Security patches: RestartJob ignores pid, argv[0]; kill runs as child UID

Security patches: RestartJob ignores pid, argv[0]; kill runs as child UID

BUG=chromium-os:7018
TEST=unit tests, install on device and verify that we can go to BWSI mode

Change-Id: Ic5eb082bf1399d961ee8a15f0a70b8bc2039f9d3

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=d6a119a

[Will Drewry, 2010-09-24 17:48:48]

LGTM w/nit


This is certainly better than -STOP, check /proc, -CONT, though if we do see pid
raciness, we may need to check the target more thoroughly :/

http://codereview.chromium.org/3493012/diff/1/8
File system_utils.cc (right):

http://codereview.chromium.org/3493012/diff/1/8#newcode30
system_utils.cc:30: if (setresuid(owner, owner, suid)) {
I believe suid==-1 will leave it unchanged.  That said, it's possible you were
running as root but didn't have a suid of 0.  It might be worth doing:

getresuid(&uid, &euid, &suid);
suid = 0;
...

Just to not hti any weird edge conditions if session_manager or any callers of
this code are run in another way.

[Chris Masone, 2010-09-24 17:58:45]

http://codereview.chromium.org/3493012/diff/1/8
File system_utils.cc (right):

http://codereview.chromium.org/3493012/diff/1/8#newcode30
system_utils.cc:30: if (setresuid(owner, owner, suid)) {
On 2010/09/24 17:48:48, Will Drewry wrote:
> I believe suid==-1 will leave it unchanged.  That said, it's possible you were
> running as root but didn't have a suid of 0.  It might be worth doing:
> 
> getresuid(&uid, &euid, &suid);
> suid = 0;
> ...
> 
> Just to not hti any weird edge conditions if session_manager or any callers of
> this code are run in another way.

Can't force it to 0, as unit tests are run as non-root.  I'll use -1 instead of
|suid|, though