On Sat, 2 Apr 2016, Philip Guenther wrote:
> I ask as I see some "this can't work" logic between the kernel and gdb for
> at least i386 and possibly others and I'd like to hear what *is* working
> that I should verify I don't break when making changes...
Here's a quick fix for the initial bug that caught my eye. On i386,
savectx() sets PCB_SAVECTX in pcb_flags. Okay, that's nice, but
a) savectx() is called from cpu_fork(), which results in the flag being
set on both parent and child, and
b) nothing clears it
Result is that it's set on *all* threads, which makes gdb's "kvm proc"
never show the frame of mi_switch(), as it thinks this pcb is always from
savectx().
Diff below fixes that by eliminating the savectx() call in cpu_fork().
Turns out that call is completely unnecessary: savectx() only updates
pcb_esp and pcb_ebp (and the PCB_SAVECTX flag), but cpu_fork() overwrites
both those values in the child's pcb, while the values in the parent are
invalid as soon as it returns.
before:
(gdb) bt
#0 0xd03bf814 in sleep_finish (sls=0x0, do_sleep=1)
at ../../../../kern/kern_synch.c:266
#1 0xd03bfddf in tsleep (ident=0xd0b8ae44, priority=288,
wmesg=0xd09e9690 "nanosleep", timo=0) at
../../../../kern/kern_synch.c:148
#2 0xd03c1faa in sys_nanosleep (p=0xd7203cdc, v=0xf5d5cf5c, retval=0xf5d5cf7c)
at ../../../../kern/kern_time.c:289
after:
(gdb) bt
#0 mi_switch () at cpu.h:208
#1 0xd03bf814 in sleep_finish (sls=0xf5d84e7c, do_sleep=1)
at ../../../../kern/kern_synch.c:266
#2 0xd03bfddf in tsleep (ident=0xd0b8ae44, priority=288,
wmesg=0xd09e9690 "nanosleep", timo=600001)
at ../../../../kern/kern_synch.c:148
#3 0xd03c1faa in sys_nanosleep (p=0xd7204b70, v=0xf5d84f5c, retval=0xf5d84f7c)
at ../../../../kern/kern_time.c:289
Note the bogus sls=0x0 in before case.
ok?
Philip Guenther
Index: vm_machdep.c
===================================================================
RCS file: /data/src/openbsd/src/sys/arch/i386/i386/vm_machdep.c,v
retrieving revision 1.63
diff -u -p -r1.63 vm_machdep.c
--- vm_machdep.c 5 May 2015 02:13:46 -0000 1.63
+++ vm_machdep.c 3 Apr 2016 07:19:50 -0000
@@ -85,13 +85,8 @@ cpu_fork(struct proc *p1, struct proc *p
p2->p_md.md_flags = p1->p_md.md_flags;
- /* Copy pcb from proc p1 to p2. */
- if (p1 == curproc) {
- /* Sync the PCB before we copy it. */
- savectx(curpcb);
- }
#ifdef DIAGNOSTIC
- else if (p1 != &proc0)
+ if (p1 != curproc && p1 != &proc0)
panic("cpu_fork: curproc");
#endif
*pcb = p1->p_addr->u_pcb;
