Magic-Mirror/gecko-dev
1
0
Fork 0
mirror of https://github.com/mozilla/gecko-dev.git synced 2024-10-08 02:14:43 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Bug 1142693 - Recognize EMSGSIZE as non-fatal on OS X for IPC sendmsg(). r=bent

Browse Source 
Loosely based on Chromium git commit 86c3d9ef4fdf, but redone to insert a
sched_yield(), because treating EMSGSIZE as if it were EAGAIN/EWOULDBLOCK
is (as the Chromium developers note) likely to act as a busy-wait for the
receiver to make progress.
This commit is contained in:
Jed Davis 2015-04-10 22:47:05 -07:00
parent ddba70e696
commit 9596875de6
1 changed files with 36 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
ipc/chromium/src/chrome/common/ipc_channel_posix.cc
View File

@ -6,6 +6,9 @@
#include <errno.h>
#include <fcntl.h>
#if defined(OS_MACOSX)
#include <sched.h>
#endif
#include <stddef.h>
#include <unistd.h>
#include <sys/types.h>
@ -724,9 +727,39 @@ bool Channel::ChannelImpl::ProcessOutgoingMessages() {
msg->file_descriptor_set()->CommitAll();
#endif
if (bytes_written < 0 && errno != EAGAIN) {
CHROMIUM_LOG(ERROR) << "pipe error: " << strerror(errno);
return false;
if (bytes_written < 0) {
switch (errno) {
case EAGAIN:
// Not an error; the sendmsg would have blocked, so return to the
// event loop and try again later.
break;
#if defined(OS_MACOSX)
// (Note: this comment is copied from https://crrev.com/86c3d9ef4fdf6;
// see also bug 1142693 comment #73.)
//
// On OS X if sendmsg() is trying to send fds between processes and
// there isn't enough room in the output buffer to send the fd
// structure over atomically then EMSGSIZE is returned.
//
// EMSGSIZE presents a problem since the system APIs can only call us
// when there's room in the socket buffer and not when there is
// "enough" room.
//
// The current behavior is to return to the event loop when EMSGSIZE
// is received and hopefull service another FD. This is however still
// technically a busy wait since the event loop will call us right
// back until the receiver has read enough data to allow passing the
// FD over atomically.
case EMSGSIZE:
// Because this is likely to result in a busy-wait, we'll try to make
// it easier for the receiver to make progress.
sched_yield();
break;
#endif
default:
CHROMIUM_LOG(ERROR) << "pipe error: " << strerror(errno);
return false;
}
}
if (static_cast<size_t>(bytes_written) != amt_to_write) {