On Tue, 2011-01-04 at 18:00 -0500, Alex Deucher wrote:
Lots of EDIDs have bad checksums or bad version numbers, but have otherwise good data in them. The current code rejects them which results in bad modes or blank screens in some cases. This patch adds a new module parameter, edid_strict, to drm.ko. It defaults to 1 (strict conformance, the current behavior), but if the user sets it to 0, it will bypass the checks and accept the EDID.
I really, really hate this. Especially for things like the CEA extension, where there's a bunch of internal length data that we trust while walking the block, so if it's wrong then we're just going to walk off the end of the EDID block and into arbitrary kernel memory.
However, I have neither the time nor the inclination to write a sufficiently paranoid permissive parser right now. So I'd kind of prefer that this set the taint flag, and I absolutely disclaim any responsibility for any security bugs that result from someone shooting themselves in the foot with this. But, that said:
Regretted-by: Adam Jackson ajax@redhat.com
- ajax