diff options
| author | Mark Johnston <markjdb@gmail.com> | 2022-01-21 19:54:05 +0000 |
|---|---|---|
| committer | Mark Johnston <markj@FreeBSD.org> | 2022-02-21 14:59:58 +0000 |
| commit | b55a7f3422d76a6765716b2b6e78967bd75199c9 (patch) | |
| tree | 79548692706cfc4e947f89fc7d95ac72814a28e8 | |
| parent | 005452c350f5f2f3e16267557f3365a198336868 (diff) | |
Fix handling of errors from dmu_write_uio_dbuf() on FreeBSD
FreeBSD's implementation of zfs_uio_fault_move() returns EFAULT when a
page fault occurs while copying data in or out of user buffers. The VFS
treats such errors specially and will retry the I/O operation (which may
have made some partial progress).
When the FreeBSD and Linux implementations of zfs_write() were merged,
the handling of errors from dmu_write_uio_dbuf() changed such that
EFAULT is not handled as a partial write. For example, when appending
to a file, the z_size field of the znode is not updated after a partial
write resulting in EFAULT.
Restore the old handling of errors from dmu_write_uio_dbuf() to fix
this. This should have no impact on Linux, which has special handling
for EFAULT already.
Reviewed-by: Andriy Gapon <avg@FreeBSD.org>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes #12964
(cherry picked from commit 063daa8350d4a78f96d1ee6550910363fd3756fb)
| -rw-r--r-- | sys/contrib/openzfs/module/zfs/zfs_vnops.c | 15 |
1 files changed, 11 insertions, 4 deletions
diff --git a/sys/contrib/openzfs/module/zfs/zfs_vnops.c b/sys/contrib/openzfs/module/zfs/zfs_vnops.c index 170e392abe93..54749810d45d 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_vnops.c +++ b/sys/contrib/openzfs/module/zfs/zfs_vnops.c @@ -323,7 +323,7 @@ out: int zfs_write(znode_t *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) { - int error = 0; + int error = 0, error1; ssize_t start_resid = zfs_uio_resid(uio); /* @@ -561,7 +561,11 @@ zfs_write(znode_t *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) continue; } #endif - if (error != 0) { + /* + * On FreeBSD, EFAULT should be propagated back to the + * VFS, which will handle faulting and will retry. + */ + if (error != 0 && error != EFAULT) { dmu_tx_commit(tx); break; } @@ -645,7 +649,7 @@ zfs_write(znode_t *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) while ((end_size = zp->z_size) < zfs_uio_offset(uio)) { (void) atomic_cas_64(&zp->z_size, end_size, zfs_uio_offset(uio)); - ASSERT(error == 0); + ASSERT(error == 0 || error == EFAULT); } /* * If we are replaying and eof is non zero then force @@ -655,7 +659,10 @@ zfs_write(znode_t *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) zp->z_size = zfsvfs->z_replay_eof; - error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); + error1 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); + if (error1 != 0) + /* Avoid clobbering EFAULT. */ + error = error1; zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, NULL, NULL); |