Refactor XDPLua to reduce its footprint

include/linux/netdevice.h

@@ -3587,10 +3587,6 @@ u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
 		    enum bpf_netdev_command cmd);
 int xdp_umem_query(struct net_device *dev, u16 queue_id);
 
-#ifdef CONFIG_XDP_LUA
-int generic_xdp_lua_install_prog(char *lua_prog);
-#endif /* CONFIG_XDP_LUA */
-
 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
 bool is_skb_forwardable(const struct net_device *dev,

include/net/xdp.h

@@ -63,30 +63,13 @@ struct xdp_rxq_info {
 	struct xdp_mem_info mem;
 } ____cacheline_aligned; /* perf critical, avoid false-sharing */
 
-#ifdef CONFIG_XDP_LUA
-struct xdplua_create_work {
-	char				lua_script[8192];
-	struct lua_State	*L;
-	struct work_struct	work;
-	spinlock_t		lock;
-	bool			init;
-};
-
-DECLARE_PER_CPU(struct xdplua_create_work, luaworks);
-#endif /* CONFIG_XDP_LUA */
-
-
 struct xdp_buff {
 	void *data;
 	void *data_end;
 	void *data_meta;
 	void *data_hard_start;
 	unsigned long handle;
 	struct xdp_rxq_info *rxq;
-#ifdef CONFIG_XDP_LUA
-	struct sk_buff *skb;
-	struct lua_State *L;
-#endif /* CONFIG_XDP_LUA */
 };
 
 struct xdp_frame {

include/net/xdplua.h

@@ -0,0 +1,77 @@
+/* XDPLua internal functions
+ * Copyright (C) 2021 Victor Nogueira <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation version 2.
+ *
+ * This program is distributed "as is" WITHOUT ANY WARRANTY of any
+ * kind, whether express or implied; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+#ifndef __LINUX_NET_XDP_LUA_H__
+#define __LINUX_NET_XDP_LUA_H__
+
+#include <lua.h>
+
+#include <linux/skbuff.h>
+#include <linux/workqueue.h>
+#include <uapi/linux/if_link.h>
+#include <linux/filter.h>
+
+struct xdp_lua_work {
+	char				script[XDP_LUA_MAX_SCRIPT_LEN];
+	struct lua_State	*L;
+	struct sk_buff		*skb;
+	struct work_struct	work;
+};
+
+DECLARE_PER_CPU(struct xdp_lua_work, luaworks);
+
+#define XDP_LUA_BPF_FUNC(name)	BPF_FUNC_lua_##name
+
+#define xdp_lua_get_skb() 		(this_cpu_ptr(&luaworks)->skb)
+#define xdp_lua_set_skb(skb) 	(this_cpu_ptr(&luaworks)->skb = skb)
+
+void generic_xdp_lua_install_prog(const char *script);
+void xdp_lua_init(void);
+
+#define __BPF_LUA_MAP_0(l, m, v, ...) l
+#define __BPF_LUA_MAP_1(l, m, v, t, a, ...) l, __BPF_MAP_1(m, v, t, a, __VA_ARGS__)
+#define __BPF_LUA_MAP_2(l, m, v, t, a, ...) l, __BPF_MAP_2(m, v, t, a, __VA_ARGS__)
+#define __BPF_LUA_MAP_3(l, m, v, t, a, ...) l, __BPF_MAP_3(m, v, t, a, __VA_ARGS__)
+#define __BPF_LUA_MAP_4(l, m, v, t, a, ...) l, __BPF_MAP_4(m, v, t, a, __VA_ARGS__)
+#define __BPF_LUA_MAP_5(l, m, v, t, a, ...) l, __BPF_MAP_5(m, v, t, a, __VA_ARGS__)
+
+#define __BPF_LUA_MAP(n, l, ...) __BPF_LUA_MAP_##n(l, __VA_ARGS__)
+
+#define LST_TYPE	lua_State
+#define LST_NAME	L
+#define LST_DECL	__BPF_DECL_ARGS(LST_TYPE*, LST_NAME)
+
+#define BPF_LUA_CALL_x(x, name, ...)					\
+	static __always_inline						       \
+	u64 ____##name(__BPF_LUA_MAP(x, LST_DECL, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__)); \
+	u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__));	       \
+	u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__))	       \
+	{								       \
+		int ret = -ENOENT;					\
+		lua_State *L = this_cpu_ptr(&luaworks)->L;		\
+		WARN_ON(!L);							\
+		if (!L)								\
+			return ret;						\
+		ret = ____##name(__BPF_LUA_MAP(x, LST_NAME, __BPF_CAST, __BPF_N, __VA_ARGS__));	\
+		return ret;								\
+	}								       \
+	static __always_inline						       \
+	u64 ____##name(__BPF_LUA_MAP(x, LST_DECL, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__))
+
+#define BPF_LUA_CALL_0(name, ...)	BPF_LUA_CALL_x(0, name, __VA_ARGS__)
+#define BPF_LUA_CALL_1(name, ...)	BPF_LUA_CALL_x(1, name, __VA_ARGS__)
+#define BPF_LUA_CALL_2(name, ...)	BPF_LUA_CALL_x(2, name, __VA_ARGS__)
+#define BPF_LUA_CALL_3(name, ...)	BPF_LUA_CALL_x(3, name, __VA_ARGS__)
+#define BPF_LUA_CALL_4(name, ...)	BPF_LUA_CALL_x(4, name, __VA_ARGS__)
+#define BPF_LUA_CALL_5(name, ...)	BPF_LUA_CALL_x(5, name, __VA_ARGS__)
+
+#endif /* __LINUX_NET_XDP_LUA_H__ */

