remote kernel gdb stack frames corrupted after breakpoint or
gallatin at cs.duke.edu
Wed Sep 15 14:53:04 PDT 2004
Kip McAtee writes:
> I'm building on 5.3-BETA4 (and 6.0 current)
> with gdb (and kgdb) 6.1.1.
> I can get into the remote kernel debugger with
> sysctl debug.kdb.enter=1
> I can set a breakpoint and continue.
> When it hits the breakpoint I get:
> Breakpoint 2, wakeup (ident=0xc0c0f9e8) at ../../../kern/kern_synch.c:253
> 253 sleepq_broadcast(ident, SLEEPQ_MSLEEP, -1);
> (kgdb) c
> Program received signal SIGSEGV, Segmentation fault.
> Looking at the backtrace implies that the stack frames
> are not displayed correctly. I'm setting:
FWIW, I see the same thing when continuing from breakpoints in
RELENG_5 via ddb. I've appended a log from my serial console.
Interestingly, dropping into the debugger via break on console works
like a charm, its only inserted breakpoints which don't work.
I think this started happening after the ddb->kdb conversion.
[send halt on serial console]
db> break sched_switch
Stopped at sched_switch+0x1: movl %esp,%ebp
sched_switch(1,0,0,0,0) at sched_switch+0x1
idle_proc(0,e3f78d48,10e90,10e98,10ea0) at idle_proc+0x143
fork_exit(c04fa329,0,e3f78d48) at fork_exit+0x80
fork_trampoline() at fork_trampoline+0x8
--- trap 0x1, eip = 0, esp = 0xe3f78d7c, ebp = 0 ---
kernel trap 12 with interrupts disabled
Fatal trap 12: page fault while in kernel mode
cpuid = 0; apic id = 00
fault virtual address = 0x0
fault code = supervisor read, page not present
instruction pointer = 0x8:0xc0527256
stack pointer = 0x10:0xe3f78c8c
frame pointer = 0x10:0xe3f78c9c
code segment = base 0x0, limit 0xfffff, type 0x1b
= DPL 0, pres 1, def32 1, gran 1
processor eflags = resume, IOPL = 0
current process = 12 (idle: cpu0)
Stopped at sched_switch+0xe: movl 0(%ebx),%edx
sched_switch(0,10,44096318,47c326eb,ffc00014) at sched_switch+0xe
end(74c085d0,b816,a3640000,c,89d0558b) at 0xc1562640
db> call cpu_reset
cpu_reset called on cpu#0
More information about the freebsd-current