1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-27 11:55:06 +00:00
freebsd/sys/kern/subr_kdb.c
Attilio Rao dc6fbf6545 * Completely Remove the option STOP_NMI from the kernel. This option
has proven to have a good effect when entering KDB by using a NMI,
but it completely violates all the good rules about interrupts
disabled while holding a spinlock in other occasions.  This can be the
cause of deadlocks on events where a normal IPI_STOP is expected.
* Adds an new IPI called IPI_STOP_HARD on all the supported architectures.
This IPI is responsible for sending a stop message among CPUs using a
privileged channel when disponible. In other cases it just does match a
normal IPI_STOP.
Right now the IPI_STOP_HARD functionality uses a NMI on ia32 and amd64
architectures, while on the other has a normal IPI_STOP effect. It is
responsibility of maintainers to eventually implement an hard stop
when necessary and possible.
* Use the new IPI facility in order to implement a new userend SMP kernel
function called stop_cpus_hard(). That is specular to stop_cpu() but
it does use the privileged channel for the stopping facility.
* Let KDB use the newly introduced function stop_cpus_hard() and leave
stop_cpus() for all the other cases
* Disable interrupts on CPU0 when starting the process of APs suspension.
* Style cleanup and comments adding

This patch should fix the reboot/shutdown deadlocks many users are
constantly reporting on mailing lists.

Please don't forget to update your config file with the STOP_NMI
option removal

Reviewed by:	jhb
Tested by:	pho, bz, rink
Approved by:	re (kib)
2009-08-13 17:09:45 +00:00

548 lines
12 KiB
C

/*-
* Copyright (c) 2004 The FreeBSD Project
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_kdb.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kdb.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/pcpu.h>
#include <sys/proc.h>
#include <sys/smp.h>
#include <sys/sysctl.h>
#include <machine/kdb.h>
#include <machine/pcb.h>
#ifdef SMP
#include <machine/smp.h>
#endif
int kdb_active = 0;
static void *kdb_jmpbufp = NULL;
struct kdb_dbbe *kdb_dbbe = NULL;
static struct pcb kdb_pcb;
struct pcb *kdb_thrctx = NULL;
struct thread *kdb_thread = NULL;
struct trapframe *kdb_frame = NULL;
KDB_BACKEND(null, NULL, NULL, NULL);
SET_DECLARE(kdb_dbbe_set, struct kdb_dbbe);
static int kdb_sysctl_available(SYSCTL_HANDLER_ARGS);
static int kdb_sysctl_current(SYSCTL_HANDLER_ARGS);
static int kdb_sysctl_enter(SYSCTL_HANDLER_ARGS);
static int kdb_sysctl_panic(SYSCTL_HANDLER_ARGS);
static int kdb_sysctl_trap(SYSCTL_HANDLER_ARGS);
static int kdb_sysctl_trap_code(SYSCTL_HANDLER_ARGS);
SYSCTL_NODE(_debug, OID_AUTO, kdb, CTLFLAG_RW, NULL, "KDB nodes");
SYSCTL_PROC(_debug_kdb, OID_AUTO, available, CTLTYPE_STRING | CTLFLAG_RD, NULL,
0, kdb_sysctl_available, "A", "list of available KDB backends");
SYSCTL_PROC(_debug_kdb, OID_AUTO, current, CTLTYPE_STRING | CTLFLAG_RW, NULL,
0, kdb_sysctl_current, "A", "currently selected KDB backend");
SYSCTL_PROC(_debug_kdb, OID_AUTO, enter, CTLTYPE_INT | CTLFLAG_RW, NULL, 0,
kdb_sysctl_enter, "I", "set to enter the debugger");
SYSCTL_PROC(_debug_kdb, OID_AUTO, panic, CTLTYPE_INT | CTLFLAG_RW, NULL, 0,
kdb_sysctl_panic, "I", "set to panic the kernel");
SYSCTL_PROC(_debug_kdb, OID_AUTO, trap, CTLTYPE_INT | CTLFLAG_RW, NULL, 0,
kdb_sysctl_trap, "I", "set to cause a page fault via data access");
SYSCTL_PROC(_debug_kdb, OID_AUTO, trap_code, CTLTYPE_INT | CTLFLAG_RW, NULL, 0,
kdb_sysctl_trap_code, "I", "set to cause a page fault via code access");
/*
* Flag indicating whether or not to IPI the other CPUs to stop them on
* entering the debugger. Sometimes, this will result in a deadlock as
* stop_cpus() waits for the other cpus to stop, so we allow it to be
* disabled. In order to maximize the chances of success, use a hard
* stop for that.