include/uapi/linux/if_link.h

@@ -927,6 +927,7 @@ enum {
 #define XDP_FLAGS_MASK			(XDP_FLAGS_UPDATE_IF_NOEXIST | \
 					 XDP_FLAGS_MODES)
 
+#define XDP_LUA_MAX_SCRIPT_LEN			8192
 /* These are stored into IFLA_XDP_ATTACHED on dump. */
 enum {
 	XDP_ATTACHED_NONE = 0,

net/core/Makefile

@@ -8,14 +8,15 @@ obj-y := sock.o request_sock.o skbuff.o datagram.o stream.o scm.o \
 
 obj-$(CONFIG_SYSCTL) += sysctl_net_core.o
 
-CFLAGS_dev.o = -Ilib/lunatik/lua/ -D_KERNEL \
-		-Ilib/luadata/
+CFLAGS_dev.o = -Ilib/lunatik/lua/ -D_KERNEL
 CFLAGS_filter.o = -Ilib/lunatik/lua/ -D_KERNEL \
 		-Ilib/luadata/ -Ilib/luaunpack/
+CFLAGS_rtnetlink.o = -Ilib/lunatik/lua/ -D_KERNEL
+CFLAGS_xdplua.o = -Ilib/lunatik/lua/ -Ilib/luadata -D_KERNEL
 obj-y		     += dev.o ethtool.o dev_addr_lists.o dst.o netevent.o \
 			neighbour.o rtnetlink.o utils.o link_watch.o filter.o \
 			sock_diag.o dev_ioctl.o tso.o sock_reuseport.o \
-			fib_notifier.o xdp.o
+			fib_notifier.o xdp.o xdplua.o
 
 obj-y += net-sysfs.o
 obj-$(CONFIG_PAGE_POOL) += page_pool.o

net/core/dev.c

@@ -72,14 +72,6 @@
  *				        - netif_rx() feedback
  */
 
-#ifdef CONFIG_XDP_LUA
-#include <lua.h>
-#include <lauxlib.h>
-#include <lstate.h>
-#include <lualib.h>
-#include <luadata.h>
-#endif /* CONFIG_XDP_LUA */
-
 #include <linux/uaccess.h>
 #include <linux/bitops.h>
 #include <linux/capability.h>
@@ -164,8 +156,9 @@
 static DEFINE_SPINLOCK(ptype_lock);
 static DEFINE_SPINLOCK(offload_lock);
 #ifdef CONFIG_XDP_LUA
-DEFINE_PER_CPU(struct xdplua_create_work, luaworks);
-#endif
+#include <net/xdplua.h>
+#undef free
+#endif /* CONFIG_XDP_LUA */
 struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
 struct list_head ptype_all __read_mostly;	/* Taps */
 static struct list_head offload_base __read_mostly;
@@ -4290,9 +4283,6 @@ static u32 netif_receive_generic_xdp(struct sk_buff *skb,
 	u32 metalen, act = XDP_DROP;
 	int hlen, off;
 	u32 mac_len;
-#ifdef CONFIG_XDP_LUA
-	struct xdplua_create_work *lw;
-#endif /* CONFIG_XDP_LUA */
 
