On Tue, Apr 30, 2013 at 10:13:00PM +1000, Dave Chinner wrote:
> Hi folks,
> The attached tarball contains the following patches and a series
> xfs_repair: validate on-disk extent count better
> xfs_logprint: print all AGI unlinked buckets
> xfs_fsr: file reads should be O_DIRECT
> xfsprogs: Die dir1 Die!
> xfsprogs: sync code to current kernel code
> xfsprogs: update libxfs to 3.9-rc1 + xfsdev
> xfsprogs: add CRC32c infrastructure
> xfsprogs: update libxlog to current kernel code
Version 2 adds:
xfs_logprint: fix continuation transactions
As the fourth patch in the series.
Other updates are folding the dirv1 xfs_db fixes into the "die dir1
die" patch, and the log read IO alignment fixes into the libxlog
Also of note is that the commit message and a coment in the libxlog
has been updated. The comment about not knowing the size of the
iclogbuf that wrote the header and hence being unable to validate
the CRCs is incorrect. The size is written to the iclog header, so
we can CRC validate the log records. However, the patch still doe
snot enable that validation as it is still unclear to me what we
should do with failures in userspace.
If we are trying to determine if the log is dirty, a CRC failure
doesn't really matter, and if we are printing the log, then a CRC
failure shouldn't stop the logprint output. Hence the error handling
is likely to be different to the kernel log recovery code, so it
needs some more thought as to the best approach to handling this
difference before enabling CRC validation....
Description: Binary data