*/
#ifdef SMP
static int kdb_stop_cpus = 1;
SYSCTL_INT(_debug_kdb, OID_AUTO, stop_cpus, CTLTYPE_INT | CTLFLAG_RW,
&kdb_stop_cpus, 0, "stop other CPUs when entering the debugger");
TUNABLE_INT("debug.kdb.stop_cpus", &kdb_stop_cpus);
#endif
/*
* Flag to indicate to debuggers why the debugger was entered.
*/
const char * volatile kdb_why = KDB_WHY_UNSET;
static int
kdb_sysctl_available(SYSCTL_HANDLER_ARGS)
{
struct kdb_dbbe *be, **iter;
char *avail, *p;
ssize_t len, sz;
int error;
sz = 0;
SET_FOREACH(iter, kdb_dbbe_set) {
be = *iter;
if (be->dbbe_active == 0)
sz += strlen(be->dbbe_name) + 1;
}
sz++;
avail = malloc(sz, M_TEMP, M_WAITOK);
p = avail;
*p = '\0';
SET_FOREACH(iter, kdb_dbbe_set) {
be = *iter;
if (be->dbbe_active == 0) {
len = snprintf(p, sz, "%s ", be->dbbe_name);
p += len;
sz -= len;
}
}
KASSERT(sz >= 0, ("%s", __func__));
error = sysctl_handle_string(oidp, avail, 0, req);
free(avail, M_TEMP);
return (error);
}
static int
kdb_sysctl_current(SYSCTL_HANDLER_ARGS)
{
char buf[16];
int error;
if (kdb_dbbe != NULL) {
strncpy(buf, kdb_dbbe->dbbe_name, sizeof(buf));
buf[sizeof(buf) - 1] = '\0';
} else
*buf = '\0';
error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
if (error != 0 || req->newptr == NULL)
return (error);
if (kdb_active)
return (EBUSY);
return (kdb_dbbe_select(buf));
}
static int
kdb_sysctl_enter(SYSCTL_HANDLER_ARGS)
{
int error, i;
error = sysctl_wire_old_buffer(req, sizeof(int));
if (error == 0) {
i = 0;
error = sysctl_handle_int(oidp, &i, 0, req);
}
if (error != 0 || req->newptr == NULL)
return (error);
if (kdb_active)
return (EBUSY);
kdb_enter(KDB_WHY_SYSCTL, "sysctl debug.kdb.enter");
return (0);
}
static int
kdb_sysctl_panic(SYSCTL_HANDLER_ARGS)
{
int error, i;
error = sysctl_wire_old_buffer(req, sizeof(int));
if (error == 0) {
i = 0;
error = sysctl_handle_int(oidp, &i, 0, req);
}
if (error != 0 || req->newptr == NULL)
return (error);
panic("kdb_sysctl_panic");
return (0);
}
static int
kdb_sysctl_trap(SYSCTL_HANDLER_ARGS)
{
int error, i;
int *addr = (int *)0x10;
error = sysctl_wire_old_buffer(req, sizeof(int));
if (error == 0) {
i = 0;
error = sysctl_handle_int(oidp, &i, 0, req);
}
if (error != 0 || req->newptr == NULL)
return (error);
return (*addr);
}
static int
kdb_sysctl_trap_code(SYSCTL_HANDLER_ARGS)
{
int error, i;
void (*fp)(u_int, u_int, u_int) = (void *)0xdeadc0de;
error = sysctl_wire_old_buffer(req, sizeof(int));
if (error == 0) {
i = 0;
error = sysctl_handle_int(oidp, &i, 0, req);
}
if (error != 0 || req->newptr == NULL)
return (error);
(*fp)(0x11111111, 0x22222222, 0x33333333);
return (0);
}
void
kdb_panic(const char *msg)
{
#ifdef SMP
stop_cpus_hard(PCPU_GET(other_cpus));
#endif
printf("KDB: panic\n");
panic(msg);
}
void
kdb_reboot(void)
{
printf("KDB: reboot requested\n");
shutdown_nice(0);
}
/*
* Solaris implements a new BREAK which is initiated by a character sequence
* CR ~ ^b which is similar to a familiar pattern used on Sun servers by the
* Remote Console.
