From: Jason Xing <kerneljasonxing@gmail.com>
To: davem@davemloft.net, edumazet@google.com, kuba@kernel.org,
pabeni@redhat.com, bjorn@kernel.org, magnus.karlsson@intel.com,
maciej.fijalkowski@intel.com, jonathan.lemon@gmail.com,
sdf@fomichev.me, ast@kernel.org, daniel@iogearbox.net,
hawk@kernel.org, john.fastabend@gmail.com, horms@kernel.org,
andrew+netdev@lunn.ch
Cc: bpf@vger.kernel.org, netdev@vger.kernel.org,
Jason Xing <kernelxing@tencent.com>
Subject: [PATCH net-next v2 1/9] xsk: introduce XDP_GENERIC_XMIT_BATCH setsockopt
Date: Mon, 25 Aug 2025 21:53:34 +0800 [thread overview]
Message-ID: <20250825135342.53110-2-kerneljasonxing@gmail.com> (raw)
In-Reply-To: <20250825135342.53110-1-kerneljasonxing@gmail.com>
From: Jason Xing <kernelxing@tencent.com>
Add a new socket option to provide an alternative to achieve a higher
overall throughput with the rest of series applied.
Init skb_cache and desc_batch when setting setsockopt with xs->mutex
protection.
skb_cache will be used to store newly allocated skb at one time in the
xmit path. desc_batch will be used to temporarily store descriptors of
pool.
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
Documentation/networking/af_xdp.rst | 11 +++++++
include/net/xdp_sock.h | 3 ++
include/uapi/linux/if_xdp.h | 1 +
net/xdp/xsk.c | 47 +++++++++++++++++++++++++++++
tools/include/uapi/linux/if_xdp.h | 1 +
5 files changed, 63 insertions(+)
diff --git a/Documentation/networking/af_xdp.rst b/Documentation/networking/af_xdp.rst
index 50d92084a49c..decb4da80db4 100644
--- a/Documentation/networking/af_xdp.rst
+++ b/Documentation/networking/af_xdp.rst
@@ -447,6 +447,17 @@ mode to allow application to tune the per-socket maximum iteration for
better throughput and less frequency of send syscall.
Allowed range is [32, xs->tx->nentries].
+XDP_GENERIC_XMIT_BATCH
+----------------------
+
+It provides an option that allows application to use batch xmit in the copy
+mode. Batch process tries to allocate a certain number skbs through bulk
+mechanism first and then send them out at one time, minimizing the number
+of grabbing/releasing a few locks (like cache lock and queue lock).
+it normally gains the overall performance improvement as observed by
+xdpsock benchmark, whereas it might increase the latency of per packet.
+The maximum value shouldn't be larger than xs->max_tx_budget.
+
XDP_STATISTICS getsockopt
-------------------------
diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index ce587a225661..c2b05268b8ad 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -70,6 +70,7 @@ struct xdp_sock {
* preventing other XSKs from being starved.
*/
u32 tx_budget_spent;
+ u32 generic_xmit_batch;
/* Statistics */
u64 rx_dropped;
@@ -89,6 +90,8 @@ struct xdp_sock {
struct mutex mutex;
struct xsk_queue *fq_tmp; /* Only as tmp storage before bind */
struct xsk_queue *cq_tmp; /* Only as tmp storage before bind */
+ struct sk_buff **skb_cache;
+ struct xdp_desc *desc_batch;
};
/*
diff --git a/include/uapi/linux/if_xdp.h b/include/uapi/linux/if_xdp.h
index 23a062781468..44cb72cd328e 100644
--- a/include/uapi/linux/if_xdp.h
+++ b/include/uapi/linux/if_xdp.h
@@ -80,6 +80,7 @@ struct xdp_mmap_offsets {
#define XDP_STATISTICS 7
#define XDP_OPTIONS 8
#define XDP_MAX_TX_SKB_BUDGET 9
+#define XDP_GENERIC_XMIT_BATCH 10
struct xdp_umem_reg {
__u64 addr; /* Start of packet data area */
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 9c3acecc14b1..e75a6e2bab83 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -1122,6 +1122,8 @@ static int xsk_release(struct socket *sock)
xskq_destroy(xs->tx);
xskq_destroy(xs->fq_tmp);
xskq_destroy(xs->cq_tmp);
+ kfree(xs->skb_cache);
+ kvfree(xs->desc_batch);
sock_orphan(sk);
sock->sk = NULL;
@@ -1456,6 +1458,51 @@ static int xsk_setsockopt(struct socket *sock, int level, int optname,
WRITE_ONCE(xs->max_tx_budget, budget);
return 0;
