On 06/21/2012 12:24 PM, MORITA Kazutaka wrote:
> At Wed, 20 Jun 2012 18:16:02 +0800,
> Liu Yuan wrote:
>> +
>> +static void do_check_repair(uint64_t oid, int nr_copies)
>> +{
>> + struct sd_vnode *tgt_vnodes[nr_copies];
>> + unsigned char sha1[nr_copies][SHA1_LEN];
>> + void *buf;
>> + int i;
>> +
>> + collie_oid_to_vnodes(vnode_list_entries, nr_vnodes,
>> + oid, nr_copies, tgt_vnodes);
>> + for (i = 0; i < nr_copies; i++) {
>> + buf = read_object_from(tgt_vnodes[i], oid);
>> + SHA1(buf, SD_DATA_OBJ_SIZE, sha1[i]);
>
> memcmp against contents of the object directly looks simpler than
> calculating their hash values.
>
>
>> + free(buf);
>> + }
>> +
>> + if (!memcmp(sha1[0], sha1[1], SHA1_LEN) &&
>> + !memcmp(sha1[0], sha1[2], SHA1_LEN))
>> + return; /* All replica consistent */
>
> This will fail if the number of replica is not three.
>
Ah, yes. I'll do a new patch to address these issue.
Thanks,
Yuan
|
|