Re: [coreboot] Issues with Supermicro H8SCM

[She, Kerry]

Hello, Samir
> -----Original Message-----
> From: Samir Ibradžić [mailto:[e-mail]]
> Sent: Monday, November 21, 2011 1:35 AM
> To: She, Kerry
> Cc: Thomas Gstädtner; Marc Jones; coreboot
> Subject: Re: [coreboot] Issues with Supermicro H8SCM
>
> Hello Kerry, Thomas, Marc,
>
> I got this board for a personal server project, as well as test platform
> to dive into coreboot. Board is H8SCM-F rev 1.02, one with BMC. Had
> build nice external FT2232 programmer for bailing out of broken flash
> situation, works great. I've experimented with recent coreboot revision,
> ea5c2b6 + Kerry's AGESA patches (thanks a bunch!), and had experienced
> same problem as Thomas, plus some more;
> Two DIMMs in dual-channel setup wont work, same issue as Thomas,
>
> 2x4GB ECC DIMMs here.
> After booting via coreboot with same Linux distro, SR5850 NorthBridge
>
> heatsink instantly gets much more hot than booting via SuperMicro BIOS.
> The temperature is MUCH higher, on idle system, though i can not measure
> this by other method than putting my finger there, and i got instantly
> BURNED when coreboot is being used. I am afraid to fry my board if it
> runs longer this way, it is really that hot.

The hardware monitor is not enable in coreboot.
> HPET is not being reported at all by coreboot ACPI, OS timer
>
> precision suffers big deal. Dont know if this is a feature or bug? Dmesg
> clearly shows the difference, all these lines are missing when booting
> via coreboot:
>
> # dmesg | grep -i hpet
> [    0.000000] ACPI: HPET 00000000cfeaa5e0 00038 (v01 062911 OEMHPET
> 20110629 MSFT 00000097)
> [    0.000000] ACPI: HPET id: 0x8300 base: 0xfed00000
> [    0.000000] hpet clockevent registered
> [    1.132106] hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0, 0
> [    1.140072] hpet0: 4 comparators, 32-bit 14.318180 MHz counter
> [    1.146104] Switching to clocksource hpet
> [    3.499557] rtc0: alarms up to one month, y3k, 114 bytes nvram, hpet
> irqs
> CPU (got 4170 HE) frequency scaling only goes as low as 1800MHz in
>
> coreboot when idle. SuperMicro BIOS easily goes all the way down to
> 800MHz. Spotted this in /proc/cpuinfo
> USB/EHCI - i haven't really tested this, but it seems minor compared
>
> to 1 & 2.

Kevin have fixed some usb-msc bugs in seabios recently,
you may need to check the seabios maillist.
> Now, i know this board is not exactly supported, most probably being an
> AGESA problem, but i would really like to do my best to help to make it
> work better if possible.
>
> @ Kerry,
>
> I was trying to analyze some AGESA code to get to the DIMM problem, but
> most of it is a mystery to me. Could you please point me to the right
> place for "DRAM training" in the code, i am willing to invest some time
> and nerves to test & investigate this, as well as NB overheating
> problem. Any plans to update those AGESA patches for this board?

We have made some update for h8scm mainboard.
In order to support new Orochi platform,
Except the mainboard code, agesa wrapper, sr56x0 cimx wrapper and
sb700/sp5100 cimx wrapper are also included in the change set.
But the patch set still pending to release to the community for review.
We need some time to get the release process done.
Hope the code would be available soon.
> One strange thing about RAM init is that the coreboot somehow thinks the
> board has 8 slots, but it has only four (there should be no channels 2 &
> 3);

AGESA support 4 channels.
But family10 processor for h8scm only have 2 channels.
> Socket 0 Channel 0 Dimm 0 found dimm: 00400745
> Socket 0 Channel 0 Dimm 1 not found dimm
> Socket 0 Channel 1 Dimm 0 found dimm: 00400947
> Socket 0 Channel 1 Dimm 1 not found dimm
> Socket 0 Channel 2 Dimm 0 not found dimm
> Socket 0 Channel 2 Dimm 1 not found dimm
> Socket 0 Channel 3 Dimm 0 not found dimm
> Socket 0 Channel 3 Dimm 1 not found dimm
>   Dct 0 Channel 0
>    RegDimmPresent:    1
>    SODimmPresent:     0
>    ChDimmValid:       1
>    DimmPlPresent:     0
>    DimmQrPresent:     0
>    DimmDrPresent:     0
>    DimmSRPresent:     1
>    Dimmx4Present:     1
>    DimmX8Present:     0
>    DimmX16Present:    0
>    DimmMirrorPresent: 0
>   Dct 1 Channel 0
>    RegDimmPresent:    1
>    SODimmPresent:     0
>    ChDimmValid:       1
>    DimmPlPresent:     0
>    DimmQrPresent:     0
>    DimmDrPresent:     0
>    DimmSRPresent:     1
>    Dimmx4Present:     1
>    DimmX8Present:     0
>    DimmX16Present:    0
>    DimmMirrorPresent: 0
> ASSERTION FAILED: file
> 'src/vendorcode/amd/agesa/f10/Proc/Mem/Main/C32/mmflowC32.c',  line 163

Maybe a dimm spd address problem, you can try following patch:
--- a/src/mainboard/supermicro/h8scm_fam10/dimmSpd.c
+++ b/src/mainboard/supermicro/h8scm_fam10/dimmSpd.c
@@ -55,8 +55,11 @@ static void sp5100_set_gpio(u8 reg, u8 out, u8 enable)
 static const UINT8 spdAddressLookup [1] [2] [2] = { // socket, channel, dimm
        /* socket 0 */
        {
-               {0xA6, 0xA4},
-               {0xA2, 0xA0},
+               /* silkprint: DIMM1A DIMM2A DIMM1B DIMM2B
+                * i2c addr:    A6     A2     A4     A0
+                */
+               {0xA4, 0xA6},
+               {0xA0, 0xA2},
        },
 };

Thanks
Kerry