The NBD spec says that clients should not try to write/trim to
an export advertised as read-only by the server. But we failed
to check that, and would allow the block layer to use NBD with
BDRV_O_RDWR even when the server is read-only, which meant we
were depending on the server sending a proper EPERM failure for
various commands, and also exposes a leaky abstraction: using
qemu-io in read-write mode would succeed on 'w -z 0 0' because
of local short-circuiting logic, but 'w 0 0' would send a
request over the wire (where it then depends on the server, and
fails at least for qemu-nbd but might pass for other NBD
implementations).
With this patch, a client MUST request read-only mode to access
a server that is doing a read-only export, or else it will get
a message like:
can't open device nbd://localhost:10809/foo: request for write access conflicts with read-only export
It is no longer possible to even attempt writes over the wire
(including the corner case of 0-length writes), because the block
layer enforces the explicit read-only request; this matches the
behavior of qcow2 when backed by a read-only POSIX file.
Fix iotest 147 to comply with the new behavior (since nbd-server-add
defaults to a read-only export, we must tell blockdev-add to set up a
read-only client).
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
---
v2: retitle; require explicit read-only request, rather than magically
setting it (includes iotest fallout) [Kevin]
---
block/nbd-client.c | 9 +++++++++
tests/qemu-iotests/147 | 1 +
2 files changed, 10 insertions(+)
diff --git a/block/nbd-client.c b/block/nbd-client.c
index de6c153328..daa4392531 100644
--- a/block/nbd-client.c
+++ b/block/nbd-client.c
@@ -697,6 +697,7 @@ int nbd_client_co_pwritev(BlockDriverState *bs, uint64_t offset,
.len = bytes,
};
+ assert(!(client->info.flags & NBD_FLAG_READ_ONLY));
if (flags & BDRV_REQ_FUA) {
assert(client->info.flags & NBD_FLAG_SEND_FUA);
request.flags |= NBD_CMD_FLAG_FUA;
@@ -717,6 +718,7 @@ int nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
.len = bytes,
};
+ assert(!(client->info.flags & NBD_FLAG_READ_ONLY));
if (!(client->info.flags & NBD_FLAG_SEND_WRITE_ZEROES)) {
return -ENOTSUP;
}
@@ -756,6 +758,7 @@ int nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
.len = bytes,
};
+ assert(!(client->info.flags & NBD_FLAG_READ_ONLY));
if (!(client->info.flags & NBD_FLAG_SEND_TRIM)) {
return 0;
}
@@ -814,6 +817,12 @@ int nbd_client_init(BlockDriverState *bs,
logout("Failed to negotiate with the NBD server\n");
return ret;
}
+ if (client->info.flags & NBD_FLAG_READ_ONLY &&
+ !bdrv_is_read_only(bs)) {
+ error_setg(errp,
+ "request for write access conflicts with read-only export");
+ return -EACCES;
+ }
if (client->info.flags & NBD_FLAG_SEND_FUA) {
bs->supported_write_flags = BDRV_REQ_FUA;
bs->supported_zero_flags |= BDRV_REQ_FUA;
diff --git a/tests/qemu-iotests/147 b/tests/qemu-iotests/147
index db34838cd0..90f40ed245 100755
--- a/tests/qemu-iotests/147
+++ b/tests/qemu-iotests/147
@@ -43,6 +43,7 @@ class NBDBlockdevAddBase(iotests.QMPTestCase):
'driver': 'raw',
'file': {
'driver': 'nbd',
+ 'read-only': True,
'server': address
} }
if export is not None:
--
2.13.6
On 11/08/2017 03:56 PM, Eric Blake wrote: > Fix iotest 147 to comply with the new behavior (since nbd-server-add > defaults to a read-only export, we must tell blockdev-add to set up a > read-only client). My bad for testing only './check -nbd' and not also './check -qcow2'; iotest 58 and 140 also need fixing. I'll post the patch for squashing while awaiting review; I'd like for this to make it into -rc1. -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
09.11.2017 00:56, Eric Blake wrote: > The NBD spec says that clients should not try to write/trim to > an export advertised as read-only by the server. But we failed > to check that, and would allow the block layer to use NBD with > BDRV_O_RDWR even when the server is read-only, which meant we > were depending on the server sending a proper EPERM failure for > various commands, and also exposes a leaky abstraction: using > qemu-io in read-write mode would succeed on 'w -z 0 0' because > of local short-circuiting logic, but 'w 0 0' would send a > request over the wire (where it then depends on the server, and > fails at least for qemu-nbd but might pass for other NBD > implementations). > > With this patch, a client MUST request read-only mode to access > a server that is doing a read-only export, or else it will get > a message like: > > can't open device nbd://localhost:10809/foo: request for write access conflicts with read-only export > > It is no longer possible to even attempt writes over the wire > (including the corner case of 0-length writes), because the block > layer enforces the explicit read-only request; this matches the > behavior of qcow2 when backed by a read-only POSIX file. > > Fix iotest 147 to comply with the new behavior (since nbd-server-add > defaults to a read-only export, we must tell blockdev-add to set up a > read-only client). > > CC: qemu-stable@nongnu.org > Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> -- Best regards, Vladimir
...
