I would be surprised if the "iscsi_full_feature" code was generating
a lot of leaks, because this code should only be run during a login
to the target, and logins should not be occurring too frequently.

A new bug 6657591 deals with one of the recently spotted memory leak,
and the iscsi team do appear to be working on this now, with a fix
expected in Nevada snv_85.
http://bugs.opensolaris.org/view_bug.do?bug_id=6657591

It looks to me like this bug was introduced in snv_56 during
the fixing of another bug:
6491077 - "daemon should use umem_cache_create and friends for speedier 
allocations."
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6491077

In function "trans_cmd_dup()" a "free()" should have been changed
 to a "umem_cache_free()", but it was missed.
http://src.opensolaris.org/source/diff/onnv/onnv-gate/usr/src/cmd/iscsi/iscsitgtd/t10_sam.c?r2=3213&r1=3126
(A good reason for having code reviews!)

I found a couple of interesting links regarding identifying memory
leaks, here:
http://developers.sun.com/solaris/articles/libumem_library.html
http://blogs.sun.com/pnayak/entry/finding_memory_leaks_within_solaris

By the way, the version of the iscsi target in S10u4
does seem to be quite old.  A bug was recently found in S10u4
that had been fixed nearly a year ago in Nevada!
Regards
Nigel Smith
 
 
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to