| Message ID | 153870028762.29072.5369530877410002226.stgit@magnolia (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | fs: fixes for serious clone/dedupe problems | expand |
On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Refactor all the reflink preparation steps into a separate helper that > we'll use to land all the upcoming fixes for insufficient input checks. If I've read the patch right, this also changes the location of the page cache truncation, right? i.e. it now happens in the xfs_reflink_remap_prep() function instead of after the remap? I think the commit message needs to mention that because it's a fix to incorrect behaviour.... I've added: -- This rework also moves the invalidation of the destination range to the prep function so that it is done before the range is remapped. This ensures that nobody can access the data in range being remapped until the remap is complete. -- Sound OK? Otherwise this looks fine. Reviewed-by: Dave Chinner <dchinner@redhat.com> -Dave. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > --- > fs/xfs/xfs_reflink.c | 96 +++++++++++++++++++++++++++++++++++++------------- > 1 file changed, 71 insertions(+), 25 deletions(-) > > > diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c > index 38f405415b88..80ca9b6793cd 100644 > --- a/fs/xfs/xfs_reflink.c > +++ b/fs/xfs/xfs_reflink.c > @@ -1195,11 +1195,33 @@ xfs_iolock_two_inodes_and_break_layout( > return 0; > } > > +/* Unlock both inodes after they've been prepped for a range clone. */ > +STATIC void > +xfs_reflink_remap_unlock( > + struct file *file_in, > + struct file *file_out) > +{ > + struct inode *inode_in = file_inode(file_in); > + struct xfs_inode *src = XFS_I(inode_in); > + struct inode *inode_out = file_inode(file_out); > + struct xfs_inode *dest = XFS_I(inode_out); > + bool same_inode = (inode_in == inode_out); > + > + xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); > + if (!same_inode) > + xfs_iunlock(src, XFS_MMAPLOCK_SHARED); > + inode_unlock(inode_out); > + if (!same_inode) > + inode_unlock_shared(inode_in); > +} > + > /* > - * Link a range of blocks from one file to another. > + * Prepare two files for range cloning. Upon a successful return both inodes > + * will have the iolock and mmaplock held, the page cache of the out file > + * will be truncated, and any leases on the out file will have been broken. > */ > -int > -xfs_reflink_remap_range( > +STATIC int > +xfs_reflink_remap_prep( > struct file *file_in, > loff_t pos_in, > struct file *file_out, > @@ -1211,19 +1233,9 @@ xfs_reflink_remap_range( > struct xfs_inode *src = XFS_I(inode_in); > struct inode *inode_out = file_inode(file_out); > struct xfs_inode *dest = XFS_I(inode_out); > - struct xfs_mount *mp = src->i_mount; > bool same_inode = (inode_in == inode_out); > - xfs_fileoff_t sfsbno, dfsbno; > - xfs_filblks_t fsblen; > - xfs_extlen_t cowextsize; > ssize_t ret; > > - if (!xfs_sb_version_hasreflink(&mp->m_sb)) > - return -EOPNOTSUPP; > - > - if (XFS_FORCED_SHUTDOWN(mp)) > - return -EIO; > - > /* Lock both files against IO */ > ret = xfs_iolock_two_inodes_and_break_layout(inode_in, inode_out); > if (ret) > @@ -1254,8 +1266,6 @@ xfs_reflink_remap_range( > if (ret) > goto out_unlock; > > - trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); > - > /* > * Clear out post-eof preallocations because we don't have page cache > * backing the delayed allocations and they'll never get freed on > @@ -1272,6 +1282,51 @@ xfs_reflink_remap_range( > if (ret) > goto out_unlock; > > + /* Zap any page cache for the destination file's range. */ > + truncate_inode_pages_range(&inode_out->i_data, pos_out, > + PAGE_ALIGN(pos_out + len) - 1); > + return 0; > +out_unlock: > + xfs_reflink_remap_unlock(file_in, file_out); > + return ret; > +} > + > +/* > + * Link a range of blocks from one file to another. > + */ > +int > +xfs_reflink_remap_range( > + struct file *file_in, > + loff_t pos_in, > + struct file *file_out, > + loff_t pos_out, > + u64 len, > + bool is_dedupe) > +{ > + struct inode *inode_in = file_inode(file_in); > + struct xfs_inode *src = XFS_I(inode_in); > + struct inode *inode_out = file_inode(file_out); > + struct xfs_inode *dest = XFS_I(inode_out); > + struct xfs_mount *mp = src->i_mount; > + xfs_fileoff_t sfsbno, dfsbno; > + xfs_filblks_t fsblen; > + xfs_extlen_t cowextsize; > + ssize_t ret; > + > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > + return -EOPNOTSUPP; > + > + if (XFS_FORCED_SHUTDOWN(mp)) > + return -EIO; > + > + /* Prepare and then clone file data. */ > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > + len, is_dedupe); > + if (ret) > + return ret; > + > + trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); > + > dfsbno = XFS_B_TO_FSBT(mp, pos_out); > sfsbno = XFS_B_TO_FSBT(mp, pos_in); > fsblen = XFS_B_TO_FSB(mp, len); > @@ -1280,10 +1335,6 @@ xfs_reflink_remap_range( > if (ret) > goto out_unlock; > > - /* Zap any page cache for the destination file's range. */ > - truncate_inode_pages_range(&inode_out->i_data, pos_out, > - PAGE_ALIGN(pos_out + len) - 1); > - > /* > * Carry the cowextsize hint from src to dest if we're sharing the > * entire source file to the entire destination file, the source file > @@ -1300,12 +1351,7 @@ xfs_reflink_remap_range( > is_dedupe); > > out_unlock: > - xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); > - if (!same_inode) > - xfs_iunlock(src, XFS_MMAPLOCK_SHARED); > - inode_unlock(inode_out); > - if (!same_inode) > - inode_unlock_shared(inode_in); > + xfs_reflink_remap_unlock(file_in, file_out); > if (ret) > trace_xfs_reflink_remap_range_error(dest, ret, _RET_IP_); > return ret; > >
On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Refactor all the reflink preparation steps into a separate helper that > we'll use to land all the upcoming fixes for insufficient input checks. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> ..... > +xfs_reflink_remap_range( > + struct file *file_in, > + loff_t pos_in, > + struct file *file_out, > + loff_t pos_out, > + u64 len, > + bool is_dedupe) > +{ > + struct inode *inode_in = file_inode(file_in); > + struct xfs_inode *src = XFS_I(inode_in); > + struct inode *inode_out = file_inode(file_out); > + struct xfs_inode *dest = XFS_I(inode_out); > + struct xfs_mount *mp = src->i_mount; > + xfs_fileoff_t sfsbno, dfsbno; > + xfs_filblks_t fsblen; > + xfs_extlen_t cowextsize; > + ssize_t ret; > + > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > + return -EOPNOTSUPP; > + > + if (XFS_FORCED_SHUTDOWN(mp)) > + return -EIO; > + > + /* Prepare and then clone file data. */ > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > + len, is_dedupe); > + if (ret) > + return ret; generic/013 indicates there's a double unlock bug here. vfs_clone_file_prep_inodes() can return zero (do nothing, but don't fail!), and when that happens xfs_reflink_remap_prep() unlocks the inodes and returns 0. This new code doesn't catch it, we do the remap on unlocked inodes, and then trip lock debugging bugs > @@ -1300,12 +1351,7 @@ xfs_reflink_remap_range( > is_dedupe); > > out_unlock: > - xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); > - if (!same_inode) > - xfs_iunlock(src, XFS_MMAPLOCK_SHARED); > - inode_unlock(inode_out); > - if (!same_inode) > - inode_unlock_shared(inode_in); > + xfs_reflink_remap_unlock(file_in, file_out); here: DEBUG_LOCKS_WARN_ON(sem->owner != get_current()) WARNING: CPU: 3 PID: 4766 at kernel/locking/rwsem.c:133 up_write+0x66/0x70 CPU: 3 PID: 4766 Comm: fsstress Not tainted 4.19.0-rc6-dgc+ #671 .... Call Trace: xfs_iunlock+0x152/0x220 xfs_reflink_remap_unlock+0x22/0x70 xfs_reflink_remap_range+0x129/0x2a0 do_clone_file_range+0x119/0x200 vfs_clone_file_range+0x35/0xa0 ioctl_file_clone+0x8a/0xa0 do_vfs_ioctl+0x2e1/0x6c0 ksys_ioctl+0x70/0x80 __x64_sys_ioctl+0x16/0x20 do_syscall_64+0x5a/0x180 entry_SYSCALL_64_after_hwframe+0x49/0xbe I'll fix it for the moment by making xfs_reflink_remap_prep() behave like vfs_clone_file_prep_inodes() - it will return 1 on success, 0 for nothing to do and < 0 for an error and catch it in this code. I note that later patches in the series change the vfs_clone_file_prep_inodes() behaviour so this behaviour is probably masked by those later changes. It's still a nasty bisect landmine, though, so I'll fix it here. Cheers, Dave.
On Fri, Oct 05, 2018 at 05:02:28PM +1000, Dave Chinner wrote: > On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > Refactor all the reflink preparation steps into a separate helper that > > we'll use to land all the upcoming fixes for insufficient input checks. > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > ..... > > > +xfs_reflink_remap_range( > > + struct file *file_in, > > + loff_t pos_in, > > + struct file *file_out, > > + loff_t pos_out, > > + u64 len, > > + bool is_dedupe) > > +{ > > + struct inode *inode_in = file_inode(file_in); > > + struct xfs_inode *src = XFS_I(inode_in); > > + struct inode *inode_out = file_inode(file_out); > > + struct xfs_inode *dest = XFS_I(inode_out); > > + struct xfs_mount *mp = src->i_mount; > > + xfs_fileoff_t sfsbno, dfsbno; > > + xfs_filblks_t fsblen; > > + xfs_extlen_t cowextsize; > > + ssize_t ret; > > + > > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > > + return -EOPNOTSUPP; > > + > > + if (XFS_FORCED_SHUTDOWN(mp)) > > + return -EIO; > > + > > + /* Prepare and then clone file data. */ > > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > > + len, is_dedupe); More than one bug. vfs_clone_file_prep_inodes() modifes the length parameter in the case of whole file reflink by way of "len == 0" on a non-zero length file. So I fixed this, too. -Dave.
On Fri, Oct 05, 2018 at 03:28:09PM +1000, Dave Chinner wrote: > On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > Refactor all the reflink preparation steps into a separate helper that > > we'll use to land all the upcoming fixes for insufficient input checks. > > If I've read the patch right, this also changes the location of the > page cache truncation, right? i.e. it now happens in the > xfs_reflink_remap_prep() function instead of after the remap? I > think the commit message needs to mention that because it's a fix to > incorrect behaviour.... Right. Sorry I forgot to put that in the changelog. > I've added: > > -- > This rework also moves the invalidation of the destination range to > the prep function so that it is done before the range is remapped. > This ensures that nobody can access the data in range being remapped > until the remap is complete. > -- > > Sound OK? Yep. --D > Otherwise this looks fine. > > Reviewed-by: Dave Chinner <dchinner@redhat.com> > > -Dave. > > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > > --- > > fs/xfs/xfs_reflink.c | 96 +++++++++++++++++++++++++++++++++++++------------- > > 1 file changed, 71 insertions(+), 25 deletions(-) > > > > > > diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c > > index 38f405415b88..80ca9b6793cd 100644 > > --- a/fs/xfs/xfs_reflink.c > > +++ b/fs/xfs/xfs_reflink.c > > @@ -1195,11 +1195,33 @@ xfs_iolock_two_inodes_and_break_layout( > > return 0; > > } > > > > +/* Unlock both inodes after they've been prepped for a range clone. */ > > +STATIC void > > +xfs_reflink_remap_unlock( > > + struct file *file_in, > > + struct file *file_out) > > +{ > > + struct inode *inode_in = file_inode(file_in); > > + struct xfs_inode *src = XFS_I(inode_in); > > + struct inode *inode_out = file_inode(file_out); > > + struct xfs_inode *dest = XFS_I(inode_out); > > + bool