[PATCH] kprobes exception notifier fix

This patch modifies the return value of kprobes exceptions notify handler. 
The kprobes exception notifier returns NOTIFY_STOP on handling
notification.  This patch helps other debuggers to co-exists with the
Kprobes.  Other debuggers registered for exceptions notification must
return NOTIFY_STOP on handling the notification.
Signed-off-by: Prasanna S Panchamukhi <prasanna@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

arch/i386/kernel/kprobes.c

@@ -267,26 +267,26 @@ int kprobe_exceptions_notify(struct notifier_block *self, unsigned long val,
 	switch (val) {
 	case DIE_INT3:
 		if (kprobe_handler(args->regs))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_DEBUG:
 		if (post_kprobe_handler(args->regs))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_GPF:
 		if (kprobe_running() &&
 		    kprobe_fault_handler(args->regs, args->trapnr))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_PAGE_FAULT:
 		if (kprobe_running() &&
 		    kprobe_fault_handler(args->regs, args->trapnr))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	default:
 		break;
 	}
-	return NOTIFY_BAD;
+	return NOTIFY_DONE;
 }
 
 int setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)

arch/i386/kernel/traps.c

@@ -459,7 +459,7 @@ asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
 	info.si_code = sicode; \
 	info.si_addr = (void __user *)siaddr; \
 	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
-						== NOTIFY_OK) \
+						== NOTIFY_STOP) \
 		return; \
 	do_trap(trapnr, signr, str, 1, regs, error_code, &info); \
 }
@@ -528,7 +528,7 @@ asmlinkage void do_general_protection(struct pt_regs * regs, long error_code)
 gp_in_kernel:
 	if (!fixup_exception(regs)) {
 		if (notify_die(DIE_GPF, "general protection fault", regs,
-				error_code, 13, SIGSEGV) == NOTIFY_OK);
+				error_code, 13, SIGSEGV) == NOTIFY_STOP);
 			return;
 		die("general protection fault", regs, error_code);
 	}
@@ -602,7 +602,7 @@ static void default_do_nmi(struct pt_regs * regs)
  
 	if (!(reason & 0xc0)) {
 		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 0, SIGINT)
-							== NOTIFY_BAD)
+							== NOTIFY_STOP)
 			return;
 #ifdef CONFIG_X86_LOCAL_APIC
 		/*
@@ -617,7 +617,7 @@ static void default_do_nmi(struct pt_regs * regs)
 		unknown_nmi_error(reason, regs);
 		return;
 	}
-	if (notify_die(DIE_NMI, "nmi", regs, reason, 0, SIGINT) == NOTIFY_BAD)
+	if (notify_die(DIE_NMI, "nmi", regs, reason, 0, SIGINT) == NOTIFY_STOP)
 		return;
 	if (reason & 0x80)
 		mem_parity_error(reason, regs);
@@ -666,7 +666,7 @@ void unset_nmi_callback(void)
 asmlinkage int do_int3(struct pt_regs *regs, long error_code)
 {
 	if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
-			== NOTIFY_OK)
+			== NOTIFY_STOP)
 		return 1;
 	/* This is an interrupt gate, because kprobes wants interrupts
 	disabled.  Normal trap handlers don't. */
@@ -707,7 +707,7 @@ asmlinkage void do_debug(struct pt_regs * regs, long error_code)
 	__asm__ __volatile__("movl %%db6,%0" : "=r" (condition));
 
 	if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code,
-					SIGTRAP) == NOTIFY_OK)
+					SIGTRAP) == NOTIFY_STOP)
 		return;
 	/* It's safe to allow irq's after DR6 has been saved */
 	if (regs->eflags & X86_EFLAGS_IF)

arch/i386/mm/fault.c

@@ -227,7 +227,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code)
 	__asm__("movl %%cr2,%0":"=r" (address));
 
 	if (notify_die(DIE_PAGE_FAULT, "page fault", regs, error_code, 14,
-					SIGSEGV) == NOTIFY_OK)
+					SIGSEGV) == NOTIFY_STOP)
 		return;
 	/* It's safe to allow irq's after cr2 has been saved */
 	if (regs->eflags & (X86_EFLAGS_IF|VM_MASK))

