189799 - JavaScript Security policy for specific site can be circumvented

Status: RESOLVED FIXED

(Core :: Security: CAPS, defect)

Description

[Martin Treusch von Buttlar]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2; MultiZilla v1.1.32 final) Gecko/20021126
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2; MultiZilla v1.1.32 final) Gecko/20021126

When using a RFC1737 encoded URL with username and password prepended with an @
to the hostname (i.e http://alice:[e-mail]/ ), the sites configuration
entry is not correctly applied.

Reproducible: Always

Steps to Reproduce:
1. restrict js-capabilities to a specific URL and check it is correctly enforced
as described in 
http://www.mozilla.org/projects/security/components/ConfigPolicy.html

2. Now, type in exactly the same URL but prepend a:b@

Actual Results:  
the formerly restricted capability is now accessible.

Expected Results:  
Mozilla should have blocked the capability regardless of the prepended @-string

Comment 1

[Mitchell Stoltz (not reading bugmail)]

Good catch! I'm making this Security-Sensitive. Martin, if you disagree, you can
uncheck the box above. Turns out the problem is the use of GetPrePath in
nsCodebasePrincipal::GetOrigin, since PrePath includes the username/password. I
think we should exclude those from security comparisons, so that foo.com and
a:[e-mail] are treated as the same origin. Patch to follow.

Comment 2

[Mitchell Stoltz (not reading bugmail)]

Attachment: Patch - ignore user:pass in host comparisons

Comment 3

[Daniel Veditz [:dveditz]]

I don't think this bug needs to be security sensitive since by default we
provide a secure configuration that doesn't depend on site lists. Sure, not good
that an extra feature we provide (unavailable in other browsers) can so easily
be defeated, but that's just a bug and isn't necessarily sensitive.

Comment 4

[Daniel Veditz [:dveditz]]

Comment on attachment 112228 [details] [diff] [review]
Patch - ignore user:pass in host comparisons

sr=dveditz

Comment 5

[Asa Dotzler [:asa]]

To request approval from drivers to land a reviewed patch use the "approval1.3b
?" flag in the patch manager rather than the blocking1.3b? flag in the bug. 

Comment 6

[Asa Dotzler [:asa]]

Comment on attachment 112228 [details] [diff] [review]
Patch - ignore user:pass in host comparisons

a=asa (on behalf of drivers) for checkin to 1.3beta.

Comment 7

[Mitchell Stoltz (not reading bugmail)]

Fix checked in. Should I patch the 1.0.x branch too?

Comment 8

[Daniel Veditz [:dveditz]]

Bugs published on the Known-vulnerabilities page long ago, removing confidential
flag.