linux-2.6-microblaze.git
21 months agodm bufio: fix some cases where the code sleeps with spinlock held
Mikulas Patocka [Wed, 10 Aug 2022 20:16:34 +0000 (16:16 -0400)]
dm bufio: fix some cases where the code sleeps with spinlock held

Commit b32d45824aa7 ("dm bufio: Add DM_BUFIO_CLIENT_NO_SLEEP flag")
added a "NO_SLEEP" mode, it replaces a mutex with a spinlock, and it
is only usable when the device is in read-only mode (because the write
path may be sleeping while holding the dm_bufio_client lock).

However, there are still two points where the code could sleep even in
read-only mode. One is in __get_unclaimed_buffer -> __make_buffer_clean.
The other is in __try_evict_buffer -> __make_buffer_clean. These functions
will call __make_buffer_clean which sleeps if the buffer is being read.

Fix these cases so that if c->no_sleep is set __make_buffer_clean
will not be called and the buffer will be skipped instead.

Fixes: b32d45824aa7 ("dm bufio: Add DM_BUFIO_CLIENT_NO_SLEEP flag")
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm writecache: fix smatch warning about invalid return from writecache_map
Mikulas Patocka [Mon, 8 Aug 2022 14:50:10 +0000 (10:50 -0400)]
dm writecache: fix smatch warning about invalid return from writecache_map

There's a smatch warning "inconsistent returns '&wc->lock'" in
dm-writecache. The reason for the warning is that writecache_map()
doesn't drop the lock on the impossible path.

Fix this warning by adding wc_unlock() after the BUG statement (so
that it will be compiled-away anyway).

Fixes: df699cc16ea5e ("dm writecache: report invalid return from writecache_map helpers")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: fix verity_parse_opt_args parsing
Mike Snitzer [Tue, 9 Aug 2022 22:07:28 +0000 (18:07 -0400)]
dm verity: fix verity_parse_opt_args parsing

Commit df326e7a0699 ("dm verity: allow optional args to alter primary
args handling") introduced a bug where verity_parse_opt_args() wouldn't
properly shift past an optional argument's additional params (by
ignoring them).

Fix this by avoiding returning with error if an unknown argument is
encountered when @only_modifier_opts=true is passed to
verity_parse_opt_args().

In practice this regressed the cryptsetup testsuite's FEC testing
because unknown optional arguments were encountered, wherey
short-circuiting ever testing FEC mode. With this fix all of the
cryptsetup testsuite's verity FEC tests pass.

Fixes: df326e7a0699 ("dm verity: allow optional args to alter primary args handling")
Reported-by: Milan Broz <gmazyland@gmail.com>>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: fix DM_VERITY_OPTS_MAX value yet again
Mike Snitzer [Tue, 9 Aug 2022 21:33:12 +0000 (17:33 -0400)]
dm verity: fix DM_VERITY_OPTS_MAX value yet again

Must account for the possibility that "try_verify_in_tasklet" is used.

This is the same issue that was fixed with commit 160f99db94322 -- it
is far too easy to miss that additional a new argument(s) require
bumping DM_VERITY_OPTS_MAX accordingly.

Fixes: 5721d4e5a9cd ("dm verity: Add optional "try_verify_in_tasklet" feature")
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm bufio: simplify DM_BUFIO_CLIENT_NO_SLEEP locking
Mike Snitzer [Sat, 6 Aug 2022 02:58:33 +0000 (22:58 -0400)]
dm bufio: simplify DM_BUFIO_CLIENT_NO_SLEEP locking

Historically none of the bufio code runs in interrupt context but with
the use of DM_BUFIO_CLIENT_NO_SLEEP a bufio client can, see: commit
5721d4e5a9cd ("dm verity: Add optional "try_verify_in_tasklet" feature")
That said, the new tasklet usecase still doesn't require interrupts be
disabled by bufio (let alone conditionally restore them).

Yet with PREEMPT_RT, and falling back from tasklet to workqueue, care
must be taken to properly synchronize between softirq and process
context, otherwise ABBA deadlock may occur. While it is unnecessary to
disable bottom-half preemption within a tasklet, we must consistently do
so in process context to ensure locking is in the proper order.

Fix these issues by switching from spin_lock_irq{save,restore} to using
spin_{lock,unlock}_bh instead. Also remove the 'spinlock_flags' member
in dm_bufio_client struct (that can be used unsafely if bufio must
recurse on behalf of some caller, e.g. block layer's submit_bio).

Fixes: 5721d4e5a9cd ("dm verity: Add optional "try_verify_in_tasklet" feature")
Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
21 months agodm verity: have verify_wq use WQ_HIGHPRI if "try_verify_in_tasklet"
Mike Snitzer [Thu, 4 Aug 2022 19:55:57 +0000 (15:55 -0400)]
dm verity: have verify_wq use WQ_HIGHPRI if "try_verify_in_tasklet"

Allow verify_wq to preempt softirq since verification in tasklet will
fall-back to using it for error handling (or if the bufio cache
doesn't have required hashes).

Suggested-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: remove WQ_CPU_INTENSIVE flag since using WQ_UNBOUND
Mike Snitzer [Thu, 4 Aug 2022 18:53:07 +0000 (14:53 -0400)]
dm verity: remove WQ_CPU_INTENSIVE flag since using WQ_UNBOUND

The documentation [1] says that WQ_CPU_INTENSIVE is "meaningless" for
unbound wq. So remove WQ_CPU_INTENSIVE from the verify_wq allocation.

1. https://www.kernel.org/doc/html/latest/core-api/workqueue.html#flags

Suggested-by: Maksym Planeta <mplaneta@os.inf.tu-dresden.de>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: only copy bvec_iter in verity_verify_io if in_tasklet
Mike Snitzer [Thu, 4 Aug 2022 17:37:53 +0000 (13:37 -0400)]
dm verity: only copy bvec_iter in verity_verify_io if in_tasklet

Avoid extra bvec_iter copy unless it is needed to allow retrying
verification, that failed from a tasklet, from a workqueue.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: optimize verity_verify_io if FEC not configured
Mike Snitzer [Thu, 4 Aug 2022 02:43:22 +0000 (22:43 -0400)]
dm verity: optimize verity_verify_io if FEC not configured

Only declare and copy bvec_iter if CONFIG_DM_VERITY_FEC is defined and
FEC enabled for the verity device.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: conditionally enable branching for "try_verify_in_tasklet"
Mike Snitzer [Mon, 25 Jul 2022 20:52:17 +0000 (16:52 -0400)]
dm verity: conditionally enable branching for "try_verify_in_tasklet"

Use jump_label to limit the need for branching unless the optional
"try_verify_in_tasklet" feature is used.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm bufio: conditionally enable branching for DM_BUFIO_CLIENT_NO_SLEEP
Mike Snitzer [Mon, 25 Jul 2022 22:26:48 +0000 (18:26 -0400)]
dm bufio: conditionally enable branching for DM_BUFIO_CLIENT_NO_SLEEP

Use jump_label to limit the need for branching unless the optional
DM_BUFIO_CLIENT_NO_SLEEP is used.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: allow optional args to alter primary args handling
Mike Snitzer [Tue, 26 Jul 2022 15:29:50 +0000 (11:29 -0400)]
dm verity: allow optional args to alter primary args handling

The previous commit ("dm verity: Add optional "try_verify_in_tasklet"
feature") imposed that CRYPTO_ALG_ASYNC mask be used even if the
optional "try_verify_in_tasklet" feature was not specified. This was
because verity_parse_opt_args() was called after handling the primary
args (due to it having data dependencies on having first parsed all
primary args).

Enhance verity_ctr() so that simple optional args, that don't have a
data dependency on primary args parsing, can alter how the primary
args are handled. In practice this means verity_parse_opt_args() gets
called twice. First with the new 'only_modifier_opts' arg set to true,
then again with it set to false _after_ parsing all primary args.

