| Message ID | 1313716221-20136-3-git-send-email-horms@verge.net.au (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Aug 19, 2011 at 10:10 AM, Simon Horman <horms@verge.net.au> wrote: > Provide separate interrupt handlers which may be used by platforms where > SDHI has three interrupt sources. > > This patch also removes the commented-out handling of CRC and other errors. > > Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> > Cc: Magnus Damm <magnus.damm@gmail.com> > Signed-off-by: Simon Horman <horms@verge.net.au> > > --- > +irqreturn_t tmio_mmc_irq(int irq, void *devid) > +{ > + struct tmio_mmc_host *host = devid; > + unsigned int ireg, status; > + > + pr_debug("MMC IRQ begin\n"); > + > + tmio_mmc_card_irq_status(host, &ireg, &status); > + if (__tmio_mmc_card_detect_irq(host, ireg, status)) > + return IRQ_HANDLED; > + if (__tmio_mmc_sdcard_irq(host, ireg, status)) > + return IRQ_HANDLED; > + > + tmio_mmc_sdio_irq(irq, devid); > > -out: > return IRQ_HANDLED; > } > EXPORT_SYMBOL(tmio_mmc_irq); Is there any particular reason for returning early in this interrupt handler? By returning early I mean the "if ... return IRQ_HANDLED" cases above. I realize the old ISR code in the driver does just this, so if the goal is to stay compatible then I guess we should keep this behavior. From my point of view it usually makes more sense to try to handle all events that may be associated with the IRQ. Thanks, / magnus -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Aug 19, 2011 at 12:09:50PM +0900, Magnus Damm wrote: > On Fri, Aug 19, 2011 at 10:10 AM, Simon Horman <horms@verge.net.au> wrote: > > Provide separate interrupt handlers which may be used by platforms where > > SDHI has three interrupt sources. > > > > This patch also removes the commented-out handling of CRC and other errors. > > > > Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> > > Cc: Magnus Damm <magnus.damm@gmail.com> > > Signed-off-by: Simon Horman <horms@verge.net.au> > > > > --- > > > +irqreturn_t tmio_mmc_irq(int irq, void *devid) > > +{ > > + struct tmio_mmc_host *host = devid; > > + unsigned int ireg, status; > > + > > + pr_debug("MMC IRQ begin\n"); > > + > > + tmio_mmc_card_irq_status(host, &ireg, &status); > > + if (__tmio_mmc_card_detect_irq(host, ireg, status)) > > + return IRQ_HANDLED; > > + if (__tmio_mmc_sdcard_irq(host, ireg, status)) > > + return IRQ_HANDLED; > > + > > + tmio_mmc_sdio_irq(irq, devid); > > > > -out: > > return IRQ_HANDLED; > > } > > EXPORT_SYMBOL(tmio_mmc_irq); > > Is there any particular reason for returning early in this interrupt > handler? By returning early I mean the "if ... return IRQ_HANDLED" > cases above. > > I realize the old ISR code in the driver does just this, so if the > goal is to stay compatible then I guess we should keep this behavior. > >From my point of view it usually makes more sense to try to handle all > events that may be associated with the IRQ. My original post had the behaviour that you suggest but Guennadi indicted that he would be much more comfortable with keeping the original behaviour as it is know to work on a wide range of hardware. -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Aug 19, 2011 at 12:30 PM, Simon Horman <horms@verge.net.au> wrote: > On Fri, Aug 19, 2011 at 12:09:50PM +0900, Magnus Damm wrote: >> On Fri, Aug 19, 2011 at 10:10 AM, Simon Horman <horms@verge.net.au> wrote: >> > Provide separate interrupt handlers which may be used by platforms where >> > SDHI has three interrupt sources. >> > >> > This patch also removes the commented-out handling of CRC and other errors. >> > >> > Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> >> > Cc: Magnus Damm <magnus.damm@gmail.com> >> > Signed-off-by: Simon Horman <horms@verge.net.au> >> > >> > --- >> >> > +irqreturn_t tmio_mmc_irq(int irq, void *devid) >> > +{ >> > + struct tmio_mmc_host *host = devid; >> > + unsigned int ireg, status; >> > + >> > + pr_debug("MMC IRQ begin\n"); >> > + >> > + tmio_mmc_card_irq_status(host, &ireg, &status); >> > + if (__tmio_mmc_card_detect_irq(host, ireg, status)) >> > + return IRQ_HANDLED; >> > + if (__tmio_mmc_sdcard_irq(host, ireg, status)) >> > + return IRQ_HANDLED; >> > + >> > + tmio_mmc_sdio_irq(irq, devid); >> > >> > -out: >> > return IRQ_HANDLED; >> > } >> > EXPORT_SYMBOL(tmio_mmc_irq); >> >> Is there any particular reason for returning early in this interrupt >> handler? By returning early I mean the "if ... return IRQ_HANDLED" >> cases above. >> >> I realize the old ISR code in the driver does just this, so if the >> goal is to stay compatible then I guess we should keep this behavior. >> >From my point of view it usually makes more sense to try to handle all >> events that may be associated with the IRQ. > > My original post had the behaviour that you suggest but Guennadi > indicted that he would be much more comfortable with keeping the original > behaviour as it is know to work on a wide range of hardware. I see, thanks for your patience... So I may remember this wrong, but for the 3 different interrupt sources I believe that the SDIO IRQ code was added by Arnd for one of the Renesas SDHI platforms. Back then Ian disliked supporting more than a single interrupt source, so for that reason the SDIO IRQ code was added on top of the common interrupt handler. We had no documentation either, so it was added in a rather random way. I recall the SDIO IRQ being handled before the other interrupt types in the common handler not last, but that's not very important. Anyway, with the fact that SDIO IRQ support was added for SDHI platforms in mind then we can assume that other platforms won't need it. Not sure if this fact will improve our situation or not. As for the two remaining interrupts, I believe they share hardware registers somehow. I guess I'm OK keeping the original behavior somehow, but I still believe it's incorrect. It may not matter very much though since it's rather unlikely that hotplug insertion or eject coincides with the data IRQs. Thanks for your help! / magnus -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Aug 19, 2011 at 01:27:41PM +0900, Magnus Damm wrote: > On Fri, Aug 19, 2011 at 12:30 PM, Simon Horman <horms@verge.net.au> wrote: > > On Fri, Aug 19, 2011 at 12:09:50PM +0900, Magnus Damm wrote: > >> On Fri, Aug 19, 2011 at 10:10 AM, Simon Horman <horms@verge.net.au> wrote: > >> > Provide separate interrupt handlers which may be used by platforms where > >> > SDHI has three interrupt sources. > >> > > >> > This patch also removes the commented-out handling of CRC and other errors. > >> > > >> > Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> > >> > Cc: Magnus Damm <magnus.damm@gmail.com> > >> > Signed-off-by: Simon Horman <horms@verge.net.au> > >> > > >> > --- > >> > >> > +irqreturn_t tmio_mmc_irq(int irq, void *devid) > >> > +{ > >> > + struct tmio_mmc_host *host = devid; > >> > + unsigned int ireg, status; > >> > + > >> > + pr_debug("MMC IRQ begin\n"); > >> > + > >> > + tmio_mmc_card_irq_status(host, &ireg, &status); > >> > + if (__tmio_mmc_card_detect_irq(host, ireg, status)) > >> > + return IRQ_HANDLED; > >> > + if (__tmio_mmc_sdcard_irq(host, ireg, status)) > >> > + return IRQ_HANDLED; > >> > + > >> > + tmio_mmc_sdio_irq(irq, devid); > >> > > >> > -out: > >> > return IRQ_HANDLED; > >> > } > >> > EXPORT_SYMBOL(tmio_mmc_irq); > >> > >> Is there any particular reason for returning early in this interrupt > >> handler? By returning early I mean the "if ... return IRQ_HANDLED" > >> cases above. > >> > >> I realize the old ISR code in the driver does just this, so if the > >> goal is to stay compatible then I guess we should keep this behavior. > >> >From my point of view it usually makes more sense to try to handle all > >> events that may be associated with the IRQ. > > > > My original post had the behaviour that you suggest but Guennadi > > indicted that he would be much more comfortable with keeping the original > > behaviour as it is know to work on a wide range of hardware. > > I see, thanks for your patience... > > So I may remember this wrong, but for the 3 different interrupt > sources I believe that the SDIO IRQ code was added by Arnd for one of > the Renesas SDHI platforms. Back then Ian disliked supporting more > than a single interrupt source, so for that reason the SDIO IRQ code > was added on top of the common interrupt handler. We had no > documentation either, so it was added in a rather random way. I recall > the SDIO IRQ being handled before the other interrupt types in the > common handler not last, but that's not very important. Anyway, with > the fact that SDIO IRQ support was added for SDHI platforms in mind > then we can assume that other platforms won't need it. Not sure if > this fact will improve our situation or not. > > As for the two remaining interrupts, I believe they share hardware > registers somehow. I guess I'm OK keeping the original behavior > somehow, but I still believe it's incorrect. It may not matter very > much though since it's rather unlikely that hotplug insertion or eject > coincides with the data IRQs. Hi Magnus, to be honest I'm unsure if the current behaviour is correct or not. But it does appear to work on a wide range of hardware. And as tmio_mmc_irq() is an implementation of the single-source ISR intended for legacy purposes it does seem reasonable to maintain the existing behaviour. If there is a need to change the behaviour of tmio_mmc_irq() in some way then I think that is best dealt with in a separate patch (series) subsequent to the current series that is intended to introduce broken-out handlers. -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Aug 19, 2011 at 1:59 PM, Simon Horman <horms@verge.net.au> wrote: > On Fri, Aug 19, 2011 at 01:27:41PM +0900, Magnus Damm wrote: >> On Fri, Aug 19, 2011 at 12:30 PM, Simon Horman <horms@verge.net.au> wrote: >> > On Fri, Aug 19, 2011 at 12:09:50PM +0900, Magnus Damm wrote: >> >> On Fri, Aug 19, 2011 at 10:10 AM, Simon Horman <horms@verge.net.au> wrote: >> >> > Provide separate interrupt handlers which may be used by platforms where >> >> > SDHI has three interrupt sources. >> >> > >> >> > This patch also removes the commented-out handling of CRC and other errors. >> >> > >> >> > Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> >> >> > Cc: Magnus Damm <magnus.damm@gmail.com> >> >> > Signed-off-by: Simon Horman <horms@verge.net.au> >> >> > >> >> > --- >> >> >> >> > +irqreturn_t tmio_mmc_irq(int irq, void *devid) >> >> > +{ >> >> > + struct tmio_mmc_host *host = devid; >> >> > + unsigned int ireg, status; >> >> > + >> >> > + pr_debug("MMC IRQ begin\n"); >> >> > + >> >> > + tmio_mmc_card_irq_status(host, &ireg, &status); >> >> > + if (__tmio_mmc_card_detect_irq(host, ireg, status)) >> >> > + return IRQ_HANDLED; >> >> > + if (__tmio_mmc_sdcard_irq(host, ireg, status)) >> >> > + return IRQ_HANDLED; >> >> > + >> >> > + tmio_mmc_sdio_irq(irq, devid); >> >> > >> >> > -out: >> >> > return IRQ_HANDLED; >> >> > } >> >> > EXPORT_SYMBOL(tmio_mmc_irq); >> >> >> >> Is there any particular reason for returning early in this interrupt >> >> handler? By returning early I mean the "if ... return IRQ_HANDLED" >> >> cases above. >> >> >> >> I realize the old ISR code in the driver does just this, so if the >> >> goal is to stay compatible then I guess we should keep this behavior. >> >> >From my point of view it usually makes more sense to try to handle all >> >> events that may be associated with the IRQ. >> > >> > My original post had the behaviour that you suggest but Guennadi >> > indicted that he would be much more comfortable with keeping the original >> > behaviour as it is know to work on a wide range of hardware. >> >> I see, thanks for your patience... >> >> So I may remember this wrong, but for the 3 different interrupt >> sources I believe that the SDIO IRQ code was added by Arnd for one of >> the Renesas SDHI platforms. Back then Ian disliked supporting more >> than a single interrupt