same_inode = (inode_in == inode_out); > > + > > + xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); > > + if (!same_inode) > > + xfs_iunlock(src, XFS_MMAPLOCK_SHARED); > > + inode_unlock(inode_out); > > + if (!same_inode) > > + inode_unlock_shared(inode_in); > > +} > > + > > /* > > - * Link a range of blocks from one file to another. > > + * Prepare two files for range cloning. Upon a successful return both inodes > > + * will have the iolock and mmaplock held, the page cache of the out file > > + * will be truncated, and any leases on the out file will have been broken. > > */ > > -int > > -xfs_reflink_remap_range( > > +STATIC int > > +xfs_reflink_remap_prep( > > struct file *file_in, > > loff_t pos_in, > > struct file *file_out, > > @@ -1211,19 +1233,9 @@ xfs_reflink_remap_range( > > struct xfs_inode *src = XFS_I(inode_in); > > struct inode *inode_out = file_inode(file_out); > > struct xfs_inode *dest = XFS_I(inode_out); > > - struct xfs_mount *mp = src->i_mount; > > bool same_inode = (inode_in == inode_out); > > - xfs_fileoff_t sfsbno, dfsbno; > > - xfs_filblks_t fsblen; > > - xfs_extlen_t cowextsize; > > ssize_t ret; > > > > - if (!xfs_sb_version_hasreflink(&mp->m_sb)) > > - return -EOPNOTSUPP; > > - > > - if (XFS_FORCED_SHUTDOWN(mp)) > > - return -EIO; > > - > > /* Lock both files against IO */ > > ret = xfs_iolock_two_inodes_and_break_layout(inode_in, inode_out); > > if (ret) > > @@ -1254,8 +1266,6 @@ xfs_reflink_remap_range( > > if (ret) > > goto out_unlock; > > > > - trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); > > - > > /* > > * Clear out post-eof preallocations because we don't have page cache > > * backing the delayed allocations and they'll never get freed on > > @@ -1272,6 +1282,51 @@ xfs_reflink_remap_range( > > if (ret) > > goto out_unlock; > > > > + /* Zap any page cache for the destination file's range. */ > > + truncate_inode_pages_range(&inode_out->i_data, pos_out, > > + PAGE_ALIGN(pos_out + len) - 1); > > + return 0; > > +out_unlock: > > + xfs_reflink_remap_unlock(file_in, file_out); > > + return ret; > > +} > > + > > +/* > > + * Link a range of blocks from one file to another. > > + */ > > +int > > +xfs_reflink_remap_range( > > + struct file *file_in, > > + loff_t pos_in, > > + struct file *file_out, > > + loff_t pos_out, > > + u64 len, > > + bool is_dedupe) > > +{ > > + struct inode *inode_in = file_inode(file_in); > > + struct xfs_inode *src = XFS_I(inode_in); > > + struct inode *inode_out = file_inode(file_out); > > + struct xfs_inode *dest = XFS_I(inode_out); > > + struct xfs_mount *mp = src->i_mount; > > + xfs_fileoff_t sfsbno, dfsbno; > > + xfs_filblks_t fsblen; > > + xfs_extlen_t cowextsize; > > + ssize_t ret; > > + > > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > > + return -EOPNOTSUPP; > > + > > + if (XFS_FORCED_SHUTDOWN(mp)) > > + return -EIO; > > + > > + /* Prepare and then clone file data. */ > > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > > + len, is_dedupe); > > + if (ret) > > + return ret; > > + > > + trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); > > + > > dfsbno = XFS_B_TO_FSBT(mp, pos_out); > > sfsbno = XFS_B_TO_FSBT(mp, pos_in); > > fsblen = XFS_B_TO_FSB(mp, len); > > @@ -1280,10 +1335,6 @@ xfs_reflink_remap_range( > > if (ret) > > goto out_unlock; > > > > - /* Zap any page cache for the destination file's range. */ > > - truncate_inode_pages_range(&inode_out->i_data, pos_out, > > - PAGE_ALIGN(pos_out + len) - 1); > > - > > /* > > * Carry the cowextsize hint from src to dest if we're sharing the > > * entire source file to the entire destination file, the source file > > @@ -1300,12 +1351,7 @@ xfs_reflink_remap_range( > > is_dedupe); > > > > out_unlock: > > - xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); > > - if (!same_inode) > > - xfs_iunlock(src, XFS_MMAPLOCK_SHARED); > > - inode_unlock(inode_out); > > - if (!same_inode) > > - inode_unlock_shared(inode_in); > > + xfs_reflink_remap_unlock(file_in, file_out); > > if (ret) > > trace_xfs_reflink_remap_range_error(dest, ret, _RET_IP_); > > return ret; > > > > > > -- > Dave Chinner > david@fromorbit.com
On Fri, Oct 05, 2018 at 07:02:42PM +1000, Dave Chinner wrote: > On Fri, Oct 05, 2018 at 05:02:28PM +1000, Dave Chinner wrote: > > On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > > > Refactor all the reflink preparation steps into a separate helper that > > > we'll use to land all the upcoming fixes for insufficient input checks. > > > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > > ..... > > > > > +xfs_reflink_remap_range( > > > + struct file *file_in, > > > + loff_t pos_in, > > > + struct file *file_out, > > > + loff_t pos_out, > > > + u64 len, > > > + bool is_dedupe) > > > +{ > > > + struct inode *inode_in = file_inode(file_in); > > > + struct xfs_inode *src = XFS_I(inode_in); > > > + struct inode *inode_out = file_inode(file_out); > > > + struct xfs_inode *dest = XFS_I(inode_out); > > > + struct xfs_mount *mp = src->i_mount; > > > + xfs_fileoff_t sfsbno, dfsbno; > > > + xfs_filblks_t fsblen; > > > + xfs_extlen_t cowextsize; > > > + ssize_t ret; > > > + > > > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > > > + return -EOPNOTSUPP; > > > + > > > + if (XFS_FORCED_SHUTDOWN(mp)) > > > + return -EIO; > > > + > > > + /* Prepare and then clone file data. */ > > > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > > > + len, is_dedupe); > > More than one bug. vfs_clone_file_prep_inodes() modifes the length > parameter in the case of whole file reflink by way of "len == 0" > on a non-zero length file. So I fixed this, too. Did your patch look something like the attached? --D diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c index 80ca9b6793cd..53158bdb1105 100644 --- a/fs/xfs/xfs_reflink.c +++ b/fs/xfs/xfs_reflink.c @@ -1219,6 +1219,7 @@ xfs_reflink_remap_unlock( * Prepare two files for range cloning. Upon a successful return both inodes * will have the iolock and mmaplock held, the page cache of the out file * will be truncated, and any leases on the out file will have been broken. + * Returns negative for error, 0 for nothing to do, and 1 for success. */ STATIC int xfs_reflink_remap_prep( @@ -1226,7 +1227,7 @@ xfs_reflink_remap_prep( loff_t pos_in, struct file *file_out, loff_t pos_out, - u64 len, + u64 *len, bool is_dedupe) { struct inode *inode_in = file_inode(file_in); @@ -1257,7 +1258,7 @@ xfs_reflink_remap_prep( goto out_unlock; ret = vfs_clone_file_prep_inodes(inode_in, pos_in, inode_out, pos_out, - &len, is_dedupe); + len, is_dedupe); if (ret <= 0) goto out_unlock; @@ -1284,8 +1285,8 @@ xfs_reflink_remap_prep( /* Zap any page cache for the destination file's range. */ truncate_inode_pages_range(&inode_out->i_data, pos_out, - PAGE_ALIGN(pos_out + len) - 1); - return 0; + PAGE_ALIGN(pos_out + *len) - 1); + return 1; out_unlock: xfs_reflink_remap_unlock(file_in, file_out); return ret; @@ -1321,8 +1322,8 @@ xfs_reflink_remap_range( /* Prepare and then clone file data. */ ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, - len, is_dedupe); - if (ret) + &len, is_dedupe); + if (ret <= 0) return ret; trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); > -Dave. > -- > Dave Chinner > david@fromorbit.com
On Fri, Oct 05, 2018 at 10:21:59AM -0700, Darrick J. Wong wrote: > On Fri, Oct 05, 2018 at 07:02:42PM +1000, Dave Chinner wrote: > > On Fri, Oct 05, 2018 at 05:02:28PM +1000, Dave Chinner wrote: > > > On Thu, Oct 04, 2018 at 05:44:47PM -0700, Darrick J. Wong wrote: > > > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > > > > > Refactor all the reflink preparation steps into a separate helper that > > > > we'll use to land all the upcoming fixes for insufficient input checks. > > > > > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > > > ..... > > > > > > > +xfs_reflink_remap_range( > > > > + struct file *file_in, > > > > + loff_t pos_in, > > > > + struct file *file_out, > > > > + loff_t pos_out, > > > > + u64 len, > > > > + bool is_dedupe) > > > > +{ > > > > + struct inode *inode_in = file_inode(file_in); > > > > + struct xfs_inode *src = XFS_I(inode_in); > > > > + struct inode *inode_out = file_inode(file_out); > > > > + struct xfs_inode *dest = XFS_I(inode_out); > > > > + struct xfs_mount *mp = src->i_mount; > > > > + xfs_fileoff_t sfsbno, dfsbno; > > > > + xfs_filblks_t fsblen; > > > > + xfs_extlen_t cowextsize; > > > > + ssize_t ret; > > > > + > > > > + if (!xfs_sb_version_hasreflink(&mp->m_sb)) > > > > + return -EOPNOTSUPP; > > > > + > > > > + if (XFS_FORCED_SHUTDOWN(mp)) > > > > + return -EIO; > > > > + > > > > + /* Prepare and then clone file data. */ > > > > + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, > > > > + len, is_dedupe); > > > > More than one bug. vfs_clone_file_prep_inodes() modifes the length > > parameter in the case of whole file reflink by way of "len == 0" > > on a non-zero length file. So I fixed this, too. > > Did your patch look something like the attached? Identical. I folded it back in once it tested ok, though. Cheers, Dave.
On Fri, Oct 05, 2018 at 03:28:09PM +1000, Dave Chinner wrote: > I've added: > > -- > This rework also moves the invalidation of the destination range to > the prep function so that it is done before the range is remapped. > This ensures that nobody can access the data in range being remapped > until the remap is complete. > -- > > Sound OK? > > Otherwise this looks fine. Given that the patch will need a respin anyway I'd rather split the move of the pagecache invalidation into a separate patch.
