From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jens Axboe Subject: Re: [PATCH RESEND] csiostor:Fix locking issues in the function csio_scsim_cleanup_io_lnode Date: Tue, 29 Dec 2015 11:27:42 -0700 Message-ID: <5682D09E.2070700@kernel.dk> References: <1451413394-21278-1-git-send-email-xerofoify@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail-ob0-f172.google.com ([209.85.214.172]:36442 "EHLO mail-ob0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753612AbbL2S1o (ORCPT ); Tue, 29 Dec 2015 13:27:44 -0500 Received: by mail-ob0-f172.google.com with SMTP id ba1so164131034obb.3 for ; Tue, 29 Dec 2015 10:27:44 -0800 (PST) In-Reply-To: <1451413394-21278-1-git-send-email-xerofoify@gmail.com> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Nicholas Krause , JBottomley@odin.com Cc: martin.petersen@oracle.com, anish@chelsio.com, hare@suse.de, davem@davemloft.net, hariprasad@chelsio.com, linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org On 12/29/2015 11:23 AM, Nicholas Krause wrote: > This fixes locking issues in the function csio_scsim_cleanup_io_lnode > by locking around the call to the function csio_csci_gather_active_ios > with the function pair spin_lock_irq/spin_unlock_irq as any function > calling this particular function must do in order to avoid concurrent > threads of execution on the passed structure pointer of type csio_hw > as this structure pointer can be shared across mutliple threads in the > kernel. > > Signed-off-by: Nicholas Krause > --- > drivers/scsi/csiostor/csio_scsi.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c > index 2c4562d..c318855 100644 > --- a/drivers/scsi/csiostor/csio_scsi.c > +++ b/drivers/scsi/csiostor/csio_scsi.c > @@ -1327,7 +1327,9 @@ csio_scsim_cleanup_io_lnode(struct csio_scsim *scm, struct csio_lnode *ln) > sld.level = CSIO_LEV_LNODE; > sld.lnode = ln; > INIT_LIST_HEAD(&ln->cmpl_q); > + spin_lock_irq(&hw->lock); > csio_scsi_gather_active_ios(scm, &sld, &ln->cmpl_q); > + spin_unlock_irq(&hw->lock); All the callers grab that lock, you've just added a deadlock instead. Nick, I've said this before and I'll say this again. Stop sending me patches. They are all untested and broken, you are wasting peoples time. -- Jens Axboe