Re: [PATCH v6 4/7] xfs: Support FS_XFLAG_ATOMICWRITES

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

>>   }
>>   
>> +static xfs_failaddr_t
>> +xfs_inode_validate_atomicwrites(
>> +	struct xfs_mount	*mp,
>> +	uint32_t		cowextsize,
>> +	uint16_t		mode,
>> +	int64_t			flags2)
>> +{
>> +	/* superblock rocompat feature flag */
>> +	if (!xfs_has_atomicwrites(mp))
>> +		return __this_address;
>> +
>> +	/* Only regular files and directories */
>> +	if (!S_ISREG(mode) && !(S_ISDIR(mode)))
>> +		return __this_address;
>> +
>> +	/* COW extsize disallowed */
>> +	if (flags2 & XFS_DIFLAG2_COWEXTSIZE)
>> +		return __this_address;
>> +
>> +	/* cowextsize must be zero */
>> +	if (cowextsize)
>> +		return __this_address;
>> +
>> +	/* reflink is disallowed */
>> +	if (flags2 & XFS_DIFLAG2_REFLINK)
>> +		return __this_address;
> 
> If we're only allowing atomic writes that are 1 fsblock or less, then
> copy on write will work correctly because CoWs are always done with
> fsblock granularity.  The ioend remap is also committed atomically.
> 
> IOWs, it's forcealign that isn't compatible with reflink and you can
> drop this incompatibility.

ok, understood

> 
>> +
>> +	return NULL;
>> +}
>> +
>>   xfs_failaddr_t
>>   xfs_dinode_verify(
>>   	struct xfs_mount	*mp,
>> @@ -663,6 +693,14 @@ xfs_dinode_verify(
>>   	    !xfs_has_bigtime(mp))
>>   		return __this_address;
>>   
>> +	if (flags2 & XFS_DIFLAG2_ATOMICWRITES) {
>> +		fa = xfs_inode_validate_atomicwrites(mp,
>> +				be32_to_cpu(dip->di_cowextsize),
> 
> Technically speaking, the space used by di_cowextsize isn't defined on
> !reflink filesystems.  The contents are supposed to be zero, but nobody
> actually checks that, so you might want to special case this:
> 
> 		fa = xfs_inode_validate_atomicwrites(mp,
> 				xfs_has_reflink(mp) ?
> 					be32_to_cpu(dip->di_cowextsize) : 0,
> 				mode, flags2);
> 
> (inasmuch as this code is getting ugly and maybe you want to use a
> temporary variable)

As discussed later, I will drop the cowextsize check (so need to pass 
this at all)

