Add seccomp sandbox for non-SFI NaCl

components/nacl/loader/nonsfi/nonsfi_main.cc

Index: components/nacl/loader/nonsfi/nonsfi_main.cc
diff --git a/components/nacl/loader/nonsfi/nonsfi_main.cc b/components/nacl/loader/nonsfi/nonsfi_main.cc
index 5293498e98891f61e9a2df33014aaf084d90a8ea..d3f4df5ab1618690d6217cbef5f41d5ecea94c92 100644
--- a/components/nacl/loader/nonsfi/nonsfi_main.cc
+++ b/components/nacl/loader/nonsfi/nonsfi_main.cc
@@ -178,8 +178,6 @@ void ServiceAccept(struct NaClDesc* port) {
 }  // namespace
 
 void MainStart(NaClHandle imc_bootstrap_handle) {
-  NaClSrpcModuleInit();

[hamaji, 2014/03/24 16:25:43]

This was called from here for historical reason, I think. Hidehiko introduced
this file first, but now this module is initialized in NaClChromeMainInit, so we
don't need this anymore.

[Mark Seaborn, 2014/03/28 01:38:25]

Calling NaClChromeMainInit() for the non-SFI code path is somewhat accidental,
though.  Was there a problem with calling NaClSrpcModuleInit() twice?  There's
nothing wrong with removing this call, but AFAICT it's not related to tightening
up the sandbox.

Could you send this for review as a separate change, please?

[hamaji, 2014/03/28 12:06:10]

I'll send a patch after https://codereview.chromium.org/209423004/ is landed,
and then return to this change.

-
   struct NaClDesc* secure_pair[2] = { NULL, NULL };
   struct NaClDesc* pair[2] = { NULL, NULL };
   CreateSecureSocketPair(secure_pair, pair);