[GIT PULL] md-6.10 20240425

[GIT PULL] md-6.10 20240425

[Song Liu]

Hi Jens, 

Please consider pulling the following changes for md-6.10 on top of your
for-6.10/block branch. These changes contain various fixes by Yu Kuai,
Li Nan, and Florian-Ewald Mueller. 

Please note that change "md: Fix overflow in is_mddev_idle" changes 
gendisk->sync_io from blkdev.h. This is only used by md, so I included 
this change in this pull request. 

Thanks,
Song




The following changes since commit 688c8b9208356eb5c3fa8047f3e35666f3049a4d:

 blk-cgroup: use group allocation/free of per-cpu counters API (2024-04-03 09:10:17 -0600)

are available in the Git repository at:

 https://git.kernel.org/pub/scm/linux/kernel/git/song/md.git tags/md-6.10-20240425

for you to fetch changes up to 9d1110f99c253ccef82e480bfe9f38a12eb797a7:

 md: don't account sync_io if iostats of the disk is disabled (2024-04-08 21:15:46 -0700)

----------------------------------------------------------------
Florian-Ewald Mueller (1):
     md: add check for sleepers in md_wakeup_thread()

Li Nan (2):
     md: Fix overflow in is_mddev_idle
     md: don't account sync_io if iostats of the disk is disabled

Yu Kuai (1):
     md/raid5: fix deadlock that raid5d() wait for itself to clear MD_SB_CHANGE_PENDING

drivers/md/md.c        | 14 ++++++++++----
drivers/md/md.h        |  5 +++--
drivers/md/raid5.c     | 15 +++------------
include/linux/blkdev.h |  2 +-
4 files changed, 17 insertions(+), 19 deletions(-)

Re: [GIT PULL] md-6.10 20240425

[Jens Axboe]

On 4/25/24 1:51 PM, Song Liu wrote:
> Hi Jens, 
> 
> Please consider pulling the following changes for md-6.10 on top of your
> for-6.10/block branch. These changes contain various fixes by Yu Kuai,
> Li Nan, and Florian-Ewald Mueller. 
> 
> Please note that change "md: Fix overflow in is_mddev_idle" changes 
> gendisk->sync_io from blkdev.h. This is only used by md, so I included 
> this change in this pull request. 

A bit dubious to just bump the counts to 64-bit, I think. You only care
about the difference, so regular integer math should suffice. In fact
you only care about the difference. Feels like a bit of a lazy fix to be
honest.

I've pulled this, but I would strongly suggest to re-visit this change
and spend a bit more time getting this right, rather than just blindly
bumping everything to a 64-bit value. I don't care about the md bits
here, just the gendisk side.

-- 
Jens Axboe

Re: [GIT PULL] md-6.10 20240425

[Song Liu]

Hi Jens,

> On Apr 25, 2024, at 1:02 PM, Jens Axboe <axboe@kernel.dk> wrote:
> 
> On 4/25/24 1:51 PM, Song Liu wrote:
>> Hi Jens, 
>> 
>> Please consider pulling the following changes for md-6.10 on top of your
>> for-6.10/block branch. These changes contain various fixes by Yu Kuai,
>> Li Nan, and Florian-Ewald Mueller. 
>> 
>> Please note that change "md: Fix overflow in is_mddev_idle" changes 
>> gendisk->sync_io from blkdev.h. This is only used by md, so I included 
>> this change in this pull request.
> 
> A bit dubious to just bump the counts to 64-bit, I think. You only care
> about the difference, so regular integer math should suffice. In fact
> you only care about the difference. Feels like a bit of a lazy fix to be
> honest.
> 
> I've pulled this, but I would strongly suggest to re-visit this change
> and spend a bit more time getting this right, rather than just blindly
> bumping everything to a 64-bit value. I don't care about the md bits
> here, just the gendisk side.

Thanks for the advice! We will look more into this and come up with a
follow-up or update. 

Song
 

Re: [GIT PULL] md-6.10 20240425

[Song Liu]

Hi Jens,

> On Apr 25, 2024, at 1:02 PM, Jens Axboe <axboe@kernel.dk> wrote:
> 
> On 4/25/24 1:51 PM, Song Liu wrote:
>> Hi Jens, 
>> 
>> Please consider pulling the following changes for md-6.10 on top of your
>> for-6.10/block branch. These changes contain various fixes by Yu Kuai,
>> Li Nan, and Florian-Ewald Mueller. 
>> 
>> Please note that change "md: Fix overflow in is_mddev_idle" changes 
>> gendisk->sync_io from blkdev.h. This is only used by md, so I included 
>> this change in this pull request.
> 
> A bit dubious to just bump the counts to 64-bit, I think. You only care
> about the difference, so regular integer math should suffice. In fact
> you only care about the difference. Feels like a bit of a lazy fix to be
> honest.
> 
> I've pulled this, but I would strongly suggest to re-visit this change
> and spend a bit more time getting this right, rather than just blindly
> bumping everything to a 64-bit value. I don't care about the md bits
> here, just the gendisk side.

Thanks for the advice! We will look more into this and come up with a
follow-up or update. 

Song