[Mesa-dev] [PATCH] mesa syncobj: don't store a pointer to the set_entry

[Ian Romanick]

On 12/09/2012 09:47 AM, Stefan Dösinger wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Am 2012-12-08 22:24, schrieb Jordan Justen:
>> This likely will fix
>> https://bugs.freedesktop.org/show_bug.cgi?id=58012 (Regression
>> since 56e95d3c)
> The crashes are gone on r300g and r600g.
>
> This is most likely unrelated, but on r600g I get broken rendering in
> Half Life 2 (just random garbage on the screen). I'm sure it doesn't
> crash though because I don't see the debugger kicking in, and hl2
> terminates properly after the timedemo. There are complaints about
> commands rejected by the kernel:
>
>
> [  760.187261] [drm:radeon_cs_ib_chunk] *ERROR* Invalid command stream !
> [  760.192898] radeon 0000:01:00.0:
> evergreen_cs_track_validate_stencil:602 stencil read bo base
> 4148500480 not aligned with 16384
> [  760.192901] radeon 0000:01:00.0: evergreen_packet3_check:2098
> invalid cmd stream 2440
>
> Note that I don't know when this started - I haven't updated Mesa on
> that box for weeks.

It sounds like you should rebase Jordan's original patches out and 
bisect.  It would be nice to know when things started going wrong.  That 
usually makes the fix obvious.