Skip to contentSkip to conversation

mremap bug and 2.4?

DISCUSSION Robert L. Harris 9 participants 14 messages Jan 5, 2004 – Jan 6, 2004

lore
Share & tools
lore

Robert L. Harris

author
#0 lore raw

Just read this on full disclosure:

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

Is it valid? No working proof of concept code has been posted so I can't test my systems. The article only lists 2.4 and 2.6. Is this 2.4.16-current, etc? Anyone have any details about versions that are safe so I/We can determine if I need to roll a new production kernel out again?

Thanks,
Robert

:wq!
---------------------------------------------------------------------------
Robert L. Harris | GPG Key ID: E344DA3B

                                         @ x-hkp://pgp.mit.edu

DISCLAIMER:
These are MY OPINIONS ALONE. I speak for no-one else.

Life is not a destination, it's a journey.
Microsoft produces 15 car pileups on the highway.
Don't stop traffic to stand and gawk at the tragedy.

#1 lore raw
Quoted 5 lines · Robert L. Harris

Just read this on full disclosure:

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

Is it valid?

Yes, Marcelo released 2.4.24 an hour ago for that reason.

Erik

Signature · 2 lines
+-- Erik Mouw -- www.harddisk-recovery.com -- +31 70 370 12 90 --
| Lab address: Delftechpark 26, 2628 XH, Delft, The Netherlands
#2 lore raw
In reply to Robert L. Harris
Quoted 11 lines · Robert L. Harris

Just read this on full disclosure:

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

Is it valid? No working proof of concept code has been posted so I can't test my systems. The article only lists 2.4 and 2.6. Is this 2.4.16-current, etc? Anyone have any details about versions that are safe so I/We can determine if I need to roll a new production kernel out again?

It is possible that the problem is exploitable. There is no known public exploit yet, however.

2.4.24 includes a fix for this (mm/mremap.c diff)

author
#3 lore raw

I love you guys. Yeah, I have to compile a new kernel, test it and push it out this week to 600 machines but atleast I don't have to wait 6 months and then hope it doesn't kill all my apps.

You guys are great, THANKS!

Robert

Thus spake Marcelo Tosatti ([email protected]):

Quoted 16 lines · Robert L. Harris
Quoted 5 lines

Just read this on full disclosure:

Quoted 7 lines

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

Is it valid? No working proof of concept code has been posted so I can't test my systems. The article only lists 2.4 and 2.6. Is this 2.4.16-current, etc? Anyone have any details about versions that are safe so I/We can determine if I need to roll a new production kernel out again?

Quoted 4 lines

It is possible that the problem is exploitable. There is no known public exploit yet, however.

2.4.24 includes a fix for this (mm/mremap.c diff)

:wq!
---------------------------------------------------------------------------
Robert L. Harris | GPG Key ID: E344DA3B

                                         @ x-hkp://pgp.mit.edu

DISCLAIMER:
These are MY OPINIONS ALONE. I speak for no-one else.

Life is not a destination, it's a journey.
Microsoft produces 15 car pileups on the highway.
Don't stop traffic to stand and gawk at the tragedy.

#5 lore raw
Quoted 5 lines · On Jan-05 2004, Mon, 18:10 +0100 Diego Calleja <[email protected]> wrote:
Quoted 3 lines · El Mon, 5 Jan 2004 13:26:23 -0200 (BRST) Marcelo Tosatti…
Quoted 2 lines · Robert L. Harris

Just read this on full disclosure:

Quoted 8 lines
Quoted 1 line
Quoted 1 line

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

[...]

Quoted 4 lines

It is possible that the problem is exploitable. There is no known public exploit yet, however.

2.4.24 includes a fix for this (mm/mremap.c diff)

It names 2.2 too. Is there a fix for 2.2?

Ask Alan. He's not following the kernel mailing list too closely these days.

Signature · 1 line
Tomas Szepe <[email protected]>
#6 lore raw

mremap() bug and 2.2?

In reply to Diego Calleja

Dear diary, on Mon, Jan 05, 2004 at 06:10:53PM CET, I got a letter, where Diego Calleja <[email protected]> told me, that...

Quoted 5 lines · El Mon, 5 Jan 2004 13:26:23 -0200 (BRST) Marcelo Tosatti…
Quoted 4 lines · Robert L. Harris
Quoted 2 lines

Just read this on full disclosure:

Quoted 8 lines
Quoted 1 line
Quoted 1 line

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

[...]

Quoted 4 lines

It is possible that the problem is exploitable. There is no known public exploit yet, however.

2.4.24 includes a fix for this (mm/mremap.c diff)

It names 2.2 too. Is there a fix for 2.2?