arch/sparc64/kernel/kprobes.c

@@ -179,26 +179,26 @@ int kprobe_exceptions_notify(struct notifier_block *self, unsigned long val,
 	switch (val) {
 	case DIE_DEBUG:
 		if (kprobe_handler(args->regs))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_DEBUG_2:
 		if (post_kprobe_handler(args->regs))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_GPF:
 		if (kprobe_running() &&
 		    kprobe_fault_handler(args->regs, args->trapnr))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	case DIE_PAGE_FAULT:
 		if (kprobe_running() &&
 		    kprobe_fault_handler(args->regs, args->trapnr))
-			return NOTIFY_OK;
+			return NOTIFY_STOP;
 		break;
 	default:
 		break;
 	}
-	return NOTIFY_BAD;
+	return NOTIFY_DONE;
 }
 
 asmlinkage void kprobe_trap(unsigned long trap_level, struct pt_regs *regs)
@@ -216,7 +216,7 @@ asmlinkage void kprobe_trap(unsigned long trap_level, struct pt_regs *regs)
 	 */
 	if (notify_die((trap_level == 0x170) ? DIE_DEBUG : DIE_DEBUG_2,
 		       (trap_level == 0x170) ? "debug" : "debug_2",
-		       regs, 0, trap_level, SIGTRAP) != NOTIFY_OK)
+		       regs, 0, trap_level, SIGTRAP) != NOTIFY_STOP)
 		bad_trap(regs, trap_level);
 }
 

arch/sparc64/kernel/traps.c

@@ -96,7 +96,7 @@ void bad_trap(struct pt_regs *regs, long lvl)
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "bad trap", regs,
-		       0, lvl, SIGTRAP) == NOTIFY_OK)
+		       0, lvl, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	if (lvl < 0x100) {
@@ -126,7 +126,7 @@ void bad_trap_tl1(struct pt_regs *regs, long lvl)
 	char buffer[32];
 	
 	if (notify_die(DIE_TRAP_TL1, "bad trap tl1", regs,
-		       0, lvl, SIGTRAP) == NOTIFY_OK)
+		       0, lvl, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	dump_tl1_traplog((struct tl1_traplog *)(regs + 1));
@@ -149,7 +149,7 @@ void instruction_access_exception(struct pt_regs *regs,
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "instruction access exception", regs,
-		       0, 0x8, SIGTRAP) == NOTIFY_OK)
+		       0, 0x8, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	if (regs->tstate & TSTATE_PRIV) {
@@ -173,7 +173,7 @@ void instruction_access_exception_tl1(struct pt_regs *regs,
 				      unsigned long sfsr, unsigned long sfar)
 {
 	if (notify_die(DIE_TRAP_TL1, "instruction access exception tl1", regs,
-		       0, 0x8, SIGTRAP) == NOTIFY_OK)
+		       0, 0x8, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	dump_tl1_traplog((struct tl1_traplog *)(regs + 1));
@@ -186,7 +186,7 @@ void data_access_exception(struct pt_regs *regs,
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "data access exception", regs,
-		       0, 0x30, SIGTRAP) == NOTIFY_OK)
+		       0, 0x30, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	if (regs->tstate & TSTATE_PRIV) {
@@ -260,7 +260,7 @@ void do_iae(struct pt_regs *regs)
 	spitfire_clean_and_reenable_l1_caches();
 
 	if (notify_die(DIE_TRAP, "instruction access exception", regs,
-		       0, 0x8, SIGTRAP) == NOTIFY_OK)
+		       0, 0x8, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	info.si_signo = SIGBUS;
@@ -292,7 +292,7 @@ void do_dae(struct pt_regs *regs)
 	spitfire_clean_and_reenable_l1_caches();
 
 	if (notify_die(DIE_TRAP, "data access exception", regs,
-		       0, 0x30, SIGTRAP) == NOTIFY_OK)
+		       0, 0x30, SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	info.si_signo = SIGBUS;
@@ -1695,7 +1695,7 @@ void do_fpe_common(struct pt_regs *regs)
 void do_fpieee(struct pt_regs *regs)
 {
 	if (notify_die(DIE_TRAP, "fpu exception ieee", regs,
-		       0, 0x24, SIGFPE) == NOTIFY_OK)
+		       0, 0x24, SIGFPE) == NOTIFY_STOP)
 		return;
 
