Skip to contentSkip to conversation

New: PCI devices missing - including USB controller. Boot fail

Bug 88451 Bjorn Helgaas 2 participants 3 messages Nov 18, 2014 1 message not in the archive PCI

lore
Share & tools
lore

Bjorn Helgaas

author maintainer
#1 lore raw
In reply to a message not in the archive

[+cc linux-pci]

Hi Roland,

Thanks for the report!

Quoted 23 lines · On Tue, Nov 18, 2014 at 5:08 AM, <[email protected]> wrote:

https://bugzilla.kernel.org/show_bug.cgi?id=88451

            Bug ID: 88451
           Summary: PCI devices missing - including USB controller. Boot
                    fail
           Product: Drivers
           Version: 2.5
    Kernel Version: 3.17 3.18rc4
          Hardware: i386
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: PCI
          Assignee: [email protected]
          Reporter: [email protected]
        Regression: No

Created attachment 157971
--> https://bugzilla.kernel.org/attachment.cgi?id=157971&action=edit
dmesg from working kernel

Quoted 8 lines

While stock debian kernel 3.2.0-4-486 and kernel 3.2.63 showed no problems, 3.17+ fails.

Apparently, all PCI devices except 0000:00:00.0 and 0000:00:12.2 are suddenly missing, including USB controller - and thus boot from USB fails.

I have taken a look at git and there seems to be a lot of PCI code rework between 3.2.63 and 3.17+, which may be an explanation

Yeah, there have definitely been a lot of changes since 3.2.63 :) I can't think of an obvious suspect, though.

Quoted 2 lines

So, im curious whats the problem and how to fix it. Too many pci bootparams to try all of them :(

Wow, very impressive screenshot console log of the failing kernel! Thanks for all the work to put that together.

I notice that you're using "acpi=off" on both kernels. Is that to work around some problem? Do you know whether it's still needed in v3.17?

Can you boot with "ignore_loglevel"? Some of the PCI probing output is at KERN_DEBUG, which makes it into dmesg, but not normally to the console. The useful part is the stuff that looks like this:

  [    0.111587] pci 0000:00:00.0: [1078:0001] type 0 class 0x000600
  [    0.111927] pci 0000:00:0f.0: [100b:0020] type 0 class 0x000200
  [    0.112886] pci 0000:00:12.0: [1078:0100] type 0 class 0x000601
  [    0.113494] pci 0000:00:12.1: [1078:0101] type 0 class 0x000680
  ...

I can tell from other driver printks that in v3.17, we did find these devices:

[ 4.12] 00:12.2
[ 4.44] 00:12.4
[ 5.73] 00:0f.0
[16.54] 00:12.3

so my suspicion is that the PCI core actually does enumerate all the devices, but for some reason ohci_hcd isn't claiming 00:13.0.

Bjorn

Quoted 15 lines

This is an older geode gx1 board (still have multiple around). for their age, they are quite power efficient (<6W) and i have some ideas to turn them into something useful.

lspci
00:00.0 Host bridge: Cyrix Corporation PCI Master
00:0f.0 Ethernet controller: National Semiconductor Corporation DP83815 (MacPhyter) Ethernet Controller
00:12.0 ISA bridge: Cyrix Corporation 5530 Legacy [Kahlua] (rev 30) 00:12.1 Bridge: Cyrix Corporation 5530 SMI [Kahlua]
00:12.2 IDE interface: Cyrix Corporation 5530 IDE [Kahlua] 00:12.3 Multimedia audio controller: Cyrix Corporation 5530 Audio [Kahlua] 00:12.4 VGA compatible controller: Cyrix Corporation 5530 Video [Kahlua] 00:13.0 USB controller: Compaq Computer Corporation ZFMicro Chipset USB (rev
06)

#2 lore raw

Hello Bjorn,

Quoted 11 lines

Gesendet: Dienstag, 18. November 2014 um 15:22 Uhr
Von: "Bjorn Helgaas" <[email protected]>
An: "[email protected]" <[email protected]>
Cc: "Roland Kletzing" <[email protected]>
Betreff: Re: [Bug 88451] New: PCI devices missing - including USB controller. Boot fail

[+cc linux-pci]

Hi Roland,

Thanks for the report!

Thanks for quick response & help !

