Re: bug in softint_execute() ?

[Andrew Doran <ad%netbsd.org@localhost>]

On Fri, Apr 11, 2008 at 07:22:04PM +0900, Takeshi Nakayama wrote:

> FYI, I got a similar strange assertion panic on sparc64 SMP box.
> 
> panic: kernel diagnostic assertion "curcpu() == ci" failed: file 
> "../../../../kern/vfs_bio.c", line 1541
> Stopped in pid 19327.1 (troff) at       netbsd:cpu_Debugger+0x4:        nop
> db{1}> t
> __kernassert(1430560, 1462168, 605, 1462290, 1, c418000) at 
> netbsd:__kernassert+0x2c
> biointr(0, 2b60400, 0, 400, 5, 27cc00) at netbsd:biointr+0xc0
> softint_overlay(2b6ac00, 0, 0, 21, 0, 27cc00) at netbsd:softint_overlay+0x3e4
> lwp_userret(10cddc40, 0, 2, ffffffffffffab00, 1, 1c14000) at 
> netbsd:lwp_userret+0x170
> trap(d1cded0, fffffffffffffffe, 14c2b4, 820082, 132a500, 80) at 
> netbsd:trap+0x408
> ?(ffffffffffffa7b6, 0, 0, 21, 27bc80, 27cc00) at 0x1008b64

This should be fixed as of now. softint_overlay binds the LWP to the current
CPU.

Andrew