 	do_fpe_common(regs);
@@ -1709,7 +1709,7 @@ void do_fpother(struct pt_regs *regs)
 	int ret = 0;
 
 	if (notify_die(DIE_TRAP, "fpu exception other", regs,
-		       0, 0x25, SIGFPE) == NOTIFY_OK)
+		       0, 0x25, SIGFPE) == NOTIFY_STOP)
 		return;
 
 	switch ((current_thread_info()->xfsr[0] & 0x1c000)) {
@@ -1728,7 +1728,7 @@ void do_tof(struct pt_regs *regs)
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "tagged arithmetic overflow", regs,
-		       0, 0x26, SIGEMT) == NOTIFY_OK)
+		       0, 0x26, SIGEMT) == NOTIFY_STOP)
 		return;
 
 	if (regs->tstate & TSTATE_PRIV)
@@ -1750,7 +1750,7 @@ void do_div0(struct pt_regs *regs)
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "integer division by zero", regs,
-		       0, 0x28, SIGFPE) == NOTIFY_OK)
+		       0, 0x28, SIGFPE) == NOTIFY_STOP)
 		return;
 
 	if (regs->tstate & TSTATE_PRIV)
@@ -1936,7 +1936,7 @@ void do_illegal_instruction(struct pt_regs *regs)
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "illegal instruction", regs,
-		       0, 0x10, SIGILL) == NOTIFY_OK)
+		       0, 0x10, SIGILL) == NOTIFY_STOP)
 		return;
 
 	if (tstate & TSTATE_PRIV)
@@ -1965,7 +1965,7 @@ void mem_address_unaligned(struct pt_regs *regs, unsigned long sfar, unsigned lo
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "memory address unaligned", regs,
-		       0, 0x34, SIGSEGV) == NOTIFY_OK)
+		       0, 0x34, SIGSEGV) == NOTIFY_STOP)
 		return;
 
 	if (regs->tstate & TSTATE_PRIV) {
@@ -1991,7 +1991,7 @@ void do_privop(struct pt_regs *regs)
 	siginfo_t info;
 
 	if (notify_die(DIE_TRAP, "privileged operation", regs,
-		       0, 0x11, SIGILL) == NOTIFY_OK)
+		       0, 0x11, SIGILL) == NOTIFY_STOP)
 		return;
 
 	if (test_thread_flag(TIF_32BIT)) {

arch/sparc64/mm/fault.c

@@ -149,7 +149,7 @@ static void unhandled_fault(unsigned long address, struct task_struct *tsk,
 	       (tsk->mm ? (unsigned long) tsk->mm->pgd :
 		          (unsigned long) tsk->active_mm->pgd));
 	if (notify_die(DIE_GPF, "general protection fault", regs,
-		       0, 0, SIGSEGV) == NOTIFY_OK)
+		       0, 0, SIGSEGV) == NOTIFY_STOP)
 		return;
 	die_if_kernel("Oops", regs);
 }
@@ -325,7 +325,7 @@ asmlinkage void do_sparc64_fault(struct pt_regs *regs)
 	fault_code = get_thread_fault_code();
 
 	if (notify_die(DIE_PAGE_FAULT, "page_fault", regs,
-		       fault_code, 0, SIGSEGV) == NOTIFY_OK)
+		       fault_code, 0, SIGSEGV) == NOTIFY_STOP)
 		return;
 
 	si_code = SEGV_MAPERR;

arch/x86_64/kernel/nmi.c

@@ -390,7 +390,8 @@ void nmi_watchdog_tick (struct pt_regs * regs, unsigned reason)
 		 */
 		alert_counter[cpu]++;
 		if (alert_counter[cpu] == 5*nmi_hz) {
-			if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_BAD) { 
+			if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT)
+							== NOTIFY_STOP) {
 				alert_counter[cpu] = 0; 
 				return;
 			} 