This allows the v->use_tasklet flag to be properly set and then used
when verity_ctr() parses the primary args and then calls
crypto_alloc_ahash() with CRYPTO_ALG_ASYNC conditionally set.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
21 months agodm verity: Add optional "try_verify_in_tasklet" feature
Nathan Huckleberry [Fri, 22 Jul 2022 09:38:23 +0000 (09:38 +0000)]
dm verity: Add optional "try_verify_in_tasklet" feature

Using tasklets for disk verification can reduce IO latency. When there
are accelerated hash instructions it is often better to compute the
hash immediately using a tasklet rather than deferring verification to
a work-queue. This reduces time spent waiting to schedule work-queue
jobs, but requires spending slightly more time in interrupt context.

If the dm-bufio cache does not have the required hashes we fallback to
the work-queue implementation. FEC is only possible using work-queue
because code to support the FEC feature may sleep.

The following shows a speed comparison of random reads on a dm-verity
device. The dm-verity device uses a 1G ramdisk for data and a 1G
ramdisk for hashes. One test was run using tasklets and one test was
run using the existing work-queue solution. Both tests were run when
the dm-bufio cache was hot. The tasklet implementation performs
significantly better since there is no time spent waiting for
work-queue jobs to be scheduled.

   READ: bw=181MiB/s (190MB/s), 181MiB/s-181MiB/s (190MB/s-190MB/s),
   io=512MiB (537MB), run=2827-2827msec
   READ: bw=23.6MiB/s (24.8MB/s), 23.6MiB/s-23.6MiB/s (24.8MB/s-24.8MB/s),
   io=512MiB (537MB), run=21688-21688msec

Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm bufio: Add DM_BUFIO_CLIENT_NO_SLEEP flag
Nathan Huckleberry [Fri, 22 Jul 2022 09:38:22 +0000 (09:38 +0000)]
dm bufio: Add DM_BUFIO_CLIENT_NO_SLEEP flag

Add an optional flag that ensures dm_bufio_client does not sleep
(primary focus is to service dm_bufio_get without sleeping). This
allows the dm-bufio cache to be queried from interrupt context.

To ensure that dm-bufio does not sleep, dm-bufio must use a spinlock
instead of a mutex. Additionally, to avoid deadlocks, special care
must be taken so that dm-bufio does not sleep while holding the
spinlock.

But again: the scope of this no_sleep is initially confined to
dm_bufio_get, so __alloc_buffer_wait_no_callback is _not_ changed to
avoid sleeping because __bufio_new avoids allocation for NF_GET.

Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm bufio: Add flags argument to dm_bufio_client_create
Nathan Huckleberry [Fri, 22 Jul 2022 09:38:21 +0000 (09:38 +0000)]
dm bufio: Add flags argument to dm_bufio_client_create

Add a flags argument to dm_bufio_client_create and update all the
callers. This is in preparation to add the DM_BUFIO_NO_SLEEP flag.

Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: fix dm-raid crash if md_handle_request() splits bio
Mike Snitzer [Wed, 20 Jul 2022 17:58:04 +0000 (13:58 -0400)]
dm: fix dm-raid crash if md_handle_request() splits bio

Commit ca522482e3eaf ("dm: pass NULL bdev to bio_alloc_clone")
introduced the optimization to _not_ perform bio_associate_blkg()'s
relatively costly work when DM core clones its bio. But in doing so it
exposed the possibility for DM's cloned bio to alter DM target
behavior (e.g. crash) if a target were to issue IO without first
calling bio_set_dev().

The DM raid target can trigger an MD crash due to its need to split
the DM bio that is passed to md_handle_request(). The split will
recurse to submit_bio_noacct() using a bio with an uninitialized
->bi_blkg. This NULL bio->bi_blkg causes blk_throtl_bio() to
dereference a NULL blkg_to_tg(bio->bi_blkg).

Fix this in DM core by adding a new 'needs_bio_set_dev' target flag that
will make alloc_tio() call bio_set_dev() on behalf of the target.
dm-raid is the only target that requires this flag. bio_set_dev()
initializes the DM cloned bio's ->bi_blkg, using bio_associate_blkg,
before passing the bio to md_handle_request().

Long-term fix would be to audit and refactor MD code to rely on DM to
split its bio, using dm_accept_partial_bio(), but there are MD raid
personalities (e.g. raid1 and raid10) whose implementation are tightly
coupled to handling the bio splitting inline.

Fixes: ca522482e3eaf ("dm: pass NULL bdev to bio_alloc_clone")
Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm raid: fix address sanitizer warning in raid_resume
Mikulas Patocka [Sun, 24 Jul 2022 18:33:52 +0000 (14:33 -0400)]
dm raid: fix address sanitizer warning in raid_resume

There is a KASAN warning in raid_resume when running the lvm test
lvconvert-raid.sh. The reason for the warning is that mddev->raid_disks
is greater than rs->raid_disks, so the loop touches one entry beyond
the allocated length.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm raid: fix address sanitizer warning in raid_status
Mikulas Patocka [Sun, 24 Jul 2022 18:31:35 +0000 (14:31 -0400)]
dm raid: fix address sanitizer warning in raid_status

There is this warning when using a kernel with the address sanitizer
and running this testsuite:
https://gitlab.com/cki-project/kernel-tests/-/tree/main/storage/swraid/scsi_raid

==================================================================
BUG: KASAN: slab-out-of-bounds in raid_status+0x1747/0x2820 [dm_raid]
Read of size 4 at addr ffff888079d2c7e8 by task lvcreate/13319
CPU: 0 PID: 13319 Comm: lvcreate Not tainted 5.18.0-0.rc3.<snip> #1
Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
Call Trace:
 <TASK>
 dump_stack_lvl+0x6a/0x9c
 print_address_description.constprop.0+0x1f/0x1e0
 print_report.cold+0x55/0x244
 kasan_report+0xc9/0x100
 raid_status+0x1747/0x2820 [dm_raid]
 dm_ima_measure_on_table_load+0x4b8/0xca0 [dm_mod]
 table_load+0x35c/0x630 [dm_mod]
 ctl_ioctl+0x411/0x630 [dm_mod]
 dm_ctl_ioctl+0xa/0x10 [dm_mod]
 __x64_sys_ioctl+0x12a/0x1a0
 do_syscall_64+0x5b/0x80

The warning is caused by reading conf->max_nr_stripes in raid_status. The
code in raid_status reads mddev->private, casts it to struct r5conf and
reads the entry max_nr_stripes.

However, if we have different raid type than 4/5/6, mddev->private
doesn't point to struct r5conf; it may point to struct r0conf, struct
r1conf, struct r10conf or struct mpconf. If we cast a pointer to one
of these structs to struct r5conf, we will be reading invalid memory
and KASAN warns about it.

Fix this bug by reading struct r5conf only if raid type is 4, 5 or 6.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: Start pr_preempt from the same starting path
Mike Christie [Sun, 17 Jul 2022 22:45:08 +0000 (17:45 -0500)]
dm: Start pr_preempt from the same starting path

pr_preempt has a similar issue as reserve where for all the
reservation types except the All Registrants ones the preempt can
create a reservation. And a follow up reservation or release needs to
go down the same path the preempt did. This has the pr_preempt work
like reserve and release where we always start from the first path in
the first group.

This commit has been tested with windows failover clustering's
validation test and libiscsi's PGR tests to check for regressions.
They both don't have tests to verify this case, so I tested it
manually.

Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: Fix PR release handling for non All Registrants
Mike Christie [Sun, 17 Jul 2022 22:45:07 +0000 (17:45 -0500)]
dm: Fix PR release handling for non All Registrants

This commit fixes a bug where we are leaving the reservation in place
even though pr_release has run and returned success.

If we have a Write Exclusive, Exclusive Access, or Write/Exclusive
Registrants only reservation, the release must be sent down the path
that is the reservation holder. The problem is multipath_prepare_ioctl
most likely selected path N for the reservation, then later when we do
the release multipath_prepare_ioctl will select a completely different
path. The device will then return success becuase the nvme and scsi
specs say to return success if there is no reservation or if the
release is sent down from a path that is not the holder. We then think
we have released the reservation.

