| Message ID | 20240623053532.857496-11-hch@lst.de (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | [01/10] xfs: fix freeing speculative preallocations for preallocated files | expand |
On Sun, Jun 23, 2024 at 07:34:55AM +0200, Christoph Hellwig wrote: > The XFS XFS_DIFLAG_APPEND maps to the VFS S_APPEND flag, which forbids > writes that don't append at the current EOF. > > But the commit originally adding XFS_DIFLAG_APPEND support (commit > a23321e766d in xfs xfs-import repository) also checked it to skip > releasing speculative preallocations, which doesn't make any sense. > > Another commit (dd9f438e3290 in the xfs-import repository) late extended later > that flag to also report these speculation preallocations which should > not exist in getbmap. > > Remove these checks as nothing XFS_DIFLAG_APPEND implies that > preallocations beyond EOF should exist, but explicitly check for > XFS_DIFLAG_APPEND in xfs_file_release to bypass the algorithm that > discard preallocations on the first close as append only file aren't files > expected to be written to only once. > > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > fs/xfs/xfs_bmap_util.c | 12 +++++------- > fs/xfs/xfs_file.c | 4 ++++ > fs/xfs/xfs_icache.c | 2 +- > 3 files changed, 10 insertions(+), 8 deletions(-) > > diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c > index 52863b784b023f..aa924d7cd32abd 100644 > --- a/fs/xfs/xfs_bmap_util.c > +++ b/fs/xfs/xfs_bmap_util.c > @@ -331,8 +331,7 @@ xfs_getbmap( > } > > if (xfs_get_extsz_hint(ip) || > - (ip->i_diflags & > - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) > + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) The last time you tried to remove XFS_DIFLAG_APPEND from this test, I noticed that there's some fstest that "fails" because the bmap output for an append-only file now stops at isize instead of maxbytes. Do you see this same regression? > max_len = mp->m_super->s_maxbytes; > else > max_len = XFS_ISIZE(ip); > @@ -524,12 +523,11 @@ xfs_can_free_eofblocks( > return false; > > /* > - * Only free real extents for inodes with persistent preallocations or > - * the append-only flag. > + * Do not free real extents in preallocated files unless the file has > + * delalloc blocks and we are forced to remove them. > */ > - if (ip->i_diflags & (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) > - if (ip->i_delayed_blks == 0) > - return false; > + if ((ip->i_diflags & XFS_DIFLAG_PREALLOC) && !ip->i_delayed_blks) > + return false; > > /* > * Do not try to free post-EOF blocks if EOF is beyond the end of the > diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c > index 1903fa5568a37d..b05822a70ea680 100644 > --- a/fs/xfs/xfs_file.c > +++ b/fs/xfs/xfs_file.c > @@ -1231,6 +1231,9 @@ xfs_file_release( > * one file after another without going back to it while keeping the > * preallocation for files that have recurring open/write/close cycles. > * > + * This heuristic is skipped for inodes with the append-only flag as > + * that flags is rather pointless for inodes written oly once. flag only --D > + * > * There is no point in freeing blocks here for open but unlinked files > * as they will be taken care of by the inactivation path soon. > * > @@ -1245,6 +1248,7 @@ xfs_file_release( > */ > if (inode->i_nlink && > (file->f_mode & FMODE_WRITE) && > + !(ip->i_diflags & XFS_DIFLAG_APPEND) && > !xfs_iflags_test(ip, XFS_EOFBLOCKS_RELEASED) && > xfs_ilock_nowait(ip, XFS_IOLOCK_EXCL)) { > if (xfs_can_free_eofblocks(ip)) { > diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c > index 9967334ea99f1a..0f07ec842b7023 100644 > --- a/fs/xfs/xfs_icache.c > +++ b/fs/xfs/xfs_icache.c > @@ -1158,7 +1158,7 @@ xfs_inode_free_eofblocks( > if (xfs_can_free_eofblocks(ip)) > return xfs_free_eofblocks(ip); > > - /* inode could be preallocated or append-only */ > + /* inode could be preallocated */ > trace_xfs_inode_free_eofblocks_invalid(ip); > xfs_inode_clear_eofblocks_tag(ip); > return 0; > -- > 2.43.0 > >
On Mon, Jun 24, 2024 at 08:54:43AM -0700, Darrick J. Wong wrote: > > > > if (xfs_get_extsz_hint(ip) || > > - (ip->i_diflags & > > - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) > > + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) > > The last time you tried to remove XFS_DIFLAG_APPEND from this test, I > noticed that there's some fstest that "fails" because the bmap output > for an append-only file now stops at isize instead of maxbytes. Do you > see this same regression? No. What test did you see it with? Any special mkfs or mount options?
On Mon, Jun 24, 2024 at 06:07:35PM +0200, Christoph Hellwig wrote: > On Mon, Jun 24, 2024 at 08:54:43AM -0700, Darrick J. Wong wrote: > > > > > > if (xfs_get_extsz_hint(ip) || > > > - (ip->i_diflags & > > > - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) > > > + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) > > > > The last time you tried to remove XFS_DIFLAG_APPEND from this test, I > > noticed that there's some fstest that "fails" because the bmap output > > for an append-only file now stops at isize instead of maxbytes. Do you > > see this same regression? > > No. What test did you see it with? Any special mkfs or mount options? IIRC /think/ it was xfs/009. No particularly special mkfs/mount options, though my memory of 10 days ago is a bit hazy now. :( --D
On Mon, Jun 24, 2024 at 10:06:18AM -0700, Darrick J. Wong wrote: > On Mon, Jun 24, 2024 at 06:07:35PM +0200, Christoph Hellwig wrote: > > On Mon, Jun 24, 2024 at 08:54:43AM -0700, Darrick J. Wong wrote: > > > > > > > > if (xfs_get_extsz_hint(ip) || > > > > - (ip->i_diflags & > > > > - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) > > > > + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) > > > > > > The last time you tried to remove XFS_DIFLAG_APPEND from this test, I > > > noticed that there's some fstest that "fails" because the bmap output > > > for an append-only file now stops at isize instead of maxbytes. Do you > > > see this same regression? > > > > No. What test did you see it with? Any special mkfs or mount options? > > IIRC /think/ it was xfs/009. No particularly special mkfs/mount > options, though my memory of 10 days ago is a bit hazy now. :( 009 was the case when I also removed the XFS_DIFLAG_PREALLOC check accidentally in the very first version.
