DRM Panic Handling Is Back To Being Talked About
During the early days of kernel mode-setting (KMS) one of the frequently talked about future improvements that could be made as a result of it were improved error messages (like Windows BSODs) in the case of problems and other improvements on that front. While patches have emerged from time to time, it still seems like functionality that's still less than fulfilled compared to the original talked about goals. Patches this week have been revived for DRM panic handling.
Independent developer Noralf Trønnes who recently has proposed a DRM text mode and is trying to revive the SimpleDRM driver has sent out revised patches of DRM panic handling support that were started a few years back by David Herrmann.
With the three patches is a way to get the output of kernel messages on the display during a kernel panic. In terms of the architecture, "I just collect kernel messages in a circular buffer until a panic situation occurs and then the messages are rendered on all the framebuffers/devices it can find...The responsibility of the driver is to provide
a framebuffer with a virtual mapping that the messages can be rendered on."
These patches appear to be a start but more work is ahead before the code would be mainlined. Intel's Daniel Vetter has so far been the main developer providing feedback and there's a lot of it when it comes to what's needed for a full-blown driver. Daniel brought up the amount of work ahead would be roughly that of a good Google Summer of Code project for lasting a couple of months. Thus don't expect any pretty panic messages on your Linux system for the near future.
Independent developer Noralf Trønnes who recently has proposed a DRM text mode and is trying to revive the SimpleDRM driver has sent out revised patches of DRM panic handling support that were started a few years back by David Herrmann.
With the three patches is a way to get the output of kernel messages on the display during a kernel panic. In terms of the architecture, "I just collect kernel messages in a circular buffer until a panic situation occurs and then the messages are rendered on all the framebuffers/devices it can find...The responsibility of the driver is to provide
a framebuffer with a virtual mapping that the messages can be rendered on."
These patches appear to be a start but more work is ahead before the code would be mainlined. Intel's Daniel Vetter has so far been the main developer providing feedback and there's a lot of it when it comes to what's needed for a full-blown driver. Daniel brought up the amount of work ahead would be roughly that of a good Google Summer of Code project for lasting a couple of months. Thus don't expect any pretty panic messages on your Linux system for the near future.
26 Comments