere, instead of using
different schemes for different places.
- Also from Christoph, some cleanups around request allocation and IO
scheduler interactions in blk-mq.
- And yet another series from Christoph, cleaning up how we handle
and do bounce buffering in the block layer.
- A blk-mq debugfs series from Bart, further improving on the support
we have for exporting internal information to aid debugging IO
hangs or stalls.
- Also from Bart, a series that cleans up the request initialization
differences across types of devices.
- A series from Goldwyn Rodrigues, allowing the block layer to return
failure if we will block and the user asked for non-blocking.
- Patch from Hannes for supporting setting loop devices block size to
that of the underlying device.
- Two series of patches from Javier, fixing various issues with
lightnvm, particular around pblk.
- A series from me, adding support for write hints. This comes with
NVMe support as well, so applications can help guide data placement
on flash to improve performance, latencies, and write
amplification.
- A series from Ming, improving and hardening blk-mq support for
stopping/starting and quiescing hardware queues.
- Two pull requests for NVMe updates. Nothing major on the feature
side, but lots of cleanups and bug fixes. From the usual crew.
- A series from Neil Brown, greatly improving the bio rescue set
support. Most notably, this kills the bio rescue work queues, if we
don't really need them.
- Lots of other little bug fixes that are all over the place"
* 'for-4.13/block' of git://git.kernel.dk/linux-block: (217 commits)
lightnvm: pblk: set line bitmap check under debug
lightnvm: pblk: verify that cache read is still valid
lightnvm: pblk: add initialization check
lightnvm: pblk: remove target using async. I/Os
lightnvm: pblk: use vmalloc for GC data buffer
lightnvm: pblk: use right metadata buffer for recovery
lightnvm: pblk: schedule if data is not ready
lightnvm: pblk: remove unused return variable
lightnvm: pblk: fix double-free on pblk init
lightnvm: pblk: fix bad le64 assignations
nvme: Makefile: remove dead build rule
blk-mq: map all HWQ also in hyperthreaded system
nvmet-rdma: register ib_client to not deadlock in device removal
nvme_fc: fix error recovery on link down.
nvmet_fc: fix crashes on bad opcodes
nvme_fc: Fix crash when nvme controller connection fails.
nvme_fc: replace ioabort msleep loop with completion
nvme_fc: fix double calls to nvme_cleanup_cmd()
nvme-fabrics: verify that a controller returns the correct NQN
nvme: simplify nvme_dev_attrs_are_visible
...
process_prepared_discard_passdown_pt1() should cleanup
dm_thin_new_mapping in cases of error.
dm_pool_inc_data_range() can fail trying to get a block reference:
metadata operation 'dm_pool_inc_data_range' failed: error = -61
When dm_pool_inc_data_range() fails, dm thin aborts current metadata
transaction and marks pool as PM_READ_ONLY. Memory for thin mapping
is released as well. However, current thin mapping will be queued
onto next stage as part of queue_passdown_pt2() or passdown_endio().
This dangling thin mapping memory when processed and accessed in
next stage will lead to device mapper crashing.
Code flow without fix:
-> process_prepared_discard_passdown_pt1(m)
-> dm_thin_remove_range()
-> discard passdown
--> passdown_endio(m) queues m onto next stage
-> dm_pool_inc_data_range() fails, frees memory m
but does not remove it from next stage queue
-> process_prepared_discard_passdown_pt2(m)
-> processes freed memory m and crashes
One such stack:
Call Trace:
[<ffffffffa037a46f>] dm_cell_release_no_holder+0x2f/0x70 [dm_bio_prison]
[<ffffffffa039b6dc>] cell_defer_no_holder+0x3c/0x80 [dm_thin_pool]
[<ffffffffa039b88b>] process_prepared_discard_passdown_pt2+0x4b/0x90 [dm_thin_pool]
[<ffffffffa0399611>] process_prepared+0x81/0xa0 [dm_thin_pool]
[<ffffffffa039e735>] do_worker+0xc5/0x820 [dm_thin_pool]
[<ffffffff8152bf54>] ? __schedule+0x244/0x680
[<ffffffff81087e72>] ? pwq_activate_delayed_work+0x42/0xb0
[<ffffffff81089f53>] process_one_work+0x153/0x3f0
[<ffffffff8108a71b>] worker_thread+0x12b/0x4b0
[<ffffffff8108a5f0>] ? rescuer_thread+0x350/0x350
[<ffffffff8108fd6a>] kthread+0xca/0xe0
[<ffffffff8108fca0>] ? kthread_park+0x60/0x60
[<ffffffff81530b45>] ret_from_fork+0x25/0x30
The fix is to first take the block ref count for discarded block and
then do a passdown discard of this block. If block ref count fails,
then bail out aborting current metadata transaction, mark pool as
PM_READ_ONLY and also free current thin mapping memory (existing error
handling code) without queueing this thin mapping onto next stage of
processing. If block ref count succeeds, then passdown discard of this
block. Discard callback of passdown_endio() will queue this thin mapping
onto next stage of processing.
