Message-ID: <7d7516a6-07b7-4882-9da2-2c192ef43039@redhat.com>
Date: Tue, 26 Aug 2025 13:31:43 +0200
From: Paolo Abeni <pabeni@...hat.com>
To: John Ousterhout <ouster@...stanford.edu>, netdev@...r.kernel.org
Cc: edumazet@...gle.com, horms@...nel.org, kuba@...nel.org
Subject: Re: [PATCH net-next v15 09/15] net: homa: create homa_rpc.h and
 homa_rpc.c

On 8/18/25 10:55 PM, John Ousterhout wrote:
> +/**
> + * homa_rpc_reap() - Invoked to release resources associated with dead
> + * RPCs for a given socket.
> + * @hsk:      Homa socket that may contain dead RPCs. Must not be locked by the
> + *            caller; this function will lock and release.
> + * @reap_all: False means do a small chunk of work; there may still be
> + *            unreaped RPCs on return. True means reap all dead RPCs for
> + *            hsk.  Will busy-wait if reaping has been disabled for some RPCs.
> + *
> + * Return: A return value of 0 means that we ran out of work to do; calling
> + *         again will do no work (there could be unreaped RPCs, but if so,
> + *         they cannot currently be reaped).  A value greater than zero means
> + *         there is still more reaping work to be done.
> + */
> +int homa_rpc_reap(struct homa_sock *hsk, bool reap_all)
> +{
> +	/* RPC Reaping Strategy:
> +	 *
> +	 * (Note: there are references to this comment elsewhere in the
> +	 * Homa code)
> +	 *
> +	 * Most of the cost of reaping comes from freeing sk_buffs; this can be
> +	 * quite expensive for RPCs with long messages.
> +	 *
> +	 * The natural time to reap is when homa_rpc_end is invoked to
> +	 * terminate an RPC, but this doesn't work for two reasons. First,
> +	 * there may be outstanding references to the RPC; it cannot be reaped
> +	 * until all of those references have been released. Second, reaping
> +	 * is potentially expensive and RPC termination could occur in
> +	 * homa_softirq when there are short messages waiting to be processed.
> +	 * Taking time to reap a long RPC could result in significant delays
> +	 * for subsequent short RPCs.
> +	 *
> +	 * Thus Homa doesn't reap immediately in homa_rpc_end. Instead, dead
> +	 * RPCs are queued up and reaping occurs in this function, which is
> +	 * invoked later when it is less likely to impact latency. The
> +	 * challenge is to do this so that (a) we don't allow large numbers of
> +	 * dead RPCs to accumulate and (b) we minimize the impact of reaping
> +	 * on latency.
> +	 *
> +	 * The primary place where homa_rpc_reap is invoked is when threads
> +	 * are waiting for incoming messages. The thread has nothing else to
> +	 * do (it may even be polling for input), so reaping can be performed
> +	 * with no latency impact on the application.  However, if a machine
> +	 * is overloaded then it may never wait, so this mechanism isn't always
> +	 * sufficient.
> +	 *
> +	 * Homa now reaps in two other places, if reaping while waiting for
> +	 * messages isn't adequate:
> +	 * 1. If too may dead skbs accumulate, then homa_timer will call
> +	 *    homa_rpc_reap.
> +	 * 2. If this timer thread cannot keep up with all the reaping to be
> +	 *    done then as a last resort homa_dispatch_pkts will reap in small
> +	 *    increments (a few sk_buffs or RPCs) for every incoming batch
> +	 *    of packets . This is undesirable because it will impact Homa's
> +	 *    performance.
> +	 *
> +	 * During the introduction of homa_pools for managing input
> +	 * buffers, freeing of packets for incoming messages was moved to
> +	 * homa_copy_to_user under the assumption that this code wouldn't be
> +	 * on the critical path. However, there is evidence that with
> +	 * fast networks (e.g. 100 Gbps) copying to user space is the
> +	 * bottleneck for incoming messages, and packet freeing takes about
> +	 * 20-25% of the total time in homa_copy_to_user. So, it may eventually
> +	 * be desirable to remove packet freeing out of homa_copy_to_user.

See skb_attempt_defer_free()

> +	 */
> +#define BATCH_MAX 20
> +	struct homa_rpc *rpcs[BATCH_MAX];
> +	struct sk_buff *skbs[BATCH_MAX];

A lot of bytes on the stack, and a quite large batch. You should probaly
decrease it.

Also it still feel suspect the need for just another tx free strategy on
top of the several existing caches.

