Message-ID: <bb1f66f4-da09-4841-a7f7-839047b1fe32@huaweicloud.com>
Date: Mon, 25 Aug 2025 19:25:16 +0800
From: Zhang Yi <yi.zhang@...weicloud.com>
To: Paul Menzel <pmenzel@...gen.mpg.de>
Cc: linux-block@...r.kernel.org, linux-raid@...r.kernel.org,
 drbd-dev@...ts.linbit.com, linux-fsdevel@...r.kernel.org,
 linux-kernel@...r.kernel.org, john.g.garry@...cle.com, hch@....de,
 martin.petersen@...cle.com, axboe@...nel.dk, yi.zhang@...wei.com,
 yukuai3@...wei.com, yangerkun@...wei.com
Subject: Re: [PATCH 1/2] md: init queue_limits->max_hw_wzeroes_unmap_sectors
 parameter

Hi, Paul!

On 8/25/2025 4:59 PM, Paul Menzel wrote:
> Dear Yi,
> 
> 
> Thank you for your patch.
> 
> Am 25.08.25 um 10:33 schrieb Zhang Yi:
>> From: Zhang Yi <yi.zhang@...wei.com>
>>
>> The parameter max_hw_wzeroes_unmap_sectors in queue_limits should be
>> equal to max_write_zeroes_sectors if it is set to a non-zero value.
> 
> Excuse my ignorance, but why?

Currently, the max_hw_wzeroes_unmap_sectors parameter is used only to
determine whether the backend device supports the "unmap write zeroes"
operation. If it is set to a non-zero value, it indicates that the
device supports this operation. It depends on the device supports the
write zeroes command, which means that the max_write_zeroes_sectors
should not be zero. However, we do not use this specific value, so the
max_hw_wzeroes_unmap_sectors can only have one of two values:
max_write_zeroes_sectors or 0, any other value is meaningless.

> 
>> However, the stacked md drivers call md_init_stacking_limits() to
>> initialize this parameter to UINT_MAX but only adjust
>> max_write_zeroes_sectors when setting limits. Therefore, this
>> discrepancy triggers a value check failure in blk_validate_limits().
>>
>> Fix this failure by explicitly setting max_hw_wzeroes_unmap_sectors to
>> zero.
> 
> In `linear_set_limits()` and `raid0_set_limits()` you set it to `mddev->chunk_sectors`. Is that intentional?

Yes, the linear and raid0 drivers can support unmap write zeroes
operation if all of the backend devices supports it, so we can initialize
it to chunk_sectors (the same to max_write_zeroes_sectors). raid1/10/5
drivers doesn't support write zeroes, so we have to set it to zero.

> 
>> Fixes: 0c40d7cb5ef3 ("block: introduce max_{hw|user}_wzeroes_unmap_sectors to queue limits")
>> Reported-by: John Garry <john.g.garry@...cle.com>
>> Closes: https://lore.kernel.org/linux-block/803a2183-a0bb-4b7a-92f1-afc5097630d2@oracle.com/
> 
> It’d be great if you added the test case to the commit message.

Yeah, I will add a test to blktests.

Thanks,
Yi.

> 
>> Signed-off-by: Zhang Yi <yi.zhang@...wei.com>
>> ---
>>   drivers/md/md-linear.c | 1 +
>>   drivers/md/raid0.c     | 1 +
>>   drivers/md/raid1.c     | 1 +
>>   drivers/md/raid10.c    | 1 +
>>   drivers/md/raid5.c     | 1 +
>>   5 files changed, 5 insertions(+)
>>
>> diff --git a/drivers/md/md-linear.c b/drivers/md/md-linear.c
>> index 5d9b08115375..3e1f165c2d20 100644
>> --- a/drivers/md/md-linear.c
>> +++ b/drivers/md/md-linear.c
>> @@ -73,6 +73,7 @@ static int linear_set_limits(struct mddev *mddev)
>>       md_init_stacking_limits(&lim);
>>       lim.max_hw_sectors = mddev->chunk_sectors;
>>       lim.max_write_zeroes_sectors = mddev->chunk_sectors;
>> +    lim.max_hw_wzeroes_unmap_sectors = mddev->chunk_sectors;
>>       lim.io_min = mddev->chunk_sectors << 9;
>>       err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY);
>>       if (err)
>> diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
>> index f1d8811a542a..419139ad7663 100644
>> --- a/drivers/md/raid0.c
>> +++ b/drivers/md/raid0.c
>> @@ -382,6 +382,7 @@ static int raid0_set_limits(struct mddev *mddev)
>>       md_init_stacking_limits(&lim);
>>       lim.max_hw_sectors = mddev->chunk_sectors;
>>       lim.max_write_zeroes_sectors = mddev->chunk_sectors;
>> +    lim.max_hw_wzeroes_unmap_sectors = mddev->chunk_sectors;
>>       lim.io_min = mddev->chunk_sectors << 9;
>>       lim.io_opt = lim.io_min * mddev->raid_disks;
>>       lim.chunk_sectors = mddev->chunk_sectors;
>> diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
>> index 408c26398321..35c6498b4917 100644
>> --- a/drivers/md/raid1.c
>> +++ b/drivers/md/raid1.c
>> @@ -3211,6 +3211,7 @@ static int raid1_set_limits(struct mddev *mddev)
>>         md_init_stacking_limits(&lim);
>>       lim.max_write_zeroes_sectors = 0;
>> +    lim.max_hw_wzeroes_unmap_sectors = 0;
>>       lim.features |= BLK_FEAT_ATOMIC_WRITES;
>>       err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY);
>>       if (err)
>> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
>> index b60c30bfb6c7..9832eefb2f15 100644
>> --- a/drivers/md/raid10.c
>> +++ b/drivers/md/raid10.c
>> @@ -4008,6 +4008,7 @@ static int raid10_set_queue_limits(struct mddev *mddev)
>>         md_init_stacking_limits(&lim);
>>       lim.max_write_zeroes_sectors = 0;
>> +    lim.max_hw_wzeroes_unmap_sectors = 0;
>>       lim.io_min = mddev->chunk_sectors << 9;
>>       lim.chunk_sectors = mddev->chunk_sectors;
>>       lim.io_opt = lim.io_min * raid10_nr_stripes(conf);
>> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
>> index 023649fe2476..e385ef1355e8 100644
>> --- a/drivers/md/raid5.c
>> +++ b/drivers/md/raid5.c
>> @@ -7732,6 +7732,7 @@ static int raid5_set_limits(struct mddev *mddev)
>>       lim.features |= BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE;
>>       lim.discard_granularity = stripe;
>>       lim.max_write_zeroes_sectors = 0;
>> +    lim.max_hw_wzeroes_unmap_sectors = 0;
>>       mddev_stack_rdev_limits(mddev, &lim, 0);
>>       rdev_for_each(rdev, mddev)
>>           queue_limits_stack_bdev(&lim, rdev->bdev, rdev->new_data_offset,

Powered by blists - more mailing lists