On Sun, 2010-12-19 at 19:07 -0700, Matthew Wilcox wrote: > On Sun, Dec 19, 2010 at 05:22:06PM -0800, Nicholas A. Bellinger wrote: > > Actually sorry, Mike Christie did already make a clarification on this > > subject here: > > > > http://marc.info/?l=linux-scsi&m=129010439421506&w=2 > > > > I had originally thought the same that session->lock should be using > > some flavour of spin_lock_irq*() as well, but apparently this is not the > > case for libiscsi. > > Right, so it seems. "the session lock is just locked in softirqs/timers" > means that it does need to be the _bh() version of spin_lock though. > > I'm actually not sure ... is it safe to use the _bh versions in BH > context? I think it is because the preempt count is nested, unlike the > _irq variants of spinlocks. > Hmmm, fair point. Merging the following incremental patch to convert session->lock to use spin_lock_bh() in iscsi_queuecommand() into lock_less-LLDs-for-38-v3: commit 744f1c119b3fb0c1a6b3c67a7b490e234d1a7e75 Author: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx> Date: Mon Dec 20 09:17:19 2010 +0000 libiscsi: Convert iscsi_queuecommand to use spin_lock_bh This patch converts iscsi_queuecommand() code to obtain struct iscsi_session->lock using spin_lock_bh() to properly handle bottom-half context operation. Reported-by: Matthew Wilcox <willy@xxxxxxxxxxxxxxx> Signed-off-by: Nicholas A. Bellinger <nab@xxxxxxxxxxxxxxx> After a quick audit of iscsi_session->lock usage, and I see that iscsi_complete_pdu(), iscsi_tmf_timedout(), iscsi_eh_cmd_timed_out(), iscsi_check_transport_timeouts() are using spin_lock(), and iscsi_session_failure() and iscsi_conn_failure() are using spin_lock_irqsave(). Mike and Hannes, would you guys mind commenting on this..? From what I can determine these should all be converted to use spin_lock_bh(), yes..? --nab -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html