Re: [PATCH v7 3/3] mmc: Add driver for LiteX's LiteSDCard interface

["Gabriel L. Somlo" <gsomlo@xxxxxxxxx>]

On Fri, Jan 07, 2022 at 10:58:40PM +0200, Andy Shevchenko wrote:
> On Fri, Jan 7, 2022 at 10:50 PM Andy Shevchenko
> <andy.shevchenko@xxxxxxxxx> wrote:
> > On Fri, Jan 7, 2022 at 7:08 PM Gabriel L. Somlo <gsomlo@xxxxxxxxx> wrote:
> > > On Fri, Jan 07, 2022 at 12:06:16PM -0500, Gabriel Somlo wrote:
> 
> ...
> 
> > > Any more ordering or devm vs. non-devm mixing violations here? If so,
> > > can you please link me to an example or some docs where I ould figure
> > > out what it is I'm still doing wrong?
> >
> > Device managed resources are attached to the instance of the device
> > object and removed in the order they have been attached to, but with
> > the caveat that they have no clue about non-managed calls in between.
> > Now you may figure out what happens. Ex.:
> >
> > probe()
> >   A
> >   devm_B
> >   C
> >   devm_D
> >
> > remove()
> >   un_C
> >   un_A
> >
> > WRONG!
> 
> For the sake of comprehensivity of the examples the right one(s) depicted below:
> 
> ->probe()
> 
> 1)
>   devm_A
>   devm_B
>   C
>   D
> 
> 2)
>   A
>   B
>   C
>   D
> 
> 3)
>   devm_A
>   devm_B
>   devm_C
>   devm_D
> 
> Hint:
> `git log --no-merges --grep devm_add_action_or_reset`

Thanks again!

As far as I can tell, I *meant* (but failed to) use `devm_request_irq()`,
which would then have justified the absence of `free_irq()` on the
probe() function's error path!

Similarly, I would no longer have to call it during remove() either:

diff --git a/drivers/mmc/host/litex_mmc.c b/drivers/mmc/host/litex_mmc.c
index d96da0bcba55..38952f169a27 100644
--- a/drivers/mmc/host/litex_mmc.c
+++ b/drivers/mmc/host/litex_mmc.c
@@ -486,8 +486,8 @@ static int litex_mmc_irq_init(struct litex_mmc_host *host)
        if (IS_ERR(host->sdirq))
                return PTR_ERR(host->sdirq);
 
-       ret = request_irq(host->irq, litex_mmc_interrupt, 0,
-                         "litex-mmc", host->mmc);
+       ret = devm_request_irq(dev, host->irq, litex_mmc_interrupt, 0,
+                              "litex-mmc", host->mmc);
        if (ret < 0) {
                dev_warn(dev, "IRQ request error %d, using polling\n", ret);
                goto use_polling;
@@ -626,20 +626,16 @@ static int litex_mmc_probe(struct platform_device *pdev)
        return 0;
 
 err:
-       if (host->irq > 0)
-               free_irq(host->irq, mmc);
        mmc_free_host(mmc);
        return ret;
 }
 
 static int litex_mmc_remove(struct platform_device *pdev)
 {
-       struct litex_mmc_host *host = dev_get_drvdata(&pdev->dev);
+       struct litex_mmc_host *host = platform_get_drvdata(pdev);
        struct mmc_host *mmc = host->mmc;
 
        mmc_remove_host(mmc);
-       if (host->irq > 0)
-               free_irq(host->irq, mmc);
        mmc_free_host(mmc);
 
        return 0;


I'll send out v8 shortly, hopeuflly with all ordering and devm-ness
issues fixed.

Thanks,
--Gabriel