Skip to contentSkip to conversation

Fix: compat_rw_copy_check_uvector() misuse in aio, readv, writev, and security keys

PATCH In review Mathieu Desnoyers 3 participants 6 messages Feb 25, 2013 – Mar 13, 2013

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix: compat_rw_copy_check_uvector() misuse in aio, readv, writev, and security keys

Looking at mm/process_vm_access.c:process_vm_rw() and comparing it to compat_process_vm_rw() shows that the compatibility code requires an explicit "access_ok()" check before calling compat_rw_copy_check_uvector(). The same difference seems to appear when we compare fs/read_write.c:do_readv_writev() to fs/compat.c:compat_do_readv_writev().

This subtle difference between the compat and non-compat requirements should probably be debated, as it seems to be error-prone. In fact, there are two others sites that use this function in the Linux kernel, and they both seem to get it wrong:

Now shifting our attention to fs/aio.c, we see that aio_setup_iocb() also ends up calling compat_rw_copy_check_uvector() through aio_setup_vectored_rw(). Unfortunately, the access_ok() check appears to be missing. Same situation for security/keys/compat.c:compat_keyctl_instantiate_key_iov().

I propose that we add the access_ok() check directly into compat_rw_copy_check_uvector(), so callers don't have to worry about it, and it therefore makes the compat call code similar to its non-compat counterpart. Place the access_ok() check in the same location where copy_from_user() can trigger a -EFAULT error in the non-compat code, so the ABI behaviors are alike on both compat and non-compat.

While we are here, fix compat_do_readv_writev() so it checks for compat_rw_copy_check_uvector() negative return values.

And also, fix a memory leak in compat_keyctl_instantiate_key_iov() error handling.

---
 fs/compat.c            |   15 +++++++--------
 mm/process_vm_access.c |    8 --------
 security/keys/compat.c |    4 ++--
 3 files changed, 9 insertions(+), 18 deletions(-)
fs/compat.c +7 -8
Index: linux/fs/compat.c
===================================================================
@@ -558,6 +558,10 @@ ssize_t compat_rw_copy_check_uvector(int
 	}
 	*ret_pointer = iov;
 
+	ret = -EFAULT;
+	if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
+		goto out;
+
 	/*
 	 * Single unix specification:
 	 * We should -EINVAL if an element length is not >= 0 and fitting an
@@ -1080,17 +1084,12 @@ static ssize_t compat_do_readv_writev(in
 	if (!file->f_op)
 		goto out;
 
-	ret = -EFAULT;
-	if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
-		goto out;
-
-	tot_len = compat_rw_copy_check_uvector(type, uvector, nr_segs,
+	ret = compat_rw_copy_check_uvector(type, uvector, nr_segs,
 					       UIO_FASTIOV, iovstack, &iov);
-	if (tot_len == 0) {
-		ret = 0;
+	if (ret <= 0)
 		goto out;
-	}
 
+	tot_len = ret;
 	ret = rw_verify_area(type, file, pos, tot_len);
 	if (ret < 0)
 		goto out;
mm/process_vm_access.c +0 -8
Index: linux/mm/process_vm_access.c
===================================================================
@@ -429,12 +429,6 @@ compat_process_vm_rw(compat_pid_t pid,
 	if (flags != 0)
 		return -EINVAL;
 
-	if (!access_ok(VERIFY_READ, lvec, liovcnt * sizeof(*lvec)))
-		goto out;
-
-	if (!access_ok(VERIFY_READ, rvec, riovcnt * sizeof(*rvec)))
-		goto out;
-
 	if (vm_write)
 		rc = compat_rw_copy_check_uvector(WRITE, lvec, liovcnt,
 						  UIO_FASTIOV, iovstack_l,
@@ -459,8 +453,6 @@ free_iovecs:
 		kfree(iov_r);
 	if (iov_l != iovstack_l)
 		kfree(iov_l);
-
-out:
 	return rc;
 }
 
security/keys/compat.c +2 -2
Index: linux/security/keys/compat.c
===================================================================
@@ -40,12 +40,12 @@ static long compat_keyctl_instantiate_ke
 					   ARRAY_SIZE(iovstack),
 					   iovstack, &iov);
 	if (ret < 0)
-		return ret;
+		goto err;
 	if (ret == 0)
 		goto no_payload_free;
 
 	ret = keyctl_instantiate_key_common(id, iov, ioc, ret, ringid);
-
+err:
 	if (iov != iovstack)
 		kfree(iov);
 	return ret;
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
#1 lore raw
Quoted 31 lines · Mathieu Desnoyers

Looking at mm/process_vm_access.c:process_vm_rw() and comparing it to compat_process_vm_rw() shows that the compatibility code requires an explicit "access_ok()" check before calling compat_rw_copy_check_uvector(). The same difference seems to appear when we compare fs/read_write.c:do_readv_writev() to fs/compat.c:compat_do_readv_writev().

