I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, which has been known to exhibit itself even in Windows if you're using the old/stock via drivers. Even if I do get the LB issues figured out, the DMA issue (as far as I can tell) is _unfixable_. Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Here's the big question:
Can someone recommend a _NON-VIA_ small form factor motherboard (mini-itx, preferrably), that has all the usual goodies (IDE,USB,VGA,serial, etc) that is known to work well with LinuxBios? Naturally, one without a VIA chipset?
On 5/9/06, Eric Poulsen eric@zyxod.com wrote:
I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, which has been known to exhibit itself even in Windows if you're using the old/stock via drivers. Even if I do get the LB issues figured out, the DMA issue (as far as I can tell) is _unfixable_. Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Wow thats a major bummer. Can you summarize up everthing you have found? If that is really true that its _unfixable_ then it probally needs to be up on the wiki. Save someone else the trouble.
What epia boards does this affect?
(mini-itx, preferrably), that has all the usual goodies (IDE,USB,VGA,serial, etc) that is known to work well with LinuxBios? Naturally, one without a VIA chipset?
Working well with LB is the kicker. You could join us on the GX1 quest. There are several GX1 boards out there. If you get one like Chris's with an external VGA chip then you won't have to mess with the VSA stuff. I think the GX1 is close to working good but it just need a few tweaks to get there.
-- Richard A. Smith
Richard Smith wrote:
On 5/9/06, Eric Poulsen eric@zyxod.com wrote:
I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, which has been known to exhibit itself even in Windows if you're using the old/stock via drivers. Even if I do get the LB issues figured out, the DMA issue (as far as I can tell) is _unfixable_. Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Wow thats a major bummer. Can you summarize up everthing you have found? If that is really true that its _unfixable_ then it probally needs to be up on the wiki. Save someone else the trouble.
Not "unfixable" per se, but no fix that I could find that actually works 100%. There a bunch of hoops you can jump through that mostly involve recompiling the kernel and turning off several features (or turning off DMA -- not an option). In the end, you'll only reduce the likelyhood of a crash.
What epia boards does this affect?
Well, there's a known issue with the VT8235 Southbridge under heavy DMA usage. typically: Heavy IDE usage + Network | PVR | Audio.
From
'http://www.viaarena.com/default.aspx?PageID=22&DCatType=3&DSCat=155', I gather that these boards have the problem:
VIA EPIA-MS10000E EPIA-M mainboards EPIA-MII mainboards EPIA-TC mainboard EPIA-PD mainboard VIA EPIA M mainboard VIA EPIA MII mainboard VIA EPIA SP mainboard VIA EPIA CL mainboard
Note that the "fix" is a new BIOS, which is useless to potential LB users. Also, the ML (my board) is not on there, but it has the EXACT same symptoms.
(mini-itx, preferrably), that has all the usual goodies (IDE,USB,VGA,serial, etc) that is known to work well with LinuxBios? Naturally, one without a VIA chipset?
Working well with LB is the kicker. You could join us on the GX1 quest. There are several GX1 boards out there. If you get one like Chris's with an external VGA chip then you won't have to mess with the VSA stuff. I think the GX1 is close to working good but it just need a few tweaks to get there.
I'll look into them ...
* Eric Poulsen eric@zyxod.com [060510 06:13]:
Note that the "fix" is a new BIOS, which is useless to potential LB users. Also, the ML (my board) is not on there, but it has the EXACT same symptoms.
Does the bios update fix the problem completely? Maybe they do some smart pci configuration thing or they load a microcode update?
If its pci configuration we might compare the pci config space with the old and the new factory bios to find out the fix.
Stefan
Stefan Reinauer wrote:
- Eric Poulsen eric@zyxod.com [060510 06:13]:
Note that the "fix" is a new BIOS, which is useless to potential LB users. Also, the ML (my board) is not on there, but it has the EXACT same symptoms.
Does the bios update fix the problem completely? Maybe they do some smart pci configuration thing or they load a microcode update?
Some people say that it works for them, other people say it's useless.
Upon further reading, there might be two distinct problems: DMA bugs in 8235/8237 and/or Linux CPU scaling (longhaul).
If its pci configuration we might compare the pci config space with the old and the new factory bios to find out the fix.
Only if the bios actually works, and there isn't any reason to believe it does. Additionally, there isn't a "fix" bios for my board. Best I can do is try the M bios and hope that it works for the ML boards.
Stefan
Richard Smith schrieb:
On 5/9/06, Eric Poulsen eric@zyxod.com wrote:
I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, which has been known to exhibit itself even in Windows if you're using the old/stock via drivers. Even if I do get the LB issues figured out, the DMA issue (as far as I can tell) is _unfixable_. Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Somewhere I have read, that disabling the acpi function in the kernel solve many problems for the epia. I use an Epia-M for myself and it works stable, without acpi. Ok, the running applications for my board don't make high loads on the cpu/ide or network.
Working well with LB is the kicker. You could join us on the GX1 quest. There are several GX1 boards out there. If you get one like Chris's with an external VGA chip then you won't have to mess with the VSA stuff. I think the GX1 is close to working good but it just need a few tweaks to get there.
My stb3036 is not the fastet thing I guess ;) It is a question of usage.
-- Richard A. Smith
Christian Sühs schrieb:
Richard Smith schrieb:
On 5/9/06, Eric Poulsen eric@zyxod.com wrote:
I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, [...] Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Somewhere I have read, that disabling the acpi function in the kernel solve many problems for the epia. I use an Epia-M for myself and it works stable, without acpi.
Somebody on this list recommended me to disable ACPI in the Linux kernel. This had solved my kernel panic on EPIA ML. Eric, please try to disable ACPI, even if it is merely a workaround.
Any ideas why kernel ACPI works with factory BIOS but panics with LinuxBIOS on VIA EPIA ML?
Thanks, Daniel.
Daniel Parthey wrote:
Christian Sühs schrieb:
Richard Smith schrieb:
On 5/9/06, Eric Poulsen eric@zyxod.com wrote:
I've been researching the IDE/DMA/Linux issue that appears to plague the VT8235 southbridge. Seems there is _no_ reliable fix for this issue, [...] Reliability is a huge thing for me (hence the use of Linux). The Epia board is bound for Ebay.
Somewhere I have read, that disabling the acpi function in the kernel solve many problems for the epia. I use an Epia-M for myself and it works stable, without acpi.
Somebody on this list recommended me to disable ACPI in the Linux kernel. This had solved my kernel panic on EPIA ML. Eric, please try to disable ACPI, even if it is merely a workaround.
Any ideas why kernel ACPI works with factory BIOS but panics with LinuxBIOS on VIA EPIA ML?
oh, there's all kinds of fun possibilities :-)
possible the ML says 'tweak this bit' and the bit is set up wrong on LB.
ACPI is Yer Basic Shoppe of Horrors
ron
-
Christian Sühs -
Daniel Parthey -
Eric Poulsen -
Richard Smith -
Ronald G Minnich -
Stefan Reinauer