I'm trying to investigate that right now. In 2.2, mremap() doesn't yet take yet the new_addr argument, therefore the "official" 2.4 fix wouldn't apply at all to it. There are four possibilities:

* The isec.pl guys just made a mistake.

* 2.2's get_unmapped_area() can return dangerous pages for len == 0, whilst the 2.4's get_unmapped_area() cannot. (I'm not sure, looking into that code right now.)

* 2.4's fix is incorrect.

* I'm missing something obvious.

Anyone has an idea?

Signature · 6 lines
				Petr "Pasky" Baudis
.
The brain is a wonderful organ; it starts working the moment you get up
in the morning, and does not stop until you get to work.
.
Stuff: http://pasky.or.cz/
#7 lore raw

mremap() bug IMHO not in 2.2

Dear diary, on Mon, Jan 05, 2004 at 07:26:07PM CET, I got a letter, where Petr Baudis <[email protected]> told me, that...

Quoted 22 lines

Dear diary, on Mon, Jan 05, 2004 at 06:10:53PM CET, I got a letter, where Diego Calleja <[email protected]> told me, that...

Quoted 5 lines · El Mon, 5 Jan 2004 13:26:23 -0200 (BRST) Marcelo Tosatti…
Quoted 4 lines · Robert L. Harris
Quoted 2 lines

Just read this on full disclosure:

Quoted 8 lines
Quoted 1 line
Quoted 1 line

http://isec.pl/vulnerabilities/isec-0013-mremap.txt

[...]

Quoted 4 lines

It is possible that the problem is exploitable. There is no known public exploit yet, however.

2.4.24 includes a fix for this (mm/mremap.c diff)

It names 2.2 too. Is there a fix for 2.2?

I'm trying to investigate that right now. In 2.2, mremap() doesn't yet take yet the new_addr argument, therefore the "official" 2.4 fix wouldn't apply at all to it. There are four possibilities:

* The isec.pl guys just made a mistake.

Quoted 7 lines

* 2.2's get_unmapped_area() can return dangerous pages for len == 0, whilst the 2.4's get_unmapped_area() cannot. (I'm not sure, looking into that code right now.)

* 2.4's fix is incorrect.

* I'm missing something obvious.

Actually, after looking at the code again, I'm now quite convinced 2.2 has not this particular vulnerability. In order for the exploit to work, you'd need mremap() to relocate you.

But mremap() won't take newaddr argument, so you can't get yourself relocated explicitly. And mremap() will not relocate yourself implicitly to some random spot neither, because since newlen is zero, it will always trigger the shrinking code, which will just munmap() and bail out.

ihaquer, any comments? Is there something we don't know about? If not, please correct your announcement.

Kind regards,

Signature · 6 lines
				Petr "Pasky" Baudis
.
The brain is a wonderful organ; it starts working the moment you get up
in the morning, and does not stop until you get to work.
.
Stuff: http://pasky.or.cz/
#8 lore raw

Re: mremap() bug IMHO not in 2.2

Quoted 4 lines · Petr Baudis

Actually, after looking at the code again, I'm now quite convinced 2.2 has not this particular vulnerability. In order for the exploit to work, you'd need mremap() to relocate you.

Can somebody tell me (in private) what the exploit is in the first place?

The thing is, I can see the VM getting confused and creating a zero-sized vma, and I agree that it shouldn't do that. The fix is trivial. But I don't see where the claimed privilege escalation comes from. A zero-sized vma isn't ever going to be _useful_, since nothing will actually find it.

So yes, it creates some confusion in the VM layer, but it all seems benign. It's clearly a bug, but where does the security problem come in?

Linus

#9 lore raw

Re: mremap() bug IMHO not in 2.2

Quoted 2 lines · Linus Torvalds

So yes, it creates some confusion in the VM layer, but it all seems benign. It's clearly a bug, but where does the security problem come in?

Just guessing, but would a zero-length vma be rounded up to a page, and thus give the attacker scribble permission on a page he shouldn't have had?

#10 lore raw

Re: mremap() bug IMHO not in 2.2

Quoted 8 lines · On Mon, 5 Jan 2004 [email protected] wrote:
Quoted 2 lines · Linus Torvalds

So yes, it creates some confusion in the VM layer, but it all seems benign. It's clearly a bug, but where does the security problem come in?

Just guessing, but would a zero-length vma be rounded up to a page, and thus give the attacker scribble permission on a page he shouldn't have had?

Almost certainly not.

