Re: [PATCH] sdio: fix suspend/resume regression

[Maxim Levitsky]

On Wed, 2010-10-13 at 09:31 +0200, Ohad Ben-Cohen wrote:
> Fix SDIO suspend/resume regression introduced by
> 4c2ef25fe0b847d2ae818f74758ddb0be1c27d8e "mmc: fix all hangs related to
> mmc/sd card insert/removal during suspend/resume":
> 
> [ 5647.295953] PM: Syncing filesystems ... done.
> [ 5647.318792] Freezing user space processes ... (elapsed 0.01 seconds) done.
> [ 5647.337048] Freezing remaining freezable tasks ... (elapsed 0.01 seconds) done.
> [ 5647.356915] Suspending console(s) (use no_console_suspend to debug)
> [ 5647.366651] pm_op(): platform_pm_suspend+0x0/0x5c returns -38
> [ 5647.366671] PM: Device pxa2xx-mci.0 failed to suspend: error -38
> [ 5647.367082] PM: Some devices failed to suspend
> 
> 4c2ef25fe0b847d2ae818f74758ddb0be1c27d8e moved the card removal/insertion
> mechanism out of MMC's suspend/resume path and into pm notifiers
> (mmc_pm_notify), and that broke SDIO's expectation that mmc_suspend_host()
> will remove the card, and squash the error, in case -ENOSYS is returned
> from the bus suspend handler (mmc_sdio_suspend() in this case).
> 
> mmc_sdio_suspend() is using this whenever at least one of the card's SDIO
> function drivers does not have suspend/resume handlers - in that case
> it is agreed to force removal of the entire card.
> 
> This patch fixes this regression by trivially bringing back that part of
> mmc_suspend_host(), which was removed by 4c2ef25fe0b847d2ae818f74758ddb0be1c27d8e.
> 
> Reported-and-tested-by: Sven Neumann <s.neumann@xxxxxxxxxxxx>
> Signed-off-by: Ohad Ben-Cohen <ohad@xxxxxxxxxx>
> Cc: Maxim Levitsky <maximlevitsky@xxxxxxxxx>
> Cc: <stable@xxxxxxxxxx>
> --
> 
> It may still be desired to further clean this area up by using the card
> removal mechanism in mmc_pm_notify() for SDIO as well.
> 
> To use mmc_pm_notify's card-removal code also for SDIO, we need it
> to check if all the SDIO functions have suspend handlers. That
> would probably make us add a new bus_ops handler (something like
> host->bus_ops->remove_card_on_suspend ?).
> 
> It's starting to get a bit complicated though, and I'm not sure it
> would make the code a lot more readable.
> 
> In addition, this would still not work for drivers like libertas sdio,
> which do have a suspend handler, but sometimes let it return -ENOSYS,
> expecting mmc_suspend_host() to remove the card and squash the error.
> For those cases, we still need the old card-removal logic in mmc_suspend_host().
> 
> This brings up a question whether libertas_sdio really needs this
> functionality; When MMC_PM_KEEP_POWER is not needed, can't it just return 0
> (and as a result the card will be powered down, but not removed) ?
> 
> Until we have an agreement on this, I suggest we at least fix the
> regression with this patch.
> 
> Thanks Sven Neumann for reporting and testing the issue and this patch.
> 
>  drivers/mmc/core/core.c |   13 +++++++++++++
>  1 files changed, 13 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c
> index c94565d..515ff39 100644
> --- a/drivers/mmc/core/core.c
> +++ b/drivers/mmc/core/core.c
> @@ -1682,6 +1682,19 @@ int mmc_suspend_host(struct mmc_host *host)
>  	if (host->bus_ops && !host->bus_dead) {
>  		if (host->bus_ops->suspend)
>  			err = host->bus_ops->suspend(host);
> +		if (err == -ENOSYS || !host->bus_ops->resume) {
This reintroduces the bug I fixed.

if the CONFIG_MMC_UNSAFE_RESUME isn't set (and that is default
unfortunately), the host->bus_ops->resume will be NULL (see core/mmc.c
mmc_ops), and therefore card will be removed, that will trigger a block
device removal, sync, and deadlock).

I actually thought about the sdio case. Sorry for breaking yet.

My idea was to move the effective suspend/resume to a pm notifier,
and the mmc_pm_notify is supposed to do the job.

Could you test why it fails?

The relevant code in mmc_pm_notify:

		if (!host->bus_ops || host->bus_ops->suspend)
			break;

		mmc_claim_host(host);

		if (host->bus_ops->remove)
			host->bus_ops->remove(host);

		mmc_detach_bus(host);
		mmc_release_host(host);
		host->pm_flags = 0;
		break;


So NULL host->bus_ops->suspend should trigger a card remove by that
function and it did here on my system without CONFIG_MMC_UNSAFE_RESUME.

I suspect that in your case, the .suspend isn't NULL, but .resume is.
Then, we just need an one liner change to mmc_pm_notify to account that
case.

Note that I don't call the  host->bus_ops->suspend(host); in
mmc_pm_notify on purpose as it is too early.

So what happens if you set .suspend to NULL? instead of -ENOSYS return?



> +			/*
> +			 * We simply "remove" the card in this case.
> +			 * It will be redetected on resume.
> +			 */
> +			if (host->bus_ops->remove)
> +				host->bus_ops->remove(host);
> +			mmc_claim_host(host);
> +			mmc_detach_bus(host);
> +			mmc_release_host(host);
> +			host->pm_flags = 0;
> +			err = 0;
> +		}
>  	}
>  	mmc_bus_put(host);
>  

Best regards,
	Maxim Levitsky

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/