This commit has us loop over each path and send a release so we can
make sure the release is executed on the correct path. It has been
tested with windows failover clustering's validation test which checks
this case, and it has been tested manually (the libiscsi PGR tests
don't have a test case for this yet, but I will be adding one).

Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: Start pr_reserve from the same starting path
Mike Christie [Sun, 17 Jul 2022 22:45:06 +0000 (17:45 -0500)]
dm: Start pr_reserve from the same starting path

When an app does a pr_reserve it will go to whatever path we happen to
be using at the time. This can result in errors when the app does a
second pr_reserve call and expects success but gets a failure because
the reserve is not done on the holder's path. This commit has us
always start trying to do reserves from the first path in the first
group.

Windows failover clustering will produce the type of pattern above.
With this commit, we will now pass its validation test for this case.

Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: Allow dm_call_pr to be used for path searches
Mike Christie [Sun, 17 Jul 2022 22:45:05 +0000 (17:45 -0500)]
dm: Allow dm_call_pr to be used for path searches

The specs state that if you send a reserve down a path that is already
the holder success must be returned and if it goes down a path that
is not the holder reservation conflict must be returned. Windows
failover clustering will send a second reservation and expects that a
device returns success. The problem for multipathing is that for an
All Registrants reservation, we can send the reserve down any path but
for all other reservation types there is one path that is the holder.

To handle this we could add PR state to dm but that can get nasty.
Look at target_core_pr.c for an example of the type of things we'd
have to track. It will also get more complicated because other
initiators can change the state so we will have to add in async
event/sense handling.

This commit, and the 3 commits that follow, tries to keep dm simple
and keep just doing passthrough. This commit modifies dm_call_pr to be
able to find the first usable path that can execute our pr_op then
return. When dm_pr_reserve is converted to dm_call_pr in the next
commit for the normal case we will use the same path for every
reserve.

Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: return early from dm_pr_call() if DM device is suspended
Mike Snitzer [Fri, 22 Jul 2022 19:31:23 +0000 (15:31 -0400)]
dm: return early from dm_pr_call() if DM device is suspended

Otherwise PR ops may be issued while the broader DM device is being
reconfigured, etc.

Fixes: 9c72bad1f31a ("dm: call PR reserve/unreserve on each underlying device")
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm thin: fix use-after-free crash in dm_sm_register_threshold_callback
Luo Meng [Thu, 14 Jul 2022 11:28:25 +0000 (19:28 +0800)]
dm thin: fix use-after-free crash in dm_sm_register_threshold_callback

Fault inject on pool metadata device reports:
  BUG: KASAN: use-after-free in dm_pool_register_metadata_threshold+0x40/0x80
  Read of size 8 at addr ffff8881b9d50068 by task dmsetup/950

  CPU: 7 PID: 950 Comm: dmsetup Tainted: G        W         5.19.0-rc6 #1
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1.fc33 04/01/2014
  Call Trace:
   <TASK>
   dump_stack_lvl+0x34/0x44
   print_address_description.constprop.0.cold+0xeb/0x3f4
   kasan_report.cold+0xe6/0x147
   dm_pool_register_metadata_threshold+0x40/0x80
   pool_ctr+0xa0a/0x1150
   dm_table_add_target+0x2c8/0x640
   table_load+0x1fd/0x430
   ctl_ioctl+0x2c4/0x5a0
   dm_ctl_ioctl+0xa/0x10
   __x64_sys_ioctl+0xb3/0xd0
   do_syscall_64+0x35/0x80
   entry_SYSCALL_64_after_hwframe+0x46/0xb0

This can be easily reproduced using:
  echo offline > /sys/block/sda/device/state
  dd if=/dev/zero of=/dev/mapper/thin bs=4k count=10
  dmsetup load pool --table "0 20971520 thin-pool /dev/sda /dev/sdb 128 0 0"

If a metadata commit fails, the transaction will be aborted and the
metadata space maps will be destroyed. If a DM table reload then
happens for this failed thin-pool, a use-after-free will occur in
dm_sm_register_threshold_callback (called from
dm_pool_register_metadata_threshold).

Fix this by in dm_pool_register_metadata_threshold() by returning the
-EINVAL error if the thin-pool is in fail mode. Also fail pool_ctr()
with a new error message: "Error registering metadata threshold".

Fixes: ac8c3f3df65e4 ("dm thin: generate event when metadata threshold passed")
Cc: stable@vger.kernel.org
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Luo Meng <luomeng12@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm writecache: count number of blocks discarded, not number of discard bios
Mikulas Patocka [Mon, 11 Jul 2022 20:31:52 +0000 (16:31 -0400)]
dm writecache: count number of blocks discarded, not number of discard bios

Change dm-writecache, so that it counts the number of blocks discarded
instead of the number of discard bios. Make it consistent with the
read and write statistics counters that were changed to count the
number of blocks instead of bios.

Fixes: e3a35d03407c ("dm writecache: add event counters")
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm writecache: count number of blocks written, not number of write bios
Mikulas Patocka [Mon, 11 Jul 2022 20:31:26 +0000 (16:31 -0400)]
dm writecache: count number of blocks written, not number of write bios

Change dm-writecache, so that it counts the number of blocks written
instead of the number of write bios. Bios can be split and requeued
using the dm_accept_partial_bio function, so counting bios caused
inaccurate results.

Fixes: e3a35d03407c ("dm writecache: add event counters")
Reported-by: Yu Kuai <yukuai1@huaweicloud.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm writecache: count number of blocks read, not number of read bios
Mikulas Patocka [Mon, 11 Jul 2022 20:30:52 +0000 (16:30 -0400)]
dm writecache: count number of blocks read, not number of read bios

Change dm-writecache, so that it counts the number of blocks read
instead of the number of read bios. Bios can be split and requeued
using the dm_accept_partial_bio function, so counting bios caused
inaccurate results.

Fixes: e3a35d03407c ("dm writecache: add event counters")
Reported-by: Yu Kuai <yukuai1@huaweicloud.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm writecache: return void from functions
Mikulas Patocka [Mon, 11 Jul 2022 20:30:27 +0000 (16:30 -0400)]
dm writecache: return void from functions

The functions writecache_map_remap_origin and writecache_bio_copy_ssd
only return a single value, thus they can be made to return void.

This helps simplify the following IO accounting changes.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm kcopyd: use __GFP_HIGHMEM when allocating pages
Mikulas Patocka [Wed, 13 Jul 2022 11:05:51 +0000 (07:05 -0400)]
dm kcopyd: use __GFP_HIGHMEM when allocating pages

dm-kcopyd doesn't access the allocated pages directly, it only passes
them to dm-io which adds them to a bio list - thus, we can allocate
the pages from high memory. This will reduce pressure on the low
memory when there are a large number of kcopyd jobs in progress.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm writecache: set a default MAX_WRITEBACK_JOBS
Mikulas Patocka [Wed, 13 Jul 2022 11:09:04 +0000 (07:09 -0400)]
dm writecache: set a default MAX_WRITEBACK_JOBS

dm-writecache has the capability to limit the number of writeback jobs
in progress. However, this feature was off by default. As such there
were some out-of-memory crashes observed when lowering the low
watermark while the cache is full.

This commit enables writeback limit by default. It is set to 256MiB or
1/16 of total system memory, whichever is smaller.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agoDocumentation: dm writecache: Render status list as list
Bagas Sanjaya [Sat, 2 Jul 2022 04:23:50 +0000 (11:23 +0700)]
Documentation: dm writecache: Render status list as list

The status list isn't rendered as list, but rather as normal paragraph,
because there is missing blank line between "Status:" line and the list.

Fix the issue by adding the blank line separator.

