Message ID | 4B4ECD81.8020205@kernel.org |
---|---|
State | Not Applicable |
Delegated to: | David Miller |
Headers | show |
Hello. Tejun Heo wrote: > Traditional IDE interface sucks in that it doesn't have a reliable IRQ > pending bit, so if the controller raises IRQ while the driver is > expecting it not to, the IRQ won't be cleared and eventually the IRQ > line will be killed by interrupt subsystem. Some controllers have > non-standard mechanism to indicate IRQ pending so that this condition > can be detected and worked around. > > This patch adds an optional operation ->sff_irq_check() which will be > called for each port from the ata_sff_interrupt() if an unexpected > interrupt is received. If the operation returns %true, > ->sff_check_status() and ->sff_irq_clear() will be cleared for the > port. Note that this doesn't mark the interrupt as handled so it > won't prevent IRQ subsystem from killing the IRQ if this mechanism > fails to clear the spurious IRQ. > > This patch also implements ->sff_irq_check() for ata_piix. Note that > this adds slight overhead to shared IRQ operation as IRQs which are > destined for other controllers will trigger extra register accesses to > check whether IDE interrupt is pending but this solves rare screaming > IRQ cases and for some curious reason also helps weird BIOS related > glitch on Samsung n130 as reported in bko#14314. > > http://bugzilla.kernel.org/show_bug.cgi?id=14314 > > Signed-off-by: Tejun Heo <tj@kernel.org> > Reported-by: Johannes Stezenbach <js@sig21.net> > Reported-by: Hans Werner <hwerner4@gmx.de> > Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> > Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com> > --- > This is generalized safe version of the previous patch which only > worked for ata_piix and marked the interrupt as handled. > > Thanks. > > drivers/ata/ata_piix.c | 18 ++++++++++++++++-- > drivers/ata/libata-sff.c | 22 ++++++++++++++++++++++ > include/linux/libata.h | 1 + > 3 files changed, 39 insertions(+), 2 deletions(-) > > Index: ata/include/linux/libata.h > =================================================================== > --- ata.orig/include/linux/libata.h > +++ ata/include/linux/libata.h > @@ -857,6 +857,7 @@ struct ata_port_operations { > unsigned int (*sff_data_xfer)(struct ata_device *dev, > unsigned char *buf, unsigned int buflen, int rw); > u8 (*sff_irq_on)(struct ata_port *); > + bool (*sff_irq_check)(struct ata_port *); > void (*sff_irq_clear)(struct ata_port *); > > void (*bmdma_setup)(struct ata_queued_cmd *qc); > Index: ata/drivers/ata/ata_piix.c > =================================================================== > --- ata.orig/drivers/ata/ata_piix.c > +++ ata/drivers/ata/ata_piix.c > @@ -173,6 +173,7 @@ static int piix_sidpr_scr_read(struct at > unsigned int reg, u32 *val); > static int piix_sidpr_scr_write(struct ata_link *link, > unsigned int reg, u32 val); > +static bool piix_irq_check(struct ata_port *ap); > #ifdef CONFIG_PM > static int piix_pci_device_suspend(struct pci_dev *pdev, pm_message_t mesg); > static int piix_pci_device_resume(struct pci_dev *pdev); > @@ -309,8 +310,13 @@ static struct scsi_host_template piix_sh > ATA_BMDMA_SHT(DRV_NAME), > }; > > -static struct ata_port_operations piix_pata_ops = { > +static struct ata_port_operations piix_base_ops = { > .inherits = &ata_bmdma32_port_ops, > + .sff_irq_check = piix_irq_check, > +}; > + > +static struct ata_port_operations piix_pata_ops = { > + .inherits = &piix_base_ops, > .cable_detect = ata_cable_40wire, > .set_piomode = piix_set_piomode, > .set_dmamode = piix_set_dmamode, > @@ -329,7 +335,7 @@ static struct ata_port_operations ich_pa > }; > > static struct ata_port_operations piix_sata_ops = { > - .inherits = &ata_bmdma32_port_ops, > + .inherits = &piix_base_ops, > }; > Not sure it was worth wasting memory on having 2 identical copies of the struct ata_port_operations... > > static struct ata_port_operations piix_sidpr_sata_ops = { > @@ -962,6 +968,14 @@ static int piix_sidpr_scr_write(struct a > return 0; > } > > +static bool piix_irq_check(struct ata_port *ap) > +{ > + if (unlikely(!ap->ioaddr.bmdma_addr)) > + return false; > + > + return ap->ops->bmdma_status(ap) & ATA_DMA_INTR; > +} > + > I'm not at all sure that old, pre-ICH controllers set this bit also for the PIO mode commands, not only for DMA. And if you didn't make such assumption, I don't see why this can't be generic and placed into libata-sff.c instead... WBR, Sergei -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello, On 01/14/2010 09:42 PM, Sergei Shtylyov wrote: >> +static bool piix_irq_check(struct ata_port *ap) >> +{ >> + if (unlikely(!ap->ioaddr.bmdma_addr)) >> + return false; >> + >> + return ap->ops->bmdma_status(ap) & ATA_DMA_INTR; >> +} >> + >> > > I'm not at all sure that old, pre-ICH controllers set this bit also > for the PIO mode commands, not only for DMA. And if you didn't make such > assumption, I don't see why this can't be generic and placed into > libata-sff.c instead... Because different controllers have different mechanisms for detecting pending IRQ? Also, I think the current round is a bit dangerous in that it might end up accessing registers while the driver is doing polling PIO. Having spurious IRQ clearing support would definitely be helpful for such conditions but it might also lead to silent data corruption which is worse. I'll update the patch series so that the mechanism only kicks in only when the controller is believed to be completely idle. Thanks.
Hello. Tejun Heo wrote: >>> +static bool piix_irq_check(struct ata_port *ap) >>> +{ >>> + if (unlikely(!ap->ioaddr.bmdma_addr)) >>> + return false; >>> + >>> + return ap->ops->bmdma_status(ap) & ATA_DMA_INTR; >>> +} >>> + >>> >>> >> I'm not at all sure that old, pre-ICH controllers set this bit also >> for the PIO mode commands, not only for DMA. And if you didn't make such >> assumption, I don't see why this can't be generic and placed into >> libata-sff.c instead... >> > > Because different controllers have different mechanisms for detecting > pending IRQ? All SFF-8038i (BMIDE) controllers have the same mechanism. They may have some additional interrupt bits though, reflecting the interrupt status in both PIO and DMA mode though. WBR, Sergei -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 01/15/2010 07:23 PM, Sergei Shtylyov wrote: >> Because different controllers have different mechanisms for detecting >> pending IRQ? > > All SFF-8038i (BMIDE) controllers have the same mechanism. They may > have some additional interrupt bits though, reflecting the interrupt > status in both PIO and DMA mode though. Oh, yeah, I was thinking about modern piixs where the bit works as a true IRQ pending bit regardless of command state (it works while even idle). I don't think the original BMIDE IRQ pending bit would be too useful for spurious IRQ detection. It's interlocked with DMA transfer protocol and unless the controller is horribly broken it won't be too useful for spurious IRQ detection. Thanks.