It's more likely that one of the two functions that walk through _all_ the vma's (fork() and exit()) simply knows that a vma can never be zero-length, and uses a

	addr = vma->vm_start;
	do {
		...
		addr += PAGE_SIZE;
	} while (addr < vma->vm_end);

kind of loop - which means that either fork() or exit() would copy or release one page too many.

The only page that should matter is likely the one at 0xC0000000, where there can be extra complications from the fact that we use 4MB pages for the kernel, so when fork/exit tries to walk the page table, it would get bogus results.

Still, I'd expect that to lead to a triple fault (and thus a reboot) rather than any elevation of privileges..

Interesting, in any case. Good catch from whoever found it.

Linus

#11 lore raw

Re: mremap() bug IMHO not in 2.2

Quoted 7 lines · On Jan-05 2004, Mon, 16:08 -0800 Linus Torvalds <[email protected]> wrote:

The only page that should matter is likely the one at 0xC0000000, where there can be extra complications from the fact that we use 4MB pages for the kernel, so when fork/exit tries to walk the page table, it would get bogus results.

Still, I'd expect that to lead to a triple fault (and thus a reboot) rather than any elevation of privileges..

Hmmm... so what about non-x86?

Quoted 1 line

Interesting, in any case. Good catch from whoever found it.

Impressive, yes.

Signature · 1 line
Tomas Szepe <[email protected]>
#12 lore raw

Re: mremap() bug IMHO not in 2.2

In reply to Linus Torvalds
Quoted 7 lines · Linus Torvalds

The only page that should matter is likely the one at 0xC0000000, where there can be extra complications from the fact that we use 4MB pages for the kernel, so when fork/exit tries to walk the page table, it would get bogus results.

This is right, the proof-of-concept exploit to be found on full-disclosure exactly uses that memory address.

Quoted 3 lines

Still, I'd expect that to lead to a triple fault (and thus a reboot) rather than any elevation of privileges..

I agree with Linus. I tested the POC-exploit here on Linux 2.4.22-rc2 and Linux 2.4.23 and everything it does is to simply reboot the box. As for Linux 2.6.0-test9, I get something like a hangup (the same sound is played again and again and only reset helps).

I actually am not sure whether this should be called 'local privlige escalation' or rather 'possibility for Denial of Service attacks'.

Quoted 4 lines

Interesting, in any case. Good catch from whoever found it.

Linus
-

Signature · 4 lines
  .''`.   Martin Loschwitz           Debian GNU/Linux developer
 : :'  :  [email protected]        [email protected]
 `. `'`   http://www.madkiss.org/    people.debian.org/~madkiss/
   `-     Use Debian GNU/Linux 3.0!  See http://www.debian.org/
#13 lore raw

mremap() bug indeed not in 2.2 (confirmed)

In reply to Petr Baudis

Dear diary, on Mon, Jan 05, 2004 at 11:55:08PM CET, I got a letter, where Petr Baudis <[email protected]> told me, that...

Quoted 11 lines

Dear diary, on Mon, Jan 05, 2004 at 07:26:07PM CET, I got a letter, where Petr Baudis <[email protected]> told me, that...

Quoted 9 lines

Dear diary, on Mon, Jan 05, 2004 at 06:10:53PM CET, I got a letter, where Diego Calleja <[email protected]> told me, that...

Quoted 1 line

It names 2.2 too. Is there a fix for 2.2?

I'm trying to investigate that right now. In 2.2, mremap() doesn't yet take yet the new_addr argument, therefore the "official" 2.4 fix wouldn't apply at all to it. There are four possibilities:

* The isec.pl guys just made a mistake.

..snip..

Quoted 3 lines

Actually, after looking at the code again, I'm now quite convinced 2.2 has not this particular vulnerability. In order for the exploit to work, you'd need mremap() to relocate you.

..snip..

Quoted 2 lines

ihaquer, any comments? Is there something we don't know about? If not, please correct your announcement.

It seems to be indeed so. This was just posted to bugtraq & co:

Hi,

our initial posting contains a mistake about the vulnerability of the 2.2 kernel series. Since the 2.2 kernel series doesn't support the MREMAP_FIXED flag it is NOT vulnerable. The source states "MREMAP_FIXED option added 5-Dec-1999" but it didn't make into recent 2.2.x. We apologize for inconvenience.

--
Paul Starzetz
iSEC Security Research
http://isec.pl/

Here you go. And I don't need to worry about my 2.2.25-running pets ;-).

Kind regards,

Signature · 6 lines
				Petr "Pasky" Baudis
.
The brain is a wonderful organ; it starts working the moment you get up
in the morning, and does not stop until you get to work.
.
Stuff: http://pasky.or.cz/