From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Vasquez Subject: Re: still BUG's for smp_processor_id() on interrupt Date: Tue, 8 Feb 2005 22:44:35 -0800 Message-ID: <20050209064435.GA4897@plap.qlogic.org> References: <20050209010929.GA2933@us.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Received: from avexch02.qlogic.com ([198.70.193.200]:18289 "EHLO avexch01.qlogic.com") by vger.kernel.org with ESMTP id S261623AbVBIGoN (ORCPT ); Wed, 9 Feb 2005 01:44:13 -0500 Content-Disposition: inline In-Reply-To: <20050209010929.GA2933@us.ibm.com> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Patrick Mansfield Cc: linux-scsi@vger.kernel.org, Andrew Morton On Tue, 08 Feb 2005, Patrick Mansfield wrote: > I'm still getting lots of BUG's for the smp_processor_id, but via the > interrupt function. > It's during the driver's init-time polling for interrupts... > I am running the latest bk, it has your patch to qla_os.c: > > [elm3b79 qla2xxx]$ grep smp_proc qla_os.c > if (_smp_processor_id() == ha->last_irq_cpu || was_empty) > > I'm running on a NUMAQ (sometimes has funky latencies). There are no > errors for simple IO (dd if=/dev/sda of=/dev/null bs=64k). There are about > 40 disk drives attached. > Argg -- that was careless... Again, the smp_processor_id() is used only as a heuristic. The attached patch should quash the noise. Thanks. Signed-off-by: Andrew Vasquez ===== drivers/scsi/qla2xxx/qla_isr.c 1.23 vs edited ===== --- 1.23/drivers/scsi/qla2xxx/qla_isr.c 2005-02-04 10:24:27 -08:00 +++ edited/drivers/scsi/qla2xxx/qla_isr.c 2005-02-08 22:41:08 -08:00 @@ -91,7 +91,7 @@ } spin_unlock_irqrestore(&ha->hardware_lock, flags); - ha->last_irq_cpu = smp_processor_id(); + ha->last_irq_cpu = _smp_processor_id(); ha->total_isr_cnt++; if (test_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags) && @@ -200,7 +200,7 @@ } spin_unlock_irqrestore(&ha->hardware_lock, flags); - ha->last_irq_cpu = smp_processor_id(); + ha->last_irq_cpu = _smp_processor_id(); ha->total_isr_cnt++; if (test_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags) &&