Re: Some problems with KDB/scheduler?

Robert Watson wrote:

>I've been having some problems with KDB this evening; right now what I'm
>bumping into appears to be a problem with interrupts and KDB/DDB.  An
>example below.
>
>(debugger entered using sysctl debug.kdb.enter=1)
>
>db
>tiger-3# [thread 100071]
>Breakpoint at   vlan_start:     pushl   %ebp
>db> cont
>KDB: stack backtrace:
>mi_switch(6,c5457300,c5457450,c5457300,e7868c48) at mi_switch+0x2ce
>maybe_preempt(c5457300,e7868c40,c06033ce,197,c080c288) at
>maybe_preempt+0xf9
>sched_add(c5457300) at sched_add+0x26a
>


ithread_schedule() doesn't call sched_add().. It calls setrunqueue() 
which calls
sched_add().

this looks like something that was reported a few days ago where
it looks like kdb/ddb is skipping frames.

>ithread_schedule(c5455500,c5404a80,12,0,c554da00) at ithread_schedule+0x8a
>intr_execute_handlers(c53fc5e0,e7868ca8,12,e7868cec,c0793e03) at
>intr_execute_ha
>ndlers+0x65
>lapic_handle_intr(42) at lapic_handle_intr+0x23
>Xapic_isr2() at Xapic_isr2+0x33
>--- interrupt, eip = 0xc0a27b71, esp = 0xe7868cec, ebp = 0xe7868cec ---
>acpi_cpu_c1(e7868d04,c0619321,c08d1520,1,fffffffe) at acpi_cpu_c1+0x5
>acpi_cpu_idle(c05f8328,c5403600,e7868d34,c05f811e,0) at acpi_cpu_idle+0xe1
>idle_proc(0,e7868d48,0,c05f8328,0) at idle_proc+0x65
>fork_exit(c05f8328,0,e7868d48) at fork_exit+0x7e
>fork_trampoline() at fork_trampoline+0x8
>--- trap 0x1, eip = 0, esp = 0xe7868d7c, ebp = 0 ---
>panic: mi_switch: did not reenter debugger
>cpuid = 0
>boot() called on cpu#0
>Uptime: 1m26s
>
>_______________________________________________
>
>  
>