Hi all,
I have a few questions about the supermicro/h8dmr linuxBIOS port in specific, and the process for enabling vga support in general. Apologies if these questions are too noobish in nature.
I can successfully extract the VGA ROM from a supermicro h8dmr-i2 using either /dev/mem with offsets obtained from /proc/iomem, or via amideco. Both methods yield files of the same size, but diff reports a difference. Since I wasn't sure about which one to use. I've alternately prepended both images to an appropriate linuxbios.rom. Regardless of which VGA ROM is used, the machine fails to boot (fans spin up for a moment, then stop, no serial or vga output) after I flash the resultant image. When configured to use serial console, without the VGA ROM, the system boots and works as expected.
Something I've been wondering about is the "option ROM_SIZE" entry in the target's Config.lb: The VGA ROM is 44kB, and the SST Flash chip is 1024kB, but when I set ROM_SIZE to (1024*1024)-(44*1024) or 1003520, the final concatenated ROM weighs in at 1020kB, which flashrom will refuse to write. If instead I pad ROM_SIZE by 4kB, making it 1007616, I get a 1MB which flashrom will accept. Am I doing something silly here, or has anyone else seen this behavior?
Per the VGA Howto, I've un-commented the appropriate device section in src/mainboard/supermicro/h8dmr/Config.lb, and enabled both CONFIG_CONSOLE_VGA and CONFIG_PCI_ROM_RUN in the target's Config.lb.
Any help is greatly appreciated, thanks, Patrick.
Hi Patrick,
On Fri, May 04, 2007 at 02:23:27PM -0700, Patrick McMurchie wrote:
I have a few questions about the supermicro/h8dmr linuxBIOS port in specific, and the process for enabling vga support in general. Apologies if these questions are too noobish in nature.
Don't worry about that. There are no stupid questions - only stupid answers :)
I can successfully extract the VGA ROM from a supermicro h8dmr-i2 using either /dev/mem with offsets obtained from /proc/iomem, or via amideco. Both methods yield files of the same size, but diff reports a difference.
I've just tried extracting the VGA rom from the downloadable BIOS, and got a 44KB image. I would probably try that one first - sometimes there is strangeness when trying to recover the VGA rom from a running system.
Since I wasn't sure about which one to use. I've alternately prepended both images to an appropriate linuxbios.rom. Regardless of which VGA ROM is used, the machine fails to boot (fans spin up for a moment, then stop, no serial or vga output) after I flash the resultant image. When configured to use serial console, without the VGA ROM, the system boots and works as expected.
Something I've been wondering about is the "option ROM_SIZE" entry in the target's Config.lb: The VGA ROM is 44kB, and the SST Flash chip is 1024kB, but when I set ROM_SIZE to (1024*1024)-(44*1024) or 1003520, the final concatenated ROM weighs in at 1020kB, which flashrom will refuse to write. If instead I pad ROM_SIZE by 4kB, making it 1007616, I get a 1MB which flashrom will accept. Am I doing something silly here, or has anyone else seen this behavior?
Hmmm, this is strange, and perhaps the cause of the problem. I've just built the h8dmr target with ROM_SIZE set to 1003520, and the image it generated was exactly that size. So I'm not seeing this behavior here, in other words. Prepending the VGA rom gives me an image that is exactly 1024KB, as expected. I can't test it of course since I don't have the hardware :)
I guess my first question is, are you building from the main LinuxBIOS V2 tree, or from the patches that Yinghai sent to the list months ago? You should work from the tree, since Yinghai's MCP55 patch is now sufficiently merged.
Per the VGA Howto, I've un-commented the appropriate device section in src/mainboard/supermicro/h8dmr/Config.lb,
I was looking at that file, and I'm not sure what you had to un-comment? Nothing was marked as VGA.
and enabled both CONFIG_CONSOLE_VGA and CONFIG_PCI_ROM_RUN in the target's Config.lb.
Yeah, those are enabled by default in the tree.
Thanks, Ward.
Thanks for the respnonse, things are going much better now.
On Sat, 2007-05-05 at 22:05 -0400, Ward Vandewege wrote: [snip...]
Hmmm, this is strange, and perhaps the cause of the problem. I've just built the h8dmr target with ROM_SIZE set to 1003520, and the image it generated was exactly that size. So I'm not seeing this behavior here, in other words. Prepending the VGA rom gives me an image that is exactly 1024KB, as expected. I can't test it of course since I don't have the hardware :)
I guess my first question is, are you building from the main LinuxBIOS V2 tree, or from the patches that Yinghai sent to the list months ago? You should work from the tree, since Yinghai's MCP55 patch is now sufficiently merged.
I'm using un-patched source strait from svn. I checked out a fresh copy to retrace my steps, and things are working now. My buildrom line was missing the "failover" argument, which accounts for the missing 4k :). Silly me.
Per the VGA Howto, I've un-commented the appropriate device section in src/mainboard/supermicro/h8dmr/Config.lb,
I was looking at that file, and I'm not sure what you had to un-comment? Nothing was marked as VGA.
"device pci 6.0" in the pci_domain 0 section. It looks like I had misread the vga howto and calculated the offset of the /end/ of the VGA ROM, in relation to the 4GB boudary, whoops.
and enabled both CONFIG_CONSOLE_VGA and CONFIG_PCI_ROM_RUN in the target's Config.lb.
Yeah, those are enabled by default in the tree.
Ah, so they are.
On another note, when booted from LinuxBIOS the MAC addresses of the onboard NICs are different. They appear to hail from a block assigned to Tyan. Looking at the docs, this appears to be normal and the suggested solution is to set them in the OS. Can this be configured in the lxbios build process, as well? I worry about using etherboot to pxe-boot multiple servers at the same time.
Perhaps that's unfounded, though-- If I flash the same lxbios image to a second system, will the MACs be the same as on the first?
Thanks, Patrick.
-
Patrick McMurchie -
Ward Vandewege