Re: [PATCH 3/4] xfs_logprint: fix some unaligned accesses

[Eric Sandeen <sandeen@sandeen.net>]

On 10/9/15 8:24 AM, Brian Foster wrote:
> On Thu, Oct 08, 2015 at 07:25:50PM -0500, Eric Sandeen wrote:
>> This routine had a fair bit of gyration to avoid unaligned accesses,
>> but didn't fix them all.  Fix some more spotted at runtime by libubsan.
>>
>> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
>> ---
>>  logprint/log_misc.c |   18 +++++++++++++++---
>>  repair/btree.c      |    1 +
>>  2 files changed, 16 insertions(+), 3 deletions(-)
>>
>> diff --git a/logprint/log_misc.c b/logprint/log_misc.c
>> index d76145c..6cd249a 100644
>> --- a/logprint/log_misc.c
>> +++ b/logprint/log_misc.c
>> @@ -325,7 +325,11 @@ xlog_print_trans_buffer(char **ptr, int len, int *i, int num_ops)
>>  		}
>>  		super_block = 0;
>>  	} else if (be32_to_cpu(*(__be32 *)(*ptr)) == XFS_AGI_MAGIC) {
>> -		agi = (xfs_agi_t *)(*ptr);
>> +		struct xfs_agi agi_s;
>> +
>> +		/* memmove because *ptr may not be 8-byte aligned */
>> +		memmove(&agi_s, *ptr, sizeof(struct xfs_agi));
>> +		agi = &agi_s;
> 
> Nit: could we either define the new variables in the same scope as the
> pointer (either here or at the top of the function), or just ditch the
> pointers altogether?

Let me see how that looks, sure.

>>  		printf(_("AGI Buffer: XAGI  "));
>>  		/*
>>  		 * v4 filesystems only contain the fields before the uuid.
> ...
>> diff --git a/repair/btree.c b/repair/btree.c
>> index 66fb40b..e31e67a 100644
>> --- a/repair/btree.c
>> +++ b/repair/btree.c
>> @@ -230,6 +230,7 @@ btree_get_next(
>>  	}
>>  	if (level == 0) {
>>  		if (key) {
>> +		/* XXXX what if index past MAX?  What if no next? */
> 
> Unintentional hunk?

Yeah, dammit, I thought I removed that, sorry.

Thanks,
-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs