usb stack freezes Dell Latitude D810

[bryanv]

Get a Dell Latitude D810

Boot Haiku

if the usb_hid device driver is installed, the machine will freeze (this is the first USB device that inits the bus_manager)

Removing that will get you further along the boot process. Any attempt to access USB will instantly freeze the machine, no KDL.

[bryanv]

Replying to bryanv:

Get a Dell Latitude D810

Boot Haiku

if the usb_hid device driver is installed, the machine will freeze (this is the first USB device that inits the bus_manager)

Removing that will get you further along the boot process. Any attempt to access USB will instantly freeze the machine, no KDL.

I should add that removing all USB drivers resulted in a perfectly working system.

[mmlr]

Hi Bryan

Could you check if there is relevant information in the serial debug output and post it here if you find something? I need to know whether some actions are initiated or if it's something with actually loading the stack and drivers.

Regards Michael

[korli]

Bryan, could you also make a "listdev" and give info about USB controllers found ?

[mmlr]

Could you retest with the latest changes? We've changed the usb_hid and the devfs rescan sections quite extensively in the last week. Maybe it fixed or at least changed the behaviour with that?

[bryanv]

I wish I could.

However, I no longer have this laptop. It was a machine from my employer, and I've since changed jobs.

I'll see what I can do about finding someone to test it though..

:-(

[scottmc]

Anyone else with a Dell Latitude D810 that can do some testing on this?

[Adek336]

On a Dell Latitude C800, it fails to boot most of the time (if it boots, it works fine). It either hangs after the third boot icon with F12 not working; blanks the screen and then KDLs (which I know it does, because typing "reb<cr>" reboots the laptop without pressing F12); or gives me an NMI either in picasso (in vesa add-on) or in net_server.

[Adek336]

To boot, I have to type continue in the kdl prompt after a NMI.

[mmlr]

The NMI is most probably not related to USB. The hang at boot might be though. Can you please enable on-screen debug output and check what the last few messages are that get printed when it hangs?

[Adek336]

corresponding part of the syslog when boot succeeds

[Adek336]

I obviously didn't see which boot icon appeared at that moment but I suspect this is what I was supposed to reproduce. Note that this problem happens occasionally, generally it happens only once in many boots. F12 didn't work. When on screen debug is enabled, at the beginning of the boot process, after only a few messages, I always get an NMI in main2(), but continue can be used to ignore it.

ints output under a successfully booted Haiku

...
int 11, enabled 3, handled 17912, unhandled 0
     firewire:fwohci_intr ...
     3com:intr_wrapper ...
     uhci: InterruptHandler_4UHCPIPv ...
...

[Adek336]

On my desktop computer I've also seen Haiku freeze after the third icon, although very rarely, only a few times.

[Adek336]

So this problem seems to occur to two rather different hardware settings. I think it is possible, that this bug happens to much more users, but they just ignore the problem and restart the computer after a freeze during boot.

I suggest adding the following to trunk

src/system/kernel/int.c:
221   -               dprintf("More than 99%% interrupts of vector %d are unhandled\n",
221   +               panic("More than 99%% interrupts of vector %d are unhandled\n",

so we could verify if it happens only for a handful, or for a vast group of users.

Also, it would give us more debugging options than now, when it freezes hard.

And I suspect that when the "More than" message appears, the system crawls and pretty much doesn't work anyway, so panicing it shouldn't harm anyone.

[Adek336]

More than 99%: backtrace

[Adek336]

More than 99%: teams, threads

[Adek336]

Can't continue through the panic, though. It should panic() only the first time it finds out that more than 99% interrupts are unhandled, and later on only dprintf the message instead of panicing over and over.

[scottmc]

Is this issue still with us? Do you still have the D810?

[tqh]

No feedback in 21 months, closing.