Mike Hommey wrote:
Hi Tony,
On Wed, Jun 10, 2009 at 03:59:52AM +0100, Tony Hoyle wrote:
Package: vmfs-tools Version: 0.1.1-1 Severity: important
On powerpc, any file over 2MB returns essentially random data.
This is testable by running md5sum across any file larger than this
- it comes up with different results each time.
The same package installed on amd64 appears to behave correctly, so
this is a platform specific issue, possibly due to endinness.
Does this happen with debugvmfs of vmfs-fuse ? I can't reproduce with
debugvmfs on ppc.
Here's the output I'm getting from a simple ms5sum:
# md5sum sol-10-u6-ga1-x86-dvd.iso sol-10-u6-ga1-x86-dvd.iso
sol-10-u6-ga1-x86-dvd.iso
7ac54248f85cd22764dcbd4c503e81ea sol-10-u6-ga1-x86-dvd.iso
20a9814af68cdf2cc1749fdf540437f5 sol-10-u6-ga1-x86-dvd.iso
b5134e4d505a5cd369912349610895da sol-10-u6-ga1-x86-dvd.iso
None of these are correct!
debugvmfs also does not work:
# debugvmfs /dev/sdc1 cat sol-10-u6-ga1-x86-dvd.iso >/tmp/sol.iso
VMFS: i/o spanned over several extents is unsupported
VMFS: i/o spanned over several extents is unsupported
VMFS: i/o spanned over several extents is unsupported
...
repeated about 1000 times
Presumably the driver is not handling this condition and silently
corrupting the files.
The same command executed on amd64 with the exact same vmfs image gives
no errors and produces a file that appears correct.
# debugvmfs /dev/sdb1 cat sol-10-u6-ga1-x86-dvd.iso >/tmp/sol.iso
# md5sum /tmp/sol.iso
ae8e9aa4a1a3f3fd8846a4caef9f09e2 /tmp/sol.iso
I didn't do anything special to create the drive, and both machines have
the same version of everything as far as I can tell - kernel, libc
version, etc. The only difference (and I'll admit it's a biggie) is the
different processor. It was created on esxi 3.5 (I think.. the drive it
was on failed hence the requirement to extract the data, but 4.0 was
only released last month and I definately haven't run any updates in
that time).
Unfortunately this drive has to go back into the machine it came from to
be reused.. I'll look at preserving the contents but as it's about 200gb
of data that's going to be difficult. Hope I've given you enough to
work with.
Tony
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
