re: kern/40487 (thinkpad r50e can't poweroff)

[matthew green <mrg%eterna.com.au@localhost>]

i understand what is going on.  the 82801DB has been integrated
into the same ic the 82855GM is on, and speedstep should be
ignored there.

(the "enable speed step" bits in the current driver are reserved
bits in the docs i see for 82801DB.. ugh.)


the patch below ignores speedstep if we find a 82855GM on the
system, and should address this problem somewhat sanely.

i've tested it on the pentium-4m system this code was originally
tested on 5 years ago, and it seems to work there still.


.mrg.


Index: ichlpcib.c
===================================================================
RCS file: /cvsroot/src/sys/arch/x86/pci/ichlpcib.c,v
retrieving revision 1.14
diff -p -r1.14 ichlpcib.c
*** ichlpcib.c  13 Oct 2008 15:10:51 -0000      1.14
--- ichlpcib.c  27 Jan 2009 02:48:17 -0000
*************** error:
*** 542,547 ****
--- 542,549 ----
  /*
   * Linux driver says that SpeedStep on older chipsets cause
   * lockups on Dell Inspiron 8000 and 8100.
+  * It should also not be enabled on systems with the 82855GM
+  * Hub, which typically have an EST-enabled CPU.
   */
  static int
  speedstep_bad_hb_check(struct pci_attach_args *pa)
*************** speedstep_bad_hb_check(struct pci_attach
*** 551,556 ****
--- 553,561 ----
            PCI_REVISION(pa->pa_class) < 5)
                return 1;
  
+       if (PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_INTEL_82855GM_MCH)
+               return 1;
+ 
        return 0;
  }