On Thu, Oct 23, 2014 at 05:49:30PM +0100, Marc Zyngier wrote: > Commit 18dcf433f3de (AHCI: Optimize single IRQ interrupt processing) > switched the single IRQ case of the AHCI driver to use threaded > threaded interrupts. > > During this conversion, only the IRQF_SHARED flag was provided. The net > effect of this in the presence of level interrupts is that the > interrupt will not be masked during the execution of the treaded > handler, leading to a screaming interrupt if the thread is not > scheduled quickly enough, specially in error conditions: [...] > Not good. The culprit is a missing IRQF_ONESHOT flag, which addition > solves the problem entierly. The number of interrupts drops radically: I worried how many devices hanging off the same IRQ line in legacy mode would fail to initialize after this update due to IRQF_ONESHOT mismatch. Given that I did not observe nor anticipate this kind of problem and LKP test shows a perf drop I am thinking may be we need to revert 18dcf43 for now. Tejun? > [Booting Debian Jessie to a prompt] > Before fix: > 88: 3562 0 GIC 88 ahci-sunxi > After fix: > 88: 1992 0 GIC 88 ahci-sunxi > > Tested on a A20 board (ahci-sunxi). > > Cc: Alexander Gordeev <agordeev@xxxxxxxxxx> > Cc: Tejun Heo <tj@xxxxxxxxxx> > Cc: linux-ide@xxxxxxxxxxxxxxx > Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx> > --- > drivers/ata/libahci.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c > index 5eb61c9..49c649f 100644 > --- a/drivers/ata/libahci.c > +++ b/drivers/ata/libahci.c > @@ -2492,7 +2492,7 @@ static int ahci_host_activate_single_irq(struct ata_host *host, int irq, > return rc; > > rc = devm_request_threaded_irq(host->dev, irq, ahci_single_irq_intr, > - ahci_thread_fn, IRQF_SHARED, > + ahci_thread_fn, IRQF_SHARED | IRQF_ONESHOT, > dev_driver_string(host->dev), host); > if (rc) > return rc; > -- > 2.0.4 > -- Regards, Alexander Gordeev agordeev@xxxxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
