Skip to contentSkip to conversation

splice: fix user pointer access in get_iovec_page_array()

PATCH In review Pekka J Enberg 6 participants 10 messages Feb 10, 2008 – Feb 11, 2008 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 splice: fix user pointer access in get_iovec_page_array()

From: Bastian Blank <[email protected]>

The commit 8811930dc74a503415b35c4a79d14fb0b408a361 ("splice: missing user pointer access verification") added access_ok() to copy_from_user_mmap_sem() which only ensures we can copy the struct iovecs from userspace to the kernel but we also must check whether we can access the actual memory region pointed to by the struct iovec to close the local root exploit.

---
Bastian, can I have your Signed-off-by for this, please? Oliver, Niki, can you please confirm this closes the hole?

 fs/splice.c |    3 +++
 1 file changed, 3 insertions(+)
fs/splice.c +3 -0
Index: linux-2.6/fs/splice.c
===================================================================
@@ -1237,6 +1237,9 @@ static int get_iovec_page_array(const st
 		if (unlikely(!base))
 			break;
 
+		if (unlikely(!access_ok(VERIFY_READ, base, len)))
+			break;
+
 		/*
 		 * Get this base offset and number of pages, then map
 		 * in the user pages.
#1 lore raw
Quoted 8 lines · Pekka J Enberg

From: Bastian Blank <[email protected]>

The commit 8811930dc74a503415b35c4a79d14fb0b408a361 ("splice: missing user pointer access verification") added access_ok() to copy_from_user_mmap_sem() which only ensures we can copy the struct iovecs from userspace to the kernel but we also must check whether we can access the actual memory region pointed to by the struct iovec to close the local root exploit.

Quoted 4 lines
Quoted 14 lines

Index: linux-2.6/fs/splice.c
===================================================================

fs/splice.c +3 -0
@@ -1237,6 +1237,9 @@ static int get_iovec_page_array(const st
 		if (unlikely(!base))
 			break;
 
+		if (unlikely(!access_ok(VERIFY_READ, base, len)))
+			break;
+
 		/*
 		 * Get this base offset and number of pages, then map
 		 * in the user pages.
Signature · 2 lines
Those who hate and fight must stop themselves -- otherwise it is not stopped.
		-- Spock, "Day of the Dove", stardate unknown
#2 lore raw

----8<----

Linux pancs 2.6.22.17-opt2-cve2 #1 SMP Sun Feb 10 16:22:37 CET 2008 i686 GNU/Linux
-----------------------------------
Linux vmsplice Local Root Exploit
By qaaz
-----------------------------------
[+] mmap: 0x0 .. 0x1000
[+] page: 0x0
[+] page: 0x20
[+] mmap: 0x4000 .. 0x5000
[+] page: 0x4000
[+] page: 0x4020
[+] mmap: 0x1000 .. 0x2000
[+] page: 0x1000
[+] mmap: 0xb7f2d000 .. 0xb7f5f000
[-] vmsplice: Bad address

-----

oliver@pancs:/tmp$ uname -a && ./2623_2624_root_exploit
Linux pancs 2.6.22.17-opt2-cve2 #1 SMP Sun Feb 10 16:22:37 CET 2008
i686 GNU/Linux
-----------------------------------
 Linux vmsplice Local Root Exploit
 By qaaz
-----------------------------------
[+] addr: 0xc01112e9
[-] wtf

---->8----

Quoted 38 lines · Bastian Blank
Quoted 4 lines · Pekka J Enberg

From: Bastian Blank <[email protected]>

The commit 8811930dc74a503415b35c4a79d14fb0b408a361 ("splice: missing user pointer access verification") added access_ok() to

copy_from_user_mmap_sem()

Quoted 1 line

which only ensures we can copy the struct iovecs from userspace to the

kernel

Quoted 1 line

but we also must check whether we can access the actual memory region

pointed

Quoted 6 lines

to by the struct iovec to close the local root exploit.

Quoted 5 lines

Index: linux-2.6/fs/splice.c
===================================================================

fs/splice.c static int get_iovec_page_array(const st · line 1245

Quoted 9 lines
fs/splice.c +3 -0
@@ -1237,6 +1237,9 @@ static int get_iovec_page_array(const st
 		if (unlikely(!base))
 			break;
 
+		if (unlikely(!access_ok(VERIFY_READ, base, len)))
+			break;
+
 		/*
 		 * Get this base offset and number of pages, then map
 		 * in the user pages.

--
Those who hate and fight must stop themselves -- otherwise it is not stopped.
-- Spock, "Day of the Dove", stardate unknown

Signature · 2 lines
Thanks,
Oliver
#3 lore raw
In reply to Pekka J Enberg
Quoted 8 lines · Pekka J Enberg

From: Bastian Blank <[email protected]>

The commit 8811930dc74a503415b35c4a79d14fb0b408a361 ("splice: missing user pointer access verification") added access_ok() to copy_from_user_mmap_sem() which only ensures we can copy the struct iovecs from userspace to the kernel but we also must check whether we can access the actual memory region pointed to by the struct iovec to close the local root exploit.

Quoted 7 lines

---
Bastian, can I have your Signed-off-by for this, please? Oliver, Niki, can you please confirm this closes the hole?

Pekka, I confirm that it also closes the hole once backported to 2.6.22.

