Re: [PATCH] base/platform: fix panic when probe function is NULL

[Wilck, Martin]

Hello Uwe,

thanks for your review.

> This may cause a panic later. For example, inserting the tpm_tis
> > driver with parameter "force=1" (i.e. registering tpm_tis as a platform
> > driver) will panic in tpmm_chip_alloc() because dev->driver is NULL:
> > 
> >      chip->cdev.owner = chip->pdev->driver->owner;
> 
> This sounds like a separate issue though. Looking at init_tis there is:
> 
>         rc = platform_driver_register(&tis_drv);
>         if (rc < 0)
>                 return rc;
>         pdev = platform_device_register_simple("tpm_tis", -1, NULL, 0);
>         if (IS_ERR(pdev)) {
>                 rc = PTR_ERR(pdev);
>                 goto err_dev;
>         }
>         rc = tpm_tis_init(&pdev->dev, &tis_default_info, NULL);
> 
> tpm_tis_init calls tpmm_chip_alloc which barfs when pdev (i.e. the return value
> of platform_device_register_simple above) isn't bound. It is not allowed
> to assume that the device is bound after the above function calls.

I agree that the TPM platform device code deserves improvement. Jason
wrote that he has already some patches available for that.

I lack the knowledge to judge whether or not tpm_is_init's assumption
was correct. But, maybe just by luck, this assumption used to be *true*
until patch b8b2c7d845d5. Driver and device were matched by name
("tpm_tis") by the platform driver probing code, and device and driver
were actually bound to each other after this sequence of calls. 

> So I'd say drop the paragraph about tpm_tis and the change is fine.

I didn't mean to blame your patch. But a note about the panic might be
helpful just in case someone else runs into the same problem. The
connection between your patch and tpm_tis loading is far from obvious.
I mentioned the panic in order to clarify that this wasn't just a
theoretical issue.

Anyway, I'll resubmit with your style hints applied and will try to find
a wording for the commit message that we can agree upon.

Best Regards,
Martin

> 
> > This patch fixes this by returning success in platform_drv_probe() if
> > "just" dev_pm_domain_attach() had failed. This restores the semantics
> > of platform_device_register_XXX() if the associated platform driver has
> > no "probe" function.
> > 
> > Fixes: b8b2c7d845d5 ("base/platform: assert that dev_pm_domain
> > callbacks are called unconditionally")
> > 
> 
> I think line breaks in the Fixes: line are frowned on. Also usually
> there is no empty line between Fixes: and S-o-b:.
> 
> > Signed-off-by: Martin Wilck <Martin.Wilck@xxxxxxxxxxxxxx>
> > ---
> >  drivers/base/platform.c | 12 ++++++++----
> >  1 file changed, 8 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/base/platform.c b/drivers/base/platform.c
> > index 1dd6d3b..c994e76 100644
> > --- a/drivers/base/platform.c
> > +++ b/drivers/base/platform.c
> > @@ -513,10 +513,14 @@ static int platform_drv_probe(struct device *_dev)
> >  		return ret;
> >  
> >  	ret = dev_pm_domain_attach(_dev, true);
> > -	if (ret != -EPROBE_DEFER && drv->probe) {
> > -		ret = drv->probe(dev);
> > -		if (ret)
> > -			dev_pm_domain_detach(_dev, true);
> > +	if (ret != -EPROBE_DEFER) {
> > +		if (drv->probe) {
> > +			ret = drv->probe(dev);
> > +			if (ret)
> > +				dev_pm_domain_detach(_dev, true);
> > +		} else
> > +			/* don't fail if just dev_pm_domain_attach failed */
> > +			ret = 0;
> 
> An else that has a } should also have a {, according to 
> checkpatch and Documentation/CodingStyle. You can write it
> alternatively as:
> 
> 	if (ret != -EPROBE_DEFER) {
> 		if (drv->probe)
> 			ret = drv->probe(dev);
> 		else
> 			ret = 0;
> 
> 		if (ret)
> 			dev_pm_domain_detach(_dev, true);
> 	}
> 
> .
> 
> Best regards
> Uwe
> 
N‹§²æ¸›yú²X¬¶ÇvØ–)Þ{.nlj·¥Š{±‘êX§¶›¡Ü}©ž²ÆzÚj:+v‰¨¾«‘êZ+€Êzf£¢·hšˆ§~†­†Ûÿû®w¥¢¸?™¨è&¢)ßf”ùy§m…á«a¶Úÿ0¶ìå