Sata errors on sil3112 - kernel not booting

Please find a way to grab dmesg from this failed boot (maybe using a usb stick).

After searching for usb-devices I get more kernel errors (I'm not sure which ones because I'm not behind my own pc now) the kernel dumps some errors and hangs. I am able to reboot with the sysrq(s u b)-keys but I don't know how to get a prompt to get the dmesg-output for the sata drives.

Is there a way to get a prompt before the system hangs with kernel parameters or something like that? Or maybe with use of the sysrq-keys?

I'm getting the same thing, my hardware is a HP/Compaq nc8230 Laptop. Generic Kernel 2.6.17 works fine and notice that it does not load any SATA modules. But when I try any 2.6.19 Kernel I get the above. I compiled a custom without SATA and it will boot just fine.

I do have a USB stick let me know what I need to do to log dmesg to it and I'll get it posted ASAP.

I managed to get some error-messages on paper by writing them down ;) I can't give the dump, but I do have the messages that appear before the crash:

ata2.01: qc timeout (cmd 0xef)
ata2.01: failed to set xfermode (err_mask=0x4)
ata2.01: limiting speed to UDMA/25
ata2.01: failed to recover some devices, retrying in 5 secs
ata: abnormal status 0x7f on port 0x177
ata: abnormal status 0x7f on port 0x177

After that, it says the same, but than for PIO-mode instead of UDMA/25.

And than after reading some files the kernel crashes on "/lib/kref.c:32 kref_get" into a soft lockup with a dump.

This is all I could get by hand, hope it is useful :)

Today saw a new build of the kernel, thought I would give it another try and got the same error.

But I figured out what the problem was (I'm a dork) with the 2.6.17 sources it did not detect the sata drives and configured the grub kernel line with hda1, but with 2.6.19 it now detects the sata. It looks like the pkg is not configuring grub correctly had to change the kernel line to kernel /boot/vmlinuz-2.6.19-7-generic root=/dev/sda1 ro quiet splash instead of root=/dev/hda1.

Please retest against 2.6.20-2 when it is available in the feisty archive.

I tried the new kernel and it looks like the problem has been solved because the seagate errata fix is now applied. See the attached dmesg output. I haven't done a full benchmark but the CPU usage seems to be less than it was in previous kernels and accessing the drive seems faster although it still isn't as fast as in windows:

sudo hdparm -tT /dev/sdb
/dev/sdb:
 Timing cached reads: 1074 MB in 2.00 seconds = 536.44 MB/sec
 Timing buffered disk reads: 72 MB in 3.02 seconds = 23.85 MB/sec

I have to note that I installed a new harddisk for my root partition (WD on sda/ata1.00) since I reported the bug, and I now use the seagate as a secondary drive but I don't think this influenced the bug since it was about the mod15 workaround and I can access the drive fine now.

This has also cleared for me as well. I'm having issues with fglrx but that's a separate issue...

Thanks