Panic on boot (AHCI and/or VM related?)

[pieterpan]

Using the latest revision (r22589) to test to see if #1417 was fixed. (it is, btw) Have a core2duo laptop with ahci. Unfortunately now I'm greeted with a PANIC. Maybe VM or AHCI related, but AHCI has worked well before.

PANIC: vm_page_fault: unhandled page fault in kernel space at 0x31, ip 0x802c42c1

See attached bt. Sorry, no serial debug log, as the laptop doesn't have serial port.

[pieterpan]

back trace

[pieterpan]

Forgot to mention that it continues to boot if I enable console output

[marcusoverhagen]

This appears to be the same cause as bug #1434 The device manager loads the driver and passes an invalid cookie pointer

[korli]

Actually the problem resides in the device manager. KDiskDeviceManager::InitialDeviceScan(). Then the devfs triggers a device scan with "/dev", probe_for_driver_modules() includes the ahci pci device in the scan because its type "/dev/disk" starts with "/dev" (if it was an exact type match, it wouldn't find the disk devfs directory anyway). The ahci driver attachs its device to the ahci pci device. Then the devfs triggers a device scan with "/dev/disk", probe_for_driver_modules() includes the ahci pci device in the scan because its type "/dev/disk" starts with "/dev/disk". The ahci driver attachs again to the same ahci pci device.

A possible fix would be to check whether a pci device has already a child device attached, and avoid a new scan in probe_for_driver_modules() or register_supporting_child_devices() when node->children is not null.

Thoughts ?

[korli]

Fixed in r22726.

[marcusoverhagen]

Unfortunately r22726 didn't really fix it. It still happenes with the revision from last week, same backtrace.

[tqh]

Seeing same things as well on r23739. I'm using a AMD64 X2 on a SATA-drive in SATA-mode. Panics after 20-30s. Actually this was on a few revisions earlier. This one (where mmlr fixed a bit on sem) I think reported 'Panic: Double Fault' in wm, but I suspect it will do both.

But at least it boots...