Skip to contentSkip to conversation

Fix check for default backing files in bdrv_reopen_prepare()

PATCH In review Alberto Garcia 2 participants 7 messages Apr 11, 2019 – Apr 13, 2019

lore
Share & tools
lore
#0 lore raw

cover Fix check for default backing files in bdrv_reopen_prepare()

Hi,

Patch 1 is the fix for the bug reported by Max here:

https://lists.gnu.org/archive/html/qemu-block/2019-04/msg00293.html

Patch 2 fixes a different (but slightly related) issue that I found while preparing the first patch.

Regards,

Berto

Alberto Garcia (2):
block: Fix check for default backing files in bdrv_reopen_prepare()
block: Clear the backing_file fields on detach

 block.c                | 6 +++++-
 tests/qemu-iotests/245 | 2 +-
 2 files changed, 6 insertions(+), 2 deletions(-)
Signature · 1 line
2.11.0
#1 lore raw

PATCH 1/2 block: Fix check for default backing files in bdrv_reopen_prepare()

x-blockdev-reopen requires that the 'backing' parameter is specified when an image has a backing file attached or when there is a default backing file in the image metadata.

The latter can be checked by reading bs->auto_backing_file, but bdrv_reopen_prepare() is using bs->backing_file.

This bug should be detected by iotest 245, but unfortunately the test expectation is wrong so it must be corrected as well.

---
 block.c                | 3 ++-
 tests/qemu-iotests/245 | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)
block.c +2 -1
index 16615bc876..e496999e2f 100644
@@ -3634,7 +3634,8 @@ int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue,
      * its metadata. Otherwise the 'backing' option can be omitted.
      */
     if (drv->supports_backing && reopen_state->backing_missing &&
-        (backing_bs(reopen_state->bs) || reopen_state->bs->backing_file[0])) {
+        (backing_bs(reopen_state->bs) ||
+         reopen_state->bs->auto_backing_file[0])) {
         error_setg(errp, "backing is missing for '%s'",
                    reopen_state->bs->node_name);
         ret = -EINVAL;
tests/qemu-iotests/245 +1 -1
index 7891a210c1..9784ca3ced 100644
@@ -716,7 +716,7 @@ class TestBlockdevReopen(iotests.QMPTestCase):
 
         # Detach hd2 from hd0.
         self.reopen(opts, {'backing': None})
-        self.reopen(opts, {}, "backing is missing for 'hd0'")
+        self.reopen(opts)
 
         # Remove both hd0 and hd2
         result = self.vm.qmp('blockdev-del', conv_keys = True, node_name = 'hd0')
Signature · 1 line
2.11.0
#2 lore raw

PATCH 2/2 block: Clear the backing_file fields on detach

In reply to Alberto Garcia
---
 block.c | 3 +++
 1 file changed, 3 insertions(+)
block.c +3 -0
index e496999e2f..e9c0c9e2ad 100644
@@ -1085,6 +1085,9 @@ static void bdrv_backing_detach(BdrvChild *c)
     error_free(parent->backing_blocker);
     parent->backing_blocker = NULL;
 
+    parent->backing_file[0] = '\0';
+    parent->backing_format[0] = '\0';
+
     bdrv_child_cb_detach(c);
 }
 
Signature · 1 line
2.11.0
#3 lore raw
In reply to Alberto Garcia
Quoted 8 lines · Alberto Garcia

Hi,

Patch 1 is the fix for the bug reported by Max here:

   https://lists.gnu.org/archive/html/qemu-block/2019-04/msg00293.html

Patch 2 fixes a different (but slightly related) issue that I found while preparing the first patch.

I think the real problem is that bs->backing_file is not a cache for bs->backing->bs->filename.

In fact, every user of bs->backing_file expects it to be something different. Some expect it to be a cache for bs->backing->bs->filename, some expect it to be what the image header says (i.e., it if’s a relative path, it’s relative to the overlay), some expect it to be what the image header says, but relative paths to be translated so they are relative to qemu’s CWD.

All of this should be cleaned up and this is what patch 7 in my "block:
Deal with filters" series does:

http://lists.nongnu.org/archive/html/qemu-block/2019-04/msg00308.html

Max

#4 lore raw
Quoted 13 lines · Max Reitz
Quoted 2 lines

Patch 2 fixes a different (but slightly related) issue that I found while preparing the first patch.

I think the real problem is that bs->backing_file is not a cache for bs->backing->bs->filename.

In fact, every user of bs->backing_file expects it to be something different. Some expect it to be a cache for bs->backing->bs->filename, some expect it to be what the image header says (i.e., it if’s a relative path, it’s relative to the overlay), some expect it to be what the image header says, but relative paths to be translated so they are relative to qemu’s CWD.

Quoted 4 lines

All of this should be cleaned up and this is what patch 7 in my "block:
Deal with filters" series does:

http://lists.nongnu.org/archive/html/qemu-block/2019-04/msg00308.html

Ok, you can leave out the second patch then. The first one should still be correct, right?

Berto

#5 lore raw
Quoted 14 lines · Alberto Garcia
Quoted 13 lines · Max Reitz
Quoted 2 lines

Patch 2 fixes a different (but slightly related) issue that I found while preparing the first patch.

I think the real problem is that bs->backing_file is not a cache for bs->backing->bs->filename.

In fact, every user of bs->backing_file expects it to be something different. Some expect it to be a cache for bs->backing->bs->filename, some expect it to be what the image header says (i.e., it if’s a relative path, it’s relative to the overlay), some expect it to be what the image header says, but relative paths to be translated so they are relative to qemu’s CWD.

Quoted 7 lines
Quoted 4 lines

All of this should be cleaned up and this is what patch 7 in my "block:
Deal with filters" series does:

http://lists.nongnu.org/archive/html/qemu-block/2019-04/msg00308.html

Ok, you can leave out the second patch then. The first one should still be correct, right?

I just think it’s unnecessary because as of my series both backing_file and auto_backing_file serve the purpose.

Max

#6 lore raw
Quoted 5 lines · Max Reitz
Quoted 2 lines

Ok, you can leave out the second patch then. The first one should still be correct, right?

I just think it’s unnecessary because as of my series both backing_file and auto_backing_file serve the purpose.

Ok then!

Berto