*
* Note that this function may be called from almost anywhere, with interrupts
* disabled and with unknown locks held, so it must not access data other than
* its arguments. Its up to the caller to ensure that the state variable is
* consistent.
*/
#define KEY_CR 13 /* CR '\r' */
#define KEY_TILDE 126 /* ~ */
#define KEY_CRTLB 2 /* ^B */
#define KEY_CRTLP 16 /* ^P */
#define KEY_CRTLR 18 /* ^R */
int
kdb_alt_break(int key, int *state)
{
int brk;
brk = 0;
switch (*state) {
case 0:
if (key == KEY_CR)
*state = 1;
break;
case 1:
if (key == KEY_TILDE)
*state = 2;
break;
case 2:
if (key == KEY_CRTLB)
brk = KDB_REQ_DEBUGGER;
else if (key == KEY_CRTLP)
brk = KDB_REQ_PANIC;
else if (key == KEY_CRTLR)
brk = KDB_REQ_REBOOT;
*state = 0;
}
return (brk);
}
/*
* Print a backtrace of the calling thread. The backtrace is generated by
* the selected debugger, provided it supports backtraces. If no debugger
* is selected or the current debugger does not support backtraces, this
* function silently returns.
*/
void
kdb_backtrace(void)
{
if (kdb_dbbe != NULL && kdb_dbbe->dbbe_trace != NULL) {
printf("KDB: stack backtrace:\n");
kdb_dbbe->dbbe_trace();
}
}
/*
* Set/change the current backend.
*/
int
kdb_dbbe_select(const char *name)
{
struct kdb_dbbe *be, **iter;
SET_FOREACH(iter, kdb_dbbe_set) {
be = *iter;
if (be->dbbe_active == 0 && strcmp(be->dbbe_name, name) == 0) {
kdb_dbbe = be;
return (0);
}
}
return (EINVAL);
}
/*
* Enter the currently selected debugger. If a message has been provided,
* it is printed first. If the debugger does not support the enter method,
* it is entered by using breakpoint(), which enters the debugger through
* kdb_trap(). The 'why' argument will contain a more mechanically usable
* string than 'msg', and is relied upon by DDB scripting to identify the
* reason for entering the debugger so that the right script can be run.
*/
void
kdb_enter(const char *why, const char *msg)
{
if (kdb_dbbe != NULL && kdb_active == 0) {
if (msg != NULL)
printf("KDB: enter: %s\n", msg);
kdb_why = why;
breakpoint();
kdb_why = KDB_WHY_UNSET;
}
}
/*
* Initialize the kernel debugger interface.
*/
void
kdb_init(void)
{
struct kdb_dbbe *be, **iter;
int cur_pri, pri;
kdb_active = 0;
kdb_dbbe = NULL;
cur_pri = -1;
SET_FOREACH(iter, kdb_dbbe_set) {
be = *iter;
pri = (be->dbbe_init != NULL) ? be->dbbe_init() : -1;
be->dbbe_active = (pri >= 0) ? 0 : -1;
if (pri > cur_pri) {
cur_pri = pri;
kdb_dbbe = be;
}
}
if (kdb_dbbe != NULL) {
printf("KDB: debugger backends:");
SET_FOREACH(iter, kdb_dbbe_set) {
be = *iter;
if (be->dbbe_active == 0)
printf(" %s", be->dbbe_name);
}
printf("\n");
printf("KDB: current backend: %s\n",
kdb_dbbe->dbbe_name);
}
}
/*
* Handle contexts.
