[PATCH bpf-next 0/2] xsk: XSKMAP lookup improvements

[PATCH bpf-next 0/2] xsk: XSKMAP lookup improvements

[Maciej Fijalkowski]

Hi,

This small set consists of two patches from Bjorn and myself which are
optimizing the XSKMAP lookups.  In first patch, Bjorn converts the array
of pointers to struct xdp_sock to the flexible array member of XSKMAP
itself. Second patch, based on Bjorn's work, implements the
map_gen_lookup() for XSKMAP.

Based on the XDP program from tools/lib/bpf/xsk.c where
bpf_map_lookup_elem() is explicitly called, this work yields the 5%
improvement for xdpsock's rxdrop scenario.

Thanks!


Björn Töpel (1):
  xsk: store struct xdp_sock as a flexible array member of the XSKMAP

Maciej Fijalkowski (1):
  xsk: implement map_gen_lookup() callback for XSKMAP

 kernel/bpf/xskmap.c | 74 +++++++++++++++++++++++++--------------------
 1 file changed, 42 insertions(+), 32 deletions(-)

-- 
2.20.1

[PATCH bpf-next 1/2] xsk: store struct xdp_sock as a flexible array member of the XSKMAP

[Maciej Fijalkowski]

From: Björn Töpel <bjorn.topel@intel.com>

Prior this commit, the array storing XDP socket instances were stored
in a separate allocated array of the XSKMAP. Now, we store the sockets
as a flexible array member in a similar fashion as the arraymap. Doing
so, we do less pointer chasing in the lookup.

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 kernel/bpf/xskmap.c | 57 ++++++++++++++++++++-------------------------
 1 file changed, 25 insertions(+), 32 deletions(-)

diff --git a/kernel/bpf/xskmap.c b/kernel/bpf/xskmap.c
index 82a1ffe15dfa..86e3027d4605 100644
--- a/kernel/bpf/xskmap.c
+++ b/kernel/bpf/xskmap.c
@@ -11,9 +11,9 @@
 
 struct xsk_map {
 	struct bpf_map map;
-	struct xdp_sock **xsk_map;
 	struct list_head __percpu *flush_list;
 	spinlock_t lock; /* Synchronize map updates */
+	struct xdp_sock *xsk_map[];
 };
 
 int xsk_map_inc(struct xsk_map *map)
@@ -80,9 +80,11 @@ static void xsk_map_sock_delete(struct xdp_sock *xs,
 
 static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
 {
+	int cpu, err, numa_node = bpf_map_attr_numa_node(attr);
+	struct bpf_map_memory mem;
 	struct xsk_map *m;
-	int cpu, err;
-	u64 cost;
+	u32 max_entries;
+	u64 cost, size;
 
 	if (!capable(CAP_NET_ADMIN))
 		return ERR_PTR(-EPERM);
@@ -92,44 +94,36 @@ static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
 	    attr->map_flags & ~(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY))
 		return ERR_PTR(-EINVAL);
 
-	m = kzalloc(sizeof(*m), GFP_USER);
-	if (!m)
-		return ERR_PTR(-ENOMEM);
+	max_entries = attr->max_entries;
 
-	bpf_map_init_from_attr(&m->map, attr);
-	spin_lock_init(&m->lock);
+	size = sizeof(*m) + max_entries * sizeof(m->xsk_map[0]);
+	cost = size + sizeof(struct list_head) * num_possible_cpus();
 
-	cost = (u64)m->map.max_entries * sizeof(struct xdp_sock *);
-	cost += sizeof(struct list_head) * num_possible_cpus();
+	err = bpf_map_charge_init(&mem, cost);
+	if (err < 0)
+		return ERR_PTR(err);
 
-	/* Notice returns -EPERM on if map size is larger than memlock limit */
-	err = bpf_map_charge_init(&m->map.memory, cost);
-	if (err)
-		goto free_m;
+	m = bpf_map_area_alloc(size, numa_node);
+	if (!m) {
+		bpf_map_charge_finish(&mem);
+		return ERR_PTR(-ENOMEM);
+	}
 
