[Bacula-users] Hash options ???

[John L. <jlo...@um...>]

>From various places online I see that these are the options and signature
comparison values for the various hash methods.

1 : SHA1
2 : SHA256
3 : SHA512
5 : MD5
6 : XXHASH64
7 : XXH3_64
8 : XXH3_128

But, then diving into the code (15.0.2) I see that there are issues with
SHA512 and that's disabled in the code, in some places and not in others.
I also see that for XXH3_64 the code says it was never implemented, and
again disabled in some sections of the code, but not others.

So, I'd like to use a Signature of: XXH3_128, that seems to pass the muster
for the "Signature =" line in the Fileset.  However in the "Accurate ="
line, the "8" is being rejected, as in "Accurate = pino8" and "Verify =
pins8".

Am I reading the code wrong and "8" is not the value associated with
XXH3_128?

Also, of note, that none of this is updated in the documentation.  There
are mentions in the ChangeLog about various things being added, but in the
appropriate section (FileSet Resource) they aren't there, or are completely
wrong.


   - XXHASH is mentioned as a Signature option in the docs, yet that
   specific string can't be found as an option in the source.
   - There is no mention in the docs of XXH3_64 or XXH3_128.  In the code
   both are acceptable for the Signature, yet they fail validation.

Best I can see is that the options are not included in scan_include_options
within src/dird/inc_conf.c.  At line 281:
   /* Check if the options are correct */
   switch(keyword) {
   case INC_KW_VERIFY:
      fs_options = "ipnugsamcd5123:V"; /* From dird/verify.c */
      break;
   case INC_KW_BASEJOB:
   case INC_KW_ACCURATE:
      fs_options = "oipnugsamMcdA5123:JC"; /* From filed/accurate.c
accurate_check_file() */
      break;
   default:
      break;

I suspect the first fs_options should be:
        fs_options = "ipnugsamcd5123678:V"; /* From dird/verify.c */
and, the second fs_options should be:
        fs_options = "oipnugsamMcdA5123678:JC"; /* From filed/accurate.c
accurate_check_file() */


Though, in src/filed/accurate.c in the accurate_check_file function, at
line no 640, in this chunk:
   /* Go through opts once again, this time check only for checksum-related
opts */
   for (char *p=opts; *p; p++) {
      switch (*p) {
      case '5':                /* compare MD5 */
      case '1':                /* compare SHA1 */
      case '2':                /* compare SHA256 */
      case '3':                /* compare SHA512 */
      case '6':                /* compare XXHASH64 */
      case '7':                /* compare XXH3_64 */
      case '8':                /* compare XXH3_128 */
         if (ff_pkt->type != FT_LNKSAVED &&
               (S_ISREG(ff_pkt->statp.st_mode) &&
                ff_pkt->flags & (FO_MD5|FO_SHA1|FO_SHA256|FO_SHA512))) {
            checksum = true;
         }
         break;
      case 'o':
         only_changed  = true;
         break;
      default:
         break;
      }


I'm thinking this line:
                ff_pkt->flags & (FO_MD5|FO_SHA1|FO_SHA256|FO_SHA512))) {
Is probably supposed to read:
                ff_pkt->flags &
(FO_MD5|FO_SHA1|FO_SHA256|FO_SHA512|O_XXHASH64|FO_XXH3_64|FO_XXH3_128)))
{


I'm pondering making these changes and seeing what happens.

-John

-- 
-- 
- Adaptability -- Analytical --- Ideation ---- Input ----- Belief -
-------------------------------------------------------------------
         John M. Lockard |  U of Michigan - School of Information
          Unix Sys Admin |      Suite 205 | 309 Maynard Street
      jlo...@um... |        Ann Arbor, MI  48104-2211
 www.umich.edu/~jlockard <http://www.umich.edu/%7Ejlockard> |
734-615-8776 | 734-763-9677 FAX
-------------------------------------------------------------------
- The University of Michigan will never ask you for your password -