Skip to content

L
linux
Switch branch/tag
  1. 28 Aug, 2020 11 commits
    • Sagi Grimberg's avatar
      nvme-rdma: fix reset hang if controller died in the middle of a reset · 2362acb6
      Sagi Grimberg authored 
      If the controller becomes unresponsive in the middle of a reset, we
will hang because we are waiting for the freeze to complete, but that
cannot happen since we have commands that are inflight holding the
q_usage_counter, and we can't blindly fail requests that times out.

So give a timeout and if we cannot wait for queue freeze before
unfreezing, fail and have the error handling take care how to
proceed (either schedule a reconnect of remove the controller).
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      2362acb6
    • Sagi Grimberg's avatar
      nvme-rdma: fix timeout handler · 0475a8dc
      Sagi Grimberg authored 
      When a request times out in a LIVE state, we simply trigger error
recovery and let the error recovery handle the request cancellation,
however when a request times out in a non LIVE state, we make sure to
complete it immediately as it might block controller setup or teardown
and prevent forward progress.

However tearing down the entire set of I/O and admin queues causes
freeze/unfreeze imbalance (q->mq_freeze_depth) because and is really
an overkill to what we actually need, which is to just fence controller
teardown that may be running, stop the queue, and cancel the request if
it is not already completed.

Now that we have the controller teardown_lock, we can safely serialize
request cancellation. This addresses a hang caused by calling extra
queue freeze on controller namespaces, causing unfreeze to not complete
correctly.
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Reviewed-by: default avatarJames Smart <james.smart@broadcom.com>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      0475a8dc
    • Sagi Grimberg's avatar
      nvme-rdma: serialize controller teardown sequences · 5110f402
      Sagi Grimberg authored 
      In the timeout handler we may need to complete a request because the
request that timed out may be an I/O that is a part of a serial sequence
of controller teardown or initialization. In order to complete the
request, we need to fence any other context that may compete with us
and complete the request that is timing out.

In this case, we could have a potential double completion in case
a hard-irq or a different competing context triggered error recovery
and is running inflight request cancellation concurrently with the
timeout handler.

Protect using a ctrl teardown_lock to serialize contexts that may
complete a cancelled request due to error recovery or a reset.
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Reviewed-by: default avatarJames Smart <james.smart@broadcom.com>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      5110f402
    • Sagi Grimberg's avatar
      nvme-tcp: fix reset hang if controller died in the middle of a reset · e5c01f4f
      Sagi Grimberg authored 
      If the controller becomes unresponsive in the middle of a reset, we will
hang because we are waiting for the freeze to complete, but that cannot
happen since we have commands that are inflight holding the
q_usage_counter, and we can't blindly fail requests that times out.

So give a timeout and if we cannot wait for queue freeze before
unfreezing, fail and have the error handling take care how to proceed
(either schedule a reconnect of remove the controller).
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      e5c01f4f
    • Sagi Grimberg's avatar
      nvme-tcp: fix timeout handler · 236187c4
      Sagi Grimberg authored 
      When a request times out in a LIVE state, we simply trigger error
recovery and let the error recovery handle the request cancellation,
however when a request times out in a non LIVE state, we make sure to
complete it immediately as it might block controller setup or teardown
and prevent forward progress.

However tearing down the entire set of I/O and admin queues causes
freeze/unfreeze imbalance (q->mq_freeze_depth) because and is really
an overkill to what we actually need, which is to just fence controller
teardown that may be running, stop the queue, and cancel the request if
it is not already completed.

Now that we have the controller teardown_lock, we can safely serialize
request cancellation. This addresses a hang caused by calling extra
queue freeze on controller namespaces, causing unfreeze to not complete
correctly.
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      236187c4
    • Sagi Grimberg's avatar
      nvme-tcp: serialize controller teardown sequences · d4d61470
      Sagi Grimberg authored 
      In the timeout handler we may need to complete a request because the
request that timed out may be an I/O that is a part of a serial sequence
of controller teardown or initialization. In order to complete the
request, we need to fence any other context that may compete with us
and complete the request that is timing out.

In this case, we could have a potential double completion in case
a hard-irq or a different competing context triggered error recovery
and is running inflight request cancellation concurrently with the
timeout handler.

Protect using a ctrl teardown_lock to serialize contexts that may
complete a cancelled request due to error recovery or a reset.
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      d4d61470
    • Sagi Grimberg's avatar
      nvme: have nvme_wait_freeze_timeout return if it timed out · 7cf0d7c0
      Sagi Grimberg authored 
      Users can detect if the wait has completed or not and take appropriate
actions based on this information (e.g. weather to continue
initialization or rather fail and schedule another initialization
attempt).
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      7cf0d7c0
    • Sagi Grimberg's avatar
      nvme-fabrics: don't check state NVME_CTRL_NEW for request acceptance · d7144f5c
      Sagi Grimberg authored 
      NVME_CTRL_NEW should never see any I/O, because in order to start
initialization it has to transition to NVME_CTRL_CONNECTING and from
there it will never return to this state.
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      d7144f5c
    • Ziye Yang's avatar
      nvmet-tcp: Fix NULL dereference when a connect data comes in h2cdata pdu · a6ce7d7b
      Ziye Yang authored 
      When handling commands without in-capsule data, we assign the ttag
assuming we already have the queue commands array allocated (based
on the queue size information in the connect data payload). However
if the connect itself did not send the connect data in-capsule we
have yet to allocate the queue commands,and we will assign a bogus
ttag and suffer a NULL dereference when we receive the corresponding
h2cdata pdu.

Fix this by checking if we already allocated commands before
dereferencing it when handling h2cdata, if we didn't, its for sure a
connect and we should use the preallocated connect command.
Signed-off-by: default avatarZiye Yang <ziye.yang@intel.com>
Signed-off-by: default avatarSagi Grimberg <sagi@grimberg.me>
      a6ce7d7b
    • Jens Axboe's avatar
      Merge branch 'md-fixes' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md into block-5.9 · a433d721
      Jens Axboe authored 
      Pull MD fix from Song.

* 'md-fixes' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md:
  md/raid5: make sure stripe_size as power of two
      a433d721
    • Yufen Yu's avatar
      md/raid5: make sure stripe_size as power of two · 6af10a33
      Yufen Yu authored 
      Commit 3b5408b9 ("md/raid5: support config stripe_size by sysfs
entry") make stripe_size as a configurable value. It just requires
stripe_size as multiple of 4KB.

In fact, we should make sure stripe_size as power of two. Otherwise,
stripe_shift which is the result of ilog2 can not represent the real
stripe_size. Then, stripe_hash() and stripe_hash_locks_hash() may
get unexpected value.

Fixes: 3b5408b9 ("md/raid5: support config stripe_size by sysfs entry")
Signed-off-by: default avatarYufen Yu <yuyufen@huawei.com>
Signed-off-by: default avatarSong Liu <songliubraving@fb.com>
      6af10a33
  3. 26 Aug, 2020 2 commits
  5. 21 Aug, 2020 23 commits
  7. 18 Aug, 2020 2 commits
    • Dmitry Monakhov's avatar
      bfq: fix blkio cgroup leakage v4 · 2de791ab
      Dmitry Monakhov authored 
      Changes from v1:
    - update commit description with proper ref-accounting justification

commit db37a34c ("block, bfq: get a ref to a group when adding it to a service tree")
introduce leak forbfq_group and blkcg_gq objects because of get/put
imbalance.
In fact whole idea of original commit is wrong because bfq_group entity
can not dissapear under us because it is referenced by child bfq_queue's
entities from here:
 -> bfq_init_entity()
    ->bfqg_and_blkg_get(bfqg);
    ->entity->parent = bfqg->my_entity

 -> bfq_put_queue(bfqq)
    FINAL_PUT
    ->bfqg_and_blkg_put(bfqq_group(bfqq))
    ->kmem_cache_free(bfq_pool, bfqq);