Fixes: 48debafe4f2fea ("dm: add writecache target")
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agoDocumentation: dm writecache: add blank line before optional parameters
Mauro Carvalho Chehab [Sat, 2 Jul 2022 11:07:35 +0000 (12:07 +0100)]
Documentation: dm writecache: add blank line before optional parameters

Otherwise this warning occurs:
  Documentation/admin-guide/device-mapper/writecache.rst:23: WARNING: Unexpected indentation.

Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm snapshot: fix typo in snapshot_map() comment
Zhang Jiaming [Fri, 1 Jul 2022 08:09:59 +0000 (16:09 +0800)]
dm snapshot: fix typo in snapshot_map() comment

Signed-off-by: Zhang Jiaming <jiaming@nfschina.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm raid: remove redundant "the" in parse_raid_params() comment
Jiang Jian [Tue, 21 Jun 2022 11:32:34 +0000 (19:32 +0800)]
dm raid: remove redundant "the" in parse_raid_params() comment

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm cache: fix typo in 2 comment blocks
Steven Lung [Tue, 21 Jun 2022 07:12:59 +0000 (15:12 +0800)]
dm cache: fix typo in 2 comment blocks

Replace neccessarily with necessarily.

Signed-off-by: Steven Lung <1030steven@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm verity: fix checkpatch close brace error
JeongHyeon Lee [Wed, 15 Jun 2022 00:51:51 +0000 (09:51 +0900)]
dm verity: fix checkpatch close brace error

Resolves: ERROR: else should follow close brace '}'

Signed-off-by: JeongHyeon Lee <jhs2.lee@samsung.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm table: rename dm_target variable in dm_table_add_target()
Mike Snitzer [Tue, 5 Jul 2022 20:29:09 +0000 (16:29 -0400)]
dm table: rename dm_target variable in dm_table_add_target()

Rename from "tgt" to "ti" so that all of dm-table.c code uses the same
naming for dm_target variables.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm table: audit all dm_table_get_target() callers
Mike Snitzer [Tue, 5 Jul 2022 20:12:27 +0000 (16:12 -0400)]
dm table: audit all dm_table_get_target() callers

All callers of dm_table_get_target() are expected to do proper bounds
checking on the index they pass.

Move dm_table_get_target() to dm-core.h to make it extra clear that only
DM core code should be using it. Switch it to be inlined while at it.

Standardize all DM core callers to use the same for loop pattern and
make associated variables as local as possible. Rename some variables
(e.g. s/table/t/ and s/tgt/ti/) along the way.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm table: remove dm_table_get_num_targets() wrapper
Mike Snitzer [Tue, 5 Jul 2022 18:00:36 +0000 (14:00 -0400)]
dm table: remove dm_table_get_num_targets() wrapper

More efficient and readable to just access table->num_targets directly.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: add two stage requeue mechanism
Ming Lei [Fri, 24 Jun 2022 14:12:55 +0000 (22:12 +0800)]
dm: add two stage requeue mechanism

Commit 61b6e2e5321d ("dm: fix BLK_STS_DM_REQUEUE handling when dm_io
represents split bio") reverted DM core's bio splitting back to using
bio_split()+bio_chain() because it was found that otherwise DM's
BLK_STS_DM_REQUEUE would trigger a live-lock waiting for bio
completion that would never occur.

Restore using bio_trim()+bio_inc_remaining(), like was done in commit
7dd76d1feec7 ("dm: improve bio splitting and associated IO
accounting"), but this time with proper handling for the above
scenario that is covered in more detail in the commit header for
61b6e2e5321d.

Solve this issue by adding a two staged dm_io requeue mechanism that
uses the new dm_bio_rewind() via dm_io_rewind():

1) requeue the dm_io into the requeue_list added to struct
   mapped_device, and schedule it via new added requeue work. This
   workqueue just clones the dm_io->orig_bio (which DM saves and
   ensures its end sector isn't modified). dm_io_rewind() uses the
   sectors and sectors_offset members of the dm_io that are recorded
   relative to the end of orig_bio: dm_bio_rewind()+bio_trim() are
   then used to make that cloned bio reflect the subset of the
   original bio that is represented by the dm_io that is being
   requeued.

2) the 2nd stage requeue is same with original requeue, but
   io->orig_bio points to new cloned bio (which matches the requeued
   dm_io as described above).

This allows DM core to shift the need for bio cloning from bio-split
time (during IO submission) to the less likely BLK_STS_DM_REQUEUE
handling (after IO completes with that error).

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: add dm_bio_rewind() API to DM core
Ming Lei [Fri, 24 Jun 2022 14:12:52 +0000 (22:12 +0800)]
dm: add dm_bio_rewind() API to DM core

Commit 7759eb23fd98 ("block: remove bio_rewind_iter()") removed
a similar API for the following reasons:
    ```
    It is pointed that bio_rewind_iter() is one very bad API[1]:

    1) bio size may not be restored after rewinding

    2) it causes some bogus change, such as 5151842b9d8732 (block: reset
    bi_iter.bi_done after splitting bio)

    3) rewinding really makes things complicated wrt. bio splitting

    4) unnecessary updating of .bi_done in fast path

    [1] https://marc.info/?t=153549924200005&r=1&w=2

    So this patch takes Kent's suggestion to restore one bio into its original
    state via saving bio iterator(struct bvec_iter) in bio_integrity_prep(),
    given now bio_rewind_iter() is only used by bio integrity code.
    ```
However, saving off a copy of the 32 bytes bio->bi_iter in case rewind
needed isn't efficient because it bloats per-bio-data for what is an
unlikely case. That suggestion also ignores the need to restore
crypto and integrity info.

Add dm_bio_rewind() API for a specific use-case that is much more narrow
than the previous more generic rewind code that was reverted:

1) most bios have a fixed end sector since bio split is done from front
   of the bio, if driver just records how many sectors between current
   bio's start sector and the original bio's end sector, the original
   position can be restored. Keeping the original bio's end sector
   fixed is a _hard_ requirement for this interface!

2) if a bio's end sector won't change (usually bio_trim() isn't
   called, or in the case of DM it preserves original bio), user can
   restore the original position by storing sector offset from the
   current ->bi_iter.bi_sector to bio's end sector; together with
   saving bio size, only 8 bytes is needed to restore to original
   bio.

3) DM's requeue use case: when BLK_STS_DM_REQUEUE happens, DM core
   needs to restore to an "original bio" which represents the current
   dm_io to be requeued (which may be a subset of the original bio).
   By storing the sector offset from the original bio's end sector and
   dm_io's size, dm_bio_rewind() can restore such original bio. See
   commit 7dd76d1feec7 ("dm: improve bio splitting and associated IO
   accounting") for more details on how DM does this. Leveraging this,
   allows DM core to shift the need for bio cloning from bio-split
   time (during IO submission) to the less likely BLK_STS_DM_REQUEUE
   handling (after IO completes with that error).

4) Unlike the original rewind API, dm_bio_rewind() doesn't add .bi_done
   to bvec_iter and there is no effect on the fast path.

Implement dm_bio_rewind() by factoring out clear helpers that it calls:
dm_bio_integrity_rewind, dm_bio_crypt_rewind and dm_bio_rewind_iter.

DM is able to ensure that dm_bio_rewind() is used safely but, given
the constraint that the bio's end must never change, other
hypothetical future callers may not take the same care. So make
dm_bio_rewind() and all supporting code local to DM to avoid risk of
hypothetical abuse. A "dm_" prefix was added to all functions to avoid
any namespace collisions.

Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
22 months agodm: improve BLK_STS_DM_REQUEUE and BLK_STS_AGAIN handling
Ming Lei [Fri, 24 Jun 2022 14:12:53 +0000 (22:12 +0800)]
dm: improve BLK_STS_DM_REQUEUE and BLK_STS_AGAIN handling

If either BLK_STS_DM_REQUEUE or BLK_STS_AGAIN is returned for POLLED
io, we requeue the original bio into deferred list and kick md->wq to
re-submit it to block layer.

