Re: [PATCH RFC 09/10] xfs: Update atomic write max size

[John Garry <john.g.garry@oracle.com>]

On 05/02/2025 19:41, Darrick J. Wong wrote:
> On Tue, Feb 04, 2025 at 12:01:26PM +0000, John Garry wrote:
>> Now that CoW-based atomic writes are supported, update the max size of an
>> atomic write.
>>
>> For simplicity, limit at the max of what the mounted bdev can support in
>> terms of atomic write limits. Maybe in future we will have a better way
>> to advertise this optimised limit.
>>
>> In addition, the max atomic write size needs to be aligned to the agsize.
>> Currently when attempting to use HW offload, we  just check that the
>> mapping startblock is aligned. However, that is just the startblock within
>> the AG, and the AG may not be properly aligned to the underlying block
>> device atomic write limits.
>>
>> As such, limit atomic writes to the greatest power-of-2 which fits in an
>> AG, so that aligning to the startblock will be mean that we are also
>> aligned to the disk block.

Right, "startblock" is a bit vague

> 
> I don't understand this sentence -- what are we "aligning to the
> startblock"?  I think you're saying that you want to limit the size of
> untorn writes to the greatest power-of-two factor of the agsize so that
> allocations for an untorn write will always be aligned compatibly with
> the alignment requirements of the storage for an untorn write?

Yes, that's it. I'll borrow your wording :)

> 
>> Signed-off-by: John Garry <john.g.garry@oracle.com>
>> ---
>>   fs/xfs/xfs_iops.c  |  7 ++++++-
>>   fs/xfs/xfs_mount.c | 28 ++++++++++++++++++++++++++++
>>   fs/xfs/xfs_mount.h |  1 +
>>   3 files changed, 35 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
>> index ea79fb246e33..95681d6c2bcd 100644
>> --- a/fs/xfs/xfs_iops.c
>> +++ b/fs/xfs/xfs_iops.c
>> @@ -606,12 +606,17 @@ xfs_get_atomic_write_attr(
>>   	unsigned int		*unit_min,
>>   	unsigned int		*unit_max)
>>   {
>> +	struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
>> +	struct xfs_mount	*mp = ip->i_mount;
>> +
>>   	if (!xfs_inode_can_atomicwrite(ip)) {
>>   		*unit_min = *unit_max = 0;
>>   		return;
>>   	}
>>   
>> -	*unit_min = *unit_max = ip->i_mount->m_sb.sb_blocksize;
>> +	*unit_min = ip->i_mount->m_sb.sb_blocksize;
>> +	*unit_max =  min_t(unsigned int, XFS_FSB_TO_B(mp, mp->awu_max),
>> +					target->bt_bdev_awu_max);
>>   }
>>   
>>   static void
>> diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
>> index 477c5262cf91..4e60347f6b7e 100644
>> --- a/fs/xfs/xfs_mount.c
>> +++ b/fs/xfs/xfs_mount.c
>> @@ -651,6 +651,32 @@ xfs_agbtree_compute_maxlevels(
>>   	levels = max(levels, mp->m_rmap_maxlevels);
>>   	mp->m_agbtree_maxlevels = max(levels, mp->m_refc_maxlevels);
>>   }
>> +static inline void
>> +xfs_mp_compute_awu_max(
> 
> xfs_compute_awu_max() ?

ok

> 
>> +	struct xfs_mount	*mp)
>> +{
>> +	xfs_agblock_t		agsize = mp->m_sb.sb_agblocks;
>> +	xfs_agblock_t		awu_max;
>> +
>> +	if (!xfs_has_reflink(mp)) {
>> +		mp->awu_max = 1;
>> +		return;
>> +	}
>> +
>> +	/*
>> +	 * Find highest power-of-2 evenly divisible into agsize and which
>> +	 * also fits into an unsigned int field.
>> +	 */
>> +	awu_max = 1;
>> +	while (1) {
>> +		if (agsize % (awu_max * 2))
>> +			break;
>> +		if (XFS_FSB_TO_B(mp, awu_max * 2) > UINT_MAX)
>> +			break;
>> +		awu_max *= 2;
>> +	}
>> +	mp->awu_max = awu_max;
> 
> I think you need two awu_maxes here -- one for the data device, and
> another for the realtime device.
How about we just don't support rtdev initially for this CoW-based 
method, i.e. stick at 1x FSB awu max?

 >  The rt computation is probably more
 > complex since I think it's the greatest power of two that fits in the rt
 > extent size if it isn't a power of two;> or the greatest power of 
two> that fits in the rtgroup if rtgroups are enabled; or probably just no
 > limit otherwise.
 >

Thanks,
John