Re: [libvirt] [PATCHv3 07/36] storage: Switch metadata crawler to use storage driver to get unique path

On 06/09/14 16:07, Roman Bogorodskiy wrote: > Peter Krempa wrote: > >> On 06/07/14 20:35, Roman Bogorodskiy wrote: >>> Peter Krempa wrote: >>> >>>> Use the virStorageFileGetUniqueIdentifier() function to get a unique >>>> identifier regardless of the target storage type instead of relying on >>>> canonicalize_path(). >>>> >>>> A new function that checks whether we support a given image is >>>> introduced to avoid errors for unimplemented backends. >>>> --- >>>> >>>> Notes: >>>> Version 3: >>>> - fixed typo >>>> - ACKed by Eric >>>> >>>> src/storage/storage_driver.c | 77 +++++++++++++++++++++++++++++++------------- >>>> 1 file changed, 54 insertions(+), 23 deletions(-) >>>> >>>> diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c >>>> index f92a553..5c4188f 100644 >>>> --- a/src/storage/storage_driver.c >>>> +++ b/src/storage/storage_driver.c >> >>>> @@ -3112,49 +3135,63 @@ virStorageFileGetMetadataRecurse(virStorageSourcePtr src, >>>> int fd; >>>> int ret = -1; >>>> struct stat st; >>>> + const char *uniqueName; >>>> virStorageSourcePtr backingStore = NULL; >>>> int backingFormat; >>>> >>>> - VIR_DEBUG("path=%s canonPath=%s dir=%s format=%d uid=%d gid=%d probe=%d", >>>> - src->path, canonPath, NULLSTR(src->relDir), src->format, >>>> + VIR_DEBUG("path=%s dir=%s format=%d uid=%d gid=%d probe=%d", >>>> + src->path, NULLSTR(src->relDir), src->format, >>>> (int)uid, (int)gid, allow_probe); >>>> >>>> - if (virHashLookup(cycle, canonPath)) { >>>> - virReportError(VIR_ERR_INTERNAL_ERROR, >>>> - _("backing store for %s is self-referential"), >>>> - src->path); >>>> + /* exit if we can't load information about the current image */ >>>> + if (!virStorageFileSupportsBackingChainTraversal(src)) >>>> + return 0; >>> >>> After this change I get virstrogetest failing on FreeBSD, which doesn't >>> support any of the file storage backends currently: >>> >>> >>> Program received signal SIGSEGV, Segmentation fault. >>> [Switching to Thread 806c06400 (LWP 100061/virstoragetest)] >>> 0x0000000000410088 in mymain () at virstoragetest.c:937 >>> 937 TEST_LOOKUP(3, "qcow2", chain->backingStore->path, >>> chain->backingStore, >>> Current language: auto; currently minimal >>> (gdb) p chain >>> $1 = 0x806c7b100 >>> (gdb) p chain->backingStore >>> $2 = 0x0 >>> (gdb) >>> >>> >> >> Hmm, so FreeBSD doesn't currently compile the storage driver? We should >> probably look into enabling it. All the stuff that was done by the code >> was compiling just fine on FreeBSD previously so enabling just the local >> filesystem storage backend should work well. I'll have a look what that >> would include. > > virstorage.c test calls testStorageFileGetMetadata() which calls > virStorageFileGetMetadata(). > > Inside of that, the call sequence is: > > virStorageFileGetMetadata -> virStorageFileGetMetadataRecurse > > virStorageFileGetMetadataRecurse then checks > !virStorageFileSupportsBackingChainTraversal() and returns 0. > > virStorageFileSupportsBackingChainTraversal tries to initialise backend > using virStorageFileBackendForTypeInternal(). > > virStorageFileBackendForTypeInternal loops through fileBackends which > looks this way: > > static virStorageFileBackendPtr fileBackends[] = { > #if WITH_STORAGE_FS > &virStorageFileBackendFile, > &virStorageFileBackendBlock, > #endif > #if WITH_STORAGE_GLUSTER > &virStorageFileBackendGluster, > #endif > NULL > }; > > FreeBSD doesn't support neither WITH_STORAGE_GLUSTER nor WITH_STORAGE_FS. So > we end up having chain->backingStore == NULL. > > PS I may be wrong here as it's my first experience with the storage code > and I haven't yet spend much time on it. Yes, the problem is that VIR_STORAGE_FS isn't enabled on freebsd. I don't know whether it's worth analyzing the code to find out if it can be enabled on freebsd or not. (the needed portions certainly can be enabled, but some of the other portions might be problematic.

At any rate, I'm working on a refactor of the test that will allow to skip the test (and remove a few ugly pieces of code)