On Mon, Jun 24, 2024 at 07:22:53PM +0200, Christoph Hellwig wrote: > On Mon, Jun 24, 2024 at 10:06:18AM -0700, Darrick J. Wong wrote: > > On Mon, Jun 24, 2024 at 06:07:35PM +0200, Christoph Hellwig wrote: > > > On Mon, Jun 24, 2024 at 08:54:43AM -0700, Darrick J. Wong wrote: > > > > > > > > > > if (xfs_get_extsz_hint(ip) || > > > > > - (ip->i_diflags & > > > > > - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) > > > > > + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) > > > > > > > > The last time you tried to remove XFS_DIFLAG_APPEND from this test, I > > > > noticed that there's some fstest that "fails" because the bmap output > > > > for an append-only file now stops at isize instead of maxbytes. Do you > > > > see this same regression? > > > > > > No. What test did you see it with? Any special mkfs or mount options? > > > > IIRC /think/ it was xfs/009. No particularly special mkfs/mount > > options, though my memory of 10 days ago is a bit hazy now. :( > > 009 was the case when I also removed the XFS_DIFLAG_PREALLOC check > accidentally in the very first version. Aha! /me sneezes for the 10,000th time today and says With the typos fixed, Reviewed-by: Darrick J. Wong <djwong@kernel.org> --D
diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c index 52863b784b023f..aa924d7cd32abd 100644 --- a/fs/xfs/xfs_bmap_util.c +++ b/fs/xfs/xfs_bmap_util.c @@ -331,8 +331,7 @@ xfs_getbmap( } if (xfs_get_extsz_hint(ip) || - (ip->i_diflags & - (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) + (ip->i_diflags & XFS_DIFLAG_PREALLOC)) max_len = mp->m_super->s_maxbytes; else max_len = XFS_ISIZE(ip); @@ -524,12 +523,11 @@ xfs_can_free_eofblocks( return false; /* - * Only free real extents for inodes with persistent preallocations or - * the append-only flag. + * Do not free real extents in preallocated files unless the file has + * delalloc blocks and we are forced to remove them. */ - if (ip->i_diflags & (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) - if (ip->i_delayed_blks == 0) - return false; + if ((ip->i_diflags & XFS_DIFLAG_PREALLOC) && !ip->i_delayed_blks) + return false; /* * Do not try to free post-EOF blocks if EOF is beyond the end of the diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index 1903fa5568a37d..b05822a70ea680 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -1231,6 +1231,9 @@ xfs_file_release( * one file after another without going back to it while keeping the * preallocation for files that have recurring open/write/close cycles. * + * This heuristic is skipped for inodes with the append-only flag as + * that flags is rather pointless for inodes written oly once. + * * There is no point in freeing blocks here for open but unlinked files * as they will be taken care of by the inactivation path soon. * @@ -1245,6 +1248,7 @@ xfs_file_release( */ if (inode->i_nlink && (file->f_mode & FMODE_WRITE) && + !(ip->i_diflags & XFS_DIFLAG_APPEND) && !xfs_iflags_test(ip, XFS_EOFBLOCKS_RELEASED) && xfs_ilock_nowait(ip, XFS_IOLOCK_EXCL)) { if (xfs_can_free_eofblocks(ip)) { diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 9967334ea99f1a..0f07ec842b7023 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -1158,7 +1158,7 @@ xfs_inode_free_eofblocks( if (xfs_can_free_eofblocks(ip)) return xfs_free_eofblocks(ip); - /* inode could be preallocated or append-only */ + /* inode could be preallocated */ trace_xfs_inode_free_eofblocks_invalid(ip); xfs_inode_clear_eofblocks_tag(ip); return 0;
The XFS XFS_DIFLAG_APPEND maps to the VFS S_APPEND flag, which forbids writes that don't append at the current EOF. But the commit originally adding XFS_DIFLAG_APPEND support (commit a23321e766d in xfs xfs-import repository) also checked it to skip releasing speculative preallocations, which doesn't make any sense. Another commit (dd9f438e3290 in the xfs-import repository) late extended that flag to also report these speculation preallocations which should not exist in getbmap. Remove these checks as nothing XFS_DIFLAG_APPEND implies that preallocations beyond EOF should exist, but explicitly check for XFS_DIFLAG_APPEND in xfs_file_release to bypass the algorithm that discard preallocations on the first close as append only file aren't expected to be written to only once. Signed-off-by: Christoph Hellwig <hch@lst.de> --- fs/xfs/xfs_bmap_util.c | 12 +++++------- fs/xfs/xfs_file.c | 4 ++++ fs/xfs/xfs_icache.c | 2 +- 3 files changed, 10 insertions(+), 8 deletions(-)