Fix several iotests to comply with the new behavior (since
qemu-nbd of an internal snapshot, as well as nbd-server-add over QMP,
default to a read-only export, we must tell blockdev-add/qemu-io to
set up a read-only client).
Signed-off-by: Eric Blake <eblake@redhat.com>
---
Will squash when applying this series to the NBD queue.
---
tests/qemu-iotests/058 | 8 ++++----
tests/qemu-iotests/140 | 4 ++--
2 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/tests/qemu-iotests/058 b/tests/qemu-iotests/058
index 2253c6a6d1..5eb8784669 100755
--- a/tests/qemu-iotests/058
+++ b/tests/qemu-iotests/058
@@ -117,15 +117,15 @@ _export_nbd_snapshot sn1
echo
echo "== verifying the exported snapshot with patterns, method 1 =="
-$QEMU_IO_NBD -c 'read -P 0xa 0x1000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
-$QEMU_IO_NBD -c 'read -P 0xb 0x2000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
+$QEMU_IO_NBD -r -c 'read -P 0xa 0x1000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
+$QEMU_IO_NBD -r -c 'read -P 0xb 0x2000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
_export_nbd_snapshot1 sn1
echo
echo "== verifying the exported snapshot with patterns, method 2 =="
-$QEMU_IO_NBD -c 'read -P 0xa 0x1000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
-$QEMU_IO_NBD -c 'read -P 0xb 0x2000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
+$QEMU_IO_NBD -r -c 'read -P 0xa 0x1000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
+$QEMU_IO_NBD -r -c 'read -P 0xb 0x2000 0x1000' "$nbd_snapshot_img" | _filter_qemu_io
$QEMU_IMG convert "$TEST_IMG" -l sn1 -O qcow2 "$converted_image"
diff --git a/tests/qemu-iotests/140 b/tests/qemu-iotests/140
index f89d0d6789..a8fc95145c 100755
--- a/tests/qemu-iotests/140
+++ b/tests/qemu-iotests/140
@@ -78,7 +78,7 @@ _send_qemu_cmd $QEMU_HANDLE \
'arguments': { 'device': 'drv' }}" \
'return'
-$QEMU_IO_PROG -f raw -c 'read -P 42 0 64k' \
+$QEMU_IO_PROG -f raw -r -c 'read -P 42 0 64k' \
"nbd+unix:///drv?socket=$TEST_DIR/nbd" 2>&1 \
| _filter_qemu_io | _filter_nbd
@@ -87,7 +87,7 @@ _send_qemu_cmd $QEMU_HANDLE \
'arguments': { 'device': 'drv' }}" \
'return'
-$QEMU_IO_PROG -f raw -c close \
+$QEMU_IO_PROG -f raw -r -c close \
"nbd+unix:///drv?socket=$TEST_DIR/nbd" 2>&1 \
| _filter_qemu_io | _filter_nbd
--
2.13.6
09.11.2017 01:23, Eric Blake wrote: > ... > > Fix several iotests to comply with the new behavior (since > qemu-nbd of an internal snapshot, as well as nbd-server-add over QMP, > default to a read-only export, we must tell blockdev-add/qemu-io to > set up a read-only client). > > Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> -- Best regards, Vladimir
© 2016 - 2026 Red Hat, Inc.