After some searching I couldn't find a topic concerning this, so I open a new topic:
These are the specs:
AMD Athlon64 3700+
2x Maxtor SATA 6L0080M0 Disks
I've freshly installed Debian (etch, sarge/stable wouldn't install because it cannot use the sata_nv driver during installation) for x86, not AMD64 due to the software intended to run on the machine.
First of, the driver for the onboard NIC is giving problems as I read on several forums. This problem however reflects throughout the systems Network cards and thus also affected the 3com card. After commenting out the onboard nic in /etc/network/interfaces the problem with networking was resolved. The problem was simply that for some reason at any random moment all network traffic seemed to stop and die completely. No errors, nothing found. Unloading and Reloading the drivers did not work, only a reboot would make the network work again.
Second, the disk-issue's. I've got 2 maxtor SATA disks, not affected with the firmware problem as stated at the Maxtor support site. Independent they work great, I can perform massive read/write operations on each disk without problems. However, the moment I start copy data from one disk to the other, the system starts behaving odly. This is what happens:
The system starts copy'ing nicely, but then at random intervals just locks up for approximately a minute or so. (Any command that uses a disk wil lock, non-disk (i.e. cached) commands don't)
Then, on the First TTY, it then gives this message:
0xc8 is DMA-Read and 0xca is DMA-Write if I recall correctly.
ata1: command 0xc8 timeout, stat 0x50 host_stat 0x24
ata2: command 0xca timeout, stat 0x50 host_stat 0x24
Sometimes it succesfully copy's a lot of data before this behaviour reoccurs, but mostly it just keeps popping up directly again and again, causing any disk-to-disk transfer to drop to the wopping copy speed of arround 1K/s
1K/s is the avarage speed /proc/mdstat reported after 3 hours of rebuilding
I first encountered this behaviour when I created a Software RAID1 array. After investigating, it doesn't just occur on resyncing the mirror drive, but with virtually every disk-to-disk transfer. I can recreate this behaviour at any given moment by simply start copy'ing stuff from disk 1 to disk 2 and vice versa.
I'm using a freshly build 188.8.131.52 kernel with nVidia SATA and nVidia chipset support installed in it (In multiple fashions, currently at install #5 and back to the kernel I compiled first time, since that is the only one working properly to get the system to boot from SATA).
The nVidia 'chipset' drivers that can be downloaded, don't appear to include any SATA/chipset drivers, but only contain Audio/LAN driver.
The system works perfectly if I use only 1 drive.
Anyone have similar problems?