[mojo-edk] Sync broker is incompatible with OSX. |
|||
Issue descriptionOn OSX, ChannelPosix has control messages to track the transfer of fds due to a bug where OSX forgets about fds in transit. The sync broker uses Channel (and hence ChannelPosix) on the host side, but does its own IO on the client side because it needs to be blocking and ChannelPosix needs an IO loop. That custom IO code doesn't handle the control messages. So the bug is twofild: 1. Do we need the sync broker to create shared memory on OSX? 2. If so, how can the client broker code be harmonized with ChannelPosix?
,
Mar 2 2016
The following revision refers to this bug: https://chrome-internal.googlesource.com/bling/chromium.git/+/1b2d1c9f119d12e861e202f39353cc5fb33f982a commit 1b2d1c9f119d12e861e202f39353cc5fb33f982a Author: amistry <amistry@chromium.org> Date: Tue Mar 01 07:14:59 2016
,
Mar 15 2016
I've written a test and can finally answer #1. No, we don't need the broker to create shared memory in Mojo... as long as we use mach shared memory. You can't create posix shared memory inside the sandbox. So, as long as we can use mach shared memory on OSX, we don't need the sync broker at all. Once I've switched shm to mach on osx, I can close this as WontFix.
,
Mar 18 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/018a4e46c477ada7307ed5c7dad7eb58b7da75b5 commit 018a4e46c477ada7307ed5c7dad7eb58b7da75b5 Author: rockot <rockot@chromium.org> Date: Fri Mar 18 07:34:30 2016 [mojo-edk] Stop using the sync broker on OS X The sync broker is only needed for sandboxed shared buffer allocation on POSIX, but it's not needed for Mach shared buffers. This stops using the broker on OS X. BUG= 591742 , 590979 Review URL: https://codereview.chromium.org/1809363002 Cr-Commit-Position: refs/heads/master@{#381908} [modify] https://crrev.com/018a4e46c477ada7307ed5c7dad7eb58b7da75b5/mojo/edk/system/node_controller.cc [modify] https://crrev.com/018a4e46c477ada7307ed5c7dad7eb58b7da75b5/mojo/edk/system/node_controller.h
,
Mar 18 2016
|
|||
►
Sign in to add a comment |
|||
Comment 1 by bugdroid1@chromium.org
, Mar 1 2016