[PATCH] xfs: fix endianness error when checking log block crc on big endian platforms

[Brian Foster]

On Tue, Jan 26, 2016 at 10:05:00AM -0800, Darrick J. Wong wrote:
> Since the checksum function and the field are both __le32, don't
> perform endian conversion when comparing the two.  This fixes mount
> failures on ppc64 in 4.5-rc1.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong at oracle.com>
> ---

So somehow this first got munged in commit:

  b94fb2d1 xfs: refactor and open code log record crc check

... which was intended to only be a refactoring patch. Instead, it
removes the following line:

-       if (crc != rhead->h_crc) {

... and replaces it with:

+       if (crc != le32_to_cpu(rhead->h_crc)) {

No idea how that happened. Maybe I was just looking at the log message
code and had a thinko. Anyways, it was followed up with commit:

	6528250b xfs: support a crc verification only log record pass

... which adds the code you've fixed here and probably just copied the
busted check.

>  fs/xfs/xfs_log_recover.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
> index ac80f39..52dd946 100644
> --- a/fs/xfs/xfs_log_recover.c
> +++ b/fs/xfs/xfs_log_recover.c
> @@ -4502,7 +4502,7 @@ xlog_recover_process(
>  	 * know precisely what failed.
>  	 */
>  	if (pass == XLOG_RECOVER_CRCPASS) {
> -		if (rhead->h_crc && crc != le32_to_cpu(rhead->h_crc))
> +		if (rhead->h_crc && crc != rhead->h_crc)
>  			return -EFSBADCRC;
>  		return 0;
>  	}

There's another check a few lines after this that is also broken, based
on the first commit referenced above. Do you want me to send a v2 patch
or would you like to? I'll run some ppc64 testing either way...

Brian