Improve the handling in the following way:

1) Factor out dm_handle_requeue() for handling dm_io requeue.

2) Unify handling for BLK_STS_DM_REQUEUE and BLK_STS_AGAIN: clear
   REQ_POLLED for BLK_STS_DM_REQUEUE too, for the sake of simplicity,
   given BLK_STS_DM_REQUEUE is very unusual.

3) Queue md->wq explicitly in dm_handle_requeue(), so requeue handling
   becomes more robust.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
23 months agodm: refactor dm_md_mempool allocation
Christoph Hellwig [Wed, 8 Jun 2022 06:34:09 +0000 (08:34 +0200)]
dm: refactor dm_md_mempool allocation

The current split between dm_table_alloc_md_mempools and
dm_alloc_md_mempools is rather arbitrary, so merge the two
into one easy to follow function.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
23 months agodm: unexport dm_get_reserved_rq_based_ios
Christoph Hellwig [Wed, 8 Jun 2022 06:34:08 +0000 (08:34 +0200)]
dm: unexport dm_get_reserved_rq_based_ios

dm_get_reserved_rq_based_ios is only used in the core dm code, so
remove the export.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
23 months agoblock: simplify disk_set_independent_access_ranges
Christoph Hellwig [Wed, 29 Jun 2022 06:20:13 +0000 (08:20 +0200)]
block: simplify disk_set_independent_access_ranges

Lift setting disk->ia_ranges from disk_register_independent_access_ranges
into disk_set_independent_access_ranges, and make the behavior the same
for the registered vs non-registered queue cases.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Link: https://lore.kernel.org/r/20220629062013.1331068-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: move ->ia_ranges from the request_queue to the gendisk
Christoph Hellwig [Wed, 29 Jun 2022 06:20:12 +0000 (08:20 +0200)]
block: move ->ia_ranges from the request_queue to the gendisk

Independent access ranges only matter for file system I/O and are only
valid with a registered gendisk, so move them there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Link: https://lore.kernel.org/r/20220629062013.1331068-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: remove "select BLK_RQ_IO_DATA_LEN" from BLK_CGROUP_IOCOST dependency
Ying Sun [Wed, 29 Jun 2022 06:24:09 +0000 (14:24 +0800)]
block: remove "select BLK_RQ_IO_DATA_LEN" from BLK_CGROUP_IOCOST dependency

The configuration item BLK_RQ_IO_DATA_LEN is not declared in the kernel.
Select BLK_RQ_IO_DATA_LEN is meaningless which could be removed.

Signed-off-by: Ying Sun <sunying@nj.iscas.ac.cn>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220629062409.19458-1-sunying@nj.iscas.ac.cn
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-mq: cleanup disk sysfs registration
Christoph Hellwig [Tue, 28 Jun 2022 17:18:50 +0000 (19:18 +0200)]
blk-mq: cleanup disk sysfs registration

Pass a gendisk to the sysfs register/unregister functions and give
them descriptive names.  Also move the unregistration helper next
to the one doing the registration.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-mq: rename blk_mq_sysfs_{,un}register
Christoph Hellwig [Tue, 28 Jun 2022 17:18:49 +0000 (19:18 +0200)]
blk-mq: rename blk_mq_sysfs_{,un}register

Add a _hctx postfix to better describe what the functions do, match
the debugfs equivalents and release the old names for functions that
should be using them.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: remove the extra gendisk reference in __blk_mq_register_dev
Christoph Hellwig [Tue, 28 Jun 2022 17:18:48 +0000 (19:18 +0200)]
block: remove the extra gendisk reference in __blk_mq_register_dev

kobject_add already grabs a reference to the parent, no need to have
another one.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: use default groups to register the queue attributes
Christoph Hellwig [Tue, 28 Jun 2022 17:18:47 +0000 (19:18 +0200)]
block: use default groups to register the queue attributes

Set up the default_groups for blk_queue_ktype instead of manually calling
sysfs_create_group.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: remove a superflous queue kobject reference
Christoph Hellwig [Tue, 28 Jun 2022 17:18:46 +0000 (19:18 +0200)]
block: remove a superflous queue kobject reference

kobject_add already adds a reference to the parent that is dropped
on deletion, so don't bother grabbing another one.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: simplify blktrace sysfs attribute creation
Christoph Hellwig [Tue, 28 Jun 2022 17:18:45 +0000 (19:18 +0200)]
block: simplify blktrace sysfs attribute creation

Add the trace attributes to the default gendisk attributes, just like
we already do for partitions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220628171850.1313069-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: remove blk_cleanup_disk
Christoph Hellwig [Sun, 19 Jun 2022 06:05:52 +0000 (08:05 +0200)]
block: remove blk_cleanup_disk

blk_cleanup_disk is nothing but a trivial wrapper for put_disk now,
so remove it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: simplify disk shutdown
Christoph Hellwig [Sun, 19 Jun 2022 06:05:51 +0000 (08:05 +0200)]
block: simplify disk shutdown

Set the queue dying flag and call blk_mq_exit_queue from del_gendisk for
all disks that do not have separately allocated queues, and thus remove
the need to call blk_cleanup_queue for them.

Rename blk_cleanup_disk to blk_mq_destroy_queue to make it clear that
this function is intended only for separately allocated blk-mq queues.

This saves an extra queue freeze for devices without a separately
allocated queue.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: stop setting the nomerges flags in blk_cleanup_queue
Christoph Hellwig [Sun, 19 Jun 2022 06:05:50 +0000 (08:05 +0200)]
block: stop setting the nomerges flags in blk_cleanup_queue

These flags only apply to file system I/O, and all file system I/O is
already drained by del_gendisk and thus can't be in progress when
blk_cleanup_queue is called.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: remove QUEUE_FLAG_DEAD
Christoph Hellwig [Sun, 19 Jun 2022 06:05:49 +0000 (08:05 +0200)]
block: remove QUEUE_FLAG_DEAD

Disallow setting the blk-mq state on any queue that is already dying as
setting the state even then is a bad idea, and remove the now unused
QUEUE_FLAG_DEAD flag.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agomtip32xx: fix device removal
Christoph Hellwig [Sun, 19 Jun 2022 06:05:48 +0000 (08:05 +0200)]
mtip32xx: fix device removal

Use the proper helper to mark a surpise removal, remove the gendisk as
soon as possible when removing the device and implement the ->free_disk
callback to ensure the private data is alive as long as the gendisk has
references.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agomtip32xx: remove the device_status debugfs file
Christoph Hellwig [Sun, 19 Jun 2022 06:05:47 +0000 (08:05 +0200)]
mtip32xx: remove the device_status debugfs file

This file is a huge mess that iterates over all devices and is in the
way of fixing the device removal in this driver, so remove it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-mq: blk_mq_tag_busy is no need to return a value
Liu Song [Sat, 25 Jun 2022 15:15:21 +0000 (23:15 +0800)]
blk-mq: blk_mq_tag_busy is no need to return a value

Currently "blk_mq_tag_busy" return value has no effect, so adjust it.
Some code implementations have also been adjusted to enhance
readability.