Willy

#4 lore raw

Linux pancs 2.6.22.17-opt2-cve2 #1 SMP Sun Feb 10 16:22:37 CET 2008 i686 GNU/Linux
-----------------------------------
Linux vmsplice Local Root Exploit
By qaaz
-----------------------------------
[+] mmap: 0x0 .. 0x1000
[+] page: 0x0
[+] page: 0x20
[+] mmap: 0x4000 .. 0x5000
[+] page: 0x4000
[+] page: 0x4020
[+] mmap: 0x1000 .. 0x2000
[+] page: 0x1000
[+] mmap: 0xb7f95000 .. 0xb7fc7000
[-] vmsplice: Bad address
-----------------------------------
Linux vmsplice Local Root Exploit
By qaaz
-----------------------------------
[+] addr: 0xc01112e9
[-] wtf

the patch is good for 2.6.22.y

Quoted 25 lines · Willy Tarreau
Quoted 4 lines · Pekka J Enberg

From: Bastian Blank <[email protected]>

The commit 8811930dc74a503415b35c4a79d14fb0b408a361 ("splice: missing user pointer access verification") added access_ok() to

copy_from_user_mmap_sem()

Quoted 1 line

which only ensures we can copy the struct iovecs from userspace to the

kernel

Quoted 1 line

but we also must check whether we can access the actual memory region

pointed

Quoted 9 lines

to by the struct iovec to close the local root exploit.

---
Bastian, can I have your Signed-off-by for this, please? Oliver, Niki, can you please confirm this closes the hole?

Pekka, I confirm that it also closes the hole once backported to 2.6.22.

Willy

Signature · 2 lines
Thanks,
Oliver
#5 lore raw

PATCH 1/1 vmsplice exploit fix (was: splice: fix user pointer access in get_iovec_page_array)

In reply to Pekka J Enberg

Kudos to all involved in the rapid response. But.

Information on patching this vulnerability is not available front and center in many of the places you would expect: kernel.org front page, debian.org front page, covered on planet.debian.org but without a pointer to the patch, and so on. So this post provides a subject line for Google to find, and for good measure mentions the word vulnerability.

Also,

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=464953

I think many users would first go to kernel.org on a day like today, as I did. Nothing to see there. We could do a way better job of getting the word out.

Patch attached as posted above by Pekka. For the mortals among us:

   cd linux-2.6.recent && patch <fix.vmsplice.exploit.patch -p1

Regards,

Daniel

---
Bastian, can I have your Signed-off-by for this, please? Oliver, Niki, can you please confirm this closes the hole?

 fs/splice.c |    3 +++
 1 file changed, 3 insertions(+)
fs/splice.c +3 -0
Index: linux-2.6/fs/splice.c
===================================================================
@@ -1237,6 +1237,9 @@ static int get_iovec_page_array(const st
 		if (unlikely(!base))
 			break;
 
+		if (unlikely(!access_ok(VERIFY_READ, base, len)))
+			break;
+
 		/*
 		 * Get this base offset and number of pages, then map
 		 * in the user pages.
#6 lore raw

Re: [stable] [PATCH] vmsplice exploit fix (was: splice: fix user pointer access in get_iovec_page_array)

Quoted 8 lines · Daniel Phillips

Kudos to all involved in the rapid response. But.

Information on patching this vulnerability is not available front and center in many of the places you would expect: kernel.org front page, debian.org front page, covered on planet.debian.org but without a pointer to the patch, and so on. So this post provides a subject line for Google to find, and for good measure mentions the word vulnerability.

All currently active Linux kernel versions are now released with a fix for this problem. We have released them through our normal channels, with the needed information as to what the problem is, a pointer to the CVE number, and the patch itself.

I don't think there's much more we need to do here, do you?

thanks,

greg k-h

#7 lore raw

Re: [stable] [PATCH] vmsplice exploit fix (was: splice: fix user pointer access in get_iovec_page_array)

Quoted 1 line · Greg KH

I don't think there's much more we need to do here, do you?

No amount of information about such a thing is too much information. We do not need to do any more, but we can do more.

It is news, how about an entry in the kernel.org news section? How about a security alerts section on kernel.org? There is a lot that can be done to support people who want to patch their kernels as quickly as possible.

Regards,

Daniel

#8 lore raw

Re: [PATCH] vmsplice exploit fix (was: splice: fix user pointer access in get_iovec_page_array)

In reply to Daniel Phillips

Hi Daniel,

Quoted 3 lines · Daniel Phillips

I think many users would first go to kernel.org on a day like today, as I did. Nothing to see there. We could do a way better job of getting the word out.

Any suggestions what I could have done better here? As soon as Linus merged the patch, I sent it to [email protected] and expected information to spread from there. I did also follow up on some vendor bugzillas (Debian, RHEL, Fedora) to make sure they were aware of the merged patch (they were).

Pekka

#9 lore raw

Re: [PATCH] vmsplice exploit fix (was: splice: fix user pointer access in get_iovec_page_array)

Quoted 8 lines · Pekka Enberg

Hi Daniel,

Quoted 3 lines · Daniel Phillips

I think many users would first go to kernel.org on a day like today, as I did. Nothing to see there. We could do a way better job of getting the word out.

Any suggestions what I could have done better here?

You did your part flawlessly as far as I can see. Really, your original subject line was accurate, and lkml should not be the place to go to find the patch, but it is.

Quoted 5 lines

As soon as Linus
merged the patch, I sent it to [email protected] and expected information to spread from there. I did also follow up on some vendor bugzillas (Debian, RHEL, Fedora) to make sure they were aware of the merged patch (they were).

And they are no doubt getting kernel revs out at this very moment, that part of the machine works great. It is just the part about getting out information to users who want to patch their own kernels that could be done better.

Regards,

Daniel