Debian Patches

Status for virt-firmware/25.10-2

Patch Description Author Forwarded Bugs Origin Last update
varstore-endianness.patch There's a bug in virt/firmware/varstore/edk2.py when
reading varstore images, - it uses native endiannes
instead of little-endian.

def parse_volume(self):
...
(vlen, sig, attr, hlen, csum, xoff, rev, blocks, blksize) = \
struct.unpack_from("=QLLHHHxBLL", self.filedata, offset + 32)
logging.debug('vol=%s vlen=0x%x rev=%d blocks=%d*%d (0x%x)',
guids.name(guid), vlen, rev,
blocks, blksize, blocks * blksize)
if sig != 0x4856465f:
logging.error('%s: not a firmware volume', self.filename)
sys.exit(1)

When reading AAVMF_VARS.fd, on a little-endian system, this results in:

DEBUG:root:vol=guid:NvData vlen=0xc0000 rev=2 blocks=3*262144 (0xc0000)

The same code and input, when run on s390x, results in:

DEBUG:root:vol=guid:NvData vlen=0xc0000000000 rev=2 blocks=50331648*1024 (0xc00000000)
ERROR:root:AAVMF_VARS.fd: not a firmware volume

The unpack_from() call should have "<" (little-endian) as the first symbol
instead of "=" (native-endian), since it is reading an image defined as
having little-endian characteristics.

See two edk2 debci runs, on amd64 (sucessful) and on s390x (failed):
https://ci.debian.net/packages/e/edk2/testing/amd64/66712410/
https://ci.debian.net/packages/e/edk2/testing/s390x/66711962/

There are other places in this file where the same "=" (native) conversion
is used, - these should probably be fixed too.

Attached is a small patch (untested yet) which should fix this.

Thanks,

/mjt

diff --git a/virt/firmware/varstore/edk2.py b/virt/firmware/varstore/edk2.py
index 33b2738..54db6e1 100644		 Michael Tokarev <mjt@tls.msk.ru> yes debian upstream https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1121761#5

All known versions for source package 'virt-firmware'

Links