> +	int num_skbs, num_rpcs;
> +	struct homa_rpc *rpc;
> +	struct homa_rpc *tmp;
> +	int i, batch_size;
> +	int skbs_to_reap;
> +	int result = 0;
> +	int rx_frees;
> +
> +	/* Each iteration through the following loop will reap
> +	 * BATCH_MAX skbs.
> +	 */
> +	skbs_to_reap = hsk->homa->reap_limit;
> +	while (skbs_to_reap > 0 && !list_empty(&hsk->dead_rpcs)) {
> +		batch_size = BATCH_MAX;
> +		if (!reap_all) {
> +			if (batch_size > skbs_to_reap)
> +				batch_size = skbs_to_reap;
> +			skbs_to_reap -= batch_size;
> +		}
> +		num_skbs = 0;
> +		num_rpcs = 0;
> +		rx_frees = 0;
> +
> +		homa_sock_lock(hsk);
> +		if (atomic_read(&hsk->protect_count)) {
> +			homa_sock_unlock(hsk);
> +			if (reap_all)
> +				continue;
> +			return 0;
> +		}
> +
> +		/* Collect buffers and freeable RPCs. */
> +		list_for_each_entry_safe(rpc, tmp, &hsk->dead_rpcs,
> +					 dead_links) {
> +			int refs;
> +
> +			/* Make sure that all outstanding uses of the RPC have
> +			 * completed. We can only be sure if the reference
> +			 * count is zero when we're holding the lock. Note:
> +			 * it isn't safe to block while locking the RPC here,
> +			 * since we hold the socket lock.
> +			 */
> +			if (homa_rpc_try_lock(rpc)) {
> +				refs = atomic_read(&rpc->refs);
> +				homa_rpc_unlock(rpc);
> +			} else {
> +				refs = 1;
> +			}
> +			if (refs != 0)
> +				continue;
> +			rpc->magic = 0;
> +
> +			/* For Tx sk_buffs, collect them here but defer
> +			 * freeing until after releasing the socket lock.
> +			 */
> +			if (rpc->msgout.length >= 0) {
> +				while (rpc->msgout.packets) {
> +					skbs[num_skbs] = rpc->msgout.packets;
> +					rpc->msgout.packets = homa_get_skb_info(
> +						rpc->msgout.packets)->next_skb;
> +					num_skbs++;
> +					rpc->msgout.num_skbs--;
> +					if (num_skbs >= batch_size)
> +						goto release;
> +				}
> +			}
> +
> +			/* In the normal case rx sk_buffs will already have been
> +			 * freed before we got here. Thus it's OK to free
> +			 * immediately in rare situations where there are
> +			 * buffers left.
> +			 */
> +			if (rpc->msgin.length >= 0 &&
> +			    !skb_queue_empty_lockless(&rpc->msgin.packets)) {
> +				rx_frees += skb_queue_len(&rpc->msgin.packets);
> +				__skb_queue_purge(&rpc->msgin.packets);
> +			}
> +
> +			/* If we get here, it means all packets have been
> +			 *  removed from the RPC.
> +			 */
> +			rpcs[num_rpcs] = rpc;
> +			num_rpcs++;
> +			list_del(&rpc->dead_links);
> +			WARN_ON(refcount_sub_and_test(rpc->msgout.skb_memory,
> +						      &hsk->sock.sk_wmem_alloc));
> +			if (num_rpcs >= batch_size)
> +				goto release;
> +		}
> +
> +		/* Free all of the collected resources; release the socket
> +		 * lock while doing this.
> +		 */
> +release:
> +		hsk->dead_skbs -= num_skbs + rx_frees;
> +		result = !list_empty(&hsk->dead_rpcs) &&
> +				(num_skbs + num_rpcs) != 0;
> +		homa_sock_unlock(hsk);
> +		homa_skb_free_many_tx(hsk->homa, skbs, num_skbs);
> +		for (i = 0; i < num_rpcs; i++) {
> +			rpc = rpcs[i];
> +
> +			if (unlikely(rpc->msgin.num_bpages))
> +				homa_pool_release_buffers(rpc->hsk->buffer_pool,
> +							  rpc->msgin.num_bpages,
> +							  rpc->msgin.bpage_offsets);
> +			if (rpc->msgin.length >= 0) {
> +				while (1) {
> +					struct homa_gap *gap;
> +
> +					gap = list_first_entry_or_null(
> +							&rpc->msgin.gaps,
> +							struct homa_gap,
> +							links);
> +					if (!gap)
> +						break;
> +					list_del(&gap->links);
> +					kfree(gap);
> +				}
> +			}
> +			if (rpc->peer) {
> +				homa_peer_release(rpc->peer);
> +				rpc->peer = NULL;
> +			}
> +			rpc->state = 0;
> +			kfree(rpc);
> +		}
> +		homa_sock_wakeup_wmem(hsk);

Here num_rpcs can be zero, and you can have spurius wake-ups

> +/**
> + * homa_rpc_hold() - Increment the reference count on an RPC, which will
> + * prevent it from being freed until homa_rpc_put() is called. References
> + * are taken in two situations:
> + * 1. An RPC is going to be manipulated by a collection of functions. In
> + *    this case the top-most function that identifies the RPC takes the
> + *    reference; any function that receives an RPC as an argument can
> + *    assume that a reference has been taken on the RPC by some higher
> + *    function on the call stack.
> + * 2. A pointer to an RPC is stored in an object for use later, such as
> + *    an interest. A reference must be held as long as the pointer remains
> + *    accessible in the object.
> + * @rpc:      RPC on which to take a reference.
> + */
> +static inline void homa_rpc_hold(struct homa_rpc *rpc)
> +{
> +	atomic_inc(&rpc->refs);

`refs` should be a reference_t, since is uses as such.

/P

Powered by blists - more mailing lists