Skip to contentSkip to conversation

fork: fail on non-zero higher 32 bits of args.exit_signal

PATCH In review Eugene Syromiatnikov 4 participants 10 messages Sep 10, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fork: fail on non-zero higher 32 bits of args.exit_signal

Previously, higher 32 bits of exit_signal fields were lost when copied to the kernel args structure (that uses int as a type for the respective field). Fail with EINVAL if these are set as it looks like there's no sane reason to accept them.

* kernel/fork.c (copy_clone_args_from_user): Fail with -EINVAL if args.exit_signal converted to unsigned int is not equal to the original value.

---
 kernel/fork.c | 3 +++
 1 file changed, 3 insertions(+)
kernel/fork.c +3 -0
index 2852d0e..fcbc4d5 100644
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;
+
 	*kargs = (struct kernel_clone_args){
 		.flags		= args.flags,
 		.pidfd		= u64_to_user_ptr(args.pidfd),
Signature · 1 line
2.1.4
Reviewed-by on v1 1/1
#1 lore raw
Quoted 10 lines · Eugene Syromiatnikov

Previously, higher 32 bits of exit_signal fields were lost when copied to the kernel args structure (that uses int as a type for the respective field). Fail with EINVAL if these are set as it looks like there's no sane reason to accept them.

* kernel/fork.c (copy_clone_args_from_user): Fail with -EINVAL if args.exit_signal converted to unsigned int is not equal to the original value.

Signature · 1 line
ldv
#2 lore raw
In reply to Eugene Syromiatnikov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Oleg.

#3 lore raw
Quoted 15 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

Quoted 8 lines
	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

Eugene, can you please update the patch to use valid signal and keep it as a separate patch from the cleanup and selftest patches?

Christian

#4 lore raw
Quoted 25 lines · Christian Brauner
Quoted 15 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

Quoted 8 lines
	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Quoted 6 lines

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

Eugene, can you please update the patch to use valid signal and keep it as a separate patch from the cleanup and selftest patches?

I'll then pick this up quickly so we can get this in before 5.3 is out.

Thanks!
Christian

#5 lore raw
Quoted 23 lines · Christian Brauner
Quoted 22 lines · Christian Brauner
Quoted 21 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

Quoted 6 lines
Quoted 6 lines
Quoted 2 lines

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

So we could do your check in copy_clone_args_from_user(), and then we do another valid_signal() check in clone3_args_valid()? We could do the latter in copy_clone_args_from_user() too but it's nicer to do it along the other checks in clone3_args_valid().

Christian

#6 lore raw
Quoted 27 lines · Christian Brauner
Quoted 26 lines · Christian Brauner
Quoted 25 lines · Christian Brauner
Quoted 23 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

So we could do your check in copy_clone_args_from_user(), and then we do another valid_signal() check in clone3_args_valid()? We could do the latter in copy_clone_args_from_user() too but it's nicer to do it along the other checks in clone3_args_valid().

There's also a discrepancy between CSIGNAL (0xff) and _NSIG, used in valid_signal (which is between 32 and 128, depending on architecture), it seems it doesn't break thread_group_leader, but definitely allows passing some invalid signal numbers via legacy clone-like syscalls—I'm not sure if that's important.

Quoted 1 line

Christian

#7 lore raw
In reply to Christian Brauner
Quoted 27 lines · Christian Brauner
Quoted 26 lines · Christian Brauner
Quoted 25 lines · Christian Brauner
Quoted 23 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

So we could do your check in copy_clone_args_from_user(), and then we do another valid_signal() check in clone3_args_valid()? We could do the latter in copy_clone_args_from_user() too but it's nicer to do it along the other checks in clone3_args_valid().

I am fine either way. Sure, we can add valid_signal() into clone3_args_valid(), but then I'd ask to simplify the "overflow" check above. Something like

	if (args.exit_signal > UINT_MAX)
		return -EINVAL;

looks much more readable to me.

Or we can simply do

	if (args.exit_signal & ~((u64)CSIGNAL))
		return -EINVAL;

in copy_clone_args_from_user() and forget about all problems.

Up to Eugene and you.

Oleg.

#8 lore raw
Quoted 46 lines · Oleg Nesterov
Quoted 27 lines · Christian Brauner
Quoted 26 lines · Christian Brauner
Quoted 25 lines · Christian Brauner
Quoted 23 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

So we could do your check in copy_clone_args_from_user(), and then we do another valid_signal() check in clone3_args_valid()? We could do the latter in copy_clone_args_from_user() too but it's nicer to do it along the other checks in clone3_args_valid().

I am fine either way. Sure, we can add valid_signal() into clone3_args_valid(), but then I'd ask to simplify the "overflow" check above. Something like

	if (args.exit_signal > UINT_MAX)
		return -EINVAL;

looks much more readable to me.

Quoted 6 lines

Or we can simply do

	if (args.exit_signal & ~((u64)CSIGNAL))
		return -EINVAL;

in copy_clone_args_from_user() and forget about all problems.

Both are fine with me. The latter might have the advantage that we catch both legacy clone and clone3. I think Eugene prefers this as well.

Christian

#9 lore raw
Quoted 49 lines · Christian Brauner
Quoted 48 lines · Oleg Nesterov
Quoted 27 lines · Christian Brauner
Quoted 26 lines · Christian Brauner
Quoted 25 lines · Christian Brauner
Quoted 23 lines · Oleg Nesterov

kernel/fork.c noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, · line 2566

Quoted 9 lines · Eugene Syromiatnikov
kernel/fork.c +2 -0
@@ -2562,6 +2562,9 @@ noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
 	if (copy_from_user(&args, uargs, size))
 		return -EFAULT;
 
+	if (unlikely(((unsigned int)args.exit_signal) != args.exit_signal))
+		return -EINVAL;

Hmm. Unless I am totally confused you found a serious bug...

Without CLONE_THREAD/CLONE_PARENT copy_process() blindly does

	p->exit_signal = args->exit_signal;

the valid_signal(sig) check in do_notify_parent() mostly saves us, but we must not allow child->exit_signal < 0, if nothing else this breaks thread_group_leader().

And afaics this patch doesn't fix this? I think we need the valid_signal() check...

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

Thanks for sending this patch so quickly after our conversation yesterday, Eugene!
We definitely want valid_signal() to verify the signal is ok.

So we could do your check in copy_clone_args_from_user(), and then we do another valid_signal() check in clone3_args_valid()? We could do the latter in copy_clone_args_from_user() too but it's nicer to do it along the other checks in clone3_args_valid().

I am fine either way. Sure, we can add valid_signal() into clone3_args_valid(), but then I'd ask to simplify the "overflow" check above. Something like

	if (args.exit_signal > UINT_MAX)
		return -EINVAL;

looks much more readable to me.

Or we can simply do

Quoted 7 lines
Quoted 4 lines
	if (args.exit_signal & ~((u64)CSIGNAL))
		return -EINVAL;

in copy_clone_args_from_user() and forget about all problems.

Both are fine with me. The latter might have the advantage that we catch both legacy clone and clone3. I think Eugene prefers this as well.

Unfortunately, it doesn't. I think, the best place for the check is either in _do_fork or copy_process itself; however, it's quite messy as that way it's detached from the other checks, but, at the same time, there are a lot of code paths (like the one in arch/x86/ia32/sys_ia32.c), and it's kinda obscure that the caller of _do_fork has to check that exit_syscall is positive itself.

Quoted 1 line

Christian