On Sat, Dec 11, 1999 at 05:12:16PM +0100, Martin Husemann wrote:
> panic: softdep_disk_write_complete: lock is held
[...]
> #16 0xf012ad1c in log ()
> #17 0xf022b5f4 in softdep_disk_write_complete ()
> #18 0xf0140f73 in biodone ()
> #19 0xf01e1ea6 in wddone ()
> #20 0xf01e4529 in wdc_ata_bio_done ()
> #21 0xf01e4454 in wdc_ata_bio_intr ()
> #22 0xf0108074 in wdcintr ()
> #23 0xf01ed42e in pciide_compat_intr ()
> #24 0xf0101868 in Xintr14 ()
[...]

A bio (disk) interrupt occurred, but the softdep code thinks that
bio interrupts are blocked (this is the "lock" it's talking about).
This means that something in the softdep code or called by the
softdep code enables bio interrupts wrongly (i.e. a splx() call that's
wrong), or that the softdep code does not "release" the "lock" somewhere.

Did you ever get "WARNING: SPL NOT LOWERED ON SYSCALL EXIT" (only
printed if you have DIAGNOSTIC enabled)?

- Frank