Quoted 27 lines · On Tue, Nov 18, 2014 at 5:08 AM, <[email protected]> wrote:
Quoted 26 lines

https://bugzilla.kernel.org/show_bug.cgi?id=88451

            Bug ID: 88451
           Summary: PCI devices missing - including USB controller. Boot
                    fail
           Product: Drivers
           Version: 2.5
    Kernel Version: 3.17 3.18rc4
          Hardware: i386
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: PCI
          Assignee: [email protected]
          Reporter: [email protected]
        Regression: No

Created attachment 157971
--> https://bugzilla.kernel.org/attachment.cgi?id=157971&action=edit
dmesg from working kernel

While stock debian kernel 3.2.0-4-486 and kernel 3.2.63 showed no problems, 3.17+ fails.

Quoted 8 lines
Quoted 5 lines

Apparently, all PCI devices except 0000:00:00.0 and 0000:00:12.2 are suddenly missing, including USB controller - and thus boot from USB fails.

I have taken a look at git and there seems to be a lot of PCI code rework between 3.2.63 and 3.17+, which may be an explanation

Yeah, there have definitely been a lot of changes since 3.2.63 :) I can't think of an obvious suspect, though.

No wonder.

It is not an pci issue , as i just found out some minutes ago. Having some time today for doing stupid trial&error testing, as i`m having a bad cold :-P So i was wrong assigning this to linux-pci.

The answer is simple:

Apparently, the pci code portion has been separated from ohci_hcd, so it seems ohci_hcd is not responsible anymore for an usb controller sitting on the pci bus, but there`s a separate module
now: ohci_pci

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/usb/host/ohci-hcd.c?id=2621d0119e574f12496c4ab731265d5777cb6a18

I found this by chance as i statically compiled usb into the kernel and suddenly it worked again, as i set CONFIG_USB_OHCI_PCI=y

In dmesg i then could see that ohci_pci now was jumping in - and that was the problem:

I was simply missing that module in my initrd, as it was not needed before :-P

Quoted 5 lines
Quoted 2 lines

So, im curious whats the problem and how to fix it. Too many pci bootparams to try all of them :(

Wow, very impressive screenshot console log of the failing kernel! Thanks for all the work to put that together

Oh, that was piece of cake. just 5mins of grabbing screenshots from a video and stitching them together. The harder part was to find boot_delay and lpj kernel params and make them work (to slow down output, as my lcd+cam were not fast enough for a clean picture), as with that applied, the kernel needs a long time to give a sign of life at all - which made me think it didn`t work on the first try...

Quoted 3 lines

I notice that you're using "acpi=off" on both kernels. Is that to work around some problem? Do you know whether it's still needed in v3.17?

Yes, iirc, i have used that for long because booting had issues without that.

Now it "works" - at least it does not hang anymore with acpi on:

[    0.000000] ACPI: Early table checksum verification disabled
[    0.000000] ACPI BIOS Error (bug): A valid RSDP was not found (20140724/tbxfroot-211)
[    0.344469] ACPI: Interpreter disabled.

This hardware does not have an ordinary bios, so i suspect it also has no acpi at all. Will need to find a way for proper power-off, though....

Quoted 8 lines

Can you boot with "ignore_loglevel"? Some of the PCI probing output is at KERN_DEBUG, which makes it into dmesg, but not normally to the console. The useful part is the stuff that looks like this:

  [    0.111587] pci 0000:00:00.0: [1078:0001] type 0 class 0x000600
  [    0.111927] pci 0000:00:0f.0: [100b:0020] type 0 class 0x000200
  [    0.112886] pci 0000:00:12.0: [1078:0100] type 0 class 0x000601
  [    0.113494] pci 0000:00:12.1: [1078:0101] type 0 class 0x000680

Ah, that`s the reason why i did not see it and what made me suspect that there is something wrong with pci. I was not aware that dmesg on console and dmesg on disk can differ. Good hint!

Quoted 2 lines

so my suspicion is that the PCI core actually does enumerate all the devices, but for some reason ohci_hcd isn't claiming 00:13.0.

As told above.

I`m putting this (including dmesg from 3.17) to bugzilla and closing it.

Thanks again - and sorry for the noise.

regards
Roland

author maintainer
#3 lore raw

On Tue, Nov 18, 2014 at 8:42 AM, Roland Kletzing <[email protected]> wrote:
...

Quoted 7 lines

The answer is simple:

Apparently, the pci code portion has been separated from ohci_hcd, so it seems ohci_hcd is not responsible anymore for an usb controller sitting on the pci bus, but there`s a separate module
now: ohci_pci

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/usb/host/ohci-hcd.c?id=2621d0119e574f12496c4ab731265d5777cb6a18

...

Quoted 1 line

I`m putting this (including dmesg from 3.17) to bugzilla and closing it.

Great, thanks for tracking this down!

Bjorn