arch/x86_64/kernel/traps.c

@@ -437,7 +437,8 @@ static void do_trap(int trapnr, int signr, char *str,
 #define DO_ERROR(trapnr, signr, str, name) \
 asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
 { \
-	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) == NOTIFY_BAD) \
+	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
+							== NOTIFY_STOP) \
 		return; \
 	do_trap(trapnr, signr, str, regs, error_code, NULL); \
 }
@@ -450,7 +451,8 @@ asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
 	info.si_errno = 0; \
 	info.si_code = sicode; \
 	info.si_addr = (void __user *)siaddr; \
-	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) == NOTIFY_BAD) \
+	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
+							== NOTIFY_STOP) \
 		return; \
 	do_trap(trapnr, signr, str, regs, error_code, &info); \
 }
@@ -471,7 +473,8 @@ DO_ERROR(18, SIGSEGV, "reserved", reserved)
 asmlinkage void *do_##name(struct pt_regs * regs, long error_code) \
 { \
 	struct pt_regs *pr = ((struct pt_regs *)(current->thread.rsp0))-1; \
-	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) == NOTIFY_BAD) \
+	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
+							== NOTIFY_STOP) \
 		return regs; \
 	if (regs->cs & 3) { \
 		memcpy(pr, regs, sizeof(struct pt_regs)); \
@@ -565,7 +568,8 @@ asmlinkage void default_do_nmi(struct pt_regs * regs)
 	unsigned char reason = inb(0x61);
 
 	if (!(reason & 0xc0)) {
-		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 0, SIGINT) == NOTIFY_BAD)
+		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 0, SIGINT)
+								== NOTIFY_STOP)
 			return;
 #ifdef CONFIG_X86_LOCAL_APIC
 		/*
@@ -580,7 +584,7 @@ asmlinkage void default_do_nmi(struct pt_regs * regs)
 		unknown_nmi_error(reason, regs);
 		return;
 	}
-	if (notify_die(DIE_NMI, "nmi", regs, reason, 0, SIGINT) == NOTIFY_BAD)
+	if (notify_die(DIE_NMI, "nmi", regs, reason, 0, SIGINT) == NOTIFY_STOP)
 		return; 
 	if (reason & 0x80)
 		mem_parity_error(reason, regs);
@@ -676,7 +680,7 @@ asmlinkage void *do_debug(struct pt_regs * regs, unsigned long error_code)
 clear_TF:
 	/* RED-PEN could cause spurious errors */
 	if (notify_die(DIE_DEBUG, "debug2", regs, condition, 1, SIGTRAP) 
-	    != NOTIFY_BAD)
+								!= NOTIFY_STOP)
 	regs->eflags &= ~TF_MASK;
 	return regs;	
 }

include/linux/notifier.h

@@ -29,6 +29,10 @@ extern int notifier_call_chain(struct notifier_block **n, unsigned long val, voi
 #define NOTIFY_OK		0x0001		/* Suits me */
 #define NOTIFY_STOP_MASK	0x8000		/* Don't call further */
 #define NOTIFY_BAD		(NOTIFY_STOP_MASK|0x0002)	/* Bad/Veto action	*/
+/*
+ * Clean way to return from the notifier and stop further calls.
+ */
+#define NOTIFY_STOP		(NOTIFY_OK|NOTIFY_STOP_MASK)
 
 /*
  *	Declared notifiers so far. I can imagine quite a few more chains

kernel/kprobes.c

@@ -25,6 +25,8 @@
  *		hlists and exceptions notifier as suggested by Andi Kleen.
  * 2004-July	Suparna Bhattacharya <suparna@in.ibm.com> added jumper probes
  *		interface to access function arguments.
+ * 2004-Sep	Prasanna S Panchamukhi <prasanna@in.ibm.com> Changed Kprobes
+ *		exceptions notifier to be first on the priority list.
  */
 #include <linux/kprobes.h>
 #include <linux/spinlock.h>
@@ -108,6 +110,7 @@ void unregister_kprobe(struct kprobe *p)
 
 static struct notifier_block kprobe_exceptions_nb = {
 	.notifier_call = kprobe_exceptions_notify,
+	.priority = 0x7fffffff /* we need to notified first */
 };
 
 int register_jprobe(struct jprobe *jp)