-	err = -ENOMEM;
+	bpf_map_init_from_attr(&m->map, attr);
+	bpf_map_charge_move(&m->map.memory, &mem);
+	spin_lock_init(&m->lock);
 
 	m->flush_list = alloc_percpu(struct list_head);
-	if (!m->flush_list)
-		goto free_charge;
+	if (!m->flush_list) {
+		bpf_map_charge_finish(&m->map.memory);
+		bpf_map_area_free(m);
+		return ERR_PTR(-ENOMEM);
+	}
 
 	for_each_possible_cpu(cpu)
 		INIT_LIST_HEAD(per_cpu_ptr(m->flush_list, cpu));
 
-	m->xsk_map = bpf_map_area_alloc(m->map.max_entries *
-					sizeof(struct xdp_sock *),
-					m->map.numa_node);
-	if (!m->xsk_map)
-		goto free_percpu;
 	return &m->map;
-
-free_percpu:
-	free_percpu(m->flush_list);
-free_charge:
-	bpf_map_charge_finish(&m->map.memory);
-free_m:
-	kfree(m);
-	return ERR_PTR(err);
 }
 
 static void xsk_map_free(struct bpf_map *map)
@@ -139,8 +133,7 @@ static void xsk_map_free(struct bpf_map *map)
 	bpf_clear_redirect_map(map);
 	synchronize_net();
 	free_percpu(m->flush_list);
-	bpf_map_area_free(m->xsk_map);
-	kfree(m);
+	bpf_map_area_free(m);
 }
 
 static int xsk_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
-- 
2.20.1

[PATCH bpf-next 2/2] xsk: implement map_gen_lookup() callback for XSKMAP

[Maciej Fijalkowski]

Inline the xsk_map_lookup_elem() via implementing the map_gen_lookup()
callback. This results in emitting the bpf instructions in place of
bpf_map_lookup_elem() helper call and better performance of bpf
programs.

Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
---
 kernel/bpf/xskmap.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/kernel/bpf/xskmap.c b/kernel/bpf/xskmap.c
index 86e3027d4605..0103405f639a 100644
--- a/kernel/bpf/xskmap.c
+++ b/kernel/bpf/xskmap.c
@@ -165,6 +165,22 @@ struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, u32 key)
 	return xs;
 }
 
+static u32 xsk_map_gen_lookup(struct bpf_map *map, struct bpf_insn *insn_buf)
+{
+	const int ret = BPF_REG_0, mp = BPF_REG_1, index = BPF_REG_2;
+	struct bpf_insn *insn = insn_buf;
+
+	*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
+	*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 5);
+	*insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(sizeof(struct xsk_sock *)));
+	*insn++ = BPF_ALU64_IMM(BPF_ADD, mp, offsetof(struct xsk_map, xsk_map));
+	*insn++ = BPF_ALU64_REG(BPF_ADD, ret, mp);
+	*insn++ = BPF_LDX_MEM(BPF_DW, ret, ret, 0);
+	*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
+	*insn++ = BPF_MOV64_IMM(ret, 0);
+	return insn - insn_buf;
+}
+
 int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp,
 		       struct xdp_sock *xs)
 {
@@ -305,6 +321,7 @@ const struct bpf_map_ops xsk_map_ops = {
 	.map_free = xsk_map_free,
 	.map_get_next_key = xsk_map_get_next_key,
 	.map_lookup_elem = xsk_map_lookup_elem,
+	.map_gen_lookup = xsk_map_gen_lookup,
 	.map_lookup_elem_sys_only = xsk_map_lookup_elem_sys_only,
 	.map_update_elem = xsk_map_update_elem,
 	.map_delete_elem = xsk_map_delete_elem,
-- 
2.20.1

Re: [PATCH bpf-next 1/2] xsk: store struct xdp_sock as a flexible array member of the XSKMAP

[Jakub Kicinski]

On Thu, 24 Oct 2019 09:19:30 +0200, Maciej Fijalkowski wrote:
> @@ -92,44 +94,36 @@ static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
>  	    attr->map_flags & ~(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY))
>  		return ERR_PTR(-EINVAL);
>  
> -	m = kzalloc(sizeof(*m), GFP_USER);
> -	if (!m)
> -		return ERR_PTR(-ENOMEM);
> +	max_entries = attr->max_entries;
>  
> -	bpf_map_init_from_attr(&m->map, attr);
> -	spin_lock_init(&m->lock);
> +	size = sizeof(*m) + max_entries * sizeof(m->xsk_map[0]);

Maybe the array_size() I suggested to Toke was disputable, but this is
such an struct_size()...

Otherwise you probably need an explicit (u64) cast?