Bug 205521 – Network process sandboxes should not include 'common.sb' or 'system.sb'

• Component: WebKit Misc.
• Status: RESOLVED FIXED

[Brent Fulgham]

The WebKit Networking sandboxes include ‘common.sb’ on iOS, and 'system.sb' on macOS. This enables lots of things we don’t want.

This patch replaces the 'include' with a copy/paste of the contents of the relevant sandbox include file. I removed definitions that were not referenced in the existing Network sandbox, but did not otherwise edit the contents. There are duplicates and redundancies after this patch, which I will remove as a follow-up step once we confirm that this has no regressions.

[Brent Fulgham]

<rdar://problem/58095870>

[Brent Fulgham]

Created attachment 386257 [details]
Patch

[Darin Adler]

Comment on attachment 386257 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=386257&action=review

> Source/WebKit/Resources/SandboxProfiles/ios/com.apple.WebKit.Networking.sb:585
> -(allow mach-lookup
> +(allow mach-lookup (with report) (with telemetry)

These changes are not mentioned in the change log. Accidentally included in this patch?

[Brent Fulgham]

Comment on attachment 386257 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=386257&action=review

>> Source/WebKit/Resources/SandboxProfiles/ios/com.apple.WebKit.Networking.sb:585
>> +(allow mach-lookup (with report) (with telemetry)
> 
> These changes are not mentioned in the change log. Accidentally included in this patch?

I should have mentioned them. I wanted to capture telemetry on a few things we are not sure we need.

[Brent Fulgham]

The Mac-debug-wk1 test failures cannot be due to this change, since the Sandbox is not used in WK1 builds at all.

[Brent Fulgham]

Created attachment 386913 [details]
Patch

[Brent Fulgham]

The API test failures cannot be related, because the iOS Simulator does not use or honor sandbox rules. We do not run API tests on device (at least in EWS).

[Brent Fulgham]

Created attachment 387018 [details]
Patch

[Per Arne Vollan]

Comment on attachment 387018 [details]
Patch

R=me.

[WebKit Commit Bot]

Comment on attachment 387018 [details]
Patch

Clearing flags on attachment: 387018

Committed r254174: <https://trac.webkit.org/changeset/254174>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Truitt Savell]

It looks like the changes in https://trac.webkit.org/changeset/254174/webkit

broke 80 tests on Catalina wk2.

Tracking in https://bugs.webkit.org/show_bug.cgi?id=205932

[Brent Fulgham]

(In reply to Truitt Savell from comment #12)
> It looks like the changes in https://trac.webkit.org/changeset/254174/webkit
> 
> broke 80 tests on Catalina wk2.
> 
> Tracking in https://bugs.webkit.org/show_bug.cgi?id=205932

Truitt, can you roll it out while I investigate? These failures make it look like webrtc requires graphics features in the network process, which is unexpected!

[Truitt Savell]

Reverted r254174 for reason:

Broke 80 tests on Catalina

Committed r254204: <https://trac.webkit.org/changeset/254204>

[Brent Fulgham]

Oh, I'm wrong! This is about sysctl-read values that were allowed by the global access call, but were NOT blocked by the global blocking declaration.

Sandbox: com.apple.WebKit(35422) deny(1) sysctl-read net.routetable.0.0.3.0

I'll see if any others come up.

[Brent Fulgham]

Committed r254209: <https://trac.webkit.org/changeset/254209>