diff mbox series

fs: remove accidental overflow during wraparound check

Message ID 20240507-b4-sio-vfs_fallocate-v1-1-322f84b97ad5@google.com (mailing list archive)
State Superseded
Headers show
Series fs: remove accidental overflow during wraparound check | expand

Commit Message

Justin Stitt May 7, 2024, 11:17 p.m. UTC 
Running syzkaller with the newly enabled signed integer overflow
sanitizer produces this report:

[  195.401651] ------------[ cut here ]------------
[  195.404808] UBSAN: signed-integer-overflow in ../fs/open.c:321:15
[  195.408739] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long')
[  195.414683] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
[  195.420138] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[  195.425804] Call Trace:
[  195.427360]  <TASK>
[  195.428791]  dump_stack_lvl+0x93/0xd0
[  195.431150]  handle_overflow+0x171/0x1b0
[  195.433640]  vfs_fallocate+0x459/0x4f0
...
[  195.490053] ------------[ cut here ]------------
[  195.493146] UBSAN: signed-integer-overflow in ../fs/open.c:321:61
[  195.497030] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long)
[  195.502940] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
[  195.508395] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[  195.514075] Call Trace:
[  195.515636]  <TASK>
[  195.517000]  dump_stack_lvl+0x93/0xd0
[  195.519255]  handle_overflow+0x171/0x1b0
[  195.521677]  vfs_fallocate+0x4cb/0x4f0
[  195.524033]  __x64_sys_fallocate+0xb2/0xf0

Historically, the signed integer overflow sanitizer did not work in the
kernel due to its interaction with `-fwrapv` but this has since been
changed [1] in the newest version of Clang. It was re-enabled in the
kernel with Commit 557f8c582a9ba8ab ("ubsan: Reintroduce signed overflow
sanitizer").

Let's use the check_add_overflow helper to first verify the addition
stays within the bounds of its type (long long); then we can use that
sum for the following check.

Link: https://github.com/llvm/llvm-project/pull/82432 [1]
Closes: https://github.com/KSPP/linux/issues/356
Cc: linux-hardening@vger.kernel.org
Signed-off-by: Justin Stitt <justinstitt@google.com>
---
I wonder, though, why isn't loff_t an unsigned type? We have plently of
checks to ensure they are positive:

	if (offset < 0 || len <= 0)
		return -EINVAL;
	...
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))

... are there ABI concerns?

Here's the syzkaller reproducer:
r0 = openat(0xffffffffffffff9c, &(0x7f0000000040)='./file1\x00', 0x42, 0x0)
fallocate(r0, 0x10, 0x7fffffffffffffff, 0x2000807fffff7)

... which was used against Kees' tree here (v6.8rc2):
https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/log/?h=wip/v6.9-rc2/unsigned-overflow-sanitizer

... with this config:
https://gist.github.com/JustinStitt/824976568b0f228ccbcbe49f3dee9bf4
---
 fs/open.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)


---
base-commit: 0106679839f7c69632b3b9833c3268c316c0a9fc
change-id: 20240507-b4-sio-vfs_fallocate-7b5223ba3a81

Best regards,
--
Justin Stitt <justinstitt@google.com>

Comments

Al Viro May 8, 2024, 12:10 a.m. UTC | #1 
On Tue, May 07, 2024 at 11:17:57PM +0000, Justin Stitt wrote:

> I wonder, though, why isn't loff_t an unsigned type?

Consider
	lseek(fd, -10, SEEK_CUR)

PS: the above is *not* an endorsement of the proposed patch or
KASAN overflow nonsense in general.
Kees Cook May 8, 2024, 12:11 a.m. UTC | #2 
On Tue, May 07, 2024 at 11:17:57PM +0000, Justin Stitt wrote:
> Running syzkaller with the newly enabled signed integer overflow
> sanitizer produces this report:
> 
> [  195.401651] ------------[ cut here ]------------
> [  195.404808] UBSAN: signed-integer-overflow in ../fs/open.c:321:15
> [  195.408739] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long')
> [  195.414683] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
> [  195.420138] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [  195.425804] Call Trace:
> [  195.427360]  <TASK>
> [  195.428791]  dump_stack_lvl+0x93/0xd0
> [  195.431150]  handle_overflow+0x171/0x1b0
> [  195.433640]  vfs_fallocate+0x459/0x4f0
> ...
> [  195.490053] ------------[ cut here ]------------
> [  195.493146] UBSAN: signed-integer-overflow in ../fs/open.c:321:61
> [  195.497030] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long)
> [  195.502940] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
> [  195.508395] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [  195.514075] Call Trace:
> [  195.515636]  <TASK>
> [  195.517000]  dump_stack_lvl+0x93/0xd0
> [  195.519255]  handle_overflow+0x171/0x1b0
> [  195.521677]  vfs_fallocate+0x4cb/0x4f0
> [  195.524033]  __x64_sys_fallocate+0xb2/0xf0
> 
> Historically, the signed integer overflow sanitizer did not work in the
> kernel due to its interaction with `-fwrapv` but this has since been
> changed [1] in the newest version of Clang. It was re-enabled in the
> kernel with Commit 557f8c582a9ba8ab ("ubsan: Reintroduce signed overflow
> sanitizer").
> 
> Let's use the check_add_overflow helper to first verify the addition
> stays within the bounds of its type (long long); then we can use that
> sum for the following check.
> 
> Link: https://github.com/llvm/llvm-project/pull/82432 [1]
> Closes: https://github.com/KSPP/linux/issues/356
> Cc: linux-hardening@vger.kernel.org
> Signed-off-by: Justin Stitt <justinstitt@google.com>

