Johan et. al., I'm still having those ire producing dma_timer_expiry problems. I rebuilt my kernel with Johan's (claims to have 2.6.x working with SATA flawlessly) kernel's ".config" and couldn't even get that booted. Aside from that, I even switched to SuSe's new 64-bit 2.6.4-29-smp kernel and that didn't work. Right now (w/2.6.4-29-smp) under load, I get hda: dma_timer_expiry: dma status == 0x24 hda: DMA interrupt recovery hda: lost interrupt from dmesg w/15 second bus reset. I've changed from sata_via to libata (I think) and still am having problems. My /etc/sysconfig/kernel line is INITRD_MODULES="libata scsi_mod sd_mod reiserfs" where removing "sata_via" didn't help any. Also, I tried turning off (any of) Samsung's acoustic management tuning. Right now hdparm -Tt shows: /dev/hda: Timing buffer-cache reads: 1964 MB in 2.00 seconds = 981.66 MB/sec Timing buffered disk reads: 174 MB in 3.02 seconds = 57.68 MB/sec I just noticed that throughout the whole process the device has always been /dev/hda2 and not /dev/sda2 as it should have been with libata, right? So, maybe removing "sata_via" from the /etc/sysconfig/kernel file (and remaking initrd) didn't do anything for me (indicating sata_via's built into the kernel?). Ok, so I'm remaking 2.6.5-rc2 with 'sata_via' as module, will remove it from the /etc/sysconfig/kernel file and reboot. -- minutes later -- Ok, sata_via's been built as a module. I've left it OUT of the INITRD (dmesg agrees) but still /dev/hda2 exists (and was booted from). Still /dev/hda2 not /dev/sda2. Not sata modules recognized in 'lsmod' or in dmesg. I'm almost 100% sure I'll get the error soon. Ideas and comments are always welcome! Thanks! -- Jason -- Jason Vertrees BSCB Graduate Student @ UTMB, Galveston javertre@utmb.edu :: http://www.bscb.utmb.edu
On 04/01/2004 02:49 PM, Jason Vertrees wrote:
Ok, sata_via's been built as a module. I've left it OUT of the INITRD (dmesg agrees) but still /dev/hda2 exists (and was booted from). Still /dev/hda2 not /dev/sda2. Not sata modules recognized in 'lsmod' or in dmesg. I'm almost 100% sure I'll get the error soon.
Ideas and comments are always welcome!
According to one of my keeper messages from this list, you need both sata_via and libata. Try adding sata_via, keeeping libata and the others, running mkinitrd, and trying that. HTH. -- Joe Morris New Tribes Mission Email Address: Joe_Morris@ntm.org Web Address: http://www.mydestiny.net/~joe_morris Registered Linux user 231871 God said, I AM that I AM. I say, by the grace of God, I am what I am.
--- Jason Vertrees <javertre@utmb.edu> wrote:
I'm still having those ire producing dma_timer_expiry problems. I rebuilt my kernel with Johan's (claims to have 2.6.x working with SATA flawlessly) kernel's ".config" and couldn't even get that booted.
I am now running 2.6.4 keeping the same config that I sent to you and I have had no problems since I managed to switch over from the IDE driver to libata. Sadly, I have still not been able to clarify exactly which settings made my kernel load sata_via/libata driver. I might be wrong, but I believe that there are fundamental problems with the IDE driver in the combination of dual CPU's and disks on both channels. I think this means that as long as this problems is not explicitly fixed, you will not be able to get things working without switching to sata_via/libata. Jason, I think maybe you could try and explain your boot problems in more detail, and maybe me or someone else on this list could guide you to get that config booted. This could very well be the key to solve your problem.
Right now (w/2.6.4-29-smp) under load, I get
hda: dma_timer_expiry: dma status == 0x24 hda: DMA interrupt recovery hda: lost interrupt
This is the very symptoms I had.
from dmesg w/15 second bus reset. I've changed from sata_via to libata (I think) and still am having problems. My /etc/sysconfig/kernel line is
INITRD_MODULES="libata scsi_mod sd_mod reiserfs"
You need both libata and via_sata - together they will get your VIA card going with the libata driver. Taking a step out in deep waters: The driver you are supposed to "disable" is the IDE driver that causes your SATA drives to become visible as /dev/hda (or hde or hdg or whatever). I *think* (someone more knowledgeable might just correct me here) that it is some VIA IDE driver that in 2.6.x manages to detect the VT8237 SATA controller - and this is the cause of the problems. I am now booting my kernel without any initrd - the config I use add libata and sata_via to the kernel and that's it. As a side note: When I tried booting some distributions installation disk with a 2.4.x on it, I got my SATA drives detected as /dev/sda and sdb without doing anything. From other messages on the net I have gathered that other people have seen the same: 2.6.x may detect SATA as IDE and causes problems on SMP machines. Like I said somewhere in the middle: aim to be able to boot a kernel based on the config I sent to you and see if you can get your drives to become /dev/sd. /Johan __________________________________ Do you Yahoo!? Yahoo! Small Business $15K Web Design Giveaway http://promotions.yahoo.com/design_giveaway/
participants (3)
-
Jason Vertrees -
Joe Morris (NTM) -
Johan Backlund