380994 - (CVE-2007-3073) Fix for bug 367428 lets through escaped slashes on Linux (windows too on trunk)

Status: RESOLVED FIXED

(Core :: Networking, defect, P1)

Description

[Boris Zbarsky [:bzbarsky]]

STEPS TO REPRODUCE:

1)  Open browser on Linux
2)  Load resource:///..%2F..%2F..%2F in it

EXPECTED RESULTS:  Don't see file list for an ancestor of the install directory

ACTUAL RESULTS: I'm looking at a listing of ~bzbarsky (since the browser is installed under ~bzbarsky/mozilla/nightly/install-dir).

Comment 1

[Benjamin Smedberg]

This was reported publicly, so opening.

Comment 2

[Boris Zbarsky [:bzbarsky]]

It was?  Where?

Comment 3

[Benjamin Smedberg]

bug 367428 comment 16

Comment 4

[Benjamin Smedberg]

Taking for now, but biesi if you have time for this I'd much appreciate it.

Comment 5

[Daniel Veditz [:dveditz]]

taking to get a fix in.

Comment 6

[Daniel Veditz [:dveditz]]

On trunk it looks like windows can get fooled by ..%2f.. too.

Comment 7

[Ben Bucksch (:BenB)]

resource:///..%2F..%2F..%2F..%2F/etc/host.conf

Comment 8

[georgi - hopefully not receiving bugspam]

macosx on ppc seems partially affected - can traverse up to /Volumes (when started from .dmg)

Comment 9

[Mike Beltzner [:beltzner, not reading bugmail]]

Moving bugs that aren't beta 4 blockers to target final release.

Comment 10

[Samuel Sidler (old account; do not CC)]

Dan says this should block. It's blocking on branch as well.

Comment 11

[Mike Schroepfer]

Dveditz you on this?

Comment 12

[Mike Schroepfer]

Given this is sg:low/dos taking off the blocker list - we'd certainly take a patch..

Comment 13

[Boris Zbarsky [:bzbarsky]]

This isn't quite limited to a dos.  It has potential for data leakage, since any website can link to resource: scripts and stylesheets and then try to glean data from the result.  For example, if you happen to have any interesting data stored in JSON on your hard drive, you lose.

I agree that it's not stop-ship, but I think it should be a very high priority for fixing as soon as we possibly can....

Comment 14

[Damon Sicore (:damons)]

Not blocking, but wanted1.9.0.x+, based on comment #13. 

Comment 15

[Daniel Veditz [:dveditz]]

Attachment: coalesce escaped slashes too

The line numbers in the patch are for the 1.8 branch, but it applies cleanly to cvs trunk and mozilla-central

Comment 16

[Benjamin Smedberg]

Comment on attachment 335516 [details] [diff] [review]
coalesce escaped slashes too

Well, this will work but will almost certainly cause multiple allocations for URIs over 40 chars... can you at least SetCapacity on spec before entering the loop?

Also, this needs a unit-test... you should be able to add a trivial one to http://mxr.mozilla.org/mozilla-central/source/netwerk/test/unit/test_bug337744.js or write a new test copying that code.

Comment 17

[Daniel Veditz [:dveditz]]

Attachment: updated, with testcase

Comment 18

[Daniel Veditz [:dveditz]]

Worked a fix for related bug 394075 into this patch.

Comment 19

[Christian :Biesinger (don't email me, ping me on IRC)]

Comment on attachment 335650 [details] [diff] [review]
updated, with testcase

+        if (*src == '%' && *(src+1) == '2') {

*(src+1) mighht not be valid. You need to check that src+1 != end.

Similar on the next line

(an nsACString isn't necessarily nullterminated)

I also don't think that appending character-by-character is particularly efficient code... but that probably doesn't matter much. (other code keeps track of how many characters were left unchanged and append them in once call. cf. http://mxr.mozilla.org/mozilla-central/source/xpcom/io/nsEscape.cpp#516)

I'd suggest spaces around your + operators but this file doesn't really seem to do that

r+sr=biesi if you fix the src+1 thing

Comment 20

[georgi - hopefully not receiving bugspam]

> *(src+1) mighht not be valid.

so may be:
char ch = *(src+2);

Comment 21

[Daniel Veditz [:dveditz]]

Attachment: Address biesi's comments

Comment 22

[Samuel Sidler (old account; do not CC)]

Comment on attachment 335671 [details] [diff] [review]
Address biesi's comments

Approved for 1.8.1.17. a=ss

Comment 23

[Samuel Sidler (old account; do not CC)]

Comment on attachment 335671 [details] [diff] [review]
Address biesi's comments

And approved for 1.9.0.2 as well. a=ss

Comment 24

[Daniel Veditz [:dveditz]]

Fix checked in to 1.8.1 and 1.9.0 branches. Will seek retroactive r= from biesi to make sure this is what he wanted

Comment 25

[georgi - hopefully not receiving bugspam]

+           if (*(src+2) == 'f' || *(src+***1***) == 'F')
+             ch = '/';
+           else if (*(src+2) == 'e' || *(src+2) == 'E')

is the only |+1| on purpose ?

