(fwd) Re: Bug#289770: kernel-image-2.6.10-1-686: 2.6.10 fails to set up DMA on my IBM thinkpad

dma on a thinkpad 600E worked for 2.6.8, but didn't since 2.6.9 also non working 2.6.10.

-hda: 12594960 sectors (6448 MB) w/460KiB Cache, CHS=13328/15/63, UDMA(33) +hda: 12594960 sectors (6448 MB) w/460KiB Cache, CHS=13328/15/63

could you please look at the following ide dma bug report for piix in debian -> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=289770
full dmesg of both 2.6.8 and 2.6.10 are posted there.

thanks maks

Incidentally, this may be the same bug as Debian bug #284558:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=284558

which involves more recent PIIX hardware.

Phil

http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt

"something" steals resource needed for IDE driver so no DMA

this one is an user error, ide_generic driver shouldn't be used at the same time as piix driver

sorry, not an user error, it looks like a bug in distribution (ide_generic shouldn't be loaded unconditionally!)

Does the kernel know what the 'something' is? Should I be routing around in /proc looking for something specific?

cheers,

Phil

http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt

Philip and Bartlomiej,

I've been tracking an almost identical problem on my IA64 platform with a bk pull that is about a month old.

I tracked it down to a duplicate vector request of 0x38 = 56 by the mmtimer -- take a look at /proc/interrupts, /proc/ioports, /proc/iomem and see if anything has requested the specific area that your CD is trying to get.

My output looked like:

SGIIOC4: IDE controller at PCI slot 0000:01:01.0, revision 79
ide0: BM-DMA at 0xc00001080c200140-0xc00001080c200163 Probing IDE interface ide0...
hda: MATSHITADVD-ROM SR-8588, ATAPI CD/DVD-ROM drive
ide0: Disabled unable to get IRQ 56.
Probing IDE interface ide0...
hda: MATSHITADVD-ROM SR-8588, ATAPI CD/DVD-ROM drive
hda: IRQ probe failed (0x4ffffffe)
Probing IDE interface ide1...
ide1: Wait for ready failed before probe !
Probing IDE interface ide2...
ide2: Wait for ready failed before probe !
Probing IDE interface ide3...
ide3: Wait for ready failed before probe !
Probing IDE interface ide4...
ide4: Wait for ready failed before probe !
Probing IDE interface ide5...
ide5: Wait for ready failed before probe !
ide0: DISABLED, NO IRQ

Bartlomiej, just a thought:

I think that the output of the failures of the probes for non-existant interfaces ide1-ide5 could be suppressed, no?

What would you think if we only output the "failed before probe" messages for interfaces that return -EBUSY from wait_hwif_ready? It would make the system boot a lot "prettier":)

P.

yes

care to make a patch?

Interrupts and iomem seem pretty much the same, but ioports looks like the following:

bucephalus:~# diff -W 80 --side-by-side 2.6.8/ioports 2.6.10/ioports
0000-001f : dma1                        0000-001f : dma1
0020-0021 : pic1                        0020-0021 : pic1
0040-005f : timer                     | 0040-0043 : timer0
                                      > 0050-0053 : timer1
0060-006f : keyboard                    0060-006f : keyboard
0070-0077 : rtc                         0070-0077 : rtc
0080-008f : dma page reg                0080-008f : dma page reg
00a0-00a1 : pic2                        00a0-00a1 : pic2
00c0-00df : dma2                        00c0-00df : dma2
00f0-00ff : fpu                         00f0-00ff : fpu
01f0-01f7 : ide0                        01f0-01f7 : ide0
0200-0207 : ns558-pnp                   0200-0207 : ns558-pnp
03bc-03be : parport0                    03bc-03be : parport0
03c0-03df : vga+                        03c0-03df : vga+
03f6-03f6 : ide0                        03f6-03f6 : ide0
03f8-03ff : serial                      03f8-03ff : serial
04d0-04d1 : pnp 00:0a                   04d0-04d1 : pnp 00:0a
0530-0533 : Crystal audio controller    0530-0533 : Crystal audio
controller
0cf8-0cff : PCI conf1                   0cf8-0cff : PCI conf1
1000-100f : 0000:00:07.1              <
  1000-1007 : ide0                    <
15e0-15ef : pnp 00:0a                   15e0-15ef : pnp 00:0a
4000-40ff : PCI CardBus #02             4000-40ff : PCI CardBus #02
4400-44ff : PCI CardBus #02             4400-44ff : PCI CardBus #02
4800-48ff : PCI CardBus #06             4800-48ff : PCI CardBus #06
4c00-4cff : PCI CardBus #06             4c00-4cff : PCI CardBus #06
8400-841f : 0000:00:07.2                8400-841f : 0000:00:07.2
  8400-841f : uhci_hcd                    8400-841f : uhci_hcd
d000-dfff : PCI Bus #01                 d000-dfff : PCI Bus #01
                                      > ef00-ef3f : 0000:00:07.3
efa0-efbf : 0000:00:07.3                efa0-efbf : 0000:00:07.3
  ef00-efaf : pnp 00:0a               | fcf0-fcff : 0000:00:07.1
    ef00-ef3f : 0000:00:07.3          |   fcf8-fcff : pnp 00:1c
fcf8-fcff : pnp 00:1c                 <

cheers,

Phil

http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt

/proc/ioports for both kernels should tells us more...

Further investigation reveals that turning off CONFIG_PNPBIOS solves the problem.

With PNPBIOS set, the kernel seems to think that the io range is in use regardless of whether quickboot is set in the BIOS or not.

cheers,

Phil

http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt

I have Toshiba Satellite 320CDT.
I compiled 2.6.10 vanilla kernel with CONFIG_PNPBIOS disabled. All IDE drivers was compiled into the kernel.
I can't still enable DMA for my harrddisk.
This is my earlier crying:

http://marc.theaimsgroup.com/?t=110789166000004&r=1&w=2

What can I do?

Best regards

Marcin Kuk

Addendum:

# diff 2.6.10/ioports nopnp/ioports
10a11

12,13d12
< 0200-0207 : ns558-pnp
< 03bc-03be : parport0
16,17d14
< 03f8-03ff : serial
< 04d0-04d1 : pnp 00:0a
20d16
< 15e0-15ef : pnp 00:0a
31c27
< fcf8-fcff : pnp 00:1c
---

(The nopnp kernel doesn't have the ns558 joystick driver compiled in)

cheers,

Phil

http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt