[Intel-gfx] [PATCH v2] drm/i915/guc: Fixing error code for WOPCM initialization

[Michal Wajdeczko]

On Wed, 06 Mar 2019 09:45:17 +0100, Chris Wilson  
<chris at chris-wilson.co.uk> wrote:

> Quoting Michal Wajdeczko (2019-03-06 08:41:20)
>> On Wed, 06 Mar 2019 01:30:11 +0100, Sujaritha Sundaresan
>> <sujaritha.sundaresan at intel.com> wrote:
>>
>> > Replacing the -E2BIG error code return for WOPCM
>> > initialization with -ENODEV. This will prevent the pci from
>>
>> s/pci/CI ?
>>
>> > picking this up as a warning during fault injection testing.
>> >
>> > v2: change the final return code in i915_pci_probe() to ENODEV
>> >     instead of the specific wopcm change. - Daniele
>>
>> s/wopcm/WOPCM
>>
>> >
>> > Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
>> > Cc: Michal Wajdeczko <michal.wajdeczko at intel.com>
>> > Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan at intel.com>
>> > ---
>> >  drivers/gpu/drm/i915/i915_pci.c | 2 +-
>> >  1 file changed, 1 insertion(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/gpu/drm/i915/i915_pci.c
>> > b/drivers/gpu/drm/i915/i915_pci.c
>> > index c42c5ccf38fe..f962b5c0b3c1 100644
>> > --- a/drivers/gpu/drm/i915/i915_pci.c
>> > +++ b/drivers/gpu/drm/i915/i915_pci.c
>> > @@ -778,7 +778,7 @@ static int i915_pci_probe(struct pci_dev *pdev,
>> > const struct pci_device_id *ent)
>> >       err = i915_driver_load(pdev, ent);
>> >       if (err)
>> > -             return err;
>> > +             return i915_error_injected() ? -ENODEV : err;
>> >       if (i915_inject_load_failure()) {
>> >               i915_pci_remove(pdev);
>>
>> Hmm, since this is quite big deviation from original approach, maybe
>> you should sent it as new patch (with updated title and commit message)
>> as now this covers all injected errors (not just E2BIG from WOPCM)
>
> And I'm not sure we want to do that. When we do failure injection, we
> pick the errno and part of the test is making sure that gets reported
> unadulterated.
>
> Now doing an if (i915_error_injected() && !err) err = -EINVAL; makes
> sense to catch places where we've eaten that error and so breaking the
> test.

This will not work today, as at least in one place - i915_gem_init - we
inject -EIO which later we try to replace with success. And for that case
we may rather want to add

	if (i915_error_injected() && !err)
		i915_pci_remove(pdev);
		return -ENODEV;
	}

~Michal