| Message ID | 20081104192357.2b0dbe22@redhat.com |
|---|---|
| State | RFC, archived |
| Delegated to: | Jeff Garzik |
| Headers | show |
Chuck Ebbert wrote: > via-rhine: prevent oops when requesting an IRQ > > via-rhine requests an IRQ before it's ready to handle an interrupt. > It oopses when CONFIG_DEBUG_SHIRQ is enabled. > > https://bugzilla.redhat.com/show_bug.cgi?id=469303 > > --- > > NOTE: UNTESTED: Is it okay to init the hardware before requesting the > IRQ, or should that be done afterward? First of all, _ideally_ your interrupt handler should be able to any state of software initialization, once registers are mapped. So I would first concentrate on "hardening" the interrupt handler, if feasible. It is certainly a common technique to disable interrupts somehow, during initialization of the hardware. Note, though, that doing things before request_irq() is no real guarantee you are out of danger -- you might be on a shared PCI irq, and your init causes that interrupt to "scream" even though your driver has not registered the irq. Jeff -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, 06 Nov 2008 01:07:41 -0500 Jeff Garzik <jeff@garzik.org> wrote: > Chuck Ebbert wrote: > > via-rhine: prevent oops when requesting an IRQ > > > > via-rhine requests an IRQ before it's ready to handle an interrupt. > > It oopses when CONFIG_DEBUG_SHIRQ is enabled. > > > > https://bugzilla.redhat.com/show_bug.cgi?id=469303 > > > > --- > > > > NOTE: UNTESTED: Is it okay to init the hardware before requesting the > > IRQ, or should that be done afterward? > > First of all, _ideally_ your interrupt handler should be able to any > state of software initialization, once registers are mapped. So I would > first concentrate on "hardening" the interrupt handler, if feasible. > > It is certainly a common technique to disable interrupts somehow, during > initialization of the hardware. Note, though, that doing things before > request_irq() is no real guarantee you are out of danger -- you might be > on a shared PCI irq, and your init causes that interrupt to "scream" > even though your driver has not registered the irq. > The problem with this driver is that it doesn't call alloc_ring() before request_irq() and then it oopses in the interrupt handler trying to access the ring. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Chuck Ebbert wrote: > On Thu, 06 Nov 2008 01:07:41 -0500 > Jeff Garzik <jeff@garzik.org> wrote: > >> Chuck Ebbert wrote: >>> via-rhine: prevent oops when requesting an IRQ >>> >>> via-rhine requests an IRQ before it's ready to handle an interrupt. >>> It oopses when CONFIG_DEBUG_SHIRQ is enabled. >>> >>> https://bugzilla.redhat.com/show_bug.cgi?id=469303 >>> >>> --- >>> >>> NOTE: UNTESTED: Is it okay to init the hardware before requesting the >>> IRQ, or should that be done afterward? >> First of all, _ideally_ your interrupt handler should be able to any >> state of software initialization, once registers are mapped. So I would >> first concentrate on "hardening" the interrupt handler, if feasible. >> >> It is certainly a common technique to disable interrupts somehow, during >> initialization of the hardware. Note, though, that doing things before >> request_irq() is no real guarantee you are out of danger -- you might be >> on a shared PCI irq, and your init causes that interrupt to "scream" >> even though your driver has not registered the irq. >> > > The problem with this driver is that it doesn't call alloc_ring() before > request_irq() and then it oopses in the interrupt handler trying to access > the ring. Allocation should definitely occur prior to request_irq() Jeff -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Index: linux-2.6.27.noarch/drivers/net/via-rhine.c =================================================================== --- linux-2.6.27.noarch.orig/drivers/net/via-rhine.c +++ linux-2.6.27.noarch/drivers/net/via-rhine.c @@ -1141,24 +1141,27 @@ static int rhine_open(struct net_device void __iomem *ioaddr = rp->base; int rc; - rc = request_irq(rp->pdev->irq, &rhine_interrupt, IRQF_SHARED, dev->name, - dev); + rc = alloc_ring(dev); if (rc) return rc; - if (debug > 1) - printk(KERN_DEBUG "%s: rhine_open() irq %d.\n", - dev->name, rp->pdev->irq); - - rc = alloc_ring(dev); - if (rc) { - free_irq(rp->pdev->irq, dev); - return rc; - } alloc_rbufs(dev); alloc_tbufs(dev); rhine_chip_reset(dev); init_registers(dev); + + rc = request_irq(rp->pdev->irq, &rhine_interrupt, IRQF_SHARED, dev->name, + dev); + if (rc) { + free_tbufs(dev); + free_rbufs(dev); + free_ring(dev); + return rc; + } + + if (debug > 1) + printk(KERN_DEBUG "%s: rhine_open() irq %d.\n", + dev->name, rp->pdev->irq); if (debug > 2) printk(KERN_DEBUG "%s: Done rhine_open(), status %4.4x " "MII status: %4.4x.\n",