faraphel/chrony
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ntp: fix handling of socket errors with error queue

Browse source 
In the next Linux version the recvmmsg() system call will be probably
fixed to not return socket errors (e.g. due to ICMP) when reading from
the error queue.

The NTP I/O code assumed this was the correct behavior. When the system
call is fixed, a socket error on a client socket will cause chronyd to
enter a busy loop consuming the CPU until the receive timeout is reached
(8 seconds by default).

Use getsockopt(SO_ERROR) to clear the socket error when reading from the
error queue failed.
This commit is contained in:
Miroslav Lichvar 2018-04-04 09:07:10 +02:00
parent fd1e80802f
commit 19f3ab2225
2 changed files with 15 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
ntp_io.c
View file

@ -717,6 +717,20 @@ read_from_socket(int sock_fd, int event, void *anything)
#endif
if (status < 0) {
#ifdef HAVE_LINUX_TIMESTAMPING
/* If reading from the error queue failed, the exception should be
for a socket error. Clear the error to avoid a busy loop. */
if (flags & MSG_ERRQUEUE) {
int error = 0;
socklen_t len = sizeof (error);
if (getsockopt(sock_fd, SOL_SOCKET, SO_ERROR, &error, &len))
DEBUG_LOG("Could not get SO_ERROR");
if (error)
errno = error;
}
#endif
DEBUG_LOG("Could not receive from fd %d : %s", sock_fd,
strerror(errno));
return;

2
sys_linux.c
View file

@ -497,7 +497,7 @@ SYS_Linux_EnableSystemCallFilter(int level)
SCMP_SYS(lseek), SCMP_SYS(rename), SCMP_SYS(stat), SCMP_SYS(stat64),
SCMP_SYS(statfs), SCMP_SYS(statfs64), SCMP_SYS(unlink),
/* Socket */
SCMP_SYS(bind), SCMP_SYS(connect), SCMP_SYS(getsockname),
SCMP_SYS(bind), SCMP_SYS(connect), SCMP_SYS(getsockname), SCMP_SYS(getsockopt),
SCMP_SYS(recvfrom), SCMP_SYS(recvmmsg), SCMP_SYS(recvmsg),
SCMP_SYS(sendmmsg), SCMP_SYS(sendmsg), SCMP_SYS(sendto),
/* TODO: check socketcall arguments */