[RFC net-next v2 1/9] net: napi: Add napi_storage

From: Joe Damato
Date: Sun Sep 08 2024 - 12:09:05 EST


Add a persistent NAPI storage area for NAPI configuration to the core.
Drivers opt-in to setting the storage for a NAPI by passing an index
when calling netif_napi_add_storage.

napi_storage is allocated in alloc_netdev_mqs, freed in free_netdev
(after the NAPIs are deleted), and set to 0 when napi_enable is called.

Signed-off-by: Joe Damato <jdamato@xxxxxxxxxx>
---
.../networking/net_cachelines/net_device.rst | 1 +
include/linux/netdevice.h | 34 +++++++++++++++++++
net/core/dev.c | 18 +++++++++-
3 files changed, 52 insertions(+), 1 deletion(-)

diff --git a/Documentation/networking/net_cachelines/net_device.rst b/Documentation/networking/net_cachelines/net_device.rst
index 22b07c814f4a..a82751c88d18 100644
--- a/Documentation/networking/net_cachelines/net_device.rst
+++ b/Documentation/networking/net_cachelines/net_device.rst
@@ -106,6 +106,7 @@ rx_handler_func_t* rx_handler read_mostly
void* rx_handler_data read_mostly -
struct_netdev_queue* ingress_queue read_mostly -
struct_bpf_mprog_entry tcx_ingress - read_mostly sch_handle_ingress
+struct napi_storage* napi_storage - read_mostly napi_complete_done
struct_nf_hook_entries* nf_hooks_ingress
unsigned_char broadcast[32]
struct_cpu_rmap* rx_cpu_rmap
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index b47c00657bd0..54da1c800e65 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -342,6 +342,14 @@ struct gro_list {
*/
#define GRO_HASH_BUCKETS 8

+/*
+ * Structure for per-NAPI storage
+ */
+struct napi_storage {
+ u64 gro_flush_timeout;
+ u32 defer_hard_irqs;
+};
+
/*
* Structure for NAPI scheduling similar to tasklet but with weighting
*/
@@ -377,6 +385,8 @@ struct napi_struct {
struct list_head dev_list;
struct hlist_node napi_hash_node;
int irq;
+ int index;
+ struct napi_storage *napi_storage;
};

enum {
@@ -2009,6 +2019,9 @@ enum netdev_reg_state {
* @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
* where the clock is recovered.
*
+ * @napi_storage: An array of napi_storage structures containing per-NAPI
+ * settings.
+ *
* FIXME: cleanup struct net_device such that network protocol info
* moves out.
*/
@@ -2087,6 +2100,7 @@ struct net_device {
#ifdef CONFIG_NET_XGRESS
struct bpf_mprog_entry __rcu *tcx_ingress;
#endif
+ struct napi_storage *napi_storage;
__cacheline_group_end(net_device_read_rx);

char name[IFNAMSIZ];
@@ -2648,6 +2662,24 @@ netif_napi_add_tx_weight(struct net_device *dev,
netif_napi_add_weight(dev, napi, poll, weight);
}

+/**
+ * netif_napi_add_storage - initialize a NAPI context and set storage area
+ * @dev: network device
+ * @napi: NAPI context
+ * @poll: polling function
+ * @weight: the poll weight of this NAPI
+ * @index: the NAPI index
+ */
+static inline void
+netif_napi_add_storage(struct net_device *dev, struct napi_struct *napi,
+ int (*poll)(struct napi_struct *, int), int weight,
+ int index)
+{
+ napi->index = index;
+ napi->napi_storage = &dev->napi_storage[index];
+ netif_napi_add_weight(dev, napi, poll, weight);
+}
+
/**
* netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
* @dev: network device
@@ -2683,6 +2715,8 @@ void __netif_napi_del(struct napi_struct *napi);
*/
static inline void netif_napi_del(struct napi_struct *napi)
{
+ napi->napi_storage = NULL;
+ napi->index = -1;
__netif_napi_del(napi);
synchronize_net();
}
diff --git a/net/core/dev.c b/net/core/dev.c
index 22c3f14d9287..ca90e8cab121 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6719,6 +6719,9 @@ void napi_enable(struct napi_struct *n)
if (n->dev->threaded && n->thread)
new |= NAPIF_STATE_THREADED;
} while (!try_cmpxchg(&n->state, &val, new));
+
+ if (n->napi_storage)
+ memset(n->napi_storage, 0, sizeof(*n->napi_storage));
}
EXPORT_SYMBOL(napi_enable);

@@ -11054,6 +11057,8 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
unsigned int txqs, unsigned int rxqs)
{
struct net_device *dev;
+ size_t napi_storage_sz;
+ unsigned int maxqs;

BUG_ON(strlen(name) >= sizeof(dev->name));

@@ -11067,6 +11072,9 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
return NULL;
}

+ WARN_ON_ONCE(txqs != rxqs);
+ maxqs = max(txqs, rxqs);
+
dev = kvzalloc(struct_size(dev, priv, sizeof_priv),
GFP_KERNEL_ACCOUNT | __GFP_RETRY_MAYFAIL);
if (!dev)
@@ -11141,6 +11149,11 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
if (!dev->ethtool)
goto free_all;

+ napi_storage_sz = array_size(maxqs, sizeof(*dev->napi_storage));
+ dev->napi_storage = kvzalloc(napi_storage_sz, GFP_KERNEL_ACCOUNT);
+ if (!dev->napi_storage)
+ goto free_all;
+
strscpy(dev->name, name);
dev->name_assign_type = name_assign_type;
dev->group = INIT_NETDEV_GROUP;
@@ -11202,6 +11215,8 @@ void free_netdev(struct net_device *dev)
list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
netif_napi_del(p);

+ kvfree(dev->napi_storage);
+
ref_tracker_dir_exit(&dev->refcnt_tracker);
#ifdef CONFIG_PCPU_DEV_REFCNT
free_percpu(dev->pcpu_refcnt);
@@ -11979,7 +11994,8 @@ static void __init net_dev_struct_check(void)
#ifdef CONFIG_NET_XGRESS
CACHELINE_ASSERT_GROUP_MEMBER(struct net_device, net_device_read_rx, tcx_ingress);
#endif
- CACHELINE_ASSERT_GROUP_SIZE(struct net_device, net_device_read_rx, 104);
+ CACHELINE_ASSERT_GROUP_MEMBER(struct net_device, net_device_read_rx, napi_storage);
+ CACHELINE_ASSERT_GROUP_SIZE(struct net_device, net_device_read_rx, 112);
}

/*
--
2.25.1