Index: ata/include/linux/libata.h =================================================================== --- ata.orig/include/linux/libata.h +++ ata/include/linux/libata.h @@ -857,6 +857,7 @@ struct ata_port_operations { unsigned int (*sff_data_xfer)(struct ata_device *dev, unsigned char *buf, unsigned int buflen, int rw); u8 (*sff_irq_on)(struct ata_port *); + bool (*sff_irq_check)(struct ata_port *); void (*sff_irq_clear)(struct ata_port *); void (*bmdma_setup)(struct ata_queued_cmd *qc); Index: ata/drivers/ata/ata_piix.c =================================================================== --- ata.orig/drivers/ata/ata_piix.c +++ ata/drivers/ata/ata_piix.c @@ -173,6 +173,7 @@ static int piix_sidpr_scr_read(struct at unsigned int reg, u32 *val); static int piix_sidpr_scr_write(struct ata_link *link, unsigned int reg, u32 val); +static bool piix_irq_check(struct ata_port *ap); #ifdef CONFIG_PM static int piix_pci_device_suspend(struct pci_dev *pdev, pm_message_t mesg); static int piix_pci_device_resume(struct pci_dev *pdev); @@ -309,8 +310,13 @@ static struct scsi_host_template piix_sh ATA_BMDMA_SHT(DRV_NAME), }; -static struct ata_port_operations piix_pata_ops = { +static struct ata_port_operations piix_base_ops = { .inherits = &ata_bmdma32_port_ops, + .sff_irq_check = piix_irq_check, +}; + +static struct ata_port_operations piix_pata_ops = { + .inherits = &piix_base_ops, .cable_detect = ata_cable_40wire, .set_piomode = piix_set_piomode, .set_dmamode = piix_set_dmamode, @@ -329,7 +335,7 @@ static struct ata_port_operations ich_pa }; static struct ata_port_operations piix_sata_ops = { - .inherits = &ata_bmdma32_port_ops, + .inherits = &piix_base_ops, }; static struct ata_port_operations piix_sidpr_sata_ops = { @@ -962,6 +968,14 @@ static int piix_sidpr_scr_write(struct a return 0; } +static bool piix_irq_check(struct ata_port *ap) +{ + if (unlikely(!ap->ioaddr.bmdma_addr)) + return false; + + return ap->ops->bmdma_status(ap) & ATA_DMA_INTR; +} + #ifdef CONFIG_PM static int piix_broken_suspend(void) { Index: ata/drivers/ata/libata-sff.c =================================================================== --- ata.orig/drivers/ata/libata-sff.c +++ ata/drivers/ata/libata-sff.c @@ -1778,6 +1778,28 @@ irqreturn_t ata_sff_interrupt(int irq, v handled |= ata_sff_host_intr(ap, qc); } + /* + * If no port was expecting IRQ but the controller is actually + * asserting IRQ line, nobody cared will ensue. Check IRQ + * pending status if available and clear spurious IRQ. + */ + if (!handled) { + for (i = 0; i < host->n_ports; i++) { + struct ata_port *ap = host->ports[i]; + + if (!ap->ops->sff_irq_check || + !ap->ops->sff_irq_check(ap)) + continue; + + if (printk_ratelimit()) + ata_port_printk(ap, KERN_INFO, + "clearing spurious IRQ\n"); + + ap->ops->sff_check_status(ap); + ap->ops->sff_irq_clear(ap); + } + } + spin_unlock_irqrestore(&host->lock, flags); return IRQ_RETVAL(handled);
Traditional IDE interface sucks in that it doesn't have a reliable IRQ pending bit, so if the controller raises IRQ while the driver is expecting it not to, the IRQ won't be cleared and eventually the IRQ line will be killed by interrupt subsystem. Some controllers have non-standard mechanism to indicate IRQ pending so that this condition can be detected and worked around. This patch adds an optional operation ->sff_irq_check() which will be called for each port from the ata_sff_interrupt() if an unexpected interrupt is received. If the operation returns %true, ->sff_check_status() and ->sff_irq_clear() will be cleared for the port. Note that this doesn't mark the interrupt as handled so it won't prevent IRQ subsystem from killing the IRQ if this mechanism fails to clear the spurious IRQ. This patch also implements ->sff_irq_check() for ata_piix. Note that this adds slight overhead to shared IRQ operation as IRQs which are destined for other controllers will trigger extra register accesses to check whether IDE interrupt is pending but this solves rare screaming IRQ cases and for some curious reason also helps weird BIOS related glitch on Samsung n130 as reported in bko#14314. http://bugzilla.kernel.org/show_bug.cgi?id=14314 Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Johannes Stezenbach <js@sig21.net> Reported-by: Hans Werner <hwerner4@gmx.de> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com> --- This is generalized safe version of the previous patch which only worked for ata_piix and marked the interrupt as handled. Thanks. drivers/ata/ata_piix.c | 18 ++++++++++++++++-- drivers/ata/libata-sff.c | 22 ++++++++++++++++++++++ include/linux/libata.h | 1 + 3 files changed, 39 insertions(+), 2 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html