On Wed, Mar 5, 2014 at 8:25 PM, Markus Mayer <markus.mayer@xxxxxxxxxx> wrote: > Since sdhci_bcm_kona_sd_card_emulate() can be called from interrupt > context, we must use a spinlock instead of a mutex. This change > essentially restores the way things worked before the change to > mutexes in http://www.spinics.net/lists/linux-mmc/msg20276.html. > > Without this change, we see "scheduling while atomic" errors. > > Signed-off-by: Markus Mayer <markus.mayer@xxxxxxxxxx> > --- So digging back in memory (since I no longer have the broadcom-internal email threads on this topic), I think I recall what the problem is. card_event is being called directly from mmc_gpio_cd_irqt, which is atomic context. My plan for this was to delay the card_event callback into the mmc_rescan worker, as that would have 2 nice results: remove the callback from atomic context which it doesn't need to be in, and put it in line with the debounced rescan (which I think it should be). I made mention of that option in this thread (https://lkml.org/lkml/2013/8/19/539) but seems I never followed up with code... Something along the lines of the (untested) code below might do the trick: diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c index 098374b..d4ffba8 100644 --- a/drivers/mmc/core/core.c +++ b/drivers/mmc/core/core.c @@ -2421,6 +2421,11 @@ void mmc_rescan(struct work_struct *work) container_of(work, struct mmc_host, detect.work); int i; + if (host->trigger_card_event && host->ops->card_event) { + host->ops->card_event(host); + host->trigger_card_event = 0; + } + if (host->rescan_disable) return; diff --git a/drivers/mmc/core/slot-gpio.c b/drivers/mmc/core/slot-gpio.c index 46596b71..fc648ae 100644 --- a/drivers/mmc/core/slot-gpio.c +++ b/drivers/mmc/core/slot-gpio.c @@ -29,9 +29,7 @@ static irqreturn_t mmc_gpio_cd_irqt(int irq, void *dev_id) /* Schedule a card detection after a debounce timeout */ struct mmc_host *host = dev_id; - if (host->ops->card_event) - host->ops->card_event(host); - + host->trigger_card_event = 1; mmc_detect_change(host, msecs_to_jiffies(200)); return IRQ_HANDLED; diff --git a/include/linux/mmc/host.h b/include/linux/mmc/host.h index 99f5709..0b6bddf 100644 --- a/include/linux/mmc/host.h +++ b/include/linux/mmc/host.h @@ -322,6 +322,8 @@ struct mmc_host { int rescan_disable; /* disable card detection */ int rescan_entered; /* used with nonremovable devices */ + int trigger_card_event; /* card_event necessary */ + struct mmc_card *card; /* device attached to this host */ wait_queue_head_t wq; Of course there are (a few) more users of card_event callback than bcm_kona that need to be confirmed ok such a change Thanks, csd -- 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
