file-posix: Handle undetectable alignment
In some cases buf_align or request_alignment cannot be detected:
1. With Gluster, buf_align cannot be detected since the actual I/O is
done on Gluster server, and qemu buffer alignment does not matter.
Since we don't have alignment requirement, buf_align=1 is the best
value.
2. With local XFS filesystem, buf_align cannot be detected if reading
from unallocated area. In this we must align the buffer, but we don't
know what is the correct size. Using the wrong alignment results in
I/O error.
3. With Gluster backed by XFS, request_alignment cannot be detected if
reading from unallocated area. In this case we need to use the
correct alignment, and failing to do so results in I/O errors.
4. With NFS, the server does not use direct I/O, so both buf_align cannot
be detected. In this case we don't need any alignment so we can use
buf_align=1 and request_alignment=1.
These cases seems to work when storage sector size is 512 bytes, because
the current code starts checking align=512. If the check succeeds
because alignment cannot be detected we use 512. But this does not work
for storage with 4k sector size.
To determine if we can detect the alignment, we probe first with
align=1. If probing succeeds, maybe there are no alignment requirement
(cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we
don't have any way to tell, we treat this as undetectable alignment. If
probing with align=1 fails with EINVAL, but probing with one of the
expected alignments succeeds, we know that we found a working alignment.
Practically the alignment requirements are the same for buffer
alignment, buffer length, and offset in file. So in case we cannot
detect buf_align, we can use request alignment. If we cannot detect
request alignment, we can fallback to a safe value. To use this logic,
we probe first request alignment instead of buf_align.
Here is a table showing the behaviour with current code (the value in
parenthesis is the optimal value).
Case Sector buf_align (opt) request_alignment (opt) result
======================================================================
1 512 512 (1) 512 (512) OK
1 4096 512 (1) 4096 (4096) FAIL
----------------------------------------------------------------------
2 512 512 (512) 512 (512) OK
2 4096 512 (4096) 4096 (4096) FAIL
----------------------------------------------------------------------
3 512 512 (1) 512 (512) OK
3 4096 512 (1) 512 (4096) FAIL
----------------------------------------------------------------------
4 512 512 (1) 512 (1) OK
4 4096 512 (1) 512 (1) OK
Same cases with this change:
Case Sector buf_align (opt) request_alignment (opt) result
======================================================================
1 512 512 (1) 512 (512) OK
1 4096 4096 (1) 4096 (4096) OK
----------------------------------------------------------------------
2 512 512 (512) 512 (512) OK
2 4096 4096 (4096) 4096 (4096) OK
----------------------------------------------------------------------
3 512 4096 (1) 4096 (512) OK
3 4096 4096 (1) 4096 (4096) OK
----------------------------------------------------------------------
4 512 4096 (1) 4096 (1) OK
4 4096 4096 (1) 4096 (1) OK
I tested that provisioning VMs and copying disks on local XFS and
Gluster with 4k bytes sector size work now, resolving bugs [1],[2].
I tested also on XFS, NFS, Gluster with 512 bytes sector size.
[1] https://bugzilla.redhat.com/1737256
[2] https://bugzilla.redhat.com/1738657
Signed-off-by: Nir Soffer <nsoffer@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
diff --git a/block/file-posix.c b/block/file-posix.c
index 4479cc7..b8b4dad 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -323,6 +323,7 @@
BDRVRawState *s = bs->opaque;
char *buf;
size_t max_align = MAX(MAX_BLOCKSIZE, getpagesize());
+ size_t alignments[] = {1, 512, 1024, 2048, 4096};
/* For SCSI generic devices the alignment is not really used.
With buffered I/O, we don't have any restrictions. */
@@ -349,25 +350,38 @@
}
#endif
- /* If we could not get the sizes so far, we can only guess them */
- if (!s->buf_align) {
+ /*
+ * If we could not get the sizes so far, we can only guess them. First try
+ * to detect request alignment, since it is more likely to succeed. Then
+ * try to detect buf_align, which cannot be detected in some cases (e.g.
+ * Gluster). If buf_align cannot be detected, we fallback to the value of
+ * request_alignment.
+ */
+
+ if (!bs->bl.request_alignment) {
+ int i;
size_t align;
- buf = qemu_memalign(max_align, 2 * max_align);
- for (align = 512; align <= max_align; align <<= 1) {
- if (raw_is_io_aligned(fd, buf + align, max_align)) {
- s->buf_align = align;
+ buf = qemu_memalign(max_align, max_align);
+ for (i = 0; i < ARRAY_SIZE(alignments); i++) {
+ align = alignments[i];
+ if (raw_is_io_aligned(fd, buf, align)) {
+ /* Fallback to safe value. */
+ bs->bl.request_alignment = (align != 1) ? align : max_align;
break;
}
}
qemu_vfree(buf);
}
- if (!bs->bl.request_alignment) {
+ if (!s->buf_align) {
+ int i;
size_t align;
- buf = qemu_memalign(s->buf_align, max_align);
- for (align = 512; align <= max_align; align <<= 1) {
- if (raw_is_io_aligned(fd, buf, align)) {
- bs->bl.request_alignment = align;
+ buf = qemu_memalign(max_align, 2 * max_align);
+ for (i = 0; i < ARRAY_SIZE(alignments); i++) {
+ align = alignments[i];
+ if (raw_is_io_aligned(fd, buf + align, max_align)) {
+ /* Fallback to request_aligment. */
+ s->buf_align = (align != 1) ? align : bs->bl.request_alignment;
break;
}
}
