[git commit] mdev: re-initialize if uevent messages were lost
Denys Vlasenko
vda.linux at googlemail.com
Sun Nov 22 13:33:58 UTC 2020
commit: https://git.busybox.net/busybox/commit/?id=d560030548cb237f9ef09a82fd6f372836655306
branch: https://git.busybox.net/busybox/commit/?id=refs/heads/master
If the netlink read() failed with ENOBUFS we know that we have missed at
least one message due to a socket receive buffer overrun. The only way
how to recover is to drop the old socket, open a fresh one and make a
cold-plug scan of the current system state.
Signed-off-by: Jan Klötzke <jan at kloetzke.net>
Signed-off-by: Denys Vlasenko <vda.linux at googlemail.com>
---
util-linux/mdev.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/util-linux/mdev.c b/util-linux/mdev.c
index 4c00a2d7e..d4f06c128 100644
--- a/util-linux/mdev.c
+++ b/util-linux/mdev.c
@@ -1199,6 +1199,16 @@ static void daemon_loop(char *temp, int fd)
len = safe_read(fd, netbuf, sizeof(netbuf) - 1);
if (len < 0) {
+ if (errno == ENOBUFS) {
+ /*
+ * We ran out of socket receive buffer space.
+ * Start from scratch.
+ */
+ dbg1s("uevent overrun, rescanning");
+ close(fd);
+ fd = daemon_init(temp);
+ continue;
+ }
bb_simple_perror_msg_and_die("read");
}
end = netbuf + len;
More information about the busybox-cvs
mailing list