Hi Nicholas, Thanks for your information, I just tested the MCS feature, actually if just two connections per session, this issue will not happen. 2016-02-26 14:19 GMT+08:00 Nicholas A. Bellinger <nab@xxxxxxxxxxxxxxx>: > Hi Ronghua, > > On Wed, 2016-02-24 at 13:17 +0800, ronghua li wrote: >> Hi all, >> Now I use 3.10 kernel, and set the MaxConnection in session to 8 for >> MCS use, then use windows iscsi initiator to login all these 8 >> connections. > > Wow, that's a lot of connections per session! > > Btw just for reference, using 2x connections per session will easily > saturate the dual e1000e 1 GB/sec ports in your system. > > Anything above 2x for this configuration won't give a performance gain, > and the extra threads + TCP connections + context switches would likely > hurt overall efficiency. > >> after these steps, I use 'echo "0" tpgt_1/enable' to >> disable the target by configFS, then crash happen, call stack is >> attached. >> >> I found that in iscsit_free_session, >> iscsit_cause_connection_reinstatement was called and the input >> parameter "sleep" is set to 1, I think it want to close these >> connections one by one; but when iscsit_close_connection called, it >> want to send SIGINT to all the connections in this session by >> iscsit_stop_session(sess, 0, 0), is there any problem about this? >> When I set the "sleep" parameter of >> iscsit_cause_connection_reinstatement to “0” in iscsit_free_session, >> or annotate the iscsit_stop_session call in iscsit_close_connection, >> this crash will not happen. > > This looks like the following bug that was fixed in July 2015: > > iscsi-target: Fix use-after-free during TPG session shutdown > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/target?id=417c20a9b > > Note this patch has a stable tag for v3.1+, but it's unclear if the > 3.10.0-229.4.2.el7.x86_64 distro kernel you're running with contains the > bug-fix from upstream v3.10.y > > Adding Andy CC' to confirm. > > Thanks for reporting! > -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
