Re: [PATCH v2] mmc: pxamci: Simplify pxamci_probe() error handling using devm APIs

[Rakuram Eswaran <rakuram.e96@gmail.com>]

On Thu, 16 Oct 2025 at 14:20, Uwe Kleine-König <u.kleine-koenig@baylibre.com> wrote:
>
> On Wed, Oct 15, 2025 at 12:16:57AM +0530, Rakuram Eswaran wrote:
> > This patch refactors pxamci_probe() to use devm-managed resource
> > allocation (e.g. devm_dma_request_chan()) and dev_err_probe() for
> > improved readability and automatic cleanup on probe failure.
> >
> > This eliminates redundant NULL assignments and manual release logic.
> >
> > This issue was originally reported by Smatch:
> > drivers/mmc/host/pxamci.c:709 pxamci_probe() warn: passing zero to 'PTR_ERR'
> >
> > The warning occurred because a pointer was set to NULL before using
> > PTR_ERR(), leading to PTR_ERR(0) and an incorrect 0 return value.
> > This refactor eliminates that condition while improving overall
> > error handling robustness.
> >
> > Reported-by: kernel test robot <lkp@intel.com>
> > Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
> > Closes: https://lore.kernel.org/r/202510041841.pRlunIfl-lkp@intel.com/
> > Fixes: 58c40f3faf742c ("mmc: pxamci: Use devm_mmc_alloc_host() helper")
> > Suggested-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
> > Signed-off-by: Rakuram Eswaran <rakuram.e96@gmail.com>
> > ---
> >
> > Changes since v1:
> > Following Uwe Kleine-König’s suggestion:
> > - Replaced dma_request_chan() with devm_dma_request_chan() to make DMA
> >   channel allocation devm-managed and avoid manual release paths.
> > - Used dev_err_probe() for improved error reporting and consistent
> >   probe failure handling.
> > - Removed redundant NULL assignments and obsolete goto-based cleanup logic.
> > - Updated commit message to better describe the intent of the change.
> >
> > Testing note:
> > I do not have access to appropriate hardware for runtime testing.
> > Any help verifying on actual hardware would be appreciated.
> >
> > Build and Analysis:
> > This patch was compiled against the configuration file reported by
> > 0day CI in the above link (config: s390-randconfig-r071-20251004) using
> > `s390x-linux-gnu-gcc (Ubuntu 14.2.0-19ubuntu2) 14.2.0`.
> >
> > Static analysis was performed with Smatch to ensure the reported warning
> > no longer reproduces after applying this fix.
> >
> > Command used for verification:
> >   ARCH=s390 CROSS_COMPILE=s390x-linux-gnu- \
> >   ~/project/smatch/smatch_scripts/kchecker ./drivers/mmc/host/pxamci.c
> >
> >  drivers/mmc/host/pxamci.c | 57 +++++++++++++++------------------------
> >  1 file changed, 21 insertions(+), 36 deletions(-)
> >
> > diff --git a/drivers/mmc/host/pxamci.c b/drivers/mmc/host/pxamci.c
> > index 26d03352af63..d03388f64934 100644
> > --- a/drivers/mmc/host/pxamci.c
> > +++ b/drivers/mmc/host/pxamci.c
> > @@ -652,11 +652,14 @@ static int pxamci_probe(struct platform_device *pdev)
> >       host->clkrt = CLKRT_OFF;
> > 
> >       host->clk = devm_clk_get(dev, NULL);
> > -     if (IS_ERR(host->clk)) {
> > -             host->clk = NULL;
> > -             return PTR_ERR(host->clk);
> > -     }
> > +     if (IS_ERR(host->clk))
> > +             return dev_err_probe(dev, PTR_ERR(host->clk),
> > +                                  "Failed to acquire clock\n");
> >
> > +     /*
> > +      * Note that the return value of clk_get_rate() is only valid
> > +      * if the clock is enabled.
> > +      */
>
> The intention of this comment in my WIP suggestion was to point out
> another thing to fix as the precondition for calling clk_get_rate()
> isn't asserted. If you don't want to address this (which is fine),
> drop the comment (or improve my wording to make it more obvious that
> there is something to fix).
>

Hi Uwe,

Sorry for the delayed reply as I was in vacation. 

Ah, got it — I’ll drop the clk_get_rate() comment since it was only a reminder
from your WIP suggestion.

Just to confirm, are you referring to adding a call to clk_prepare_enable()
before clk_get_rate()? I can move the clk_get_rate() call after
clk_prepare_enable(), or drop the comment entirely.

If my understanding is correct, I’ll keep v3 focused on the current set of
fixes and handle the clk_get_rate() precondition (by moving it after
clk_prepare_enable()) in a follow-up patch. That should keep the scope of each
change clean and review-friendly.

> > -out:
> > -     if (host->dma_chan_rx)
> > -             dma_release_channel(host->dma_chan_rx);
> > -     if (host->dma_chan_tx)
> > -             dma_release_channel(host->dma_chan_tx);
>
> I was lazy in my prototype patch and didn't drop the calls to
> dma_release_channel() in pxamci_remove(). For a proper patch this is
> required though.
>
> To continue the quest: Now that I looked at pxamci_remove(): `mmc` is
> always non-NULL, so the respective check can be dropped.
>

Understood. Since pxamci_remove() is only called after successful allocation
and initialization in probe(), mmc will always be a valid pointer. I’ll drop
the if (mmc) check in v3 as it can never be NULL in normal operation, and
remove the dma_release_channel() calls as well.

I’ve prepared a preview of the v3 patch incorporating your previous comments.
Before sending it out formally, I wanted to share it with you to confirm that
the updates look good — especially the cleanup changes in pxamci_remove() and
the dropped clk_get_rate() comment.

static void pxamci_remove(struct platform_device *pdev)
{
	struct mmc_host *mmc = platform_get_drvdata(pdev);
	struct pxamci_host *host = mmc_priv(mmc);

	mmc_remove_host(mmc);

	if (host->pdata && host->pdata->exit)
		host->pdata->exit(&pdev->dev, mmc);

	pxamci_stop_clock(host);
	writel(TXFIFO_WR_REQ|RXFIFO_RD_REQ|CLK_IS_OFF|STOP_CMD|
			END_CMD_RES|PRG_DONE|DATA_TRAN_DONE,
			host->base + MMC_I_MASK);

	dmaengine_terminate_all(host->dma_chan_rx);
	dmaengine_terminate_all(host->dma_chan_tx);
}

Please let me know if anything still needs adjustment before v3.

Best Regards,
Rakuram