On 7/04/20 10:21 am, Baolin Wang wrote: > On Tue, Apr 7, 2020 at 2:38 PM Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote: >> >> On 3/04/20 10:05 am, Baolin Wang wrote: >>> The SD host controller can process one request in the atomic context if >>> the card is nonremovable, which means we can submit next request in the >>> irq hard handler when using the MMC host software queue to reduce the >>> latency. Thus this patch adds a new API request_atomic() for the host >>> controller, as well as adding support for host software queue to submit >>> a request by the new request_atomic() API. >>> >>> Moreover there is an unusual case that the card is busy when trying to >>> send a command, and we can not polling the card status in interrupt >>> context by using request_atomic() to dispatch requests. Thus we should >>> queue a work to try again in the non-atomic context in case the host >>> releases the busy signal later. >>> >>> Suggested-by: Adrian Hunter <adrian.hunter@xxxxxxxxx> >>> Signed-off-by: Baolin Wang <baolin.wang7@xxxxxxxxx> >> >> >> One minor point below, otherwise: >> >> Acked-by: Adrian Hunter <adrian.hunter@xxxxxxxxx> >> >>> --- >>> drivers/mmc/host/mmc_hsq.c | 29 ++++++++++++++++++++++++++++- >>> drivers/mmc/host/mmc_hsq.h | 1 + >>> include/linux/mmc/host.h | 3 +++ >>> 3 files changed, 32 insertions(+), 1 deletion(-) >>> >>> diff --git a/drivers/mmc/host/mmc_hsq.c b/drivers/mmc/host/mmc_hsq.c >>> index b90b2c9..a57f802 100644 >>> --- a/drivers/mmc/host/mmc_hsq.c >>> +++ b/drivers/mmc/host/mmc_hsq.c >>> @@ -16,11 +16,20 @@ >>> #define HSQ_NUM_SLOTS 64 >>> #define HSQ_INVALID_TAG HSQ_NUM_SLOTS >>> >>> +static void mmc_hsq_retry_handler(struct work_struct *work) >>> +{ >>> + struct mmc_hsq *hsq = container_of(work, struct mmc_hsq, retry_work); >>> + struct mmc_host *mmc = hsq->mmc; >>> + >>> + mmc->ops->request(mmc, hsq->mrq); >>> +} >>> + >>> static void mmc_hsq_pump_requests(struct mmc_hsq *hsq) >>> { >>> struct mmc_host *mmc = hsq->mmc; >>> struct hsq_slot *slot; >>> unsigned long flags; >>> + int ret = 0; >>> >>> spin_lock_irqsave(&hsq->lock, flags); >>> >>> @@ -42,7 +51,24 @@ static void mmc_hsq_pump_requests(struct mmc_hsq *hsq) >>> >>> spin_unlock_irqrestore(&hsq->lock, flags); >>> >>> - mmc->ops->request(mmc, hsq->mrq); >>> + if (mmc->ops->request_atomic) >>> + ret = mmc->ops->request_atomic(mmc, hsq->mrq); >>> + else >>> + mmc->ops->request(mmc, hsq->mrq); >>> + >>> + /* >>> + * If returning BUSY from request_atomic(), which means the card >>> + * may be busy now, and we should change to non-atomic context to >>> + * try again for this unusual case, to avoid time-consuming operations >>> + * in the atomic context. >>> + * >>> + * Note: we just give a warning for other error cases, since the host >>> + * driver will handle them. >>> + */ >>> + if (ret == -EBUSY) >>> + schedule_work(&hsq->retry_work); >>> + else >>> + WARN_ON_ONCE(ret && ret != -EBUSY); >> >> 'ret != -EBUSY' is redundant because it is always true in the 'else' clause. > > Ah, Yes, thanks for pointing this out and I will fix it ine next version. > > By the way, could you help to review patch 2 and 3 in this patch set? Thanks. > I'd like to handle the inhibit wait differently. I will make some patches for that and send them out.