[PATCH 2/8] mdev: increase netlink buffer sizes

[Jan Klötzke]

The socket receive buffer turned out to be too small for real world
systems. Use the same size as udevd to be on the safe side. As this is
just a limit and the memory is not allocated by the kernel until really
needed there is actually no memory wasted.

Signed-off-by: Jan Klötzke <jan at kloetzke.net>
---
 util-linux/mdev.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/util-linux/mdev.c b/util-linux/mdev.c
index 207a112c1..4e3299d22 100644
--- a/util-linux/mdev.c
+++ b/util-linux/mdev.c
@@ -1155,9 +1155,21 @@ static void initial_scan(char *temp)
 
 #if ENABLE_FEATURE_MDEV_DAEMON
 
-/* uevent applet uses 16k buffer, and mmaps it before every read */
-# define BUFFER_SIZE (2 * 1024)
-# define RCVBUF (2 * 1024 * 1024)
+/*
+ * The kernel (as of v5.4) will pass up to 32 environment variables with a
+ * total of 2kiB on each event. On top of that the action string and device
+ * path are added. Using a 3kiB buffer for the event should suffice in any
+ * case.
+ *
+ * As far as the socket receive buffer size is concerned 2MiB proved to be too
+ * small (see [1]). Udevd seems to use a whooping 128MiB. The socket receive
+ * buffer size is just a resource limit. The buffers are allocated lazily so
+ * the memory is not wasted.
+ *
+ * [1] http://lists.busybox.net/pipermail/busybox/2019-December/087665.html
+ */
+# define BUFFER_SIZE (3 * 1024)
+# define RCVBUF (128 * 1024 * 1024)
 # define MAX_ENV 32
 
 static void daemon_loop(char *temp, int fd)
-- 
2.20.1