Skip to contentSkip to conversation

kretprobe spinlock recursive remove

PATCH In review bibo,mao 2 participants 3 messages Mar 21, 2006

lore
Share & tools
lore

bibo,mao

author
#0 lore raw

PATCH 1/1 kretprobe spinlock recursive remove

In recent linux kernel version, kretprobe in IA32 is implemented in kretprobe_trampoline. And break trap code is removed from retprobe_trampoline, instead trampoline_handler is called directly. Currently if kretprobe hander hit one trap which causes another kretprobe, there will be SPINLOCK recursive bug. This patch fixes this, and will skip trap during kretprobe handler execution. This patch is based on 2.6.16-rc6-mm2.

i386/kernel/kprobes.c renamed from i386/kernel/kprobes.c.bak +4 -1
@@ -390,8 +390,11 @@ fastcall void *__kprobes trampoline_hand
 			/* another task is sharing our hash bucket */
                         continue;
 
-		if (ri->rp && ri->rp->handler)
+		if (ri->rp && ri->rp->handler){
+			__get_cpu_var(current_kprobe) = &ri->rp->kp;
 			ri->rp->handler(ri, regs);
+			__get_cpu_var(current_kprobe) = NULL;
+		}
 
 		orig_ret_address = (unsigned long)ri->ret_addr;
 		recycle_rp_inst(ri);
#1 lore raw
Quoted 8 lines · bibo,mao

In recent linux kernel version, kretprobe in IA32 is implemented in kretprobe_trampoline. And break trap code is removed from retprobe_trampoline, instead trampoline_handler is called directly. Currently if kretprobe hander hit one trap which causes another kretprobe, there will be SPINLOCK recursive bug. This patch fixes this, and will skip trap during kretprobe handler execution. This patch is based on 2.6.16-rc6-mm2.

What is "recent linux kernel"? 2.6.16? This patch does apply to 2.6.16 so I assume that's what you were referring to?

If you're referring to the kretprobes patches in -mm then which one introduced the problem?

If you're referring to 2.6.16 then is this problem sufficiently serious to warrant inclusion of this patch in 2.6.16.1?

Please remember to include Signed-off-by: tags.

i386/kernel/kprobes.c

Quoted 2 lines
i386/kernel/kprobes.c renamed from i386/kernel/kprobes.c.bak +0 -0

Please prepare patches in `patch -p1' form.

i386/kernel/kprobes.c fastcall void *__kprobes trampoline_hand · line 400

Quoted 13 lines
i386/kernel/kprobes.c renamed from i386/kernel/kprobes.c.bak +4 -1
@@ -390,8 +390,11 @@ fastcall void *__kprobes trampoline_hand
  			/* another task is sharing our hash bucket */
                          continue;
 
-		if (ri->rp && ri->rp->handler)
+		if (ri->rp && ri->rp->handler){
+			__get_cpu_var(current_kprobe) = &ri->rp->kp;
  			ri->rp->handler(ri, regs);
+			__get_cpu_var(current_kprobe) = NULL;
+		}
 
  		orig_ret_address = (unsigned long)ri->ret_addr;
  		recycle_rp_inst(ri);

Your email client appears to be altering the patches in some manner. It required `patch -l' to make this apply.

author
#2 lore raw
Quoted 8 lines · Andrew Morton
Quoted 1 line · bibo,mao

In recent linux kernel version, kretprobe in IA32 is implemented in kretprobe_trampoline. And break trap code is removed from retprobe_trampoline, instead trampoline_handler is called directly. Currently if kretprobe hander hit one trap which causes another kretprobe, there will be SPINLOCK recursive bug. This patch fixes this, and will skip trap during kretprobe handler execution. This patch is based on 2.6.16-rc6-mm2.

What is "recent linux kernel"? 2.6.16? This patch does apply to 2.6.16 so I assume that's what you were referring to?

If you're referring to the kretprobes patches in -mm then which one introduced the problem?

Sorry, I did not clarify it out, it is againt kretprobe-kretprobe-booster.patch in -mm tree.

Quoted 34 lines

If you're referring to 2.6.16 then is this problem sufficiently serious to warrant inclusion of this patch in 2.6.16.1?

Please remember to include Signed-off-by: tags.

i386/kernel/kprobes.c

Quoted 2 lines
i386/kernel/kprobes.c renamed from i386/kernel/kprobes.c.bak +0 -0

Please prepare patches in `patch -p1' form.

i386/kernel/kprobes.c fastcall void *__kprobes trampoline_hand · line 400

Quoted 13 lines
i386/kernel/kprobes.c renamed from i386/kernel/kprobes.c.bak +4 -1
@@ -390,8 +390,11 @@ fastcall void *__kprobes trampoline_hand
  			/* another task is sharing our hash bucket */
                          continue;
 
-		if (ri->rp && ri->rp->handler)
+		if (ri->rp && ri->rp->handler){
+			__get_cpu_var(current_kprobe) = &ri->rp->kp;
  			ri->rp->handler(ri, regs);
+			__get_cpu_var(current_kprobe) = NULL;
+		}
 
  		orig_ret_address = (unsigned long)ri->ret_addr;
  		recycle_rp_inst(ri);

Your email client appears to be altering the patches in some manner. It required `patch -l' to make this apply.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/