Re: [PATCH net v10 01/14] rxrpc: Fix lack of short-send handling in rxrpc_kernel_send_data()

Paolo Abeni [off-list ref] wrote:

quoted

-	mutex_lock(&call->user_mutex);
+	for (;;) {
+		mutex_lock(&call->user_mutex);

FTR coccinelle can't digest the lock inside the loop:

https://netdev-ctrl.bots.linux.dev/build/cocci/results/823484/stderr

but AFAICS the report is a false positive, and I can't think of an easy
way to avoid it.

I wonder if I should have a separate lock for sending, once sendmsg() has
decided that that's what it's going to do.  The lock would be held across
sleeps for buffer space, but would need to be interruptible in case the call
is aborted (or some other terminal condition occurs).

David