*/
void *
kdb_jmpbuf(jmp_buf new)
{
void *old;
old = kdb_jmpbufp;
kdb_jmpbufp = new;
return (old);
}
void
kdb_reenter(void)
{
if (!kdb_active || kdb_jmpbufp == NULL)
return;
longjmp(kdb_jmpbufp, 1);
/* NOTREACHED */
}
/*
* Thread related support functions.
*/
struct pcb *
kdb_thr_ctx(struct thread *thr)
{
#if defined(SMP) && defined(KDB_STOPPEDPCB)
struct pcpu *pc;
#endif
if (thr == curthread)
return (&kdb_pcb);
#if defined(SMP) && defined(KDB_STOPPEDPCB)
SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
if (pc->pc_curthread == thr && (stopped_cpus & pc->pc_cpumask))
return (KDB_STOPPEDPCB(pc));
}
#endif
return (thr->td_pcb);
}
struct thread *
kdb_thr_first(void)
{
struct proc *p;
struct thread *thr;
p = LIST_FIRST(&allproc);
while (p != NULL) {
if (p->p_flag & P_INMEM) {
thr = FIRST_THREAD_IN_PROC(p);
if (thr != NULL)
return (thr);
}
p = LIST_NEXT(p, p_list);
}
return (NULL);
}
struct thread *
kdb_thr_from_pid(pid_t pid)
{
struct proc *p;
p = LIST_FIRST(&allproc);
while (p != NULL) {
if (p->p_flag & P_INMEM && p->p_pid == pid)
return (FIRST_THREAD_IN_PROC(p));
p = LIST_NEXT(p, p_list);
}
return (NULL);
}
struct thread *
kdb_thr_lookup(lwpid_t tid)
{
struct thread *thr;
thr = kdb_thr_first();
while (thr != NULL && thr->td_tid != tid)
thr = kdb_thr_next(thr);
return (thr);
}
struct thread *
kdb_thr_next(struct thread *thr)
{
struct proc *p;
p = thr->td_proc;
thr = TAILQ_NEXT(thr, td_plist);
do {
if (thr != NULL)
return (thr);
p = LIST_NEXT(p, p_list);
if (p != NULL && (p->p_flag & P_INMEM))
thr = FIRST_THREAD_IN_PROC(p);
} while (p != NULL);
return (NULL);
}
int
kdb_thr_select(struct thread *thr)
{
if (thr == NULL)
return (EINVAL);
kdb_thread = thr;
kdb_thrctx = kdb_thr_ctx(thr);
return (0);
}
/*
* Enter the debugger due to a trap.
*/
int
kdb_trap(int type, int code, struct trapframe *tf)
{
register_t intr;
#ifdef SMP
int did_stop_cpus;
#endif
int handled;
if (kdb_dbbe == NULL || kdb_dbbe->dbbe_trap == NULL)
return (0);
/* We reenter the debugger through kdb_reenter(). */
if (kdb_active)
return (0);
intr = intr_disable();
#ifdef SMP
if ((did_stop_cpus = kdb_stop_cpus) != 0)
stop_cpus_hard(PCPU_GET(other_cpus));
#endif
kdb_active++;
kdb_frame = tf;
/* Let MD code do its thing first... */
kdb_cpu_trap(type, code);
makectx(tf, &kdb_pcb);
kdb_thr_select(curthread);
handled = kdb_dbbe->dbbe_trap(type, code);
kdb_active--;
#ifdef SMP
if (did_stop_cpus)
restart_cpus(stopped_cpus);
#endif
intr_restore(intr);
return (handled);
}