According to xfstest generic/240, applications seem to expect direct I/O
writes to either complete as a whole or to fail; short direct I/O writes
are apparently not appreciated. This means that when only part of an
asynchronous direct I/O write succeeds, we can either fail the entire
write, or we can wait for the partial write to complete and retry the
remaining write using buffered I/O. The old __blockdev_direct_IO helper
has code for waiting for partial writes to complete; the new
iomap_dio_rw iomap helper does not.
The above mentioned fallback mode is used by gfs2, which doesn't allow
block allocations under direct I/O to avoid taking cluster-wide
exclusive locks. As a consequence, an asynchronous direct I/O write to
a file range that ends in a hole will result in a short write. When
that happens, we want to complete the remaining write as buffered I/O.
To allow that, change iomap_dio_rw to wait for short direct I/O writes
like __blockdev_direct_IO does instead of returning -EIOCBQUEUED.
This makes xfstest generic/240 work on gfs2.
Signed-off-by: Andreas Gruenbacher <agruenba@xxxxxxxxxx>
Cc: Dave Chinner <dchinner@xxxxxxxxxx>
---
fs/iomap.c | 22 ++++++++++++----------
1 file changed, 12 insertions(+), 10 deletions(-)
diff --git a/fs/iomap.c b/fs/iomap.c
index 27d97a290623..fcd3e26aadb7 100644
--- a/fs/iomap.c
+++ b/fs/iomap.c
@@ -720,6 +720,7 @@ struct iomap_dio {
atomic_t ref;
unsigned flags;
int error;
+ bool wait_for_completion;
union {
/* used during submission and for synchronous completion: */
@@ -821,9 +822,8 @@ static void iomap_dio_bio_end_io(struct bio *bio)
iomap_dio_set_error(dio, blk_status_to_errno(bio->bi_status));
if (atomic_dec_and_test(&dio->ref)) {
- if (is_sync_kiocb(dio->iocb)) {
+ if (dio->wait_for_completion) {
struct task_struct *waiter = dio->submit.waiter;
-
WRITE_ONCE(dio->submit.waiter, NULL);
wake_up_process(waiter);
} else if (dio->flags & IOMAP_DIO_WRITE) {
@@ -1014,13 +1014,12 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
dio->end_io = end_io;
dio->error = 0;
dio->flags = 0;
+ dio->wait_for_completion = is_sync_kiocb(iocb);
dio->submit.iter = iter;
- if (is_sync_kiocb(iocb)) {
- dio->submit.waiter = current;
- dio->submit.cookie = BLK_QC_T_NONE;
- dio->submit.last_queue = NULL;
- }
+ dio->submit.waiter = current;
+ dio->submit.cookie = BLK_QC_T_NONE;
+ dio->submit.last_queue = NULL;
if (iov_iter_rw(iter) == READ) {
if (pos >= dio->i_size)
@@ -1057,7 +1056,7 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
dio_warn_stale_pagecache(iocb->ki_filp);
ret = 0;
- if (iov_iter_rw(iter) == WRITE && !is_sync_kiocb(iocb) &&
+ if (iov_iter_rw(iter) == WRITE && !dio->wait_for_completion &&
!inode->i_sb->s_dio_done_wq) {
ret = sb_init_dio_done_wq(inode->i_sb);
if (ret < 0)
@@ -1072,8 +1071,10 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
iomap_dio_actor);
if (ret <= 0) {
/* magic error code to fall back to buffered I/O */
- if (ret == -ENOTBLK)
+ if (ret == -ENOTBLK) {
+ dio->wait_for_completion = true;
ret = 0;
+ }
break;
}
pos += ret;
@@ -1086,8 +1087,9 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
if (ret < 0)
iomap_dio_set_error(dio, ret);
+ smp_mb__before_atomic();
if (!atomic_dec_and_test(&dio->ref)) {
- if (!is_sync_kiocb(iocb))
+ if (!dio->wait_for_completion)
return -EIOCBQUEUED;
for (;;) {
--
2.17.0