Comment 26

[Christian :Biesinger (don't email me, ping me on IRC)]

Comment on attachment 335671 [details] [diff] [review]
Address biesi's comments

+           if (*(src+2) == 'f' || *(src+1) == 'F')

I agree with georgi, that should be +2

maybe the test should verify the final URL's spec, instead of just verifying that it doesn't contain ..?

Comment 27

[Daniel Veditz [:dveditz]]

Attachment: incremental patch on top of what's checked in

Comment 28

[Mike Beltzner [:beltzner, not reading bugmail]]

This needs to land both on the branch and on the GECKO190_20080827_RELBRANCH

Comment 29

[Mike Beltzner [:beltzner, not reading bugmail]]

Er, make that both branches (1.8.1 and 1.9) and GECKO190_20080827_RELBRANCH

Comment 30

[Daniel Veditz [:dveditz]]

incremental fix checked into 1.8, 1.9 and _RELBRANCH. Filed bug 452470 for improving the testcase.

Comment 31

[juan becerra [:juanb]]

Verified on latest build candidates for 20017 and 3.0.2. When I load resource:///..%2F..%2F..%2F I see the contents of these directories: 

On 20016
Index of file:///Users/user/Desktop/Firefox.app/Contents/MacOS/../../../
Index of file:///home/mozilla/Desktop/firefox/../../../
WinXP/Vista (nothing happens)

On 20017
Index of file:///Users/user/Desktop/Firefox.app/Contents/MacOS/
Index of file:///home/mozilla/Desktop/firefox/
Index of file:///C:/Program Files/Mozilla Firefox/

On 3.0.1
Index of file:///Users/user/Desktop/Firefox.app/Contents/MacOS/../../../
Index of file:///home/mozilla/Desktop/firefox/../../../
WinXP/Vista (nothing happens)

On 3.0.2
Index of file:///Users/user/Desktop/Firefox.app/Contents/MacOS/
Index of file:///home/mozilla/Desktop/firefox/
Index of file:///C:/Program Files/Mozilla Firefox/

Comment 32

[Alexander Sack]

Comment on attachment 335671 [details] [diff] [review]
Address biesi's comments

a=asac for 1.8.0.15 (with attachment 335741 [details] [diff] [review])

Comment 33

[Alexander Sack]

Comment on attachment 335741 [details] [diff] [review]
incremental patch on top of what's checked in

a=asac for 1.8.0.15 (with attachment 335741 [details] [diff] [review] and iterator backport patch)

Comment 34

[Alexander Sack]

Attachment: iterator hack (on top) for 1.8.0

biesi, you think this 1.8.0 fix is too ugly for the sake of keeping the other patches the unmodified?

Comment 35

[Alexander Sack]

Attachment: iterator hack (on top) for 1.8.0

err, here the one that was meant to be attached :)

Comment 36

[Christian :Biesinger (don't email me, ping me on IRC)]

Comment on attachment 336251 [details] [diff] [review]
iterator hack (on top) for 1.8.0 

this is for 180 only? fine with me :)

Comment 37

[Alexander Sack]

Comment on attachment 336251 [details] [diff] [review]
iterator hack (on top) for 1.8.0 

a=asac for 1.8.0.15

thanks biesi!

Comment 38

[Marc Bejarano]

can't we resolve this as fixed, now?

Comment 39

[Daniel Veditz [:dveditz]]

I need someone to check this into mozilla-central to call it "fixed"

Comment 40

[Reed Loden [:reed]]

http://hg.mozilla.org/mozilla-central/rev/6dad95d60106

Comment 41

[Reed Loden [:reed]]

http://hg.mozilla.org/mozilla-central/rev/1eccc541661c

Comment 42

[Al Billings [:abillings - ex-MoCo]]

Verified for 1.9.0.4 with Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.0.4pre) Gecko/2008102304 GranParadiso/3.0.4pre.

Comment 43

[Benjamin Smedberg]

This landed before 1.9.1 branched

Comment 44

[Mike Beltzner [:beltzner, not reading bugmail]]

(In reply to comment #43)
> This landed before 1.9.1 branched

Adding fixed1.9.1