Re: 2.6.26-rc9-git9 doesn't boot on Macintel

[Justin Mattock]

On Wed, Jul 16, 2008 at 2:08 PM, Jack Howarth <howarth@xxxxxxxxxxxxxxxxx> wrote:
> Yunghai,
>   I had partial success with your proposed patch. The MacPro2
> identifier doesn't appear to be correct for a second generation
> MacBook Pro so I had to comment out the line...
>
>                     DMI_MATCH(DMI_PRODUCT_NAME, "MacPro2"),
>
> With that change, a patched 2.6.26-git2 kernel now uses MMCONFIG.
> However I still see the same hang. The boot messages I see on
> screen are...
>
> ACPI: bus type pci registered
> PCI: MCFG configuration 0: base f0000000 segment 0 buses 0 - 63
> PCI: MCFG area at f0000000 reserved in E820
> PCI: Using MMCONFIG at f0000000 - f3ffffff
> PCI: Using configuration type 1 for base access
> ACPI: EC: EC description table is found, configuring boot EC
> ACPI: EC: non-query interrupt received, switching to interrupt mode
> ACPI: BIOS_OSI(Linux) query ignored via DMI
> ACPI: Interpreter enabled
> ACPI: (S0 S3 S4 S5)
> ACPI: Using IOAPIC for interrupt routing
> ACPI: EC: GPE = 0x17, I/O: command/status = 0x66, data = 0x62
> ACPI: EC: drivers started in interrupt mode
> ACPI: PCI Root Bridge [PCI0] (0000.00)
> pci 0000:00:1f.00 : quirk region 0400-047f claimed by ICH6 ACPI/GPIO/TCO
> pci 0000.00:1f:00 : quirk region 0500-053f claimed by ICH6 GPIO
>
> ...at which the boot hangs. On the positive side, I was able to fully
> boot if I passed 'pci=noacpi' to the kernel options (which never worked
> before with 2.6.26). I can post the dmesg from a noacpi boot of the
> patched kernel to my bugzilla report tonight if it would help debug the
> issues we are still seeing with ACPI.
>   One thing I notice with the patched kernel (without noacpi) is that I
> only see buses 0 - 63. A normal boot of 2.6.25.10 on this machine (or
> 2.6.26 reportedly on a MacMini) always shows buses 0 - 255. Could this
> be related to the ACPI breakage? Let me know if I can do anything
> else to debug the ACPI issues under MMCONFIG.
>                     Jack
>
>
> On Wed, Jul 16, 2008 at 01:36:21AM -0700, Yinghai Lu wrote:
>> On Wed, Jul 16, 2008 at 12:50 AM, Justin Mattock
>> <justinmattock@xxxxxxxxx> wrote:
>> > On Wed, Jul 16, 2008 at 7:44 AM, Yinghai Lu <yhlu.kernel@xxxxxxxxx> wrote:
>> >> I looked at your dmesg from 2.6.25.10
>> >>
>> >> it tried to use mmconfig, and later fall back to conf1, because
>> >> 1. your MCFG said it support bus [0,255], and range will be
>> >> [0xf0000000, 0xffffffff]
>> >> 2. firmware only reserve [0xf0000000, 0xf4000000)
>> >>
>> >> and 2.5.25 will not check acp reserved range.
>> >>
>> >> from 2.6.26, it will check if the [0xf0000000, 0xffffffff] is reserved
>> >> via acpi dsdt _crs?
>> >> it will fail too.
>> >>
>> >> So MCFG table is broken.
>> >>
>> >> You need to get firmware update from your vendor.
>> >>
>> >> or we need to put that system on DMI check to change end_bus_number
>> >> from 0xff to 0x3f
>> >>
>> >> ps: mem=2g will not work, it will can not change mcfg table etc... you
>> >> need take out 1G ram out.
>> >>
>> >> YH
>> >>
>> >
>> > Well shit man, that sucks...
>> > how hard is it to do a DMI check?
>> > regards;
>>
>> please check attached patch... not sure the
>> DMI_MATCH(DMI_PRODUCT_NAME, "MacPro2"),  is right or not.
>>
>> YH
>
>> [PATCH] x86,pci: dmi check for mackpro 2.2 mmconf
>>
>> change the mconf bus range from [0,0xff] to to [0, 0x3f]
>> to match range [0xf0000000, 0xf4000000) in e820 tables.
>>
>> Signed-off-by: Yinghai Lu <yhlu.kernel@xxxxxxxxx>
>>
>> ---
>>  arch/x86/pci/mmconfig-shared.c |   35 +++++++++++++++++++++++++++++++++++
>>  1 file changed, 35 insertions(+)
>>
>> Index: linux-2.6/arch/x86/pci/mmconfig-shared.c
>> ===================================================================
>> --- linux-2.6.orig/arch/x86/pci/mmconfig-shared.c
>> +++ linux-2.6/arch/x86/pci/mmconfig-shared.c
>> @@ -374,6 +374,40 @@ reject:
>>
>>  static int __initdata known_bridge;
>>
>> +static int __devinit check_mmconf(const struct dmi_system_id *d)
>> +{
>> +     typeof(pci_mmcfg_config[0]) *cfg;
>> +
>> +     if ((pci_mmcfg_config_num != 1) ||
>> +         (pci_mmcfg_config == NULL) ||
>> +         (pci_mmcfg_config[0].address == 0))
>> +             return 1;
>> +
>> +     cfg = &pci_mmcfg_config[0];
>> +
>> +     if (cfg->start_bus_number == 0 && cfg->end_bus_number > 0x3f)
>> +             cfg->end_bus_number = 0x3f;
>> +
>> +     return 0;
>> +}
>> +
>> +static struct dmi_system_id __devinitdata mmconf_dmi_table[] = {
>> +        {
>> +                .callback = check_mmconf,
>> +                .ident = "Apple",
>> +                .matches = {
>> +                        DMI_MATCH(DMI_SYS_VENDOR, "Apple"),
>> +                     DMI_MATCH(DMI_PRODUCT_NAME, "MacPro2"),
>> +                },
>> +        },
>> +        {}
>> +};
>> +
>> +void __init check_mmconf_dmi(void)
>> +{
>> +        dmi_check_system(mmconf_dmi_table);
>> +}
>> +
>>  static void __init __pci_mmcfg_init(int early)
>>  {
>>       /* MMCONFIG disabled */
>> @@ -395,6 +429,7 @@ static void __init __pci_mmcfg_init(int
>>
>>       if (!known_bridge) {
>>               acpi_table_parse(ACPI_SIG_MCFG, acpi_parse_mcfg);
>> +             check_mmconf_dmi();
>>               pci_mmcfg_reject_broken(early);
>>       }
>>
>
>

Hmm Im not seeing MMCONFIG anywhere in dmesg
I'm wondering if I should apply this patch as well.
this is what I'm seeing with dmesg | grep MCFG:

[    0.000000] ACPI: MCFG 3FEF8000, 003C (r1 APPLE   Apple00        1
Loki       5F)
[    0.340000]  domain 0: span 0-1 level MC
[    0.340000]  domain 0: span 0-1 level MC
[    1.660595] EDAC MC: Ver: 2.1.0 Jul 15 2008

-- 
Justin P. Mattock
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/