Signed-off-by: Liu Song <liusong@linux.alibaba.com>
Link: https://lore.kernel.org/r/1656170121-1619-1-git-send-email-liusong@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Always initialize bio IO priority on submit
Jan Kara [Thu, 23 Jun 2022 07:48:34 +0000 (09:48 +0200)]
block: Always initialize bio IO priority on submit

Currently, IO priority set in task's IO context is not reflected in the
bio->bi_ioprio for most IO (only io_uring and direct IO set it). This
results in odd results where process is submitting some bios with one
priority and other bios with a different (unset) priority and due to
differing priorities bios cannot be merged. Make sure bio->bi_ioprio is
always set on bio submission.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-9-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Initialize bio priority earlier
Jan Kara [Thu, 23 Jun 2022 07:48:33 +0000 (09:48 +0200)]
block: Initialize bio priority earlier

Bio's IO priority needs to be initialized before we try to merge the bio
with other bios. Otherwise we could merge bios which would otherwise
receive different IO priorities leading to possible QoS issues.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-8-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-ioprio: Convert from rqos policy to direct call
Jan Kara [Thu, 23 Jun 2022 07:48:32 +0000 (09:48 +0200)]
blk-ioprio: Convert from rqos policy to direct call

Convert blk-ioprio handling from a rqos policy to a direct call from
blk_mq_submit_bio(). Firstly, blk-ioprio is not much of a rqos policy
anyway, it just needs a hook in bio submission path to set the bio's IO
priority. Secondly, the rqos .track hook gets actually called too late
for blk-ioprio purposes and introducing a special rqos hook just for
blk-ioprio looks even weirder.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-7-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-ioprio: Remove unneeded field
Jan Kara [Thu, 23 Jun 2022 07:48:31 +0000 (09:48 +0200)]
blk-ioprio: Remove unneeded field

blkcg->ioprio_set field is not really useful except for avoiding
possibly more expensive checks inside blkcg_ioprio_track(). The check
for blkcg->prio_policy being equal to POLICY_NO_CHANGE does the same
service so just remove the ioprio_set field and replace the check.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-6-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Fix handling of tasks without ioprio in ioprio_get(2)
Jan Kara [Thu, 23 Jun 2022 07:48:30 +0000 (09:48 +0200)]
block: Fix handling of tasks without ioprio in ioprio_get(2)

ioprio_get(2) can be asked to return the best IO priority from several
tasks (IOPRIO_WHO_PGRP, IOPRIO_WHO_USER). Currently the call treats
tasks without set IO priority as having priority
IOPRIO_CLASS_BE/IOPRIO_BE_NORM however this does not really reflect the
IO priority the task will get (which depends on task's nice value).

Fix the code to use the real IO priority task's IO will use. We have to
modify code for ioprio_get(IOPRIO_WHO_PROCESS) to keep returning
IOPRIO_CLASS_NONE priority for tasks without set IO priority as a
special case to maintain userspace visible API.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220623074840.5960-5-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Make ioprio_best() static
Jan Kara [Thu, 23 Jun 2022 07:48:29 +0000 (09:48 +0200)]
block: Make ioprio_best() static

Nobody outside of block/ioprio.c uses it.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-4-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Generalize get_current_ioprio() for any task
Jan Kara [Thu, 23 Jun 2022 07:48:28 +0000 (09:48 +0200)]
block: Generalize get_current_ioprio() for any task

get_current_ioprio() operates only on current task. We will need the
same functionality for other tasks as well. Generalize
get_current_ioprio() for that and also move the bulk out of the header
file because it is large enough.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-3-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Return effective IO priority from get_current_ioprio()
Jan Kara [Thu, 23 Jun 2022 07:48:27 +0000 (09:48 +0200)]
block: Return effective IO priority from get_current_ioprio()

get_current_ioprio() is used to initialize IO priority of various
requests. As such it should be returning the effective IO priority of
the task (i.e., reflecting the fact that unset IO priority should get
set based on task's CPU priority) so that the conversion is concentrated
in one place.

Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-2-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: fix default IO priority handling again
Jan Kara [Thu, 23 Jun 2022 07:48:26 +0000 (09:48 +0200)]
block: fix default IO priority handling again

Commit e70344c05995 ("block: fix default IO priority handling")
introduced an inconsistency in get_current_ioprio() that tasks without
IO context return IOPRIO_DEFAULT priority while tasks with freshly
allocated IO context will return 0 (IOPRIO_CLASS_NONE/0) IO priority.
Tasks without IO context used to be rare before 5a9d041ba2f6 ("block:
move io_context creation into where it's needed") but after this commit
they became common because now only BFQ IO scheduler setups task's IO
context. Similar inconsistency is there for get_task_ioprio() so this
inconsistency is now exposed to userspace and userspace will see
different IO priority for tasks operating on devices with BFQ compared
to devices without BFQ. Furthemore the changes done by commit
e70344c05995 change the behavior when no IO priority is set for BFQ IO
scheduler which is also documented in ioprio_set(2) manpage:

"If no I/O scheduler has been set for a thread, then by default the I/O
priority will follow the CPU nice value (setpriority(2)).  In Linux
kernels before version 2.6.24, once an I/O priority had been set using
ioprio_set(), there was no way to reset the I/O scheduling behavior to
the default. Since Linux 2.6.24, specifying ioprio as 0 can be used to
reset to the default I/O scheduling behavior."

So make sure we default to IOPRIO_CLASS_NONE as used to be the case
before commit e70344c05995. Also cleanup alloc_io_context() to
explicitely set this IO priority for the allocated IO context to avoid
future surprises. Note that we tweak ioprio_best() to maintain
ioprio_get(2) behavior and make this commit easily backportable.

CC: stable@vger.kernel.org
Fixes: e70344c05995 ("block: fix default IO priority handling")
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220623074840.5960-1-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-mq: Don't disable preemption around __blk_mq_run_hw_queue().
Sebastian Andrzej Siewior [Wed, 22 Jun 2022 08:25:54 +0000 (10:25 +0200)]
blk-mq: Don't disable preemption around __blk_mq_run_hw_queue().

__blk_mq_delay_run_hw_queue() disables preemption to get a stable
current CPU number and then invokes __blk_mq_run_hw_queue() if the CPU
number is part the mask.

__blk_mq_run_hw_queue() acquires a spin_lock_t which is a sleeping lock
on PREEMPT_RT and can't be acquired with disabled preemption.

It is not required for correctness to invoke __blk_mq_run_hw_queue() on
a CPU matching hctx->cpumask. Both (async and direct requests) can run
on a CPU not matching hctx->cpumask.

The CPU mask without disabling preemption and invoking
__blk_mq_run_hw_queue().

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/YrLSEiNvagKJaDs5@linutronix.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: bfq: Fix kernel-doc headers
Bart Van Assche [Fri, 17 Jun 2022 21:08:59 +0000 (14:08 -0700)]
block: bfq: Fix kernel-doc headers

Fix the following warnings:

block/bfq-cgroup.c:721: warning: Function parameter or member 'bfqg' not described in '__bfq_bic_change_cgroup'
block/bfq-cgroup.c:721: warning: Excess function parameter 'blkcg' description in '__bfq_bic_change_cgroup'
block/bfq-cgroup.c:870: warning: Function parameter or member 'ioprio_class' not described in 'bfq_reparent_leaf_entity'
block/bfq-cgroup.c:900: warning: Function parameter or member 'ioprio_class' not described in 'bfq_reparent_active_queues'

Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/r/20220617210859.106623-1-bvanassche@acm.org
23 months agoblock: bfq: Remove an unused function definition
Bart Van Assche [Fri, 17 Jun 2022 20:44:33 +0000 (13:44 -0700)]
block: bfq: Remove an unused function definition

This patch is the result of the analysis of a sparse report.

Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/r/20220617204433.102022-1-bvanassche@acm.org
23 months agobfq: Remove useless code in bfq_lookup_next_entity
GuoYong Zheng [Fri, 17 Jun 2022 10:28:04 +0000 (18:28 +0800)]
bfq: Remove useless code in bfq_lookup_next_entity

It is no need to judge entity is null or not here,
directly return entity is ok, so remove it.

