Return-path: Received: from ipmail05.adl6.internode.on.net ([150.101.137.143]:45083 "EHLO ipmail05.adl6.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753685AbbBZXhr (ORCPT ); Thu, 26 Feb 2015 18:37:47 -0500 Message-ID: <54EFAC4B.6080002@redmandi.dyndns.org> Date: Fri, 27 Feb 2015 10:29:15 +1100 From: Brendan McGrath MIME-Version: 1.0 To: Steven Toth CC: Mauro Carvalho Chehab , Hans Verkuil , Linux-Media Subject: [PATCHv2] [media] saa7164: use an MSI interrupt when available References: <54EE90BF.2030602@redmandi.dyndns.org> In-Reply-To: Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-media-owner@vger.kernel.org List-ID: From: Brendan McGrath [media] saa7164: use an MSI interrupt when available Enhances driver to use an MSI interrupt when available. Adds the module option 'enable_msi' (type bool) which by default is enabled. Can be set to 'N' to disable. Fixes (or can reduce the occurrence of) a crash which is most commonly reported when multiple saa7164 chips are in use. Signed-off-by: Brendan McGrath --- Thanks for the feedback Steven (and thanks for all your hard work on this driver!). And thank-you Kyle for taking the time to test this patch. Much appreciated. drivers/media/pci/saa7164/saa7164-core.c | 40 ++++++++++++++++++++++++++++++-- drivers/media/pci/saa7164/saa7164.h | 1 + 2 files changed, 39 insertions(+), 2 deletions(-) diff --git a/drivers/media/pci/saa7164/saa7164-core.c b/drivers/media/pci/saa7164/saa7164-core.c index 4b0bec3..1aff2ee 100644 --- a/drivers/media/pci/saa7164/saa7164-core.c +++ b/drivers/media/pci/saa7164/saa7164-core.c @@ -85,6 +85,11 @@ module_param(guard_checking, int, 0644); MODULE_PARM_DESC(guard_checking, "enable dma sanity checking for buffer overruns"); +static bool enable_msi = true; +module_param(enable_msi, bool, 0444); +MODULE_PARM_DESC(enable_msi, + "enable the use of an msi interrupt if available"); + static unsigned int saa7164_devcount; static DEFINE_MUTEX(devlist); @@ -1230,8 +1235,34 @@ static int saa7164_initdev(struct pci_dev *pci_dev, goto fail_irq; } - err = request_irq(pci_dev->irq, saa7164_irq, - IRQF_SHARED, dev->name, dev); + /* irq bit */ + if (enable_msi) + err = pci_enable_msi(pci_dev); + + if (!err && enable_msi) { + /* no error - so request an msi interrupt */ + err = request_irq(pci_dev->irq, saa7164_irq, 0, + dev->name, dev); + + if (err) { + /* fall back to legacy interrupt */ + printk(KERN_ERR "%s() Failed to get an MSI interrupt." + " Falling back to a shared IRQ\n", __func__); + pci_disable_msi(pci_dev); + } else { + dev->msi = true; + } + } + + if ((!enable_msi) || err) { + dev->msi = false; + /* if we have an error (i.e. we don't have an interrupt) + or msi is not enabled - fallback to shared interrupt */ + + err = request_irq(pci_dev->irq, saa7164_irq, + IRQF_SHARED, dev->name, dev); + } + if (err < 0) { printk(KERN_ERR "%s: can't get IRQ %d\n", dev->name, pci_dev->irq); @@ -1441,6 +1472,11 @@ static void saa7164_finidev(struct pci_dev *pci_dev) /* unregister stuff */ free_irq(pci_dev->irq, dev); + if (dev->msi) { + pci_disable_msi(pci_dev); + dev->msi = false; + } + mutex_lock(&devlist); list_del(&dev->devlist); mutex_unlock(&devlist); diff --git a/drivers/media/pci/saa7164/saa7164.h b/drivers/media/pci/saa7164/saa7164.h index cd1a07c..6df4b252 100644 --- a/drivers/media/pci/saa7164/saa7164.h +++ b/drivers/media/pci/saa7164/saa7164.h @@ -459,6 +459,7 @@ struct saa7164_dev { /* Interrupt status and ack registers */ u32 int_status; u32 int_ack; + u32 msi; struct cmd cmds[SAA_CMD_MAX_MSG_UNITS]; struct mutex lock; -- 1.9.1 On 27/02/15 02:12, Steven Toth wrote: >> I believe the root cause of the crash is due to a DMA/IRQ race condition. It >> most commonly occurs when the saa7164 driver is dealing with more than one >> saa7164 chip (the HVR-2200 and HVR-2250 for example have two - one for each >> tuner). Given MSI avoids DMA/IRQ race conditions - this would explain why >> the patch works as a fix. > Brendan, thanks. > > With MSI I've had some people report complete success, others still > have the issues. > > In my experience this does help with i2c timeout issues but not > completely in every case. I've also seen it with single card instances > so you descripton above is close - but not quiet accurate in all > cases. > > While I'm generally OK with changing the driver behaviour to enable > MSI by default, please add a module option to allow the behaviour to > be disabled, reverting the driver back to existing behaviour. > > Once this is done, I'll be happy to Ack it. > > Thanks again. > > - Steve >