[LU-15646] fix DOSTID printing of llog_id FIDs

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Labels: easy
• Severity: 3
• Assignee: Mikhail Pershin
• Reporter: Andreas Dilger

Description

It looks like lgl_oi.oi_fid is used in quite a number of places, but any use of lgl_oi.oi_fid looks to be incorrect (f_seq and oi_seq are in opposite order in struct ost_objid). This was broken by patch https://review.whamcloud.com/25640 "LU-9153 llog: update llog print format to use FIDs". It is only using oi_fid like PFID(&...lgl_oi.oi_fid), so this is only affecting error messages and not any functionality.

log_process_thread()) lfs02-MDT001e-osp-MDT0000: [0x3:0x1b70:0x4] Invalid record: index 16123 but expected 16122

The output "[0x3:0x1b70:0x4]" should be interpreted as FID [0x400001b70:0x3:0x0] (swap last and first components of the field, add intervening "0000" and add ":0x0" at the end) in the update_log_dir/ directory.

To print these FIDs correctly, there is a logid_to_fid() helper with the DOSTID macro, and it would make sense to have it also return the same pointer as the passed-in fid argument, so that it can be passed directly to PFID() in the error message, like:

        struct lu_fid tmp_fid;

        CERROR("%s: "DFID" Invalid record: index %u but expected %u\n",
               loghandle2name(loghandle),
               PFID(logid_to_fid(&loghandle->lgh_id, &tmp_fid)),          
               rec->lrh_index, index);

Alternately, it would be more efficient to add a new PLOGID() macro that is returning oi_seq, oi_id, 0, directly, and avoid the need for tmp_fid entirely.

Issue Links

• is duplicated by: LU-15936 DOSTID macro is not printing llog IDs properly in the logs
• is related to: LU-15644 failed llog cancel should not generate an error
• is related to: LU-15645 gap in recovery llog should not be a fatal error