Signed-off-by: GuoYong Zheng <zhenggy@chinatelecom.cn>
Link: https://lore.kernel.org/r/1655461684-19075-1-git-send-email-zhenggy@chinatelecom.cn
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: move blk_queue_get_max_sectors to blk.h
Christoph Hellwig [Tue, 14 Jun 2022 09:09:34 +0000 (11:09 +0200)]
block: move blk_queue_get_max_sectors to blk.h

blk_queue_get_max_sectors is private to the block layer, so move it out
of blkdev.h.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220614090934.570632-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: fold blk_max_size_offset into get_max_io_size
Christoph Hellwig [Tue, 14 Jun 2022 09:09:33 +0000 (11:09 +0200)]
block: fold blk_max_size_offset into get_max_io_size

Now that blk_max_size_offset has a single caller left, fold it into that
and clean up the naming convention for the local variables there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Pankaj Raghav <p.raghav@samsung.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220614090934.570632-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: cleanup variable naming in get_max_io_size
Christoph Hellwig [Tue, 14 Jun 2022 09:09:32 +0000 (11:09 +0200)]
block: cleanup variable naming in get_max_io_size

get_max_io_size has a very odd choice of variables names and
initialization patterns.  Switch to more descriptive names and more
clear initialization of them.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220614090934.570632-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: open code blk_max_size_offset in blk_rq_get_max_sectors
Christoph Hellwig [Tue, 14 Jun 2022 09:09:31 +0000 (11:09 +0200)]
block: open code blk_max_size_offset in blk_rq_get_max_sectors

blk_rq_get_max_sectors always uses q->limits.chunk_sectors as the
chunk_sectors argument, and already checks for max_sectors through the
call to blk_queue_get_max_sectors.  That means much of
blk_max_size_offset is not needed and open coding it simplifies the code.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220614090934.570632-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agodm: open code blk_max_size_offset in max_io_len
Christoph Hellwig [Tue, 14 Jun 2022 09:09:30 +0000 (11:09 +0200)]
dm: open code blk_max_size_offset in max_io_len

max_io_len always passes an explicitly non-zero chunk_sectors into
blk_max_size_offset.  That means much of blk_max_size_offset is not
needed and can be open coded to simplify the code.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mike Snitzer <snitzer@kernel.org>
Link: https://lore.kernel.org/r/20220614090934.570632-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: factor out a chunk_size_left helper
Christoph Hellwig [Tue, 14 Jun 2022 09:09:29 +0000 (11:09 +0200)]
block: factor out a chunk_size_left helper

Factor out a helper from blk_max_size_offset so that it can be reused
independently.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Pankaj Raghav <p.raghav@samsung.com>
Link: https://lore.kernel.org/r/20220614090934.570632-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Make blk_mq_get_sq_hctx() select the proper hardware queue type
Bart Van Assche [Wed, 15 Jun 2022 22:55:49 +0000 (15:55 -0700)]
block: Make blk_mq_get_sq_hctx() select the proper hardware queue type

Since the introduction of blk_mq_get_hctx_type() the operation type in
the second argument of blk_mq_get_hctx_type() matters. The introduction
of blk_mq_get_hctx_type() caused blk_mq_get_sq_hctx() to select a
hardware queue of type HCTX_TYPE_READ instead of HCTX_TYPE_DEFAULT.
Switch to hardware queue type HCTX_TYPE_DEFAULT since HCTX_TYPE_READ
should only be used for read requests.

Cc: Ming Lei <ming.lei@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220615225549.1054905-4-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Rename a blk_mq_map_queue() argument
Bart Van Assche [Wed, 15 Jun 2022 22:55:48 +0000 (15:55 -0700)]
block: Rename a blk_mq_map_queue() argument

Before the introduction of blk_mq_get_hctx_type(), blk_mq_map_queue()
only used the flags from its second argument. Since the introduction of
blk_mq_get_hctx_type(), blk_mq_map_queue() uses both the operation and
the flags encoded in that argument. Rename the second argument of
blk_mq_map_queue() to make this clear.

Cc: Christoph Hellwig <hch@lst.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220615225549.1054905-3-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblk-iocost: Simplify ioc_rqos_done()
Bart Van Assche [Wed, 15 Jun 2022 22:55:47 +0000 (15:55 -0700)]
blk-iocost: Simplify ioc_rqos_done()

Leave out the superfluous "& REQ_OP_MASK" code. The definition of req_op()
shows that that code is superfluous:

 #define req_op(req) ((req)->cmd_flags & REQ_OP_MASK)

Compile-tested only.

Cc: Tejun Heo <tj@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220615225549.1054905-2-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: Directly use ida_alloc()/free()
Bo Liu [Wed, 15 Jun 2022 08:18:16 +0000 (04:18 -0400)]
block: Directly use ida_alloc()/free()

Use ida_alloc()/ida_free() instead of
ida_simple_get()/ida_simple_remove().
The latter is deprecated and more verbose.

Signed-off-by: Bo Liu <liubo03@inspur.com>
Reviewed-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/20220615081816.4342-1-liubo03@inspur.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoiomap: add support for dma aligned direct-io
Keith Busch [Fri, 10 Jun 2022 19:58:30 +0000 (12:58 -0700)]
iomap: add support for dma aligned direct-io

Use the address alignment requirements from the block_device for direct
io instead of requiring addresses be aligned to the block size.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-12-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: relax direct io memory alignment
Keith Busch [Fri, 10 Jun 2022 19:58:29 +0000 (12:58 -0700)]
block: relax direct io memory alignment

Use the address alignment requirements from the block_device for direct
io instead of requiring addresses be aligned to the block size. User
space can discover the alignment requirements from the dma_alignment
queue attribute.

User space can specify any hardware compatible DMA offset for each
segment, but every segment length is still required to be a multiple of
the block size.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-11-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: introduce bdev_iter_is_aligned helper
Keith Busch [Fri, 10 Jun 2022 19:58:28 +0000 (12:58 -0700)]
block: introduce bdev_iter_is_aligned helper

Provide a convenient function for this repeatable coding pattern.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-10-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoiov: introduce iov_iter_aligned
Keith Busch [Fri, 10 Jun 2022 19:58:27 +0000 (12:58 -0700)]
iov: introduce iov_iter_aligned

The existing iov_iter_alignment() function returns the logical OR of
address and length. For cases where address and length need to be
considered separately, introduce a helper function that a caller can
specificy length and address masks that indicate if the iov is
unaligned.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-9-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock/bounce: count bytes instead of sectors
Keith Busch [Fri, 10 Jun 2022 19:58:26 +0000 (12:58 -0700)]
block/bounce: count bytes instead of sectors

Individual bv_len's may not be a sector size.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Reviewed-by: Pankaj Raghav <p.raghav@samsung.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-8-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock/merge: count bytes instead of sectors
Keith Busch [Fri, 10 Jun 2022 19:58:25 +0000 (12:58 -0700)]
block/merge: count bytes instead of sectors

Individual bv_len's may not be a sector size.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20220610195830.3574005-7-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: add a helper function for dio alignment
Keith Busch [Fri, 10 Jun 2022 19:58:24 +0000 (12:58 -0700)]
block: add a helper function for dio alignment

This will make it easier to add more complex acceptable alignment
criteria in the future.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-6-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: introduce bdev_dma_alignment helper
Keith Busch [Fri, 10 Jun 2022 19:58:23 +0000 (12:58 -0700)]
block: introduce bdev_dma_alignment helper

Preparing for upcoming dma_alignment users that have a block_device, but
don't need the request_queue.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-5-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: export dma_alignment attribute
Keith Busch [Fri, 10 Jun 2022 19:58:22 +0000 (12:58 -0700)]
block: export dma_alignment attribute

User space may want to know how to align their buffers to avoid
bouncing. Export the queue attribute.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-4-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock/bio: remove duplicate append pages code
Keith Busch [Fri, 10 Jun 2022 19:58:21 +0000 (12:58 -0700)]
block/bio: remove duplicate append pages code

The getting pages setup for zone append and normal IO are identical. Use
common code for each.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220610195830.3574005-3-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoblock: fix infinite loop for invalid zone append
Keith Busch [Fri, 10 Jun 2022 19:58:20 +0000 (12:58 -0700)]
block: fix infinite loop for invalid zone append

Returning 0 early from __bio_iov_append_get_pages() for the
max_append_sectors warning just creates an infinite loop since 0 means
success, and the bio will never fill from the unadvancing iov_iter. We
could turn the return into an error value, but it will already be turned
into an error value later on, so just remove the warning. Clearly no one
ever hit it anyway.