So parent entity can not disappear while child entity is in tree,
and child entities already has proper protection.
This patch revert commit db37a34c ("block, bfq: get a ref to a group when adding it to a service tree")

bfq_group leak trace caused by bad commit:
-> blkg_alloc
   -> bfq_pq_alloc
     -> bfqg_get (+1)
->bfq_activate_bfqq
  ->bfq_activate_requeue_entity
    -> __bfq_activate_entity
       ->bfq_get_entity
         ->bfqg_and_blkg_get (+1)  <==== : Note1
->bfq_del_bfqq_busy
  ->bfq_deactivate_entity+0x53/0xc0 [bfq]
    ->__bfq_deactivate_entity+0x1b8/0x210 [bfq]
      -> bfq_forget_entity(is_in_service = true)
	 entity->on_st_or_in_serv = false   <=== :Note2
	 if (is_in_service)
	     return;  ==> do not touch reference
-> blkcg_css_offline
 -> blkcg_destroy_blkgs
  -> blkg_destroy
   -> bfq_pd_offline
    -> __bfq_deactivate_entity
         if (!entity->on_st_or_in_serv) /* true, because (Note2)
		return false;
 -> bfq_pd_free
    -> bfqg_put() (-1, byt bfqg->ref == 2) because of (Note2)
So bfq_group and blkcg_gq  will leak forever, see test-case below.

##TESTCASE_BEGIN:
#!/bin/bash

max_iters=${1:-100}
#prep cgroup mounts
mount -t tmpfs cgroup_root /sys/fs/cgroup
mkdir /sys/fs/cgroup/blkio
mount -t cgroup -o blkio none /sys/fs/cgroup/blkio

# Prepare blkdev
grep blkio /proc/cgroups
truncate -s 1M img
losetup /dev/loop0 img
echo bfq > /sys/block/loop0/queue/scheduler

grep blkio /proc/cgroups
for ((i=0;i<max_iters;i++))
do
    mkdir -p /sys/fs/cgroup/blkio/a
    echo 0 > /sys/fs/cgroup/blkio/a/cgroup.procs
    dd if=/dev/loop0 bs=4k count=1 of=/dev/null iflag=direct 2> /dev/null
    echo 0 > /sys/fs/cgroup/blkio/cgroup.procs
    rmdir /sys/fs/cgroup/blkio/a
    grep blkio /proc/cgroups
done
##TESTCASE_END:

Fixes: db37a34c ("block, bfq: get a ref to a group when adding it to a service tree")
Tested-by: default avatarOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: default avatarDmitry Monakhov <dmtrmonakhov@yandex-team.ru>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
      2de791ab
    • Matthew Wilcox (Oracle)'s avatar
      block: Fix page_is_mergeable() for compound pages · d8166519
      Matthew Wilcox (Oracle) authored 
      If we pass in an offset which is larger than PAGE_SIZE, then
page_is_mergeable() thinks it's not mergeable with the previous bio_vec,
leading to a large number of bio_vecs being used.  Use a slightly more
obvious test that the two pages are compatible with each other.

Fixes: 52d52d1c ("block: only allow contiguous page structs in a bio_vec")
Signed-off-by: default avatarMatthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: default avatarMing Lei <ming.lei@redhat.com>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
      d8166519
  9. 17 Aug, 2020 2 commits
    • Ming Lei's avatar
      block: virtio_blk: fix handling single range discard request · af822aa6
      Ming Lei authored 
      1f23816b ("virtio_blk: add discard and write zeroes support") starts
to support multi-range discard for virtio-blk. However, the virtio-blk
disk may report max discard segment as 1, at least that is exactly what
qemu is doing.

So far, block layer switches to normal request merge if max discard segment
limit is 1, and multiple bios can be merged to single segment. This way may
cause memory corruption in virtblk_setup_discard_write_zeroes().

Fix the issue by handling single max discard segment in straightforward
way.

Fixes: 1f23816b ("virtio_blk: add discard and write zeroes support")
Signed-off-by: default avatarMing Lei <ming.lei@redhat.com>
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Cc: Changpeng Liu <changpeng.liu@intel.com>
Cc: Daniel Verkamp <dverkamp@chromium.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
      af822aa6
    • Ming Lei's avatar
      block: respect queue limit of max discard segment · 943b40c8
      Ming Lei authored 
      When queue_max_discard_segments(q) is 1, blk_discard_mergable() will
return false for discard request, then normal request merge is applied.
However, only queue_max_segments() is checked, so max discard segment
limit isn't respected.

Check max discard segment limit in the request merge code for fixing
the issue.

Discard request failure of virtio_blk is fixed.

Fixes: 69840466 ("block: fix the DISCARD request merge")
Signed-off-by: default avatarMing Lei <ming.lei@redhat.com>
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Cc: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
      943b40c8
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7