Escalated to Bugzilla from IssueTracker
The problem here is that there is some sort of leand in the slabcache. Specifically iy appears that the names_cache it consuming all of lowmem which results in an OOM kill. Please get me an entire /proc/slabinfo output when the OOM kill occurs. Please attach that output as well as the show_mem() output that appears on the console when the OOM kill occurs. BTW, is this a regression from RHEL4-U1 or did the same problem happen when you ran this application/test on that kernel? Larry Woodman
Let me re-write me last comment im "english" !!! The problem here is that there is some sort of leak in the slabcache. Specifically it appears that the names_cache is consuming all of lowmem which results in an OOM kill. Please get me an entire /proc/slabinfo output when the OOM kill occurs. Please attach that output as well as the show_mem() output that appears on the console when the OOM kill occurs. BTW, is this problem a regression from RHEL4-U1 or did the same problem happen when you ran this application/test on that kernel? Larry Woodman
Can we get a repeoducer from IBM here??? Larry Woodman
Created attachment 119820 [details] FSRACER testcase from IBM.
Created attachment 120367 [details] Patch for fs/namei.c This patch is more complete and based on the patch which Rick had submitted to mainline and has been accepted.