mmc: sdio: add a delay to call sdio_irq_work when sdio bus resume
PATCH v2 1/1 mmc: sdio: add a delay to call sdio_irq_work when sdio bus resume
From: Haibo Chen <[email protected]>
Find this issue on i.MX7D-sdb board with broadcom sdio wifi. When system resume, sometimes this wifi meet the tuning fail issue. All tuning command get command timeout error. This is because sdio_irq_work on system_wq was executed before the broadcom wifi driver resume. Due to broadcom wifi driver set the wifi in Sleep sate in system suspend, need to set the wifi to Wake state first. So need to make sure wifi driver resume first, then do the sdio_irq_work.
--- drivers/mmc/core/sdio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/mmc/core/sdio.c +1 -1
@@ -1047,7 +1047,7 @@ static int mmc_sdio_resume(struct mmc_host *host)if (!(host->caps2 &MMC_CAP2_SDIO_IRQ_NOTHREAD ))wake_up_process (host->sdio_irq_thread);else if (host->caps &MMC_CAP_SDIO_IRQ ) -queue_delayed_work (system_wq, &host->sdio_irq_work,0 ); +queue_delayed_work (system_wq, &host->sdio_irq_work,1 ); } out:
Signature · 1 line
2.17.1
Quoted 23 lines · On Wed, 19 Aug 2020 at 13:24, <[email protected]> wrote:
From: Haibo Chen <[email protected]>
Find this issue on i.MX7D-sdb board with broadcom sdio wifi. When system resume, sometimes this wifi meet the tuning fail issue. All tuning command get command timeout error. This is because
sdio_irq_workonsystem_wqwas executed before the broadcom wifi driver resume. Due to broadcom wifi driver set the wifi in Sleep sate in system suspend, need to set the wifi to Wake state first. So need to make sure wifi driver resume first, then do thesdio_irq_work.Fixes 51133850bce2 ("mmc: core: Fixup processing of SDIO IRQs during system suspend/resume")Signed-off-by Haibo Chen <[email protected]>--- drivers/mmc/core/sdio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/mmc/core/sdio.c +0
-0
drivers/mmc/core/sdio.c static int mmc_sdio_resume(struct mmc_host *host) · line 1050
Quoted 5 lines
drivers/mmc/core/sdio.c +1
-1@@ -1047,7 +1047,7 @@ static int mmc_sdio_resume(struct mmc_host *host)if (!(host->caps2 &MMC_CAP2_SDIO_IRQ_NOTHREAD ))wake_up_process (host->sdio_irq_thread);else if (host->caps &MMC_CAP_SDIO_IRQ ) -queue_delayed_work (system_wq, &host->sdio_irq_work,0 ); +queue_delayed_work (system_wq, &host->sdio_irq_work,1 );
This looks fragile, as there is really no guarantee that the WiFi driver is resumed prior to the SDIO irq work being executed.
Instead, it looks like you shouldn't keep the SDIO irqs enabled during system suspend, because those aren't system wakeup capable. Correct?
In other words, I think the WiFi driver should call sdio_release_irq() in the ->suspend() callback and sdio_claim_irq() in the ->resume() callback.
Quoted 6 lines
drivers/mmc/core/sdio.c +0
-1} out: --2.17.1
Kind regards
Uffe
Quoted 40 lines
-----Original Message-----
From: Ulf Hansson [mailto:[email protected]]
Sent: 2020年8月21日 15:59
To: Bough Chen <[email protected]>
Cc: Adrian Hunter <[email protected]>; [email protected]; dl-linux-imx <[email protected]>; Matthias Kaehlcke <[email protected]>; [email protected]; Doug Anderson <[email protected]>; Pali Rohár <[email protected]>
Subject: Re: [PATCH v2] mmc: sdio: add a delay to callsdio_irq_workwhen sdio bus resumedrivers/mmc/core/sdio.c static int mmc_sdio_resume(struct mmc_host
Quoted 23 lines · On Wed, 19 Aug 2020 at 13:24, <[email protected]> wrote:
From: Haibo Chen <[email protected]>
Find this issue on i.MX7D-sdb board with broadcom sdio wifi. When system resume, sometimes this wifi meet the tuning fail issue. All tuning command get command timeout error. This is because
sdio_irq_workonsystem_wqwas executed before the broadcom wifi driver resume. Due to broadcom wifi driver set the wifi in Sleep sate in system suspend, need to set the wifi to Wake state first. So need to make sure wifi driver resume first, then do thesdio_irq_work.Fixes 51133850bce2 ("mmc: core: Fixup processing of SDIO IRQs duringsystem suspend/resume")
Signed-off-by Haibo Chen <[email protected]>--- drivers/mmc/core/sdio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/mmc/core/sdio.c index renamed from drivers/mmc/core/sdio.c +0
-0
7b40553d3934..101632617f69100644drivers/mmc/core/sdio.c +0
-0@@ -1047,7 +1047,7 @@ static int mmc_sdio_resume(struct mmc_host*host)
drivers/mmc/core/sdio.c static int mmc_sdio_resume(struct mmc_host *host)
Quoted 4 lines
drivers/mmc/core/sdio.c +0
-1if (!(host->caps2 &MMC_CAP2_SDIO_IRQ_NOTHREAD ))wake_up_process (host->sdio_irq_thread);else if (host->caps &MMC_CAP_SDIO_IRQ ) - queue_delayed_work(system_wq,
Quoted 6 lines
&host->sdio_irq_work, 0);drivers/mmc/core/sdio.c static int mmc_sdio_resume(struct mmc_host · line 1051
Quoted 2 lines
drivers/mmc/core/sdio.c +2
-0+queue_delayed_work (system_wq, + &host->sdio_irq_work,1 );This looks fragile, as there is really no guarantee that the WiFi driver is resumed prior to the SDIO irq work being executed.
Yes, agree.
Quoted 6 lines
Instead, it looks like you shouldn't keep the SDIO irqs enabled during system suspend, because those aren't system wakeup capable. Correct?
In other words, I think the WiFi driver should call
sdio_release_irq()in the ->suspend() callback andsdio_claim_irq()in the ->resume() callback.
Let me double check the WIFI driver. Thanks for the suggestion!
Best Regards
Haibo Chen
Quoted 10 lines
Quoted 7 lines
drivers/mmc/core/sdio.c +0
-1} out: --2.17.1
Kind regards
Uffe