Allow chrome:// pages to load extension resources

Allow chrome:// pages to load extension resources. Also, lock down access from chrome-extension:// pages to the origin of the resource being requested.

BUG=51702
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=55656

[Aaron Boodman, 2010-08-10 19:07:42]

See also http://codereview.chromium.org/3064050. This patch is logically a
modification of that one, but it hasn't been submitted yet.

[Erik does not do reviews, 2010-08-10 19:42:02]

http://codereview.chromium.org/3121003/diff/1/2
File chrome/browser/extensions/extension_protocols.cc (right):

http://codereview.chromium.org/3121003/diff/1/2#newcode92
chrome/browser/extensions/extension_protocols.cc:92: return new
URLRequestErrorJob(request, net::ERR_ADDRESS_UNREACHABLE);
This test seems like a potentially incompatible change.  If I wrote two
extensions designed to communicate together for example.

[Aaron Boodman, 2010-08-10 19:51:15]

On 2010/08/10 19:42:02, Erik Kay wrote:
> http://codereview.chromium.org/3121003/diff/1/2
> File chrome/browser/extensions/extension_protocols.cc (right):
> 
> http://codereview.chromium.org/3121003/diff/1/2#newcode92
> chrome/browser/extensions/extension_protocols.cc:92: return new
> URLRequestErrorJob(request, net::ERR_ADDRESS_UNREACHABLE);
> This test seems like a potentially incompatible change.  If I wrote two
> extensions designed to communicate together for example.

They would still be able to communicate (eg via post message) -- they wouldn't
be able to load resources from each other. Do you think that is an important
feature? My suspicion is that this isn't done, but I haven't measured.

[cdn1, 2010-08-10 19:59:28]

+dmazzoni

On Tue, Aug 10, 2010 at 12:51 PM,  <aa@chromium.org> wrote:
> On 2010/08/10 19:42:02, Erik Kay wrote:
>>
>> http://codereview.chromium.org/3121003/diff/1/2
>> File chrome/browser/extensions/extension_protocols.cc (right):
>
>> http://codereview.chromium.org/3121003/diff/1/2#newcode92
>> chrome/browser/extensions/extension_protocols.cc:92: return new
>> URLRequestErrorJob(request, net::ERR_ADDRESS_UNREACHABLE);
>> This test seems like a potentially incompatible change.  If I wrote two
>> extensions designed to communicate together for example.
>
> They would still be able to communicate (eg via post message) -- they
> wouldn't
> be able to load resources from each other. Do you think that is an important
> feature? My suspicion is that this isn't done, but I haven't measured.
>
> http://codereview.chromium.org/3121003/show
>

[Aaron Boodman, 2010-08-10 20:36:26]

On 2010/08/10 19:51:15, Aaron Boodman wrote:
> On 2010/08/10 19:42:02, Erik Kay wrote:
> > http://codereview.chromium.org/3121003/diff/1/2
> > File chrome/browser/extensions/extension_protocols.cc (right):
> > 
> > http://codereview.chromium.org/3121003/diff/1/2#newcode92
> > chrome/browser/extensions/extension_protocols.cc:92: return new
> > URLRequestErrorJob(request, net::ERR_ADDRESS_UNREACHABLE);
> > This test seems like a potentially incompatible change.  If I wrote two
> > extensions designed to communicate together for example.
> 
> They would still be able to communicate (eg via post message) -- they wouldn't
> be able to load resources from each other. Do you think that is an important
> feature? My suspicion is that this isn't done, but I haven't measured.

I checked the data. There are valid examples of this today. Data++, again.

[Aaron Boodman, 2010-08-10 20:40:25]

ptal

[Erik does not do reviews, 2010-08-10 21:05:38]

LGTM

However, given the potential for regression here, we should also add an explicit
test to verify the legacy cases that we need to continue to support.

Potentially related, see bug 51702.  The way that bug is written, it looks like
we may already have a similar regression.