Details

    • Bug
    • Resolution: Fixed
    • Blocker
    • Lustre 2.4.0
    • Lustre 2.4.0
    • 3
    • 7715

    Description

      umask handling appears to be broken again on master. With x86_64 servers running 2.3.63-5chaos, and ppc64 clients at either 2.3.58 or 2.3.63, umask does not appear to be honored.

      vulcanlac2:/p/lscratchv/morrone$ mkdir foo
      vulcanlac2:/p/lscratchv/morrone$ touch bar
      vulcanlac2:/p/lscratchv/morrone$ ls -la
      total 298
      drwx------    3 morrone morrone   5632 Apr 12 15:51 .
      drwxr-xr-x 2121 root    root    293376 Apr 12 15:05 ..
      -rw-rw-rw-    1 morrone morrone      0 Apr 12 15:51 bar
      drwxrwxrwx    2 morrone morrone   5632 Apr 12 15:51 foo
      vulcanlac2:/p/lscratchv/morrone$ umask
      0077
      

      Servers are running ZFS.

      Attachments

        Activity

          [LU-3165] umask broken on master
          pjones Peter Jones added a comment -

          Landed for 2.4

          pjones Peter Jones added a comment - Landed for 2.4

          Lai, your patch takes care of the umask problem. Thanks

          nedbass Ned Bass (Inactive) added a comment - Lai, your patch takes care of the umask problem. Thanks
          laisiyao Lai Siyao added a comment -

          Both rr_flags_h and rr_umask are missing in mdt_rec_reint, the patch is on http://review.whamcloud.com/#change,6054

          laisiyao Lai Siyao added a comment - Both rr_flags_h and rr_umask are missing in mdt_rec_reint, the patch is on http://review.whamcloud.com/#change,6054

          It looks like mdt_rec_create.cr_umask uses the rr_padding_3 field from the template struct mdt_rec_reint, but lustre_swab_mdt_rec_reint() was never updated to swab that field.

          nedbass Ned Bass (Inactive) added a comment - It looks like mdt_rec_create.cr_umask uses the rr_padding_3 field from the template struct mdt_rec_reint, but lustre_swab_mdt_rec_reint() was never updated to swab that field.
          pjones Peter Jones added a comment -

          Lai

          Could you please look into this one?

          Thanks

          Peter

          pjones Peter Jones added a comment - Lai Could you please look into this one? Thanks Peter

          Perhaps another swabbing bug introduced in the umask handling code when it was recently reworked? An x86_64 client running lustre 2.1.4-3chaos does not see the umask problem. So either the server behaves differently for the older client, or the different is the architecture of the client.

          morrone Christopher Morrone (Inactive) added a comment - Perhaps another swabbing bug introduced in the umask handling code when it was recently reworked? An x86_64 client running lustre 2.1.4-3chaos does not see the umask problem. So either the server behaves differently for the older client, or the different is the architecture of the client.

          People

            laisiyao Lai Siyao
            morrone Christopher Morrone (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: