From: Shivaji Kant <shivajikant@google.com>
To: Stephen Hemminger <stephen@networkplumber.org>
Cc: dev@dpdk.org, Shivaji Kant <shivajikant@google.com>,
Ciara Loftus <ciara.loftus@intel.com>,
Maryam Tahhan <mtahhan@redhat.com>
Subject: [PATCH v4] net/af_xdp: enable AF_XDP program attachment mode
Date: Tue, 29 Jul 2025 16:17:47 +0000 [thread overview]
Message-ID: <20250729161747.894377-1-shivajikant@google.com> (raw)
In-Reply-To: <20250729100845.819452-1-shivajikant@google.com>
Currently, the AF_XDP PMD attaches its XDP program using a fixed
set of flags, defaulting to `XDP_FLAGS_UPDATE_IF_NOEXIST`. This lacks
flexibility for users who might need to explicitly control the XDP
mode based on their specific hardware capabilities or performance
requirements.
This patch introduces a new vdev argument, `mode`, for the `net_af_xdp`
PMD. This argument allows users to explicitly specify the desired
XDP program attachment mode:
`drv` (XDP_FLAGS_DRV_MODE),
`skb` (XDP_FLAGS_SKB_MODE),
`hw` (XDP_FLAGS_HW_MODE).
(If needed more can be added)
This change provides greater control and flexibility for users to
fine-tune AF_XDP behavior, enabling them to leverage native driver
or hardware offload capabilities when available, or fall back to
generic SKB mode for broader compatibility.
Signed-off-by: Shivaji Kant <shivajikant@google.com>
---
Changes in v2:
The attach mode flag was intended to be per device flag. So making
it part of the pmd_internals struct.
---
Changes in v3
Fix the build error for cases without RTE_NET_AF_XDP_LIBBPF_XDP_ATTACH
---
Changes in v4
Add check for invalid path argument return.
---
drivers/net/af_xdp/rte_eth_af_xdp.c | 77 ++++++++++++++++++++++++-----
1 file changed, 64 insertions(+), 13 deletions(-)
diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c b/drivers/net/af_xdp/rte_eth_af_xdp.c
index 5f65850a27..9858ac24b2 100644
--- a/drivers/net/af_xdp/rte_eth_af_xdp.c
+++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
@@ -169,6 +169,7 @@ struct pmd_internals {
int queue_cnt;
int max_queue_cnt;
int configured_queue_cnt;
+ uint mode_flag;
bool shared_umem;
char prog_path[PATH_MAX];
bool custom_prog_configured;
@@ -198,6 +199,26 @@ struct pmd_process_private {
#define ETH_AF_XDP_USE_CNI_ARG "use_cni"
#define ETH_AF_XDP_USE_PINNED_MAP_ARG "use_pinned_map"
#define ETH_AF_XDP_DP_PATH_ARG "dp_path"
+#define ETH_AF_XDP_MODE_ARG "mode"
+
+/* Define different modes for af_xdp prog to attach */
+#define ETH_AF_XDP_DRV_MODE_ARG "drv"
+#define ETH_AF_XDP_SKB_MODE_ARG "skb"
+#define ETH_AF_XDP_HW_MODE_ARG "hw"
+#define ETH_AF_XDP_NUM_MODE_ARG 3
+
+static const char * const mode_arguments[] = {
+ ETH_AF_XDP_DRV_MODE_ARG,
+ ETH_AF_XDP_SKB_MODE_ARG,
+ ETH_AF_XDP_HW_MODE_ARG,
+ NULL
+};
+
+static const unsigned int mode_flags[] = {
+ XDP_FLAGS_DRV_MODE,
+ XDP_FLAGS_SKB_MODE,
+ XDP_FLAGS_HW_MODE
+};
static const char * const valid_arguments[] = {
ETH_AF_XDP_IFACE_ARG,
@@ -210,6 +231,7 @@ static const char * const valid_arguments[] = {
ETH_AF_XDP_USE_CNI_ARG,
ETH_AF_XDP_USE_PINNED_MAP_ARG,
ETH_AF_XDP_DP_PATH_ARG,
+ ETH_AF_XDP_MODE_ARG,
NULL
};
@@ -950,14 +972,14 @@ remove_xdp_program(struct pmd_internals *internals)
uint32_t curr_prog_id = 0;
int ret;
- ret = bpf_xdp_query_id(internals->if_index, XDP_FLAGS_UPDATE_IF_NOEXIST,
+ ret = bpf_xdp_query_id(internals->if_index, internals->mode_flag,
&curr_prog_id);
if (ret != 0) {
AF_XDP_LOG_LINE(ERR, "bpf_xdp_query_id failed");
return ret;
}
- ret = bpf_xdp_detach(internals->if_index, XDP_FLAGS_UPDATE_IF_NOEXIST,
+ ret = bpf_xdp_detach(internals->if_index, internals->mode_flag,
NULL);
if (ret != 0)
AF_XDP_LOG_LINE(ERR, "bpf_xdp_detach failed");
@@ -978,14 +1000,14 @@ remove_xdp_program(struct pmd_internals *internals)
int ret;
ret = bpf_get_link_xdp_id(internals->if_index, &curr_prog_id,
- XDP_FLAGS_UPDATE_IF_NOEXIST);
+ internals->mode_flag);
if (ret != 0) {
AF_XDP_LOG_LINE(ERR, "bpf_get_link_xdp_id failed");
return ret;
}
ret = bpf_set_link_xdp_fd(internals->if_index, -1,
- XDP_FLAGS_UPDATE_IF_NOEXIST);
+ internals->mode_flag);
if (ret != 0)
AF_XDP_LOG_LINE(ERR, "bpf_set_link_xdp_fd failed");
return ret;
@@ -1305,7 +1327,7 @@ get_pinned_map(const char *dp_path, int *map_fd)
}
static int
-load_custom_xdp_prog(const char *prog_path, int if_index, struct bpf_map **map)
+load_custom_xdp_prog(const char *prog_path, int if_index, struct bpf_map **map, uint mode_flag)
{
int ret, prog_fd;
struct bpf_object *obj;
@@ -1328,7 +1350,7 @@ load_custom_xdp_prog(const char *prog_path, int if_index, struct bpf_map **map)
/* Link the program with the given network device */
ret = link_xdp_prog_with_dev(if_index, prog_fd,
- XDP_FLAGS_UPDATE_IF_NOEXIST);
+ mode_flag);
if (ret) {
AF_XDP_LOG_LINE(ERR, "Failed to set prog fd %d on interface",
prog_fd);
@@ -1679,7 +1701,7 @@ xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
cfg.rx_size = ring_size;
cfg.tx_size = ring_size;
cfg.libbpf_flags = 0;
- cfg.xdp_flags = XDP_FLAGS_UPDATE_IF_NOEXIST;
+ cfg.xdp_flags = internals->mode_flag;
cfg.bind_flags = 0;
/* Force AF_XDP socket into copy mode when users want it */
@@ -1698,7 +1720,7 @@ xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
if (!internals->custom_prog_configured) {
ret = load_custom_xdp_prog(internals->prog_path,
internals->if_index,
- &internals->map);
+ &internals->map, internals->mode_flag);
if (ret) {
AF_XDP_LOG_LINE(ERR, "Failed to load custom XDP program %s",
internals->prog_path);
@@ -2017,6 +2039,25 @@ parse_name_arg(const char *key __rte_unused,
return 0;
}
+/** parse name argument */
+static int
+parse_mode_arg(const char *key __rte_unused,
+ const char *value, void *extra_args)
+{
+ unsigned int *mode = extra_args;
+ unsigned int i;
+
+ for (i = 0; i < ETH_AF_XDP_NUM_MODE_ARG; i++) {
+ if (strcmp(value, mode_arguments[i]) == 0) {
+ *mode |= mode_flags[i];
+ return 0;
+ }
+ }
+
+ AF_XDP_LOG_LINE(ERR, "Invalid af_xdp mode, choose correct mode to attach af_xdp program.");
+ return -EINVAL;
+}
+
/** parse xdp prog argument */
static int
parse_prog_arg(const char *key __rte_unused,
@@ -2094,7 +2135,7 @@ static int
parse_parameters(struct rte_kvargs *kvlist, char *if_name, int *start_queue,
int *queue_cnt, int *shared_umem, char *prog_path,
int *busy_budget, int *force_copy, int *use_cni,
- int *use_pinned_map, char *dp_path)
+ int *use_pinned_map, char *dp_path, uint *xdp_mode)
{
int ret;
@@ -2147,6 +2188,13 @@ parse_parameters(struct rte_kvargs *kvlist, char *if_name, int *start_queue,
ret = rte_kvargs_process(kvlist, ETH_AF_XDP_DP_PATH_ARG,
&parse_prog_arg, dp_path);
+
+ if (ret < 0)
+ goto free_kvlist;
+
+ ret = rte_kvargs_process(kvlist, ETH_AF_XDP_MODE_ARG,
+ &parse_mode_arg, xdp_mode);
+
if (ret < 0)
goto free_kvlist;
@@ -2189,7 +2237,7 @@ static struct rte_eth_dev *
init_internals(struct rte_vdev_device *dev, const char *if_name,
int start_queue_idx, int queue_cnt, int shared_umem,
const char *prog_path, int busy_budget, int force_copy,
- int use_cni, int use_pinned_map, const char *dp_path)
+ int use_cni, int use_pinned_map, const char *dp_path, uint xdp_mode)
{
const char *name = rte_vdev_device_name(dev);
const unsigned int numa_node = dev->device.numa_node;
@@ -2220,6 +2268,7 @@ init_internals(struct rte_vdev_device *dev, const char *if_name,
internals->force_copy = force_copy;
internals->use_cni = use_cni;
internals->use_pinned_map = use_pinned_map;
+ internals->mode_flag = XDP_FLAGS_UPDATE_IF_NOEXIST | xdp_mode;
strlcpy(internals->dp_path, dp_path, PATH_MAX);
if (xdp_get_channels_info(if_name, &internals->max_queue_cnt,
@@ -2412,6 +2461,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
int force_copy = 0;
int use_cni = 0;
int use_pinned_map = 0;
+ uint xdp_mode = 0;
char dp_path[PATH_MAX] = {'\0'};
struct rte_eth_dev *eth_dev = NULL;
const char *name = rte_vdev_device_name(dev);
@@ -2456,7 +2506,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
if (parse_parameters(kvlist, if_name, &xsk_start_queue_idx,
&xsk_queue_cnt, &shared_umem, prog_path,
&busy_budget, &force_copy, &use_cni, &use_pinned_map,
- dp_path) < 0) {
+ dp_path, &xdp_mode) < 0) {
AF_XDP_LOG_LINE(ERR, "Invalid kvargs value");
return -EINVAL;
}
@@ -2524,7 +2574,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
eth_dev = init_internals(dev, if_name, xsk_start_queue_idx,
xsk_queue_cnt, shared_umem, prog_path,
busy_budget, force_copy, use_cni, use_pinned_map,
- dp_path);
+ dp_path, xdp_mode);
if (eth_dev == NULL) {
AF_XDP_LOG_LINE(ERR, "Failed to init internals");
return -1;
@@ -2587,4 +2637,5 @@ RTE_PMD_REGISTER_PARAM_STRING(net_af_xdp,
"force_copy=<int> "
"use_cni=<int> "
"use_pinned_map=<int> "
- "dp_path=<string> ");
+ "dp_path=<string> "
+ "mode=<string> ");
--
2.50.1.552.g942d659e1b-goog
next prev parent reply other threads:[~2025-07-29 16:17 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-07-23 7:42 [PATCH] " Shivaji Kant
2025-07-23 12:30 ` Ivan Malov
2025-07-29 8:24 ` [PATCH v2] " Shivaji Kant
2025-07-29 10:08 ` [PATCH v3] " Shivaji Kant
2025-07-29 14:48 ` Stephen Hemminger
2025-07-29 14:50 ` Stephen Hemminger
2025-07-29 16:17 ` Shivaji Kant [this message]
2025-07-29 22:10 ` [PATCH v4] " Stephen Hemminger
2025-07-30 11:41 ` [PATCH v5] " Shivaji Kant
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=20250729161747.894377-1-shivajikant@google.com \
--to=shivajikant@google.com \
--cc=ciara.loftus@intel.com \
--cc=dev@dpdk.org \
--cc=mtahhan@redhat.com \
--cc=stephen@networkplumber.org \
/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).