Code flow with fix:
-> process_prepared_discard_passdown_pt1(m)
-> dm_thin_remove_range()
-> dm_pool_inc_data_range()
--> if fails, free memory m and bail out
-> discard passdown
--> passdown_endio(m) queues m onto next stage
Cc: stable <stable@vger.kernel.org> # v4.9+
Reviewed-by: Eduardo Valentin <eduval@amazon.com>
Reviewed-by: Cristian Gafton <gafton@amazon.com>
Reviewed-by: Anchal Agarwal <anchalag@amazon.com>
Signed-off-by: Vallish Vaidyeshwara <vallish@amazon.com>
Reviewed-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Replace bi_error with a new bi_status to allow for a clear conversion.
Note that device mapper overloaded bi_error with a private value, which
we'll have to keep arround at least for now and thus propagate to a
proper blk_status_t value.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
Turn the error paramter into a pointer so that target drivers can change
the value, and make sure only DM_ENDIO_* values are returned from the
methods.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
whether blocks should migrate to/from the cache. The bio-prison-v2
interface supports this improvement by enabling direct dispatch of
work to workqueues rather than having to delay the actual work
dispatch to the DM cache core. So the dm-cache policies are much more
nimble by being able to drive IO as they see fit. One immediate
benefit from the improved latency is a cache that should be much more
adaptive to changing workloads.
- Add a new DM integrity target that emulates a block device that has
additional per-sector tags that can be used for storing integrity
information.
- Add a new authenticated encryption feature to the DM crypt target that
builds on the capabilities provided by the DM integrity target.
- Add MD interface for switching the raid4/5/6 journal mode and update
the DM raid target to use it to enable aid4/5/6 journal write-back
support.
- Switch the DM verity target over to using the asynchronous hash crypto
API (this helps work better with architectures that have access to
off-CPU algorithm providers, which should reduce CPU utilization).
- Various request-based DM and DM multipath fixes and improvements from
Bart and Christoph.
- A DM thinp target fix for a bio structure leak that occurs for each
discard IFF discard passdown is enabled.
- A fix for a possible deadlock in DM bufio and a fix to re-check the
new buffer allocation watermark in the face of competing admin changes
to the 'max_cache_size_bytes' tunable.
- A couple DM core cleanups.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJZB6vtAAoJEMUj8QotnQNaoicIALuZTLElgAzxzA28cfk1+1Ea
Gd09CfJ3M6cvk/YGUU7WwiSYIwu16yOJALG4sLcYnEmUCzvKfFPcl/RpeSJHPpYM
0aVXa6NIJw7K2r3C17toiK2DRMHYw6QU843WeWI93vBW13lDJklNJL9fM7GBEOLH
NMSNw2mAq9ajtLlnJhM3ZfhloA7/u/jektvlBO1AA3RQ5Kx1cXVXFPqN7FdRfcqp
4RuEMe9faAadlXLsj3bia5IBmF/W0Qza6JilP+NLKLWB4fm7LZDjN/k+TsHWMa9e
cGR73TgUGLMBJX+sDJy8R3oeBG9JZkFVkD7I30eCjzyhSOs/54XNYQ23EkqHJU0=
=9Ryi
-----END PGP SIGNATURE-----
Merge tag 'for-4.12/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer:
- A major update for DM cache that reduces the latency for deciding
whether blocks should migrate to/from the cache. The bio-prison-v2
interface supports this improvement by enabling direct dispatch of
work to workqueues rather than having to delay the actual work
dispatch to the DM cache core. So the dm-cache policies are much more
nimble by being able to drive IO as they see fit. One immediate
benefit from the improved latency is a cache that should be much more
adaptive to changing workloads.
- Add a new DM integrity target that emulates a block device that has
additional per-sector tags that can be used for storing integrity
information.
- Add a new authenticated encryption feature to the DM crypt target
that builds on the capabilities provided by the DM integrity target.
- Add MD interface for switching the raid4/5/6 journal mode and update
the DM raid target to use it to enable aid4/5/6 journal write-back
support.
