[LU-17451] `lctl dl` with Netlink/YAML fails with large numer of devices

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Severity: 3
• Assignee: James A Simmons
• Reporter: Tim Day

Description

See the following error:

[centos][127][~/Programming/lustre-release]$ sudo ./lustre/utils/lctl dl
  0 UP osd-zfs lustre-MDT0000-osd lustre-MDT0000-osd_UUID 10
  1 UP mgs MGS MGS 5
  2 UP mgc MGC172.31.20.187@tcp 9de548fc-f02c-405d-99b8-56dc3e23d83a 5
  3 UP mds MDS MDS_uuid 3
...
158 UP obd_test obd_name_136 obd_uuid_136 3
159 UP obd_test obd_name_137 obd_uuid_137 3
160 UP obd_test obd_name_138 obd_uuid_138 3
161 UP obd_test obd_name_139 obd_uuid_139 3
162 UP obd_test obd_name_140 obd_uuid_140 3
163 UP obd_test obd_name_141 obd_uuid_141 3
Scanner error: while scanning a simple key at line 990, column 3
could not find expected ':' at line 991, column 1
free(): invalid next size (normal)
Aborted 

It appears like whatever Netlink is returning isn't being correctly interpreted as a YAML. This only happens when using a large number of devices.

Issue Links

• is related to: LU-9680 Improve the user land to kernel space interface for lustre

Comments

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/54152
Subject: LU-17451 obd: calculate the needed skbuf size for devices
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 2f780143558216a01502d663d069754bccdf32c7

[James A Simmons]

I know what the issue is. I just pushed a patch for the library to handle the extra data but the kernel needs to be updated to set the cb->dump_min_alloc. I can push a patch for that.