I think this makes the checking more reading too. Thanks

Reviewed-by: Kees Cook <keescook@chromium.org>
Jan Kara May 9, 2024, 3:53 p.m. UTC | #3 
On Tue 07-05-24 23:17:57, Justin Stitt wrote:
> Running syzkaller with the newly enabled signed integer overflow
> sanitizer produces this report:
> 
> [  195.401651] ------------[ cut here ]------------
> [  195.404808] UBSAN: signed-integer-overflow in ../fs/open.c:321:15
> [  195.408739] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long')
> [  195.414683] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
> [  195.420138] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [  195.425804] Call Trace:
> [  195.427360]  <TASK>
> [  195.428791]  dump_stack_lvl+0x93/0xd0
> [  195.431150]  handle_overflow+0x171/0x1b0
> [  195.433640]  vfs_fallocate+0x459/0x4f0

Well, we compile the kernel with -fno-strict-overflow for a reason so I
wouldn't consider this a bug. But check_add_overflow() is easier to digest
since we don't have to worry about type details so I'm for this change.

> @@ -319,8 +320,12 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
>  	if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
>  		return -ENODEV;
>  
> -	/* Check for wrap through zero too */
> -	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
> +	/* Check for wraparound */
> +	if (check_add_overflow(offset, len, &sum))
> +		return -EFBIG;
> +
> +	/* Now, check bounds */
> +	if (sum > inode->i_sb->s_maxbytes || sum < 0)
>  		return -EFBIG;

But why do you check for sum < 0? We know from previous checks offset >= 0
&& len > 0 so unless we overflow, sum is guaranteed to be > 0.

								Honza
Justin Stitt May 9, 2024, 10:10 p.m. UTC | #4 
On Thu, May 9, 2024 at 8:53 AM Jan Kara <jack@suse.cz> wrote:
> > @@ -319,8 +320,12 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
> >       if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
> >               return -ENODEV;
> >
> > -     /* Check for wrap through zero too */
> > -     if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
> > +     /* Check for wraparound */
> > +     if (check_add_overflow(offset, len, &sum))
> > +             return -EFBIG;
> > +
> > +     /* Now, check bounds */
> > +     if (sum > inode->i_sb->s_maxbytes || sum < 0)
> >               return -EFBIG;
>
> But why do you check for sum < 0? We know from previous checks offset >= 0
> && len > 0 so unless we overflow, sum is guaranteed to be > 0.

Fair enough. I suppose with the overflow check in place we can no
longer have a sum less than zero there. If nothing else, it tells
readers of this code what the domain of (offset+len) is. I don't mind
sending a new version, though.

>
>                                                                 Honza
> --
> Jan Kara <jack@suse.com>
> SUSE Labs, CR
Jan Kara May 12, 2024, 8:05 a.m. UTC | #5 
On Thu 09-05-24 15:10:07, Justin Stitt wrote:
> On Thu, May 9, 2024 at 8:53 AM Jan Kara <jack@suse.cz> wrote:
> > > @@ -319,8 +320,12 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
> > >       if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
> > >               return -ENODEV;
> > >
> > > -     /* Check for wrap through zero too */
> > > -     if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
> > > +     /* Check for wraparound */
> > > +     if (check_add_overflow(offset, len, &sum))
> > > +             return -EFBIG;
> > > +
> > > +     /* Now, check bounds */
> > > +     if (sum > inode->i_sb->s_maxbytes || sum < 0)
> > >               return -EFBIG;
> >
> > But why do you check for sum < 0? We know from previous checks offset >= 0
> > && len > 0 so unless we overflow, sum is guaranteed to be > 0.
> 
> Fair enough. I suppose with the overflow check in place we can no
> longer have a sum less than zero there. If nothing else, it tells
> readers of this code what the domain of (offset+len) is. I don't mind
> sending a new version, though.

Well, for normal readers offset+len is always a positive number. That's
what you expect. If you see a check for offset+len < 0, you start wondering
what are you missing... only to find you miss nothing and the check is
pointless. So yes, please send a version without the pointless check.

								Honza
diff mbox series

Patch

diff --git a/fs/open.c b/fs/open.c
index ee8460c83c77..d216e69d6872 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -247,6 +247,7 @@  int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 {
 	struct inode *inode = file_inode(file);
 	long ret;
+	loff_t sum;
 
 	if (offset < 0 || len <= 0)
 		return -EINVAL;
@@ -319,8 +320,12 @@  int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 	if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
 		return -ENODEV;
 
-	/* Check for wrap through zero too */
-	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
+	/* Check for wraparound */
+	if (check_add_overflow(offset, len, &sum))
+		return -EFBIG;
+
+	/* Now, check bounds */
+	if (sum > inode->i_sb->s_maxbytes || sum < 0)
 		return -EFBIG;
 
 	if (!file->f_op->fallocate)