- Switch the DM verity target over to using the asynchronous hash
crypto API (this helps work better with architectures that have
access to off-CPU algorithm providers, which should reduce CPU
utilization).
- Various request-based DM and DM multipath fixes and improvements from
Bart and Christoph.
- A DM thinp target fix for a bio structure leak that occurs for each
discard IFF discard passdown is enabled.
- A fix for a possible deadlock in DM bufio and a fix to re-check the
new buffer allocation watermark in the face of competing admin
changes to the 'max_cache_size_bytes' tunable.
- A couple DM core cleanups.
* tag 'for-4.12/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (50 commits)
dm bufio: check new buffer allocation watermark every 30 seconds
dm bufio: avoid a possible ABBA deadlock
dm mpath: make it easier to detect unintended I/O request flushes
dm mpath: cleanup QUEUE_IF_NO_PATH bit manipulation by introducing assign_bit()
dm mpath: micro-optimize the hot path relative to MPATHF_QUEUE_IF_NO_PATH
dm: introduce enum dm_queue_mode to cleanup related code
dm mpath: verify __pg_init_all_paths locking assumptions at runtime
dm: verify suspend_locking assumptions at runtime
dm block manager: remove an unused argument from dm_block_manager_create()
dm rq: check blk_mq_register_dev() return value in dm_mq_init_request_queue()
dm mpath: delay requeuing while path initialization is in progress
dm mpath: avoid that path removal can trigger an infinite loop
dm mpath: split and rename activate_path() to prepare for its expanded use
dm ioctl: prevent stack leak in dm ioctl call
dm integrity: use previously calculated log2 of sectors_per_block
dm integrity: use hex2bin instead of open-coded variant
dm crypt: replace custom implementation of hex2bin()
dm crypt: remove obsolete references to per-CPU state
dm verity: switch to using asynchronous hash crypto API
dm crypt: use WQ_HIGHPRI for the IO and crypt workqueues
...
Now that we use the proper REQ_OP_WRITE_ZEROES operation everywhere we can
kill this hack.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
The deferred set is gone and all methods have _v2 appended to the end of
their names to allow for continued use of the original bio prison in DM
thin-provisioning.
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
We will want to have struct backing_dev_info allocated separately from
struct request_queue. As the first step add pointer to backing_dev_info
to request_queue and convert all users touching it. No functional
changes in this patch.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@fb.com>
This centralizes the checks for bios that needs to be go into the flush
state machine.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Since commit 63a4cc2486, bio->bi_rw contains flags in the lower
portion and the op code in the higher portions. This means that
old code that relies on manually setting bi_rw is most likely
going to be broken. Instead of letting that brokeness linger,
rename the member, to force old and out-of-tree code to break
at compile time instead of at runtime.
No intended functional changes in this commit.
Signed-off-by: Jens Axboe <axboe@fb.com>
The discard passdown was being issued after the block was unmapped,
which meant the block could be reprovisioned whilst the passdown discard
was still in flight.
We can only identify unshared blocks (safe to do a passdown a discard
to) once they're unmapped and their ref count hits zero. Block ref
counts are now used to guard against concurrent allocation of these
blocks that are being discarded. So now we unmap the block, issue
passdown discards, and the immediately increment ref counts for regions
that have been discarded via passed down (this is safe because
allocation occurs within the same thread). We then decrement ref counts
once the passdown discard IO is complete -- signaling these blocks may
now be allocated.
This fixes the potential for corruption that was reported here:
https://www.redhat.com/archives/dm-devel/2016-June/msg00311.html
Reported-by: Dennis Yang <dennisyang@qnap.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
To avoid confusion between REQ_OP_FLUSH, which is handled by
request_fn drivers, and upper layers requesting the block layer
perform a flush sequence along with possibly a WRITE, this patch
renames REQ_FLUSH to REQ_PREFLUSH.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Separate the op from the rq_flag_bits and have dm
set/get the bio using bio_set_op_attrs/bio_op.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This converts the block issue discard helper and users to use
the bio_set_op_attrs accessor and only pass in the operation flags
like REQ_SEQURE.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This has callers of submit_bio/submit_bio_wait set the bio->bi_rw
instead of passing it in. This makes that use the same as
generic_make_request and how we set the other bio fields.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Fixed up fs/ext4/crypto.c
Signed-off-by: Jens Axboe <axboe@fb.com>
There is little benefit to doing this but it does structure DM thinp's
code to more cleanly use the __blkdev_issue_discard() interface --
particularly in passdown_double_checking_shared_status().
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
With commit 38f2525533 ("block: add __blkdev_issue_discard") DM thinp
no longer needs to carry its own async discard method.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
