[LU-14431] error messages that do not stop with newline

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Affects Version/s: Lustre 2.12.6
• Labels: easy
• Assignee: Feng Lei
• Reporter: Li Xi

Description

When I am using grep to check how many CERROR in the codes, I found there are still a dozen of error messages that do not end with a newline. That will cause annoying messages in system log.

lnet/klnds/gnilnd/gnilnd_cb.c(2378):

			CERROR("Unexpected connstamp %#llx(%#llx expected)"
				" from %s", rxmsg->gnm_connstamp,
				found_conn->gnc_peer_connstamp,
				libcfs_nid2str(peer->gnp_nid));

lnet/klnds/gnilnd/gnilnd_conn.c(867):

			CERROR("Looking up network: device is in shutdown");

lustre/ptlrpc/service.c(731):

				CERROR("%s: invalid CPT pattern string: %s",
				       conf->psc_name, cconf->cc_pattern);

lustre/ptlrpc/ptlrpcd.c(754):

			CERROR("%s: invalid CPT pattern string: %s",
			       "ptlrpcd_cpts", ptlrpcd_cpts);

lustre/include/lustre_dlm.h(1159):

			CERROR("lock %p: delayed lvb init failed (rc %d)",
			       lock, rc);

lustre/quota/qmt_lock.c(761):

		CERROR("%s: failed to init env.", qmt->qmt_svname);

lustre/osd-ldiskfs/osd_handler.c(2377):

				CERROR("%s: unsupported checksum type of "
				       "T10PI type '%s'",
				       d->od_svname, name);

lustre/osd-ldiskfs/osd_handler.c(2382):

			CERROR("%s: unsupported T10PI type '%s'",
			       d->od_svname, name);

lustre/target/tgt_handler.c(945):

		CERROR("No target passed");

lustre/obdclass/lprocfs_status.c(78):

		CERROR("LprocFS: No memory to create <debugfs> entry %s", name);

Issue Links

• is related to: LU-14475 Make too short log messages meanful

Comments

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/41723/
Subject: LU-14431 log: Add ending newline for some messages.
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 503bf7f29a49914082480a7cdceb914e42a95dd8

[James A Simmons]

The current patch missed a few so another patch will be coming.

[Gerrit Updater]

Feng, Lei ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41723
Subject: LU-14431 log: Add ending newline for some messages.
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: ae259e4e6f8944a1e87ccb8bfdb84d039ea230cd

[Li Xi]

Li Xi, was your original intent for filing this ticket that Feng Lei have some simple task to become familiar with Lustre? It would have been good to assign it to him in that case.

Yes, the reason why I didn't push a patch instantly was that it looks like a good change for Feng Lei to getting familiar with the patching/pushing/testing/reviewing/merging process of Lustre.

Feng Lei has assigned the ticket to himself when started working on it a few days ago. As far as I know, Feng Lei already has a patch. As well, there are other error messages that need clean up and Feng Lei is working on that too.

simmonsja, I guess Anjus has not finished the patch yet? Does it make sense for Anjus to working on the other tickets in the easy list that Andreas pointed out (or anything else?).

[Andreas Dilger]

Li Xi, James, it definitely doesn't make sense to have two developers duplicating work on this simple patch.

Li Xi, was your original intent for filing this ticket that Feng Lei have some simple task to become familiar with Lustre? It would have been good to assign it to him in that case.

James, I guess the same is true for Anjus, which was confusing since the ticket was not assigned to anyone. It makes sense at this point for both Anjus and Feng Lei to submit whatever patches they each currently have, and then we can work out if there are conflicts.

There are many other tickets labelled "easy" in JIRA, so there is plenty of work to go around without having duplication of effort:
https://jira.whamcloud.com/issues/?jql=project%3DLU%20and%20labels%3Deasy%20and%20resolution%3Dunresolved

Some useful ones to work on would be:

• LU-7105
• LU-6488
• LU-4922
• LU-12515
• LU-9162
• LU-11698
• LU-10552
• LU-4315
• LU-1937
• LU-8962
• LU-8621
• LU-7495

[James A Simmons]

Anjus is working on the patch. Can we reassign the ticket back to her?

[Andreas Dilger]

All of the missing space errors will go away with single-line error strings, which is one reason why they are recommended by checkpatch. The other is that it makes fining error strings in the code much easier, since words in the error will not be split across lines. 

[Li Xi]

Missing space after "but"

lustre/osd-ldiskfs/osd_scrub.c(2671):
                                CERROR("%s: UUID has been changed, but"
                                       "failed to allocate RAM for report\n",
                                       osd_dev2name(dev));

[Li Xi]

Missing space after ":":

lustre/osp/osp_object.c(594):
                         CERROR("%s:osp_attr_get update error "DFID": rc = %d\n",
                                dev->dd_lu_dev.ld_obd->obd_name,
                                PFID(lu_object_fid(&dt->do_lu)), rc);

[Andreas Dilger]

Rather than putting comments in this ticket, it would probably be the same effort to make a patch that fixes them all.

[Li Xi]

1) Space is missing after "%d."
2) Two spaces before "Not fatal"
3) "\" is not necessary

lustre/ptlrpc/llog_client.c(56):
                CERROR("ctxt->loc_imp == NULL for context idx %d."    \
                       "Unable to complete MDS/OSS recovery,"         \
                       "but I'll try again next time.  Not fatal.\n", \
                       ctxt->loc_idx);                                \