> 
>> +				mode, flags2);
>> +		if (fa)
>> +			return fa;
>> +	}
>> +
>>   	return NULL;
>>   }
>>   
>> diff --git a/fs/xfs/libxfs/xfs_inode_util.c b/fs/xfs/libxfs/xfs_inode_util.c
>> index cc38e1c3c3e1..e59e98783bf7 100644
>> --- a/fs/xfs/libxfs/xfs_inode_util.c
>> +++ b/fs/xfs/libxfs/xfs_inode_util.c
>> @@ -80,6 +80,8 @@ xfs_flags2diflags2(
>>   		di_flags2 |= XFS_DIFLAG2_DAX;
>>   	if (xflags & FS_XFLAG_COWEXTSIZE)
>>   		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
>> +	if (xflags & FS_XFLAG_ATOMICWRITES)
>> +		di_flags2 |= XFS_DIFLAG2_ATOMICWRITES;
>>   
>>   	return di_flags2;
>>   }
>> @@ -126,6 +128,8 @@ xfs_ip2xflags(
>>   			flags |= FS_XFLAG_DAX;
>>   		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
>>   			flags |= FS_XFLAG_COWEXTSIZE;
>> +		if (ip->i_diflags2 & XFS_DIFLAG2_ATOMICWRITES)
>> +			flags |= FS_XFLAG_ATOMICWRITES;
>>   	}
>>   
>>   	if (xfs_inode_has_attr_fork(ip))
>> @@ -224,6 +228,8 @@ xfs_inode_inherit_flags2(
>>   	}
>>   	if (pip->i_diflags2 & XFS_DIFLAG2_DAX)
>>   		ip->i_diflags2 |= XFS_DIFLAG2_DAX;
>> +	if (pip->i_diflags2 & XFS_DIFLAG2_ATOMICWRITES)
>> +		ip->i_diflags2 |= XFS_DIFLAG2_ATOMICWRITES;
>>   
>>   	/* Don't let invalid cowextsize hints propagate. */
>>   	failaddr = xfs_inode_validate_cowextsize(ip->i_mount, ip->i_cowextsize,
>> diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c
>> index d95409f3cba6..dd819561d0a5 100644
>> --- a/fs/xfs/libxfs/xfs_sb.c
>> +++ b/fs/xfs/libxfs/xfs_sb.c
>> @@ -164,6 +164,8 @@ xfs_sb_version_to_features(
>>   		features |= XFS_FEAT_REFLINK;
>>   	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
>>   		features |= XFS_FEAT_INOBTCNT;
>> +	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_ATOMICWRITES)
>> +		features |= XFS_FEAT_ATOMICWRITES;
>>   	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_FTYPE)
>>   		features |= XFS_FEAT_FTYPE;
>>   	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES)
>> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
>> index aa4dbda7b536..44bee3e2b2bb 100644
>> --- a/fs/xfs/xfs_buf.c
>> +++ b/fs/xfs/xfs_buf.c
>> @@ -2060,6 +2060,8 @@ int
>>   xfs_init_buftarg(
>>   	struct xfs_buftarg		*btp,
>>   	size_t				logical_sectorsize,
>> +	unsigned int			awu_min,
>> +	unsigned int			awu_max,
>>   	const char			*descr)
>>   {
>>   	/* Set up device logical sector size mask */
>> @@ -2086,6 +2088,9 @@ xfs_init_buftarg(
>>   	btp->bt_shrinker->scan_objects = xfs_buftarg_shrink_scan;
>>   	btp->bt_shrinker->private_data = btp;
>>   	shrinker_register(btp->bt_shrinker);
>> +
>> +	btp->bt_bdev_awu_min = awu_min;
>> +	btp->bt_bdev_awu_max = awu_max;
>>   	return 0;
>>   
>>   out_destroy_io_count:
>> @@ -2102,6 +2107,7 @@ xfs_alloc_buftarg(
>>   {
>>   	struct xfs_buftarg	*btp;
>>   	const struct dax_holder_operations *ops = NULL;
>> +	unsigned int awu_min = 0, awu_max = 0;
>>   
>>   #if defined(CONFIG_FS_DAX) && defined(CONFIG_MEMORY_FAILURE)
>>   	ops = &xfs_dax_holder_operations;
>> @@ -2115,6 +2121,13 @@ xfs_alloc_buftarg(
>>   	btp->bt_daxdev = fs_dax_get_by_bdev(btp->bt_bdev, &btp->bt_dax_part_off,
>>   					    mp, ops);
>>   
>> +	if (bdev_can_atomic_write(btp->bt_bdev)) {
>> +		struct request_queue *q = bdev_get_queue(btp->bt_bdev);
>> +
>> +		awu_min = queue_atomic_write_unit_min_bytes(q);
>> +		awu_max = queue_atomic_write_unit_max_bytes(q);
>> +	}
>> +
>>   	/*
>>   	 * When allocating the buftargs we have not yet read the super block and
>>   	 * thus don't know the file system sector size yet.
>> @@ -2122,7 +2135,7 @@ xfs_alloc_buftarg(
>>   	if (xfs_setsize_buftarg(btp, bdev_logical_block_size(btp->bt_bdev)))
>>   		goto error_free;
>>   	if (xfs_init_buftarg(btp, bdev_logical_block_size(btp->bt_bdev),
>> -			mp->m_super->s_id))
>> +			awu_min, awu_max, mp->m_super->s_id))
>>   		goto error_free;
> 
> Rather than passing this into the constructor and making the xmbuf code
> pass zeroes, why not set the awu values here in xfs_alloc_buftarg just
> before returning btp?

ok, I can do that instead.

Thanks,
John