source, so for that reason the SDIO IRQ code >> was added on top of the common interrupt handler. We had no >> documentation either, so it was added in a rather random way. I recall >> the SDIO IRQ being handled before the other interrupt types in the >> common handler not last, but that's not very important. Anyway, with >> the fact that SDIO IRQ support was added for SDHI platforms in mind >> then we can assume that other platforms won't need it. Not sure if >> this fact will improve our situation or not. >> >> As for the two remaining interrupts, I believe they share hardware >> registers somehow. I guess I'm OK keeping the original behavior >> somehow, but I still believe it's incorrect. It may not matter very >> much though since it's rather unlikely that hotplug insertion or eject >> coincides with the data IRQs. > > Hi Magnus, > > to be honest I'm unsure if the current behaviour is correct or not. > But it does appear to work on a wide range of hardware. And as > tmio_mmc_irq() is an implementation of the single-source ISR intended > for legacy purposes it does seem reasonable to maintain the existing > behaviour. Sure, I agree. > If there is a need to change the behaviour of tmio_mmc_irq() in some way > then I think that is best dealt with in a separate patch (series) > subsequent to the current series that is intended to introduce broken-out > handlers. Sounds good, I agree that this should be handled separately if needed. Thanks, / magnus -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/mmc/host/tmio_mmc.h b/drivers/mmc/host/tmio_mmc.h index 1cf8db5..3020f98 100644 --- a/drivers/mmc/host/tmio_mmc.h +++ b/drivers/mmc/host/tmio_mmc.h @@ -97,6 +97,9 @@ void tmio_mmc_do_data_irq(struct tmio_mmc_host *host); void tmio_mmc_enable_mmc_irqs(struct tmio_mmc_host *host, u32 i); void tmio_mmc_disable_mmc_irqs(struct tmio_mmc_host *host, u32 i); irqreturn_t tmio_mmc_irq(int irq, void *devid); +irqreturn_t tmio_mmc_sdcard_irq(int irq, void *devid); +irqreturn_t tmio_mmc_card_detect_irq(int irq, void *devid); +irqreturn_t tmio_mmc_sdio_irq(int irq, void *devid); static inline char *tmio_mmc_kmap_atomic(struct scatterlist *sg, unsigned long *flags) diff --git a/drivers/mmc/host/tmio_mmc_pio.c b/drivers/mmc/host/tmio_mmc_pio.c index f0c7830..6275e3d 100644 --- a/drivers/mmc/host/tmio_mmc_pio.c +++ b/drivers/mmc/host/tmio_mmc_pio.c @@ -545,44 +545,20 @@ out: spin_unlock(&host->lock); } -irqreturn_t tmio_mmc_irq(int irq, void *devid) +static void tmio_mmc_card_irq_status(struct tmio_mmc_host *host, + int *ireg, int *status) { - struct tmio_mmc_host *host = devid; - struct mmc_host *mmc = host->mmc; - struct tmio_mmc_data *pdata = host->pdata; - unsigned int ireg, status; - unsigned int sdio_ireg, sdio_status; - - pr_debug("MMC IRQ begin\n"); - - status = sd_ctrl_read32(host, CTL_STATUS); - ireg = status & TMIO_MASK_IRQ & ~host->sdcard_irq_mask; + *status = sd_ctrl_read32(host, CTL_STATUS); + *ireg = *status & TMIO_MASK_IRQ & ~host->sdcard_irq_mask; - sdio_ireg = 0; - if (!ireg && pdata->flags & TMIO_MMC_SDIO_IRQ) { - sdio_status = sd_ctrl_read16(host, CTL_SDIO_STATUS); - sdio_ireg = sdio_status & TMIO_SDIO_MASK_ALL & - ~host->sdio_irq_mask; - - sd_ctrl_write16(host, CTL_SDIO_STATUS, sdio_status & ~TMIO_SDIO_MASK_ALL); - - if (sdio_ireg && !host->sdio_irq_enabled) { - pr_warning("tmio_mmc: Spurious SDIO IRQ, disabling! 0x%04x 0x%04x 0x%04x\n", - sdio_status, host->sdio_irq_mask, sdio_ireg); - tmio_mmc_enable_sdio_irq(mmc, 0); - goto out; - } - - if (mmc->caps & MMC_CAP_SDIO_IRQ && - sdio_ireg & TMIO_SDIO_STAT_IOIRQ) - mmc_signal_sdio_irq(mmc); - - if (sdio_ireg) - goto out; - } + pr_debug_status(*status); + pr_debug_status(*ireg); +} - pr_debug_status(status); - pr_debug_status(ireg); +static bool __tmio_mmc_card_detect_irq(struct tmio_mmc_host *host, + int ireg, int status) +{ + struct mmc_host *mmc = host->mmc; /* Card insert / remove attempts */ if (ireg & (TMIO_STAT_CARD_INSERT | TMIO_STAT_CARD_REMOVE)) { @@ -592,43 +568,102 @@ irqreturn_t tmio_mmc_irq(int irq, void *devid) ((ireg & TMIO_STAT_CARD_INSERT) && !mmc->card)) && !work_pending(&mmc->detect.work)) mmc_detect_change(host->mmc, msecs_to_jiffies(100)); - goto out; + return true; } - /* CRC and other errors */ -/* if (ireg & TMIO_STAT_ERR_IRQ) - * handled |= tmio_error_irq(host, irq, stat); - */ + return false; +} + +irqreturn_t tmio_mmc_card_detect_irq(int irq, void *devid) +{ + unsigned int ireg, status; + struct tmio_mmc_host *host = devid; + tmio_mmc_card_irq_status(host, &ireg, &status); + __tmio_mmc_card_detect_irq(host, ireg, status); + + return IRQ_HANDLED; +} +EXPORT_SYMBOL(tmio_mmc_card_detect_irq); + +static bool __tmio_mmc_sdcard_irq(struct tmio_mmc_host *host, + int ireg, int status) +{ /* Command completion */ if (ireg & (TMIO_STAT_CMDRESPEND | TMIO_STAT_CMDTIMEOUT)) { tmio_mmc_ack_mmc_irqs(host, TMIO_STAT_CMDRESPEND | TMIO_STAT_CMDTIMEOUT); tmio_mmc_cmd_irq(host, status); - goto out; + return true; } /* Data transfer */ if (ireg & (TMIO_STAT_RXRDY | TMIO_STAT_TXRQ)) { tmio_mmc_ack_mmc_irqs(host, TMIO_STAT_RXRDY | TMIO_STAT_TXRQ); tmio_mmc_pio_irq(host); - goto out; + return true; } /* Data transfer completion */ if (ireg & TMIO_STAT_DATAEND) { tmio_mmc_ack_mmc_irqs(host, TMIO_STAT_DATAEND); tmio_mmc_data_irq(host); - goto out; + return true; } - pr_warning("tmio_mmc: Spurious irq, disabling! " - "0x%08x 0x%08x 0x%08x\n", status, host->sdcard_irq_mask, ireg); - pr_debug_status(status); - tmio_mmc_disable_mmc_irqs(host, status & ~host->sdcard_irq_mask); + return false; +} + +irqreturn_t tmio_mmc_sdcard_irq(int irq, void *devid) +{ + unsigned int ireg, status; + struct tmio_mmc_host *host = devid; + + tmio_mmc_card_irq_status(host, &ireg, &status); + __tmio_mmc_sdcard_irq(host, ireg, status); + + return IRQ_HANDLED; +} +EXPORT_SYMBOL(tmio_mmc_sdcard_irq); + +irqreturn_t tmio_mmc_sdio_irq(int irq, void *devid) +{ + struct tmio_mmc_host *host = devid; + struct mmc_host *mmc = host->mmc; + struct tmio_mmc_data *pdata = host->pdata; + unsigned int ireg, status; + + if (!(pdata->flags & TMIO_MMC_SDIO_IRQ)) + return IRQ_HANDLED; + + status = sd_ctrl_read16(host, CTL_SDIO_STATUS); + ireg = status & TMIO_SDIO_MASK_ALL & ~host->sdcard_irq_mask; + + sd_ctrl_write16(host, CTL_SDIO_STATUS, status & ~TMIO_SDIO_MASK_ALL); + + if (mmc->caps & MMC_CAP_SDIO_IRQ && ireg & TMIO_SDIO_STAT_IOIRQ) + mmc_signal_sdio_irq(mmc); + + return IRQ_HANDLED; +} +EXPORT_SYMBOL(tmio_mmc_sdio_irq); + +irqreturn_t tmio_mmc_irq(int irq, void *devid) +{ + struct tmio_mmc_host *host = devid; + unsigned int ireg, status; + + pr_debug("MMC IRQ begin\n"); + + tmio_mmc_card_irq_status(host, &ireg, &status); + if (__tmio_mmc_card_detect_irq(host, ireg, status)) + return IRQ_HANDLED; + if (__tmio_mmc_sdcard_irq(host, ireg, status)) + return IRQ_HANDLED; + + tmio_mmc_sdio_irq(irq, devid); -out: return IRQ_HANDLED; } EXPORT_SYMBOL(tmio_mmc_irq);
Provide separate interrupt handlers which may be used by platforms where SDHI has three interrupt sources. This patch also removes the commented-out handling of CRC and other errors. Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de> Cc: Magnus Damm <magnus.damm@gmail.com> Signed-off-by: Simon Horman <horms@verge.net.au> --- * SDCARD portion tested on AP4/Mackerel * SDIO portion untested v4 * As suggested by Guennadi Liakhovetski - Use bool as return type for __tmio_mmc_sdcard_irq() and __tmio_mmc_card_detect_irq() v3 * Rebase for updated "mmc: tmio: Cache interrupt masks" * As suggested by Guennadi Liakhovetski - Do not alter logic to handle more than one interupt at once - Add missing "static" to declartion of __tmio_mmc_sdcard_irq() v2 * As suggested by Guennadi Liakhovetski - Combine 3 patches into one - Reduce the number of __tmio_..._irq() functions - Rename "...card_access..." functions as "...sdcard..." --- drivers/mmc/host/tmio_mmc.h | 3 + drivers/mmc/host/tmio_mmc_pio.c | 131 ++++++++++++++++++++++++-------------- 2 files changed, 86 insertions(+), 48 deletions(-)