On 4/18/24 1:13 PM, Kevin Brodsky wrote:
On 18/04/2024 10:39, carsten.haitzler@foss.arm.com wrote:
From: Carsten Haitzler carsten.haitzler@arm.com
In re-jigging the code I didn't return the new prop count but ketp the input prop count number. This fixes that. This make Xiorg start again.
Signed-off-by: Carsten Haitzler Carsten.Haitzler@arm.com
drivers/gpu/drm/drm_property.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/drivers/gpu/drm/drm_property.c b/drivers/gpu/drm/drm_property.c index 18435fc9df97..292ea8a4fa5f 100644 --- a/drivers/gpu/drm/drm_property.c +++ b/drivers/gpu/drm/drm_property.c @@ -513,6 +513,7 @@ int drm_mode_getproperty_ioctl(struct drm_device *dev, } }
- count_values = value_count;
Oops, clearly missed that during the review too... Applied on next, thanks! Also applied your DRM series + this patch on master, as that's apparently needed to get DRM to work in compat64 as well.
Yeah. IT's really subtle and so easily missed - I missed it to begin with and nothing I used for testing pointed out the problem (proptest, modetest, vbltest). It wasn't until I ran a full-fast stack that it was hit. And even then... It wasn't hit UNTIL I fixed some other small gotchas like this that you pointed out in my patchset. Until then my patchset happened to work by luck...
So thanks! This integration drop should be ready to go now - of course pending final tests.
Kevin
copied = 0; if (drm_property_type_is(property, DRM_MODE_PROP_ENUM) ||