[PATCH] kprobes: fix bug when probed on task and isr functions

This patch fixes a race condition where in system used to hang or sometime
crash within minutes when kprobes are inserted on ISR routine and a task
routine.

The fix has been stress tested on i386, ia64, pp64 and on x86_64.  To
reproduce the problem insert kprobes on schedule() and do_IRQ() functions
and you should see hang or system crash.
Signed-off-by: NAnil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Signed-off-by: NAnanth N Mavinakayanahalli <ananth@in.ibm.com>
Acked-by: NPrasanna S Panchamukhi <prasanna@in.ibm.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

arch/i386/kernel/kprobes.c

@@ -177,7 +177,8 @@ static int __kprobes kprobe_handler(struct pt_regs *regs)
 		   Disarm the probe we just hit, and ignore it. */
 		p = get_kprobe(addr);
 		if (p) {
-			if (kprobe_status == KPROBE_HIT_SS) {
+			if (kprobe_status == KPROBE_HIT_SS &&
+				*p->ainsn.insn == BREAKPOINT_INSTRUCTION) {
 				regs->eflags &= ~TF_MASK;
 				regs->eflags |= kprobe_saved_eflags;
 				unlock_kprobes();

arch/ia64/kernel/kprobes.c

@@ -95,6 +95,17 @@ static void __kprobes update_kprobe_inst_flag(uint template, uint  slot,
 	p->ainsn.inst_flag = 0;
 	p->ainsn.target_br_reg = 0;
 
+	/* Check for Break instruction
+ 	 * Bits 37:40 Major opcode to be zero
+	 * Bits 27:32 X6 to be zero
+	 * Bits 32:35 X3 to be zero
+	 */
+	if ((!major_opcode) && (!((kprobe_inst >> 27) & 0x1FF)) ) {
+		/* is a break instruction */
+	 	p->ainsn.inst_flag |= INST_FLAG_BREAK_INST;
+		return;
+	}
+
 	if (bundle_encoding[template][slot] == B) {
 		switch (major_opcode) {
 		  case INDIRECT_CALL_OPCODE:
@@ -542,8 +553,11 @@ static void __kprobes prepare_ss(struct kprobe *p, struct pt_regs *regs)
 	unsigned long bundle_addr = (unsigned long) &p->opcode.bundle;
 	unsigned long slot = (unsigned long)p->addr & 0xf;
 
-	/* Update instruction pointer (IIP) and slot number (IPSR.ri) */
-	regs->cr_iip = bundle_addr & ~0xFULL;
+	/* single step inline if break instruction */
+	if (p->ainsn.inst_flag == INST_FLAG_BREAK_INST)
+		regs->cr_iip = (unsigned long)p->addr & ~0xFULL;
+	else
+		regs->cr_iip = bundle_addr & ~0xFULL;
 
 	if (slot > 2)
 		slot = 0;
@@ -599,7 +613,9 @@ static int __kprobes pre_kprobes_handler(struct die_args *args)
 	if (kprobe_running()) {
 		p = get_kprobe(addr);
 		if (p) {
-			if (kprobe_status == KPROBE_HIT_SS) {
+			if ( (kprobe_status == KPROBE_HIT_SS) &&
+	 		     (p->ainsn.inst_flag == INST_FLAG_BREAK_INST)) {
+  				ia64_psr(regs)->ss = 0;
 				unlock_kprobes();
 				goto no_kprobe;
 			}

arch/ppc64/kernel/kprobes.c

@@ -102,7 +102,7 @@ static inline void prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
 	regs->msr |= MSR_SE;
 
