-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hello all,
I'm typing this message in Thunderbird from LinuxBIOS booted system. Yeah I told this last time, but I have still some issues left.
1) Single channel memory problems
I have two 512MB dimms in singe channel, because LinuxBIOS does not handle dualchannel for unbuffered Dimms. However when I'm using 512-1024MB of the RAM systems forces to random reboot when booting kernel or when in X. Booting with mem=512MB works like a charm
2) Reboot does not work. I dont know what I have wrong, but my fallback image just produces the initial string when console is initialized and nothing more. I have no clue what is wrong. I dont know what failover and fallback is precisely...
3) DMA to 0xE0000-0xEFFFF fails This is under investigation. I asked VIA about this.
4) no HPET and no ACPI, no PCI MCONFIG yet
5) I have some issues with LB itself when preparing various init files. The resource system have some serious trouble if I add my file for the PCIe bridge init. If I call the function not from init or enable methods but from general init and via pci_find_device, everything will work again. This seems like some issue with the infrastructure of LB rather than HW problem.
I may write more about my issues if only someone is willing to help ;) I can put my code online somewhere. Maybe it would be good to check it in even if the code for NB is a bit of mess.
USB/sound/SATA/IDE/PCIe/Network/Serial/Graphics/APIC works do far...
Thanks,
Rudolf
Rudolf Marek wrote:
I'm typing this message in Thunderbird from LinuxBIOS booted system. Yeah I told this last time, but I have still some issues left.
Great job!
- Reboot does not work. I dont know what I have wrong, but my fallback image just produces the
initial string when console is initialized and nothing more. I have no clue what is wrong. I dont know what failover and fallback is precisely...
<...>
I have some issues with LB itself when preparing various init files. The resource system have some serious trouble if I add my file for the PCIe bridge init. If I call the function not from init or enable methods but from general init and via pci_find_device, everything will work again. This seems like some issue with the infrastructure of LB rather than HW problem.
Both of these should be solved with ACPI tables, one of the Epia-M pages in the wiki has some good info on using iasl. Also make sure HAVE_HARD_RESET = 0 in mainboard Options.lb
I may write more about my issues if only someone is willing to help ;) I can put my code online somewhere. Maybe it would be good to check it in even if the code for NB is a bit of mess.
That's your call, but I doubt it could hurt. First check that your code is at least fairly clean and follows the coding guidelines (check the wiki). You might also want to check with Via and make sure that's okay, I don't know what your status is regarding NDAs. If you don't have one at all, you should be fine.
USB/sound/SATA/IDE/PCIe/Network/Serial/Graphics/APIC works do far...
USB works? Have you made any fixes or are you using the code I sent you as-is? And for graphics, are you using onboard or a seperate card?
Thanks, Corey
On 8/7/07, Corey Osgood corey.osgood@gmail.com wrote:
Rudolf Marek wrote:
I'm typing this message in Thunderbird from LinuxBIOS booted system.
Yeah I told
this last time, but I have still some issues left.
Great job!
- Reboot does not work. I dont know what I have wrong, but my fallback image just produces
the
initial string when console is initialized and nothing more. I have no clue what is wrong. I dont know what failover and fallback is precisely...
<...>
I have some issues with LB itself when preparing various init files. The resource system have some serious trouble if I add my file for the PCIe bridge init. If I call the function not from init or enable methods but from general init and via pci_find_device, everything will work again. This seems like some issue with the infrastructure of LB rather than HW problem.
Both of these should be solved with ACPI tables, one of the Epia-M pages in the wiki has some good info on using iasl. Also make sure HAVE_HARD_RESET = 0 in mainboard Options.lb
Whoops, that was supposed to be 2) and 4)...
And on 2, it actually tries to reboot but stops after printing the inital banner? No attempt at ram init or anything? It might be that yhlu has set up the k8 to not re-do ram init after a reboot, but I'm not sure.
-Corey
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi Corey,
Both of these should be solved with ACPI tables, one of the Epia-M pages in the wiki has some good info on using iasl. Also make sure HAVE_HARD_RESET = 0 in mainboard Options.lb
Ok have there =1 will try it.
I may write more about my issues if only someone is willing to help ;) I can put my code online somewhere. Maybe it would be good to check it in even if the code for NB is a bit of mess.
That's your call, but I doubt it could hurt. First check that your code is at least fairly clean and follows the coding guidelines (check the wiki). You might also want to check with Via and make sure that's okay, I don't know what your status is regarding NDAs. If you don't have one at all, you should be fine.
I have NDA. I cant just publish the specs in unaltered form. I must publish my code under GPL, comments in code are OK. I cannot create any other document except the source code (which describes VIA stuff).
USB/sound/SATA/IDE/PCIe/Network/Serial/Graphics/APIC works do far...
USB works? Have you made any fixes or are you using the code I sent you as-is? And for graphics, are you using onboard or a seperate card?
I think I sent you some code already too. It is ATI Radeon X300SE (rv370) in PCIe slot. So separate.
And on 2, it actually tries to reboot but stops after printing the inital banner? No attempt at ram init or anything? It might be that yhlu has set up the k8 to not re-do ram init after a reboot, but I'm not sure.
Yes just only one line...
Well lets try with = 0 first.
Rudolf
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hmm I tried to set some bits for socket 939 in fun2 0x90 and 0x94, seems that RAM works now for two dimms ;) Perhaps I need to develop some fix? I will write about this later.
USB works?
Yes USB does work. I tried again with my USB stick. Nothing needs to be programmed into the PCI regs, just enabled in SB as device. Thats all.
Rudolf
Rudolf Marek wrote:
Hmm I tried to set some bits for socket 939 in fun2 0x90 and 0x94, seems that RAM works now for two dimms ;) Perhaps I need to develop some fix? I will write about this later.
USB works?
Yes USB does work. I tried again with my USB stick. Nothing needs to be programmed into the PCI regs, just enabled in SB as device. Thats all.
Hmm, probably should have tried that first, I suppose. I did a bunch of init for it, and it currently causes a kernel panic. Thanks!
-Corey
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
hmm, probably should have tried that first, I suppose. I did a bunch of init for it, and it currently causes a kernel panic. Thanks!
It has only set some bits for the SMM by BIOS, which we dont need. So nothing is needed.
Hre is mu promised update:
Hmm I tried to set some bits for socket 939 in fun2 0x90 and 0x94, seems that RAM works now for two dimms ;) Perhaps I need to develop some fix? I will write about this later.
Just after some S3 ;)
- - PCI_ADDR(0, 0x18, 2, 0x8c), 0xff8fe08e, (0 << 20)|(0 << 8)|(0 << 4)|(0 << 0), + PCI_ADDR(0, 0x18, 2, 0x8c), 0xff8fe08e, (0 << 20)|(0 << 8)|(0 << 4)|(1 << 0), //change!
I enlarged the Write-to-Read Delay to 2 clocks as my AW is doing.
- - PCI_ADDR(0, 0x18, 2, 0x90), 0xf0000000, + PCI_ADDR(0, 0x18, 2, 0x90), 0xc0000000, + (1 << 29)| //change (4 << 25)|(0 << 24)| (0 << 23)|(0 << 22)|(0 << 21)|(0 << 20)| (1 << 19)|(0 << 18)|(1 << 17)|(0 << 16)| (2 << 14)|(0 << 13)|(0 << 12)| - - (0 << 11)|(0 << 10)|(0 << 9)|(0 << 8)| + (0 << 11)|(0 << 10)|(1 << 9)|(0 << 8)| //CHANGE DUAL DIMM ENABLE (0 << 3) |(0 << 1) |(0 << 0),
And then also bit9
Dual DIMM Enable (DualDimmEn)--Bit 9. When this bit is set, the A copy of the memory address bus is enabled, regardless of the MC0_EN (Function 2, Offset 94h) value, and the B copy of the memory bus is disabled if 939 package with 128-bit bus is not used. This bit should be set if unbuffered DIMMs are used, and two DIMM sockets are connected to the A copy of the memory address bus, as in SODIMM or 939 package configurations. See "Register Differences in Revisions of the AMD AthlonTM 64 and AMD OpteronTM Processors" on page 29 for revision information about this field.
And 29:
Upper Chip Select Mapping (UpperCSMap)--Bit 29. When this bit is set, chip select pins CS[7:4] are copies of chip select pins CS[3:0]. This bit should be set if the 939 package is used. See "Register Differences in Revisions of the AMD AthlonTM 64 and AMD OpteronTM Processors" on page 29 for revision information about this field.
After that my 2 DIMMS works fine it seems.
Rudolf
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi all,
Just an update. The "memtester" which is doing the memtesting while in OS will fail on "solid test".
Has anyone ever succeeded to run more than one dimm DDR unbuffered in single channel mode on K8?
Thanks,
Rudolf
On Thu, Aug 16, 2007 at 12:33:02AM +0200, Rudolf Marek wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi all,
Just an update. The "memtester" which is doing the memtesting while in OS will fail on "solid test".
Has anyone ever succeeded to run more than one dimm DDR unbuffered in single channel mode on K8?
I get the same problem on my ASUS A8N5X (K8 northbridge). It can be reproduced with memtest86 too (in my case it fails after a few seconds). Also, sometimes Linux fails CRC check while decompressing itself.
-
Corey Osgood -
Robert Millan -
Rudolf Marek