On 24 February 2015 at 03:42, NeilBrown <neilb@xxxxxxx> wrote: > According to section 7.1.2 of > > http://www.sandisk.com/media/File/OEM/Manuals/SD_SDIO_specsv1.pdf > > In the case where the interrupt mechanism is used to wake the host while > the card is in a low power state (i.e. no clocks), Both the card and the > host shall be placed into the 1-bit SD mode prior to stopping the clock. > > This is particularly important for the Marvell "libertas" wifi chip > in the GTA04. While in 4-bit mode it will only signal an interrupt > when the clock is running (which is why setting CLKEXTFREE is > important). > In 1-bit mode, the interrupt is asynchronous (explained in OMAP3 > TRM description of the CIRQ flag to MMCHS_STAT: > > In 1-bit mode, interrupt source is asynchronous (can be a source of > asynchronous wakeup). > In 4-bit mode, interrupt source is sampled during the interrupt > cycle. > > ) > > It is awkward to simply set 1-bit mode in ->runtime_suspend > as that will call mmc_set_ios which calls ops->set_ios(), > which will likely call pm_runtime_get_sync(), on the device that > is currently suspending. This deadlocks. > > So: > - create a work_struct to schedule setting of 1-bit mode > - introduce an 'sdio_narrowed' state flag which transitions: > 0 (normal) -> 1 (convert to 1-bit pending) -> > 2 (have switch to 1-bit mode) -> 0 (normal) > - create a function mmc_sdio_want_no_clocks() which can be called > when the driver wants to turn off clocks (presumably after an > idle timeout). This either succeeds (in 1-bit mode) or fails > and schedules the work to switch to 1-bit mode. > - when the host is claimed, if sdio_narrowed is 2, restore the > 4-bit bus > - When the host is released, if sdio_narrowed is 1, then some > caller other than our worker claimed the host first, so > clear sdio_narrowed. > > This all allows a graceful and race-free switch to 1-bit mode > before switching off the clocks, if SDIO interrupts are enabled. > > A host should call mmc_sdio_want_no_clocks() when about to turn off > clocks if sdio interrupts are enabled, and the ->disable() function > should not use a timeout (pm_runtime_put_autosuspend) if > ->sdio_narrowed is 2. > > Signed-off-by: NeilBrown <neil@xxxxxxxxxx> Hi Neil, Thanks for sending this patchset, and sorry for the delay in response. Overall I like the approach taken in this patchset, though I have some questions see below. > --- > drivers/mmc/core/core.c | 18 ++++++++++++++---- > drivers/mmc/core/sdio.c | 42 +++++++++++++++++++++++++++++++++++++++++- > include/linux/mmc/core.h | 2 ++ > include/linux/mmc/host.h | 2 ++ > 4 files changed, 59 insertions(+), 5 deletions(-) > > diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c > index 541c8903dc6b..0258fdf1a03d 100644 > --- a/drivers/mmc/core/core.c > +++ b/drivers/mmc/core/core.c > @@ -921,8 +921,14 @@ int __mmc_claim_host(struct mmc_host *host, atomic_t *abort) > wake_up(&host->wq); > spin_unlock_irqrestore(&host->lock, flags); > remove_wait_queue(&host->wq, &wait); > - if (host->ops->enable && !stop && host->claim_cnt == 1) > - host->ops->enable(host); > + if (!stop && host->claim_cnt == 1) { > + if (host->ops->enable) > + host->ops->enable(host); > + if (atomic_read(&host->sdio_narrowed) == 2) { > + sdio_enable_4bit_bus(host->card); > + atomic_set(&host->sdio_narrowed, 0); > + } > + } > return stop; > } > > @@ -941,8 +947,12 @@ void mmc_release_host(struct mmc_host *host) > > WARN_ON(!host->claimed); > > - if (host->ops->disable && host->claim_cnt == 1) > - host->ops->disable(host); > + if (host->claim_cnt == 1) { > + if (atomic_read(&host->sdio_narrowed) == 1) > + atomic_set(&host->sdio_narrowed, 0); > + if (host->ops->disable) > + host->ops->disable(host); > + } As omap_hsmmc in currently the only user of