> Date: Tue, 05 Feb 2008 13:34:07 PST
> From: John Tracy <[EMAIL PROTECTED]>
> Subject: Re: [storage-discuss] iscsitgt becoming very slow after two
> weeks uptime
> To: [email protected]
> Message-ID: <[EMAIL PROTECTED]>
> Content-Type: text/plain; charset=UTF-8
>
> The problem happened again (as expected). I did apply an IDR patch as
> recommended by Sun,
> patch IDR137189-02 for those who may be interested. After issuing a reboot, I
> then started
> hammering away at the iSCSI targets on this box again. These patches seem to
> primarily have
> to do with ZFS items. I'm thinking that this may be off the mark, as even
> while the performance
> is terrible with iSCSI mounted volumes, local IO operations on the box
> utilizing the same ZFS
> pool are as quick as should be expected.
>
> Just out of curiosity, I generated a core of the iscsitgtd and then ran mdk
> against it. This is the result. I'm not sure where to go from here:
>
> gcore 12529
> gcore: core.12529 dumped
> bash-3.00# mdb core.12529
> Loading modules: [ ld.so.1 libumem.so.1 libavl.so.1 libc.so.1 libnvpair.so.1
> libuutil.so.1 ]
<snip>
> > ::findleaks
> CACHE LEAKED BUFCTL CALLER
> 0000000000484028 239 00000000004ab620 iscsi_full_feature+0x365
iscsi_full_feature allocates memory for iSCSI Additional Header
Segment, and indeed does not free it up in some cases (see the patch
below). Nonetheless, these leaks are on error handling path and should
not show up during normal operations. Did you notice any iscsitgt
error messages in the log?
> ----------------------------------------------------------------------
> Total 239 buffers, 15296 bytes
>
>
> From what I read, this indicates the operations I'm performing are generating
> quite a few
> memory leaks within the iscsitgtd process. Is my understanding correct?
>
> These statistics are gathered before the box exhibits the problem (within an
> hour of restarting
> iscsitgt).
--
Regards,
Andrey
Index: iscsitgtd/iscsi_ffp.c
===================================================================
--- iscsitgtd/iscsi_ffp.c (revision 945)
+++ iscsitgtd/iscsi_ffp.c (working copy)
@@ -98,6 +98,7 @@
(void) snprintf(debug, sizeof (debug),
"CON%x Failed to read in AHS", c->c_num);
queue_str(c->c_mgmtq, Q_CONN_ERRS, msg_log, debug);
+ free(ahs);
return (False);
}
}
@@ -117,6 +118,8 @@
"CON%x CRC error: actual 0x%x v. calc 0x%x",
c->c_num, crc_actual, crc_calculated);
queue_str(c->c_mgmtq, Q_CONN_ERRS, msg_log, debug);
+ if (ahs != NULL)
+ free(ahs);
return (False);
}
}
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
