Re: [PATCH v3] xfs: Check for delayed allocations before setting extsize

["Darrick J. Wong" <djwong@kernel.org>]

On Mon, Oct 14, 2024 at 03:02:45PM +0530, Ojaswin Mujoo wrote:
> On Fri, Oct 11, 2024 at 09:40:57AM -0700, Darrick J. Wong wrote:
> > On Fri, Oct 11, 2024 at 09:38:30AM -0700, Darrick J. Wong wrote:
> > > On Fri, Oct 11, 2024 at 08:24:27PM +0530, Ojaswin Mujoo wrote:
> > > > Extsize is allowed to be set on files with no data in it. For this,
> > > > we were checking if the files have extents but missed to check if
> > > > delayed extents were present. This patch adds that check.
> > > > 
> > > > While we are at it, also refactor this check into a helper since
> > > > its used in some other places as well like xfs_inactive() or
> > > > xfs_ioctl_setattr_xflags()
> > > > 
> > > > **Without the patch (SUCCEEDS)**
> > > > 
> > > > $ xfs_io -c 'open -f testfile' -c 'pwrite 0 1024' -c 'extsize 65536'
> > > > 
> > > > wrote 1024/1024 bytes at offset 0
> > > > 1 KiB, 1 ops; 0.0002 sec (4.628 MiB/sec and 4739.3365 ops/sec)
> > > > 
> > > > **With the patch (FAILS as expected)**
> > > > 
> > > > $ xfs_io -c 'open -f testfile' -c 'pwrite 0 1024' -c 'extsize 65536'
> > > > 
> > > > wrote 1024/1024 bytes at offset 0
> > > > 1 KiB, 1 ops; 0.0002 sec (4.628 MiB/sec and 4739.3365 ops/sec)
> > > > xfs_io: FS_IOC_FSSETXATTR testfile: Invalid argument
> > > > 
> > > > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > > > Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
> > > 
> > > Looks good now,
> > > Reviewed-by: Darrick J. Wong <djwong@kernel.org>
> > 
> > That said, could you add a fixes tag for the xfs_ioctl_setattr_*
> > changes, please?
> 
> Actually a small doubt Darrick regarding the Fixes commit (asked inline
> below):
> 
> > 
> > --D
> > 
> > > --D
> > > 
> > > > ---
> > > >  fs/xfs/xfs_inode.c | 2 +-
> > > >  fs/xfs/xfs_inode.h | 5 +++++
> > > >  fs/xfs/xfs_ioctl.c | 4 ++--
> > > >  3 files changed, 8 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> > > > index bcc277fc0a83..19dcb569a3e7 100644
> > > > --- a/fs/xfs/xfs_inode.c
> > > > +++ b/fs/xfs/xfs_inode.c
> > > > @@ -1409,7 +1409,7 @@ xfs_inactive(
> > > >  
> > > >  	if (S_ISREG(VFS_I(ip)->i_mode) &&
> > > >  	    (ip->i_disk_size != 0 || XFS_ISIZE(ip) != 0 ||
> > > > -	     ip->i_df.if_nextents > 0 || ip->i_delayed_blks > 0))
> > > > +	     xfs_inode_has_filedata(ip)))
> > > >  		truncate = 1;
> > > >  
> > > >  	if (xfs_iflags_test(ip, XFS_IQUOTAUNCHECKED)) {
> > > > diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
> > > > index 97ed912306fd..03944b6c5fba 100644
> > > > --- a/fs/xfs/xfs_inode.h
> > > > +++ b/fs/xfs/xfs_inode.h
> > > > @@ -292,6 +292,11 @@ static inline bool xfs_is_cow_inode(struct xfs_inode *ip)
> > > >  	return xfs_is_reflink_inode(ip) || xfs_is_always_cow_inode(ip);
> > > >  }
> > > >  
> > > > +static inline bool xfs_inode_has_filedata(const struct xfs_inode *ip)
> > > > +{
> > > > +	return ip->i_df.if_nextents > 0 || ip->i_delayed_blks > 0;
> > > > +}
> > > > +
> > > >  /*
> > > >   * Check if an inode has any data in the COW fork.  This might be often false
> > > >   * even for inodes with the reflink flag when there is no pending COW operation.
> > > > diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> > > > index a20d426ef021..2567fd2a0994 100644
> > > > --- a/fs/xfs/xfs_ioctl.c
> > > > +++ b/fs/xfs/xfs_ioctl.c
> > > > @@ -481,7 +481,7 @@ xfs_ioctl_setattr_xflags(
> > > >  
> > > >  	if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
> > > >  		/* Can't change realtime flag if any extents are allocated. */
> > > > -		if (ip->i_df.if_nextents || ip->i_delayed_blks)
> > > > +		if (xfs_inode_has_filedata(ip))
> > > >  			return -EINVAL;
> > > >  
> > > >  		/*
> > > > @@ -602,7 +602,7 @@ xfs_ioctl_setattr_check_extsize(
> > > >  	if (!fa->fsx_valid)
> > > >  		return 0;
> > > >  
> > > > -	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
> > > > +	if (S_ISREG(VFS_I(ip)->i_mode) && xfs_inode_has_filedata(ip) &&
> 
> So seems like there have been lots of changes to this particular line
> mostly as a part of refactoring other areas but seems like the actual
> commit that introduced it was:
> 
>   commit e94af02a9cd7b6590bec81df9d6ab857d6cf322f
>   Author: Eric Sandeen <sandeen@sgi.com>
>   Date:   Wed Nov 2 15:10:41 2005 +1100
>   
>       [XFS] fix old xfs_setattr mis-merge from irix; mostly harmless esp if not
>       using xfs rt
> 
> Before this we were actually checking ip->i_delayed_blks correctly. So just wanted 
> to confirm that the fixes would have the above commit right?
> 
> If this looks okay I'll send a revision with this above tags:
> 
> Fixes: e94af02a9cd7 ("[XFS] fix old xfs_setattr mis-merge from irix; mostly harmless esp if not using xfs rt")

Yeah, that sounds fine.  Want to write a quick fstest to bang on
xfs_ioctl_setattr_check_extsize to force everyone to backport it? :)

--D

> Thanks,
> Ojaswin
> 
> > > >  	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
> > > >  		return -EINVAL;
> > > >  
> > > > -- 
> > > > 2.43.5
> > > > 
> > > > 
> > > 
>