This subtle difference between the compat and non-compat requirements should probably be debated, as it seems to be error-prone. In fact, there are two others sites that use this function in the Linux kernel, and they both seem to get it wrong:

Now shifting our attention to fs/aio.c, we see that aio_setup_iocb() also ends up calling compat_rw_copy_check_uvector() through aio_setup_vectored_rw(). Unfortunately, the access_ok() check appears to be missing. Same situation for security/keys/compat.c:compat_keyctl_instantiate_key_iov().

I propose that we add the access_ok() check directly into compat_rw_copy_check_uvector(), so callers don't have to worry about it, and it therefore makes the compat call code similar to its non-compat counterpart. Place the access_ok() check in the same location where copy_from_user() can trigger a -EFAULT error in the non-compat code, so the ABI behaviors are alike on both compat and non-compat.

While we are here, fix compat_do_readv_writev() so it checks for compat_rw_copy_check_uvector() negative return values.

And also, fix a memory leak in compat_keyctl_instantiate_key_iov() error handling.

Quoted 8 lines
---
 fs/compat.c            |   15 +++++++--------
 mm/process_vm_access.c |    8 --------
 security/keys/compat.c |    4 ++--
 3 files changed, 9 insertions(+), 18 deletions(-)

What ever happened to this patch? I don't see it in Linus's tree, was it not correct?

thanks,

greg k-h

#2 lore raw
Quoted 41 lines · * Greg Kroah-Hartman ([email protected]) wrote:
Quoted 31 lines · Mathieu Desnoyers

Looking at mm/process_vm_access.c:process_vm_rw() and comparing it to compat_process_vm_rw() shows that the compatibility code requires an explicit "access_ok()" check before calling compat_rw_copy_check_uvector(). The same difference seems to appear when we compare fs/read_write.c:do_readv_writev() to fs/compat.c:compat_do_readv_writev().

This subtle difference between the compat and non-compat requirements should probably be debated, as it seems to be error-prone. In fact, there are two others sites that use this function in the Linux kernel, and they both seem to get it wrong:

Now shifting our attention to fs/aio.c, we see that aio_setup_iocb() also ends up calling compat_rw_copy_check_uvector() through aio_setup_vectored_rw(). Unfortunately, the access_ok() check appears to be missing. Same situation for security/keys/compat.c:compat_keyctl_instantiate_key_iov().

I propose that we add the access_ok() check directly into compat_rw_copy_check_uvector(), so callers don't have to worry about it, and it therefore makes the compat call code similar to its non-compat counterpart. Place the access_ok() check in the same location where copy_from_user() can trigger a -EFAULT error in the non-compat code, so the ABI behaviors are alike on both compat and non-compat.

While we are here, fix compat_do_readv_writev() so it checks for compat_rw_copy_check_uvector() negative return values.

And also, fix a memory leak in compat_keyctl_instantiate_key_iov() error handling.

Quoted 8 lines
---
 fs/compat.c            |   15 +++++++--------
 mm/process_vm_access.c |    8 --------
 security/keys/compat.c |    4 ++--
 3 files changed, 9 insertions(+), 18 deletions(-)
Quoted 2 lines

What ever happened to this patch? I don't see it in Linus's tree, was it not correct?

Not sure. My guess would be that there is missing proof that I actually tested the patch by generating a OOPS splat and proof that it actually fixes it. I mainly tested that it does not break _correct_ use of aio, but I did not have enough time to created my own custom aio lib to generate the issue. So if this is what is expected, I might need a bit of help on this front. Testing odd behavior through libaio proved to non-straightforward.

Thoughts ?

Thanks,

Mathieu

Quoted 4 lines

thanks,

greg k-h

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
#3 lore raw
Quoted 3 lines · Mathieu Desnoyers

Not sure. My guess would be that there is missing proof that I actually tested the patch

Well, it was actually mostly that the early patches I saw had very tentative messages which made me go "Mathieu isn't sure" and kind of waited for more forceful messages and acks from people. It wasn't helped by the fact that Al was basically offline for a while.

That said, the patch definitely looks like the right thing to do. I'll apply it.

Linus

#4 lore raw
Quoted 2 lines · Linus Torvalds

That said, the patch definitely looks like the right thing to do. I'll apply it.

Hmm. I applied it, and then pretty much immediately realized what was problematic about it. What about the fs/bio.c one? This all felt like it was still a work-in-progress, and I'm not sure if you had more comments or patches coming along?

Anyway, this particular patch got applied. Does that obviate the need for the fs/bio.c one? I didn't follow all the call-chains...

Linus

#5 lore raw
Quoted 5 lines · * Linus Torvalds ([email protected]) wrote:
Quoted 2 lines · Linus Torvalds

That said, the patch definitely looks like the right thing to do. I'll apply it.

Quoted 7 lines

Hmm. I applied it, and then pretty much immediately realized what was problematic about it. What about the fs/bio.c one? This all felt like it was still a work-in-progress, and I'm not sure if you had more comments or patches coming along?

Anyway, this particular patch got applied. Does that obviate the need for the fs/bio.c one? I didn't follow all the call-chains...

The fs/bio.c issue is unrelated to this patch, so that separate issue still stands. I did not get any feedback on that private RFC though.

Should I repost it CCing lkml ?

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com