[Trac #437] Do we use AHCG?

[Zarro Boogs per Child]

#437: Do we use AHCG?
----------------------+-----------------------------------------------------
 Reporter:  dwmw2     |        Owner:  JordanCrouse
     Type:  defect    |       Status:  new         
 Priority:  normal    |    Milestone:  BTest-2     
Component:  hardware  |   Resolution:              
 Keywords:            |  
----------------------+-----------------------------------------------------
Comment (by tsylla):

 These are the clock gating MSRs and their values gleaned from the PRS
 (trac #109):

 {{{
 GL0  10002004 00f
 MC   20002004 003
 CPU  NI
 VG   80002004 000
 GP   a0002004 001
 DF   c0002004 155
 GL1  40002004 00f
 GLCP 4c002004 005
 PCIN 50002004 015
 GIO  54002004 000

 GL2  51012004 004
 PCIS 51002004 005
 USB  51202004 000
 ATA  51302004 000
 DIV  51402004 50554111
 AC97 51502004 005
 GLCP 51702004 004
 }}}
 From the looks of it, these are matched in-system:
 {{{
 -bash-3.1# rdmsr 0x10002004
 MSR register 0x10002004 => 00:00:00:00:00:00:00:05
 -bash-3.1# rdmsr 0x20002004
 MSR register 0x20002004 => 00:00:00:00:00:00:00:03
 -bash-3.1# rdmsr 0x80002004
 MSR register 0x80002004 => bf:e1:fa:10:b7:e3:59:b5
 -bash-3.1# rdmsr 0xa0002004
 MSR register 0xa0002004 => bf:ee:d2:e0:b7:e1:79:b5
 -bash-3.1# rdmsr 0xc0002004
 MSR register 0xc0002004 => bf:f0:b3:00:b7:e6:d9:b5
 -bash-3.1# rdmsr 0x40002004
 MSR register 0x40002004 => 00:00:00:00:00:00:00:05
 -bash-3.1# rdmsr 0x4c002004
 MSR register 0x4c002004 => 00:00:00:00:00:00:00:05
 -bash-3.1# rdmsr 0x50002004
 MSR register 0x50002004 => 00:00:00:00:00:00:00:15
 -bash-3.1# rdmsr 0x54002004
 MSR register 0x54002004 => 00:00:00:00:00:00:00:00
 -bash-3.1# rdmsr 0x51012004
 MSR register 0x51012004 => 00:00:00:00:00:00:00:04
 -bash-3.1# rdmsr 0x51002004
 MSR register 0x51002004 => 00:00:00:00:00:00:00:05
 -bash-3.1# rdmsr 0x51202004
 MSR register 0x51202004 => 00:00:00:00:00:00:00:00
 -bash-3.1# rdmsr 0x51302004
 MSR register 0x51302004 => 00:00:00:00:00:00:00:00
 -bash-3.1# rdmsr 0x51402004
 MSR register 0x51402004 => 00:00:00:00:50:55:41:11
 -bash-3.1# rdmsr 0x51502004
 MSR register 0x51502004 => 00:00:00:00:00:00:00:05
 -bash-3.1# rdmsr 0x51702004
 MSR register 0x51702004 => 00:00:00:00:00:00:00:04
 }}}

 The 'F's in the GLIUs should really be '5's in the PRS, so everything
 readable looks ok. The graphics MSRs aren't read properly; they should be
 read with a different mechanism to verify their values.

 It looks like the AHCG that can be enabled is enabled by LB, and it should
 not be touched by anything else. There is no reason to ever turn it off,
 and those that are not enabled by the PRS are disabled because of errata.
 (It would be good if AMD would publish the errata/spec update so this is
 clear)

-- 
Ticket URL: <http://dev.laptop.org/ticket/437#comment:1>
One Laptop Per Child <http://laptop.org/>