On Tue, Dec 03, 2024 at 03:52:23PM +0200, Leon Romanovsky wrote: > From: Or Har-Toov <ohartoov@xxxxxxxxxx> > > Currently, MADs sent via an agent are being forwarded directly to the > corresponding MAD QP layer. > MADs with a timeout value set and requiring a response (solicited MADs) > will be resent if the timeout expires without receiving a response. > In a congested subnet, flooding MAD QP layer with more solicited send > requests from the agent will only worsen the situation by triggering > more timeouts and therefore more retries. This explanation does not really capture what this patch is supposed to be doing. The point of solicited MADs is that they require a reply, and the purpose of this patch is to try and conserve a reply slot in the receive queue for the reply to land. Sending more requests than the kernel has reply buffers is probably going to overflow the HCA's receive queue. Jason
