[Intel-gfx] Fujitsu S6010 still woes (partially)

[Thomas Richter]

Am 08.04.2014 13:52, schrieb Daniel Vetter:
> On Tue, Apr 8, 2014 at 11:48 AM, Thomas Richter
>
> Hm, my X30 also locks up here on resume. What hack do you apply to
> make the ns2501 driver get through resume? I don't care about black
> screen, but I just wonder whether my X30 has the same issue - atm it
> hard-hangs.

Actually, the same "enable_DVO" hack that was there before, where I just 
poke the registers in the same way the script does it (see below). That 
is, whenever it detects that the DVO is "stuck", it configures the pipes 
to enable it. This is of course a hack, i.e. it would be a much cleaner 
solution if the suspend mechanism of the intel driver would configure 
the pipes first before trying to program or detect the DVO.

This is what does it:

intel_reg_write 0x02120 0x0
intel_reg_write 0x61100 0x00000c00
intel_reg_write 0x61160 0x10004084
intel_reg_write 0x6101c 0x027f01df
intel_reg_write 0x61000 0x031f027f
intel_reg_write 0x61004 0x03170287
intel_reg_write 0x61008 0x02ef028f
intel_reg_write 0x6100c 0x020c01df
intel_reg_write 0x61010 0x020401e7
intel_reg_write 0x61014 0x01eb01e9
intel_reg_write 0x71180 0x01000000
intel_reg_write 0x70188 0x00001000
intel_reg_write 0x20d8 0x10E0108
intel_reg_write 0x20dc 0x102
intel_reg_write 0x61120 0x0
intel_reg_write 0x6014 0xD0820000
intel_reg_write 0x6018 0x0
intel_reg_write 0x61140 0x80004084
intel_reg_write 0x61160 0x90004084

Actually, much of that is not required. You only need to configure the 
output on the right pipe (pipe B) and set the x2 flag for the DVO. The 
display timing (vblank,vtotal, hblank,htotal) are irrelevant.

Also, from the linux suspend mechanism, 
/usr/lib/pm-utils/sleep.d/99video is just useless or breaks more than it 
helps. I just removed it. It tries some weird workarounds that are not 
beneficial, and the driver (once corrected) should work without those. 
Instead, you can place the above into sleep.d/99video plus a bit of glue 
logic to run it on resume. This *almost* works, i.e. I do get a working 
display, but it is driven through the wrong pipe.

Greetings,
	Thomas