linux-stable.git/drivers/md/raid1.c, branch linux-5.6.y

md/raid1: introduce wait_for_serialization

2020-01-13T19:44:10+00:00

Previously, we call check_and_add_serial when serialization is
enabled for write IO, but it could allocate and free memory
back and forth.

Now, let's just get an element from memory pool with the new
function, then insert node to rb tree if no collision happens.

Signed-off-by: Guoqing Jiang 
Signed-off-by: Song Liu

md/raid1: use bucket based mechanism for IO serialization

2020-01-13T19:44:10+00:00

Since raid1 had already used bucket based mechanism to reduce
the conflict between write IO and resync IO, it is possible to
speed up performance for io serialization with refer to the
same mechanism.

To align with the barrier bucket mechanism, we created arrays
(with the same number of BARRIER_BUCKETS_NR) for spinlock, rb
tree and waitqueue. Then we can reduce lock competition with
multiple spinlocks, boost search performance with multiple rb
trees and also reduce thundering herd problem with multiple
waitqueues.

Signed-off-by: Guoqing Jiang 
Signed-off-by: Song Liu

md: introduce a new struct for IO serialization

2020-01-13T19:44:10+00:00

Obviously, IO serialization could cause the degradation of
performance a lot. In order to reduce the degradation, so a
rb interval tree is added in raid1 to speed up the check of
collision.

So, a rb root is needed in md_rdev, then abstract all the
serialize related members to a new struct (serial_in_rdev),
embed it into md_rdev.

Of course, we need to free the struct if it is not needed
anymore, so rdev/rdevs_uninit_serial are added accordingly.
And they should be called when destroty memory pool or can't
alloc memory.

And we need to consider to call mddev_destroy_serial_pool
in case serialize_policy/write-behind is disabled, bitmap
is destroyed or in __md_stop_writes.

Signed-off-by: Guoqing Jiang 
Signed-off-by: Song Liu

raid1: serialize the overlap write

2020-01-13T19:44:09+00:00

Before dispatch write bio, raid1 array which enables
serialize_policy need to check if overlap exists between
this bio and previous on-flying bios. If there is overlap,
then it has to wait until the collision is disappeared.

Signed-off-by: Guoqing Jiang 
Signed-off-by: Song Liu

md: rename wb stuffs

2020-01-13T19:44:09+00:00

Previously, wb_info_pool and wb_list stuffs are introduced
to address potential data inconsistence issue for write
behind device.

Now rename them to serial related name, since the same
mechanism will be used to address reorder overlap write
issue for raid1.

Signed-off-by: Guoqing Jiang 
Signed-off-by: Song Liu

md: raid1: check rdev before reference in raid1_sync_request func

2019-12-11T18:36:13+00:00

In raid1_sync_request func, rdev should be checked before reference.

Signed-off-by: Zhiqiang Liu 
Signed-off-by: Song Liu

md/raid1: avoid soft lockup under high load

2019-11-12T00:32:22+00:00

As all I/O is being pushed through a kernel thread the softlockup
watchdog might be triggered under high load.

Signed-off-by: Hannes Reinecke 
Signed-off-by: Song Liu

md: improve handling of bio with REQ_PREFLUSH in md_flush_request()

2019-10-24T22:22:40+00:00

If pers->make_request fails in md_flush_request(), the bio is lost. To
fix this, pass back a bool to indicate if the original make_request call
should continue to handle the I/O and instead of assuming the flush logic
will push it to completion.

Convert md_flush_request to return a bool and no longer calls the raid
driver's make_request function.  If the return is true, then the md flush
logic has or will complete the bio and the md make_request call is done.
If false, then the md make_request function needs to keep processing like
it is a normal bio. Let the original call to md_handle_request handle any
need to retry sending the bio to the raid driver's make_request function
should it be needed.

Also mark md_flush_request and the make_request function pointer as
__must_check to issue warnings should these critical return values be
ignored.

Fixes: 2bc13b83e629 ("md: batch flush requests.")
Cc: stable@vger.kernel.org # # v4.19+
Cc: NeilBrown 
Signed-off-by: David Jeffery 
Reviewed-by: Xiao Ni 
Signed-off-by: Song Liu

md/raid1: fail run raid1 array when active disk less than one

2019-09-03T21:52:03+00:00

When run test case:
  mdadm -CR /dev/md1 -l 1 -n 4 /dev/sd[a-d] --assume-clean --bitmap=internal
  mdadm -S /dev/md1
  mdadm -A /dev/md1 /dev/sd[b-c] --run --force

  mdadm --zero /dev/sda
  mdadm /dev/md1 -a /dev/sda

  echo offline > /sys/block/sdc/device/state
  echo offline > /sys/block/sdb/device/state
  sleep 5
  mdadm -S /dev/md1

  echo running > /sys/block/sdb/device/state
  echo running > /sys/block/sdc/device/state
  mdadm -A /dev/md1 /dev/sd[a-c] --run --force

mdadm run fail with kernel message as follow:
[  172.986064] md: kicking non-fresh sdb from array!
[  173.004210] md: kicking non-fresh sdc from array!
[  173.022383] md/raid1:md1: active with 0 out of 4 mirrors
[  173.022406] md1: failed to create bitmap (-5)

In fact, when active disk in raid1 array less than one, we
need to return fail in raid1_run().

Reviewed-by: NeilBrown 
Signed-off-by: Yufen Yu 
Signed-off-by: Song Liu

raid1: factor out a common routine to handle the completion of sync write

2019-08-07T17:25:02+00:00

It's just code clean-up.

Signed-off-by: Hou Tao 
Signed-off-by: Song Liu