the host_ops->enable|disable() callbacks, I wonder if this approach will work "race-free" for those hosts that don't implement these callbacks? Also, I was kind of hoping that we should be able to remove these host_ops callbacks, when omap_hsmmc has moved away from using them. Do you think that can be done? Within this context, I am also reviewing Adrian Hunter's patchset around "periodic re-tuning" [1] and to me, you guys are sharing a similar issue. The host driver's runtime PM suspend callback needs to be able to notify the mmc core to take some specific actions, and in a "race-free" manner. I wonder whether we could have one common solution to that? Do you think this approach will work for Adrian's "periodic re-tuning" as well? > > spin_lock_irqsave(&host->lock, flags); > if (--host->claim_cnt) { > diff --git a/drivers/mmc/core/sdio.c b/drivers/mmc/core/sdio.c > index 5bc6c7dbbd60..9761e4d5f49b 100644 > --- a/drivers/mmc/core/sdio.c > +++ b/drivers/mmc/core/sdio.c > @@ -288,7 +288,7 @@ static int sdio_disable_wide(struct mmc_card *card) > } > > > -static int sdio_enable_4bit_bus(struct mmc_card *card) > +int sdio_enable_4bit_bus(struct mmc_card *card) > { > int err; > > @@ -313,6 +313,45 @@ static int sdio_enable_4bit_bus(struct mmc_card *card) > return err; > } > > +static void mmc_sdio_width_work(struct work_struct *work) > +{ > + struct mmc_host *host = container_of(work, struct mmc_host, > + sdio_width_work); > + atomic_t noblock; > + > + atomic_set(&noblock, 1); > + if (__mmc_claim_host(host, &noblock)) > + return; > + if (atomic_read(&host->sdio_narrowed) != 1) { > + /* Nothing to do */ > + mmc_release_host(host); > + return; > + } > + if (sdio_disable_wide(host->card) == 0) > + atomic_set(&host->sdio_narrowed, 2); > + else > + atomic_set(&host->sdio_narrowed, 0); > + mmc_release_host(host); > +} > + > +int mmc_sdio_want_no_clocks(struct mmc_host *host) > +{ > + if (!(host->caps & MMC_CAP_SDIO_IRQ) || > + host->ios.bus_width == MMC_BUS_WIDTH_1) > + /* Safe to turn off clocks */ > + return 1; > + > + > + /* In 4-bit mode the card needs the clock > + * to deliver interrupts, so it isn't safe > + * to turn of clocks just yet > + */ > + atomic_add_unless(&host->sdio_narrowed, 1, 1); > + > + schedule_work(&host->sdio_width_work); > + return 0; > +} > +EXPORT_SYMBOL_GPL(mmc_sdio_want_no_clocks); > > /* > * Test if the card supports high-speed mode and, if so, switch to it. > @@ -1100,6 +1139,7 @@ int mmc_attach_sdio(struct mmc_host *host) > goto err; > } > > + INIT_WORK(&host->sdio_width_work, mmc_sdio_width_work); > /* > * Detect and init the card. > */ > diff --git a/include/linux/mmc/core.h b/include/linux/mmc/core.h > index 160448f920ac..faf6d1be0971 100644 > --- a/include/linux/mmc/core.h > +++ b/include/linux/mmc/core.h > @@ -212,4 +212,6 @@ struct device_node; > extern u32 mmc_vddrange_to_ocrmask(int vdd_min, int vdd_max); > extern int mmc_of_parse_voltage(struct device_node *np, u32 *mask); > > +extern int sdio_enable_4bit_bus(struct mmc_card *card); > +extern int mmc_sdio_want_no_clocks(struct mmc_host *host); > #endif /* LINUX_MMC_CORE_H */ > diff --git a/include/linux/mmc/host.h b/include/linux/mmc/host.h > index 0c8cbe5d1550..7e6a54c49a15 100644 > --- a/include/linux/mmc/host.h > +++ b/include/linux/mmc/host.h > @@ -350,6 +350,8 @@ struct mmc_host { > struct task_struct *sdio_irq_thread; > bool sdio_irq_pending; > atomic_t sdio_irq_thread_abort; > + struct work_struct sdio_width_work; > + atomic_t sdio_narrowed; /* 1==pending, 2==complete*/ > > mmc_pm_flag_t pm_flags; /* requested pm features */ > > > Kind regards Uffe [1] [PATCH V3 00/15] mmc: host: Add facility to support re-tuning http://www.spinics.net/lists/linux-mmc/msg31020.html -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
