| Summary: | X server crashes on DP hotplug: radeon_mode_hotplug() in xf86-video-ati does not enumerate second GPU | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | xorg | Reporter: | Max Staudt <bugzilla-fdo-max> | ||||||||
| Component: | Driver/Radeon | Assignee: | xf86-video-ati maintainers <xorg-driver-ati> | ||||||||
| Status: | RESOLVED FIXED | QA Contact: | Xorg Project Team <xorg-team> | ||||||||
| Severity: | major | ||||||||||
| Priority: | medium | CC: | strzol | ||||||||
| Version: | git | ||||||||||
| Hardware: | x86-64 (AMD64) | ||||||||||
| OS: | Linux (All) | ||||||||||
| Whiteboard: | |||||||||||
| i915 platform: | i915 features: | ||||||||||
| Attachments: |
|
||||||||||
|
Description
Max Staudt
2016-11-07 14:06:17 UTC
c801f9f10a5d was created in response to fdo#93415. Created attachment 127829 [details] [review] Refactor radeon_mode_hotplug Created attachment 127830 [details] [review] Fix radeon_mode_hotplug for GPU screens Do these two patches fix the problem? Yes, these patches fix the crash. However, as far as I can see, they do not allow for arbitrary number of Radeon cards in arbitrary primary/secondary configurations. So it's not a final fix... As far as I can see, instead of going through all screens, we could walk only over the outputs attached to 'scrn', or maybe restore the behavior before c801f9f10a5d and not walk over screens at all. Hoever that would necessitate a different way of enumerating outputs. Is it just an issue of giving every output a unique name? Maybe, instead of having globally incrementing counters per connector type, we could name the connectors "radeon-0-DisplayPort-1" or similar? Then we wouldn't have to count them globally. Also, having the driver name in the connector name prevents clashes with other drivers that have a similar naming scheme. You're misunderstanding. The purpose of this code is to find all connectors of a single GPU. There can only be one or two screens associated with any given GPU controlled by this driver. My patches make use of the entity shared between two screens of a single GPU to find the "other" screen, instead of iterating over all screens. It should work correctly in all cases. (In reply to Michel Dänzer from comment #5) > The purpose of this code is to find all connectors of a single GPU. Right, I mixed that up. > There can only be one or two screens associated with any given GPU > controlled by this driver. Okay, just so I understand this: I can have, say, 6 physical displays connected to one Radeon card (GPU), but I can distribute them between no more than two X screens (as in :0.0 and :0.1)? I looked at the code again, and now realized that you're taking the ScrnInfoPtr (primary_scrn) for the inner loop from pRADEONEnt, which in turn is a child of scrn. That's basically doing what I suggested (looping over scrn), so now I understand how this fixes my bug. Thank you for the quick fix! (In reply to Max Staudt from comment #6) > Okay, just so I understand this: I can have, say, 6 physical displays > connected to one Radeon card (GPU), but I can distribute them between no > more than two X screens (as in :0.0 and :0.1)? That's correct. Fixed in Git master: commit 9760ef33cba5795eddeda4d5c2fcbe2dcce21689 Author: Michel Dänzer <michel.daenzer@amd.com> Date: Tue Nov 8 13:02:43 2016 +0900 Use pRADEONEnt to find both screens of a GPU in radeon_mode_hotplug |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.