kernel ata_piix DMA issues [solved]

vogt · 2007-02-07 03:12:45

this is the first time that such a message appeared (i think after a kernel update); something is wrong, cause there are about 15 such messages (different want values) each and every boot.

Feb  3 10:47:23 localhost attempt to access beyond end of device
Feb  3 10:47:23 localhost sdb: rw=0, want=40017915, limit=39062500

but thats not too bad, because it doesn't even add a second to the boot time, and the performance hasn't noticeably changed.

but then, upgrading to 2.6.20, I got this somewhat related message:

Feb  6 21:55:19 localhost scsi2 : ata_piix
Feb  6 21:55:19 localhost ata2.00: ATAPI, max UDMA/33
Feb  6 21:55:19 localhost ata2.01: ATAPI, max UDMA/33
Feb  6 21:55:19 localhost ata2.01: device is on DMA blacklist, disabling DMA
Feb  6 21:55:19 localhost ata2.00: qc timeout (cmd 0xef)
Feb  6 21:55:19 localhost ata2.00: failed to set xfermode (err_mask=0x4)
Feb  6 21:55:19 localhost ata2.00: limiting speed to UDMA/25
Feb  6 21:55:19 localhost ata2: failed to recover some devices, retrying in 5 secs
Feb  6 21:55:19 localhost ata2.01: device is on DMA blacklist, disabling DMA
Feb  6 21:55:19 localhost ata2.00: qc timeout (cmd 0xef)
Feb  6 21:55:19 localhost ata2.00: failed to set xfermode (err_mask=0x4)
Feb  6 21:55:19 localhost ata2.00: limiting speed to PIO0
Feb  6 21:55:19 localhost ata2: failed to recover some devices, retrying in 5 secs
Feb  6 21:55:19 localhost ata2.01: device is on DMA blacklist, disabling DMA
Feb  6 21:55:19 localhost ata2.00: qc timeout (cmd 0xef)
Feb  6 21:55:19 localhost ata2.00: failed to set xfermode (err_mask=0x4)
Feb  6 21:55:19 localhost ata2.00: disabled
Feb  6 21:55:19 localhost ata2: failed to recover some devices, retrying in 5 secs
Feb  6 21:55:19 localhost ata2.01: device is on DMA blacklist, disabling DMA
Feb  6 21:55:19 localhost ata2.01: failed to set xfermode (err_mask=0x40)
Feb  6 21:55:19 localhost ata2.01: limiting speed to PIO3
Feb  6 21:55:19 localhost ata2: failed to recover some devices, retrying in 5 secs
Feb  6 21:55:19 localhost ata2.01: device is on DMA blacklist, disabling DMA
Feb  6 21:55:19 localhost ata2.01: configured for PIO3

it clearly adds a lot of time to boot up, so if DMA really isn't supported for my intel 815 chipset, then what do I tell the kernel so that it doesn't waste time checking for DMA support?

These two signs seem to be related somehow, I think.

This is all from my second hdd, that doesn't get stressed much, so I wouldn't notice a slowdown if there were one.

Last edited by vogt (2008-02-20 03:07:22)

cromo · 2007-02-08 22:04:30

I've noticed exactly same kind of messages with 2.6.20. i915 chipset here, sata drive.

vogt · 2007-02-08 23:23:13

I've thought about this, and ata2.01 would actually be the slave drive on the secondary ata thingy, which is some pretty generic lg 48x cd drive that I hardly use, so I could either remove it, or specify tho the kernel that it should use PIO3 (that it falls back to) at boot

Last edited by vogt (2007-02-08 23:23:26)

viga · 2007-06-12 15:37:44

I am having the same issue with fxmode failing to load. It really adds a long time to boot. Has any one solved this problem?

Thanks

kth5 · 2007-06-12 16:33:18

There's some chipsets that are reported broken in the kernel for pata still. Try to fall back to legacy ide until pata becomes more table, would be my suggestion. Really.

So what you would do to find out whether it's pata or not:
* reboot
* select Fallback
* in grub edit the append line to include "legacy_ide earlymodules=piix"
* "ro" should go away as fstab will not match "/dev/hda*" as oposed to "/dev/sda*" anymore

If even that doesn't speed up your disks, something is seriously fucked and I suggest you downgrade your kernels.

Kane · 2007-07-26 17:30:13

Hi guys,
I havnt updated arch for a long while because last time i tried i had similar sorts of problems like above. This also happened with other distros i use with the new kernel.
If i put a cd in the cdrom drive, the problem went away and it would boot fine (it also worked if i used to ide-legacy option), but forgetting to put a cd in the drive meant a long wait so i wanted to find a fix for it.
So here i am and after a few months wait i thought there may be some changes and somebody may have fixed it. A simple solution i found was to simply add the word irqpoll to the append line in lilo, the slow boot goes away (well it did for me anyway) and problem fixed (or at least i dont have to worry about putting a cd in the drive before boot lol)
So now im back and using an up to date arch once again
Hope this helps

Kane

vogt · 2008-02-20 03:08:07

This got solved by an update about half a year ago.

lilsirecho · 2008-02-20 03:58:56

Please examine dmesg regarding the assigned mode for your drives in the sdx mode.

bangkok_manouel · 2008-02-20 04:02:58

Hmmm last post before today is 7 months old guys

lilsirecho · 2008-02-20 04:09:31

Please read the following up-to-date URL for background on the problem.....

http://pixca.net/2008/02/03/limited-to- … ire-cable/

This is kernel 2.6.24 material...........

Problem solved? hehe

Arch Linux

#1 2007-02-07 03:12:45

kernel ata_piix DMA issues [solved]

#2 2007-02-08 22:04:30

Re: kernel ata_piix DMA issues [solved]

#3 2007-02-08 23:23:13

Re: kernel ata_piix DMA issues [solved]

#4 2007-06-12 15:37:44

Re: kernel ata_piix DMA issues [solved]

#5 2007-06-12 16:33:18

Re: kernel ata_piix DMA issues [solved]

#6 2007-07-26 17:30:13

Re: kernel ata_piix DMA issues [solved]

#7 2008-02-20 03:08:07

Re: kernel ata_piix DMA issues [solved]

#8 2008-02-20 03:58:56

Re: kernel ata_piix DMA issues [solved]

#9 2008-02-20 04:02:58

Re: kernel ata_piix DMA issues [solved]

#10 2008-02-20 04:09:31

Re: kernel ata_piix DMA issues [solved]

Board footer