 	/* Reinjected packets coming from act_mirred or similar should
 	 * not get XDP generic processing.
@@ -4335,21 +4325,11 @@ static u32 netif_receive_generic_xdp(struct sk_buff *skb,
 	rxqueue = netif_get_rxqueue(skb);
 	xdp->rxq = &rxqueue->xdp_rxq;
 #ifdef CONFIG_XDP_LUA
-	lw = this_cpu_ptr(&luaworks);
-
-	xdp->skb = skb;
-	xdp->L = lw->L;
+	xdp_lua_set_skb(skb);
 #endif /* CONFIG_XDP_LUA */
 
 	act = bpf_prog_run_xdp(xdp_prog, xdp);
 
-#ifdef CONFIG_XDP_LUA
-	if (lw->init) {
-		lw->init = false;
-		spin_unlock(&lw->lock);
-	}
-#endif /* CONFIG_XDP_LUA */
-
 	off = xdp->data - orig_data;
 	if (off > 0)
 		__skb_pull(skb, off);
@@ -5115,35 +5095,6 @@ static int generic_xdp_install(struct net_device *dev, struct netdev_bpf *xdp)
 	return ret;
 }
 
-#ifdef CONFIG_XDP_LUA
-
-static void per_cpu_xdp_lua_install(struct work_struct *w) {
-	int this_cpu = smp_processor_id();
-	struct xdplua_create_work *lw =
-		container_of(w, struct xdplua_create_work, work);
-
-	spin_lock_bh(&lw->lock);
-	if (luaL_dostring(lw->L, lw->lua_script)) {
-		pr_err(KERN_INFO "error: %s\nOn cpu: %d\n",
-			lua_tostring(lw->L, -1), this_cpu);
-	}
-	spin_unlock_bh(&lw->lock);
-}
-
-int generic_xdp_lua_install_prog(char *lua_script)
-{
-	int cpu;
-	struct xdplua_create_work *lw;
-
-	for_each_possible_cpu(cpu) {
-		lw = per_cpu_ptr(&luaworks, cpu);
-		strcpy(lw->lua_script, lua_script);
-		schedule_work_on(cpu, &lw->work);
-	}
-	return 0;
-}
-#endif /* CONFIG_XDP_LUA */
-
 static int netif_receive_skb_internal(struct sk_buff *skb)
 {
 	int ret;
@@ -9651,9 +9602,6 @@ static int __init net_dev_init(void)
 	for_each_possible_cpu(i) {
 		struct work_struct *flush = per_cpu_ptr(&flush_works, i);
 		struct softnet_data *sd = &per_cpu(softnet_data, i);
-#ifdef CONFIG_XDP_LUA
-		struct xdplua_create_work *lw = per_cpu_ptr(&luaworks, i);
-#endif
 
 		INIT_WORK(flush, flush_backlog);
 
@@ -9673,19 +9621,6 @@ static int __init net_dev_init(void)
 		init_gro_hash(&sd->backlog);
 		sd->backlog.poll = process_backlog;
 		sd->backlog.weight = weight_p;
-#ifdef CONFIG_XDP_LUA
-		lw->L = luaL_newstate();
-		WARN_ON(!lw->L);
-
-		if (!lw->L)
-			continue;
-
-		luaL_openlibs(lw->L);
-		luaL_requiref(lw->L, "data", luaopen_data, 1);
-		lua_pop(lw->L, 1);
-
-		INIT_WORK(&lw->work, per_cpu_xdp_lua_install);
-#endif /* CONFIG_XDP_LUA */
 	}
 
 	dev_boot_phase = 0;
@@ -9710,6 +9645,9 @@ static int __init net_dev_init(void)
 
 	rc = cpuhp_setup_state_nocalls(CPUHP_NET_DEV_DEAD, "net/dev:dead",
 				       NULL, dev_cpu_dead);
+#ifdef CONFIG_XDP_LUA
+	xdp_lua_init();
+#endif /* CONFIG_XDP_LUA */
 	WARN_ON(rc < 0);
 	rc = 0;
 out: