XHCI: endpoint is not initialized

[Lindenov]

This KDL appears on hrev55419. The system cannot start with USB devices pluged on, unlike on previous revisions or beta 3.

[waddlesplash]

X512 sent me a picture of this last night. It does not reproduce on QEMU unfortunately; I suspect it may be a race condition of some kind, but so far I am kind of lost as to the details.

[Lindenov]

Replying to waddlesplash:

X512 sent me a picture of this last night. It does not reproduce on QEMU unfortunately; I suspect it may be a race condition of some kind, but so far I am kind of lost as to the details.

Do you need any additional information from me?

[waddlesplash]

Is it possible for you to run "syslog | grep usb" at the KDL prompt and take a picture of that?

[Lindenov]

Replying to waddlesplash:

Is it possible for you to run "syslog | grep usb" at the KDL prompt and take a picture of that?

I noticed that occasionally the system DOES start with device plugged in. Perhaps it depends on how the system has been booted. "Cold start" shows KDL every single time. Rebooting from another system state (ex. I'm using beta 3 now) and switching to hrev55419 allows it to boot normaly.

[waddlesplash]

Can you attach a syslog from a successful cold boot? (Please delete your syslog files if doing this on an existing installation.)

[Lindenov]

It was too tricky to get system started with plugged devices, but somehow I got lucky and here is the syslog. However, it's not too much useful info in it. I'll keep hrev55419 up as long as it needed to get any kind of info, cause I'm not sure I'll be able to boot it once again without KDL appearing.

[waddlesplash]

Sorry, I meant to say, from a successful cold boot on a hrev where it works. The syslog fragment you sent here appears to be from such a prior hrev, as the message within it was removed in hrev55404, which is where the panic in the original screenshot was added.

So, this demonstrates that hrev55404 does not cause any new problems, it just exposes an issue which was already there. Please get me a full syslog from a "fully working" version if you would.

[Lindenov]

Replying to waddlesplash:

Please get me a full syslog from a "fully working" version if you would.

This is a hrev55403, the revision I used before updating to hrev55419. Full syslog comming right now.

[waddlesplash]

Unfortunately the critical parts of the syslog are cut off. Can you delete your syslogs, reboot, and attach both syslog and syslog.old?

[waddlesplash]

The critical bits are still missing; I guess they must get cut off early in the boot process. Can you edit the "kernel" settings file by uncommenting the "syslog_buffer_size" line and setting some large value (probably 512KB will be enough) and trying again?

[KapiX]

Isn't "critical parts missing" #15455?

[waddlesplash]

No. XHCI only starts after the kernel is loaded. And it looks like #15455 may be fixed already anyway.

[Lindenov]

Replying to waddlesplash:

probably 512KB will be enough

So, 524288? Ok, I did it.

[Lindenov]

i dunno but it looks like even worse than previous ones

[waddlesplash]

Now there are no XHCI errors at all! Maybe this is intermittent, somehow?

I guess the only way to get the bits we need at this point is to use on-screen paging during the boot with SMP disabled. That's pretty annoying to do, though.

[waddlesplash]

Hang on, I have a lead on this and the other USB crashes, at last.

[waddlesplash]

Please retest after hrev55429.

[Lindenov]

Replying to waddlesplash:

Please retest after hrev55429.

hrev55438 works fine

[waddlesplash]

Excellent!