Check 64 bit DMA from Disk-Controller (amd64/135040)

[Ralf Folkerts]

Am 01.02.2011 23:32, schrieb Torfinn Ingolfsen:
> Hello,
>
> On Tue, 01 Feb 2011 20:17:36 +0100
> Ralf Folkerts<ralf.folkerts at gmx.de>  wrote:
>
>> My SB600 ATA-Controller used to cause massive Problems (from disk not
>> being detected at all to disk being detected but later "trashed" during
>> buildworld) with DMA with 4G RAM. I circumvented that by setting
>> "hw.physmem=4g", as recommended by John Baldwin.
> Interesting. I have a machine[1] with the SB600, but so far it has
> mostly been running FreeBSD 6.x and 7.x / amd64.
> However, I recently installed Linux (Ubuntu) on a partition on that
> same machine, and it shows some interesting things in dmesg output:
>
> [    0.698790] ehci_hcd 0000:00:13.5: applying AMD SB600/SB700 USB freeze workaround
> [    1.448022] ata3: applying SB600 PMP SRST workaround and retrying
> [    1.449522] ata4: applying SB600 PMP SRST workaround and retrying
> [    1.619986] ata4.00: SB600 AHCI: limiting to 255 sectors per cmd
> [    1.626505] ata4.00: SB600 AHCI: limiting to 255 sectors per cmd
> [    1.632688] ata3.00: SB600 AHCI: limiting to 255 sectors per cmd
> [    1.634745] ata3.00: SB600 AHCI: limiting to 255 sectors per cmd
>
> which may or may not be related.
>
> According the the bios upgrade info for mny machine, some usb problem has been fixed
> by the latest bios upgrade (the machine used to hang when booting
> if any usb devices were plugged in).
> I haven't seen any trouble with the disk controller (but the machine has
> only had a single disk attached until recently).
>
> References:
> 1) http://sites.google.com/site/tingox/asus_m2a-vm_hdmi

Hi Torfinn,

thanks for the Feedback! The Problem seems to be Firmware-related; I got 
a Link back when I encountered the Problem and asked on this list that 
mentioned that the 64 bit DMA-issue is being caused by it being 
incorrectly initialized by the BIOS. The Release Notes for the 
BIOS-Version that -seems to have- solved the issue mentions "Update PCIE 
Initial Code". Probably ASUS did a better Job writing the H/W Specifics 
for the BIOS than ASRock.

Re. the 255 Sector-Limit: I get these messages on Linux, too. There is 
some discussion re. that here 
https://bugzilla.kernel.org/show_bug.cgi?id=9412

I think I'll connect a "test-disk" to the box, install PC-BSD on ZFS and 
then just try to compile Eclipse or OpenJDK (or both) and se, if these 
succeed or not...

Cheers,
_ralf_