}
+ case XDP_GENERIC_XMIT_BATCH:
+ {
+ struct xdp_desc *descs;
+ struct sk_buff **skbs;
+ unsigned int batch;
+ int ret = 0;
+
+ if (optlen != sizeof(batch))
+ return -EINVAL;
+ if (copy_from_sockptr(&batch, optval, sizeof(batch)))
+ return -EFAULT;
+ if (batch > xs->max_tx_budget)
+ return -EACCES;
+
+ mutex_lock(&xs->mutex);
+ if (!batch) {
+ kfree(xs->skb_cache);
+ kvfree(xs->desc_batch);
+ xs->generic_xmit_batch = 0;
+ goto out;
+ }
+
+ skbs = kmalloc(batch * sizeof(struct sk_buff *), GFP_KERNEL);
+ if (!skbs) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ descs = kvcalloc(batch, sizeof(*xs->desc_batch), GFP_KERNEL);
+ if (!skbs) {
+ kfree(skbs);
+ ret = -ENOMEM;
+ goto out;
+ }
+ if (xs->skb_cache)
+ kfree(xs->skb_cache);
+ if (xs->desc_batch)
+ kvfree(xs->desc_batch);
+
+ xs->skb_cache = skbs;
+ xs->desc_batch = descs;
+ xs->generic_xmit_batch = batch;
+out:
+ mutex_unlock(&xs->mutex);
+ return ret;
+ }
default:
break;
}
diff --git a/tools/include/uapi/linux/if_xdp.h b/tools/include/uapi/linux/if_xdp.h
index 23a062781468..44cb72cd328e 100644
--- a/tools/include/uapi/linux/if_xdp.h
+++ b/tools/include/uapi/linux/if_xdp.h
@@ -80,6 +80,7 @@ struct xdp_mmap_offsets {
#define XDP_STATISTICS 7
#define XDP_OPTIONS 8
#define XDP_MAX_TX_SKB_BUDGET 9
+#define XDP_GENERIC_XMIT_BATCH 10
struct xdp_umem_reg {
__u64 addr; /* Start of packet data area */
--
2.41.3
next prev parent reply other threads:[~2025-08-25 13:54 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-08-25 13:53 [PATCH net-next v2 0/9] xsk: improvement performance in copy mode Jason Xing
2025-08-25 13:53 ` Jason Xing [this message]
2025-08-25 13:53 ` [PATCH net-next v2 2/9] xsk: add descs parameter in xskq_cons_read_desc_batch() Jason Xing
2025-08-25 21:18 ` Maciej Fijalkowski
2025-08-26 0:10 ` Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 3/9] xsk: introduce locked version of xskq_prod_write_addr_batch Jason Xing
2025-08-25 21:42 ` Maciej Fijalkowski
2025-08-26 0:13 ` Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 4/9] xsk: extend xsk_build_skb() to support passing an already allocated skb Jason Xing
2025-08-25 21:49 ` Maciej Fijalkowski
2025-08-26 0:26 ` Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 5/9] xsk: add xsk_alloc_batch_skb() to build skbs in batch Jason Xing
2025-08-25 16:56 ` kernel test robot
2025-08-27 14:32 ` Alexander Lobakin
2025-08-28 0:38 ` Jason Xing
2025-08-28 15:28 ` Alexander Lobakin
2025-08-29 0:31 ` Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 6/9] xsk: add direct xmit in batch function Jason Xing
2025-08-25 17:34 ` Stanislav Fomichev
2025-08-26 0:27 ` Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 7/9] xsk: support batch xmit main logic Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 8/9] xsk: support generic batch xmit in copy mode Jason Xing
2025-08-25 13:53 ` [PATCH net-next v2 9/9] xsk: support dynamic xmit.more control for batch xmit Jason Xing
2025-08-25 17:44 ` [PATCH net-next v2 0/9] xsk: improvement performance in copy mode Jakub Kicinski
2025-08-26 0:01 ` Jason Xing
2025-08-26 0:29 ` Jakub Kicinski
2025-08-26 0:51 ` Jason Xing
2025-08-26 1:15 ` Jakub Kicinski
2025-08-26 1:49 ` Jason Xing
2025-08-25 21:15 ` Maciej Fijalkowski
2025-08-26 0:06 ` Jason Xing
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250825135342.53110-2-kerneljasonxing@gmail.com \
--to=kerneljasonxing@gmail.com \
--cc=andrew+netdev@lunn.ch \
--cc=ast@kernel.org \
--cc=bjorn@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=hawk@kernel.org \
--cc=horms@kernel.org \
--cc=john.fastabend@gmail.com \
--cc=jonathan.lemon@gmail.com \
--cc=kernelxing@tencent.com \
--cc=kuba@kernel.org \
--cc=maciej.fijalkowski@intel.com \
--cc=magnus.karlsson@intel.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=sdf@fomichev.me \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).