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c index 38f405415b88..80ca9b6793cd 100644 --- a/fs/xfs/xfs_reflink.c +++ b/fs/xfs/xfs_reflink.c @@ -1195,11 +1195,33 @@ xfs_iolock_two_inodes_and_break_layout( return 0; } +/* Unlock both inodes after they've been prepped for a range clone. */ +STATIC void +xfs_reflink_remap_unlock( + struct file *file_in, + struct file *file_out) +{ + struct inode *inode_in = file_inode(file_in); + struct xfs_inode *src = XFS_I(inode_in); + struct inode *inode_out = file_inode(file_out); + struct xfs_inode *dest = XFS_I(inode_out); + bool same_inode = (inode_in == inode_out); + + xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); + if (!same_inode) + xfs_iunlock(src, XFS_MMAPLOCK_SHARED); + inode_unlock(inode_out); + if (!same_inode) + inode_unlock_shared(inode_in); +} + /* - * Link a range of blocks from one file to another. + * Prepare two files for range cloning. Upon a successful return both inodes + * will have the iolock and mmaplock held, the page cache of the out file + * will be truncated, and any leases on the out file will have been broken. */ -int -xfs_reflink_remap_range( +STATIC int +xfs_reflink_remap_prep( struct file *file_in, loff_t pos_in, struct file *file_out, @@ -1211,19 +1233,9 @@ xfs_reflink_remap_range( struct xfs_inode *src = XFS_I(inode_in); struct inode *inode_out = file_inode(file_out); struct xfs_inode *dest = XFS_I(inode_out); - struct xfs_mount *mp = src->i_mount; bool same_inode = (inode_in == inode_out); - xfs_fileoff_t sfsbno, dfsbno; - xfs_filblks_t fsblen; - xfs_extlen_t cowextsize; ssize_t ret; - if (!xfs_sb_version_hasreflink(&mp->m_sb)) - return -EOPNOTSUPP; - - if (XFS_FORCED_SHUTDOWN(mp)) - return -EIO; - /* Lock both files against IO */ ret = xfs_iolock_two_inodes_and_break_layout(inode_in, inode_out); if (ret) @@ -1254,8 +1266,6 @@ xfs_reflink_remap_range( if (ret) goto out_unlock; - trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); - /* * Clear out post-eof preallocations because we don't have page cache * backing the delayed allocations and they'll never get freed on @@ -1272,6 +1282,51 @@ xfs_reflink_remap_range( if (ret) goto out_unlock; + /* Zap any page cache for the destination file's range. */ + truncate_inode_pages_range(&inode_out->i_data, pos_out, + PAGE_ALIGN(pos_out + len) - 1); + return 0; +out_unlock: + xfs_reflink_remap_unlock(file_in, file_out); + return ret; +} + +/* + * Link a range of blocks from one file to another. + */ +int +xfs_reflink_remap_range( + struct file *file_in, + loff_t pos_in, + struct file *file_out, + loff_t pos_out, + u64 len, + bool is_dedupe) +{ + struct inode *inode_in = file_inode(file_in); + struct xfs_inode *src = XFS_I(inode_in); + struct inode *inode_out = file_inode(file_out); + struct xfs_inode *dest = XFS_I(inode_out); + struct xfs_mount *mp = src->i_mount; + xfs_fileoff_t sfsbno, dfsbno; + xfs_filblks_t fsblen; + xfs_extlen_t cowextsize; + ssize_t ret; + + if (!xfs_sb_version_hasreflink(&mp->m_sb)) + return -EOPNOTSUPP; + + if (XFS_FORCED_SHUTDOWN(mp)) + return -EIO; + + /* Prepare and then clone file data. */ + ret = xfs_reflink_remap_prep(file_in, pos_in, file_out, pos_out, + len, is_dedupe); + if (ret) + return ret; + + trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out); + dfsbno = XFS_B_TO_FSBT(mp, pos_out); sfsbno = XFS_B_TO_FSBT(mp, pos_in); fsblen = XFS_B_TO_FSB(mp, len); @@ -1280,10 +1335,6 @@ xfs_reflink_remap_range( if (ret) goto out_unlock; - /* Zap any page cache for the destination file's range. */ - truncate_inode_pages_range(&inode_out->i_data, pos_out, - PAGE_ALIGN(pos_out + len) - 1); - /* * Carry the cowextsize hint from src to dest if we're sharing the * entire source file to the entire destination file, the source file @@ -1300,12 +1351,7 @@ xfs_reflink_remap_range( is_dedupe); out_unlock: - xfs_iunlock(dest, XFS_MMAPLOCK_EXCL); - if (!same_inode) - xfs_iunlock(src, XFS_MMAPLOCK_SHARED); - inode_unlock(inode_out); - if (!same_inode) - inode_unlock_shared(inode_in); + xfs_reflink_remap_unlock(file_in, file_out); if (ret) trace_xfs_reflink_remap_range_error(dest, ret, _RET_IP_); return ret;