 	/* single step inline if it is a trap variant */
-	if (IS_TW(insn) || IS_TD(insn) || IS_TWI(insn) || IS_TDI(insn))
+	if (is_trap(insn))
 		regs->nip = (unsigned long)p->addr;
 	else
 		regs->nip = (unsigned long)p->ainsn.insn;
@@ -152,7 +152,9 @@ static inline int kprobe_handler(struct pt_regs *regs)
 		   Disarm the probe we just hit, and ignore it. */
 		p = get_kprobe(addr);
 		if (p) {
-			if (kprobe_status == KPROBE_HIT_SS) {
+			kprobe_opcode_t insn = *p->ainsn.insn;
+			if (kprobe_status == KPROBE_HIT_SS &&
+					is_trap(insn)) {
 				regs->msr &= ~MSR_SE;
 				regs->msr |= kprobe_saved_msr;
 				unlock_kprobes();
@@ -192,8 +194,7 @@ static inline int kprobe_handler(struct pt_regs *regs)
 			 * trap variant, it could belong to someone else
 			 */
 			kprobe_opcode_t cur_insn = *addr;
-			if (IS_TW(cur_insn) || IS_TD(cur_insn) ||
-					IS_TWI(cur_insn) || IS_TDI(cur_insn))
+			if (is_trap(cur_insn))
 		       		goto no_kprobe;
 			/*
 			 * The breakpoint instruction was removed right
@@ -403,7 +404,7 @@ int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
 	default:
 		break;
 	}
-	preempt_enable();
+	preempt_enable_no_resched();
 	return ret;
 }
 

arch/x86_64/kernel/kprobes.c

@@ -311,7 +311,8 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
 		   Disarm the probe we just hit, and ignore it. */
 		p = get_kprobe(addr);
 		if (p) {
-			if (kprobe_status == KPROBE_HIT_SS) {
+			if (kprobe_status == KPROBE_HIT_SS &&
+				*p->ainsn.insn == BREAKPOINT_INSTRUCTION) {
 				regs->eflags &= ~TF_MASK;
 				regs->eflags |= kprobe_saved_rflags;
 				unlock_kprobes();

include/asm-ia64/kprobes.h

@@ -92,6 +92,7 @@ struct arch_specific_insn {
 	kprobe_opcode_t insn;
  #define INST_FLAG_FIX_RELATIVE_IP_ADDR		1
  #define INST_FLAG_FIX_BRANCH_REG		2
+ #define INST_FLAG_BREAK_INST			4
  	unsigned long inst_flag;
  	unsigned short target_br_reg;
 };

include/asm-ppc64/kprobes.h

@@ -42,6 +42,9 @@ typedef unsigned int kprobe_opcode_t;
 
 #define JPROBE_ENTRY(pentry)	(kprobe_opcode_t *)((func_descr_t *)pentry)
 
+#define is_trap(instr)	(IS_TW(instr) || IS_TD(instr) || \
+			IS_TWI(instr) || IS_TDI(instr))
+
 #define ARCH_SUPPORTS_KRETPROBES
 void kretprobe_trampoline(void);
 

kernel/kprobes.c

@@ -155,14 +155,36 @@ void __kprobes free_insn_slot(kprobe_opcode_t *slot)
 /* Locks kprobe: irqs must be disabled */
 void __kprobes lock_kprobes(void)
 {
+	unsigned long flags = 0;
+
+	/* Avoiding local interrupts to happen right after we take the kprobe_lock
+	 * and before we get a chance to update kprobe_cpu, this to prevent
+	 * deadlock when we have a kprobe on ISR routine and a kprobe on task
+	 * routine
+	 */
+	local_irq_save(flags);
+
 	spin_lock(&kprobe_lock);
 	kprobe_cpu = smp_processor_id();
+
+ 	local_irq_restore(flags);
 }
 
 void __kprobes unlock_kprobes(void)
 {
+	unsigned long flags = 0;
+
+	/* Avoiding local interrupts to happen right after we update
+	 * kprobe_cpu and before we get a a chance to release kprobe_lock,
+	 * this to prevent deadlock when we have a kprobe on ISR routine and
+	 * a kprobe on task routine
+	 */
+	local_irq_save(flags);
+
 	kprobe_cpu = NR_CPUS;
 	spin_unlock(&kprobe_lock);
+
+ 	local_irq_restore(flags);
 }
 
 /* You have to be holding the kprobe_lock */