Fixes: 0512a75b98f84 ("block: Introduce REQ_OP_ZONE_APPEND")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20220610195830.3574005-2-kbusch@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
23 months agoLinux 5.19-rc4
Linus Torvalds [Sun, 26 Jun 2022 21:22:10 +0000 (14:22 -0700)]
Linux 5.19-rc4

23 months agoMerge tag 'soc-fixes-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Sun, 26 Jun 2022 21:12:56 +0000 (14:12 -0700)]
Merge tag 'soc-fixes-5.19' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "A number of fixes have accumulated, but they are largely for harmless
  issues:

   - Several OF node leak fixes

   - A fix to the Exynos7885 UART clock description

   - DTS fixes to prevent boot failures on TI AM64 and J721s2

   - Bus probe error handling fixes for Baikal-T1

   - A fixup to the way STM32 SoCs use separate dts files for different
     firmware stacks

   - Multiple code fixes for Arm SCMI firmware, all dealing with
     robustness of the implementation

   - Multiple NXP i.MX devicetree fixes, addressing incorrect data in DT
     nodes

   - Three updates to the MAINTAINERS file, including Florian Fainelli
     taking over BCM283x/BCM2711 (Raspberry Pi) from Nicolas Saenz
     Julienne"

* tag 'soc-fixes-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (29 commits)
  ARM: dts: aspeed: nuvia: rename vendor nuvia to qcom
  arm: mach-spear: Add missing of_node_put() in time.c
  ARM: cns3xxx: Fix refcount leak in cns3xxx_init
  MAINTAINERS: Update email address
  arm64: dts: ti: k3-am64-main: Remove support for HS400 speed mode
  arm64: dts: ti: k3-j721s2: Fix overlapping GICD memory region
  ARM: dts: bcm2711-rpi-400: Fix GPIO line names
  bus: bt1-axi: Don't print error on -EPROBE_DEFER
  bus: bt1-apb: Don't print error on -EPROBE_DEFER
  ARM: Fix refcount leak in axxia_boot_secondary
  ARM: dts: stm32: move SCMI related nodes in a dedicated file for stm32mp15
  soc: imx: imx8m-blk-ctrl: fix display clock for LCDIF2 power domain
  ARM: dts: imx6qdl-colibri: Fix capacitive touch reset polarity
  ARM: dts: imx6qdl: correct PU regulator ramp delay
  firmware: arm_scmi: Fix incorrect error propagation in scmi_voltage_descriptors_get
  firmware: arm_scmi: Avoid using extended string-buffers sizes if not necessary
  firmware: arm_scmi: Fix SENSOR_AXIS_NAME_GET behaviour when unsupported
  ARM: dts: imx7: Move hsic_phy power domain to HSIC PHY node
  soc: bcm: brcmstb: pm: pm-arm: Fix refcount leak in brcmstb_pm_probe
  MAINTAINERS: Update BCM2711/BCM2835 maintainer
  ...

23 months agoMerge tag 'mm-hotfixes-stable-2022-06-26' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Sun, 26 Jun 2022 21:00:55 +0000 (14:00 -0700)]
Merge tag 'mm-hotfixes-stable-2022-06-26' of git://git./linux/kernel/git/akpm/mm

Pull hotfixes from Andrew Morton:
 "Minor things, mainly - mailmap updates, MAINTAINERS updates, etc.

  Fixes for this merge window:

   - fix for a damon boot hang, from SeongJae

   - fix for a kfence warning splat, from Jason Donenfeld

   - fix for zero-pfn pinning, from Alex Williamson

   - fix for fallocate hole punch clearing, from Mike Kravetz

  Fixes for previous releases:

   - fix for a performance regression, from Marcelo

   - fix for a hwpoisining BUG from zhenwei pi"

* tag 'mm-hotfixes-stable-2022-06-26' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
  mailmap: add entry for Christian Marangi
  mm/memory-failure: disable unpoison once hw error happens
  hugetlbfs: zero partial pages during fallocate hole punch
  mm: memcontrol: reference to tools/cgroup/memcg_slabinfo.py
  mm: re-allow pinning of zero pfns
  mm/kfence: select random number before taking raw lock
  MAINTAINERS: add maillist information for LoongArch
  MAINTAINERS: update MM tree references
  MAINTAINERS: update Abel Vesa's email
  MAINTAINERS: add MEMORY HOT(UN)PLUG section and add David as reviewer
  MAINTAINERS: add Miaohe Lin as a memory-failure reviewer
  mailmap: add alias for jarkko@profian.com
  mm/damon/reclaim: schedule 'damon_reclaim_timer' only after 'system_wq' is initialized
  kthread: make it clear that kthread_create_on_node() might be terminated by any fatal signal
  mm: lru_cache_disable: use synchronize_rcu_expedited
  mm/page_isolation.c: fix one kernel-doc comment

23 months agoMerge tag 'perf-tools-fixes-for-v5.19-2022-06-26' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sun, 26 Jun 2022 19:12:25 +0000 (12:12 -0700)]
Merge tag 'perf-tools-fixes-for-v5.19-2022-06-26' of git://git./linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Enable ignore_missing_thread in 'perf stat', enabling counting with
   '--pid' when threads disappear during counting session setup

 - Adjust output data offset for backward compatibility in 'perf inject'

 - Fix missing free in copy_kcore_dir() in 'perf inject'

 - Fix caching files with a wrong build ID

 - Sync drm, cpufeatures, vhost and svn headers with the kernel

* tag 'perf-tools-fixes-for-v5.19-2022-06-26' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  tools headers UAPI: Synch KVM's svm.h header with the kernel
  tools include UAPI: Sync linux/vhost.h with the kernel sources
  perf stat: Enable ignore_missing_thread
  perf inject: Adjust output data offset for backward compatibility
  perf trace beauty: Fix generation of errno id->str table on ALT Linux
  perf build-id: Fix caching files with a wrong build ID
  tools headers cpufeatures: Sync with the kernel sources
  tools headers UAPI: Sync drm/i915_drm.h with the kernel sources
  perf inject: Fix missing free in copy_kcore_dir()

23 months agoMerge tag 'for-5.19-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Sun, 26 Jun 2022 17:11:36 +0000 (10:11 -0700)]
Merge tag 'for-5.19-rc3-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - zoned relocation fixes:
      - fix critical section end for extent writeback, this could lead
        to out of order write
      - prevent writing to previous data relocation block group if space
        gets low

 - reflink fixes:
      - fix race between reflinking and ordered extent completion
      - proper error handling when block reserve migration fails
      - add missing inode iversion/mtime/ctime updates on each iteration
        when replacing extents

 - fix deadlock when running fsync/fiemap/commit at the same time

 - fix false-positive KCSAN report regarding pid tracking for read locks
   and data race

 - minor documentation update and link to new site

* tag 'for-5.19-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  Documentation: update btrfs list of features and link to readthedocs.io
  btrfs: fix deadlock with fsync+fiemap+transaction commit
  btrfs: don't set lock_owner when locking extent buffer for reading
  btrfs: zoned: fix critical section of relocation inode writeback
  btrfs: zoned: prevent allocation from previous data relocation BG
  btrfs: do not BUG_ON() on failure to migrate space when replacing extents
  btrfs: add missing inode updates on each iteration when replacing extents
  btrfs: fix race between reflinking and ordered extent completion

23 months agoMerge tag 'dma-mapping-5.19-2022-06-26' of git://git.infradead.org/users/hch/dma...
Linus Torvalds [Sun, 26 Jun 2022 17:01:40 +0000 (10:01 -0700)]
Merge tag 'dma-mapping-5.19-2022-06-26' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:

 - pass the correct size to dma_set_encrypted() when freeing memory
   (Dexuan Cui)

* tag 'dma-mapping-5.19-2022-06-26' of git://git.infradead.org/users/hch/dma-mapping:
  dma-direct: use the correct size for dma_set_encrypted()