From 8dc9e8ccbeb8ba31e64de800340cedaa6c9a8c0f Mon Sep 17 00:00:00 2001 From: Lorenzo Bianconi Date: Tue, 14 Jul 2020 15:56:38 +0200 Subject: [PATCH] UPSTREAM: bpf: cpumap: Add the possibility to attach an eBPF program to cpumap Introduce the capability to attach an eBPF program to cpumap entries. The idea behind this feature is to add the possibility to define on which CPU run the eBPF program if the underlying hw does not support RSS. Current supported verdicts are XDP_DROP and XDP_PASS. This patch has been tested on Marvell ESPRESSObin using xdp_redirect_cpu sample available in the kernel tree to identify possible performance regressions. Results show there are no observable differences in packet-per-second: $./xdp_redirect_cpu --progname xdp_cpu_map0 --dev eth0 --cpu 1 rx: 354.8 Kpps rx: 356.0 Kpps rx: 356.8 Kpps rx: 356.3 Kpps rx: 356.6 Kpps rx: 356.6 Kpps rx: 356.7 Kpps rx: 355.8 Kpps rx: 356.8 Kpps rx: 356.8 Kpps Co-developed-by: Jesper Dangaard Brouer Change-Id: Ib09a822f483667441e6b287a3ac694a82e9e3e0f Signed-off-by: Jesper Dangaard Brouer Signed-off-by: Lorenzo Bianconi Signed-off-by: Daniel Borkmann Acked-by: Jesper Dangaard Brouer Link: https://lore.kernel.org/bpf/5c9febdf903d810b3415732e5cd98491d7d9067a.1594734381.git.lorenzo@kernel.org --- include/linux/bpf.h | 6 ++ include/net/xdp.h | 5 ++ include/trace/events/xdp.h | 14 ++-- include/uapi/linux/bpf.h | 5 ++ kernel/bpf/cpumap.c | 121 +++++++++++++++++++++++++++++---- net/core/dev.c | 9 +++ tools/include/uapi/linux/bpf.h | 5 ++ 7 files changed, 148 insertions(+), 17 deletions(-) diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 676e3f6858e9..9504a16ec49c 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -1276,6 +1276,7 @@ struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); void __cpu_map_flush(void); int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, struct net_device *dev_rx); +bool cpu_map_prog_allowed(struct bpf_map *map); /* Return map's numa specified by userspace */ static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) @@ -1442,6 +1443,11 @@ static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, return 0; } +static inline bool cpu_map_prog_allowed(struct bpf_map *map) +{ + return false; +} + static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type) { diff --git a/include/net/xdp.h b/include/net/xdp.h index a0b6e79294b8..7a5f4405d251 100644 --- a/include/net/xdp.h +++ b/include/net/xdp.h @@ -103,6 +103,11 @@ struct xdp_frame { struct net_device *dev_rx; /* used by cpumap */ }; +struct xdp_cpumap_stats { + unsigned int pass; + unsigned int drop; +}; + /* Clear kernel pointers in xdp_frame */ static inline void xdp_scrub_frame(struct xdp_frame *frame) { diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h index ddb757b44f14..182b9e20cdff 100644 --- a/include/trace/events/xdp.h +++ b/include/trace/events/xdp.h @@ -177,9 +177,9 @@ DEFINE_EVENT(xdp_redirect_template, xdp_redirect_map_err, TRACE_EVENT(xdp_cpumap_kthread, TP_PROTO(int map_id, unsigned int processed, unsigned int drops, - int sched), + int sched, struct xdp_cpumap_stats *xdp_stats), - TP_ARGS(map_id, processed, drops, sched), + TP_ARGS(map_id, processed, drops, sched, xdp_stats), TP_STRUCT__entry( __field(int, map_id) @@ -188,6 +188,8 @@ TRACE_EVENT(xdp_cpumap_kthread, __field(unsigned int, drops) __field(unsigned int, processed) __field(int, sched) + __field(unsigned int, xdp_pass) + __field(unsigned int, xdp_drop) ), TP_fast_assign( @@ -197,16 +199,20 @@ TRACE_EVENT(xdp_cpumap_kthread, __entry->drops = drops; __entry->processed = processed; __entry->sched = sched; + __entry->xdp_pass = xdp_stats->pass; + __entry->xdp_drop = xdp_stats->drop; ), TP_printk("kthread" " cpu=%d map_id=%d action=%s" " processed=%u drops=%u" - " sched=%d", + " sched=%d" + " xdp_pass=%u xdp_drop=%u", __entry->cpu, __entry->map_id, __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB), __entry->processed, __entry->drops, - __entry->sched) + __entry->sched, + __entry->xdp_pass, __entry->xdp_drop) ); TRACE_EVENT(xdp_cpumap_enqueue, diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index b5230359c060..1a579b0b9379 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -227,6 +227,7 @@ enum bpf_attach_type { BPF_CGROUP_INET6_GETSOCKNAME, BPF_XDP_DEVMAP, BPF_CGROUP_INET_SOCK_RELEASE, + BPF_XDP_CPUMAP, __MAX_BPF_ATTACH_TYPE }; @@ -3867,6 +3868,10 @@ struct bpf_devmap_val { */ struct bpf_cpumap_val { __u32 qsize; /* queue size to remote target CPU */ + union { + int fd; /* prog fd on map write */ + __u32 id; /* prog id on map read */ + } bpf_prog; }; enum sk_action { diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c index 78c0386b68a4..488ec04c3523 100644 --- a/kernel/bpf/cpumap.c +++ b/kernel/bpf/cpumap.c @@ -63,6 +63,7 @@ struct bpf_cpu_map_entry { struct task_struct *kthread; struct bpf_cpumap_val value; + struct bpf_prog *prog; atomic_t refcnt; /* Control when this struct can be free'ed */ struct rcu_head rcu; @@ -82,6 +83,7 @@ static int bq_flush_to_queue(struct xdp_bulk_queue *bq); static struct bpf_map *cpu_map_alloc(union bpf_attr *attr) { + u32 value_size = attr->value_size; struct bpf_cpu_map *cmap; int err = -ENOMEM; u64 cost; @@ -92,7 +94,9 @@ static struct bpf_map *cpu_map_alloc(union bpf_attr *attr) /* check sanity of attributes */ if (attr->max_entries == 0 || attr->key_size != 4 || - attr->value_size != 4 || attr->map_flags & ~BPF_F_NUMA_NODE) + (value_size != offsetofend(struct bpf_cpumap_val, qsize) && + value_size != offsetofend(struct bpf_cpumap_val, bpf_prog.fd)) || + attr->map_flags & ~BPF_F_NUMA_NODE) return ERR_PTR(-EINVAL); cmap = kzalloc(sizeof(*cmap), GFP_USER); @@ -214,6 +218,8 @@ static void __cpu_map_ring_cleanup(struct ptr_ring *ring) static void put_cpu_map_entry(struct bpf_cpu_map_entry *rcpu) { if (atomic_dec_and_test(&rcpu->refcnt)) { + if (rcpu->prog) + bpf_prog_put(rcpu->prog); /* The queue should be empty at this point */ __cpu_map_ring_cleanup(rcpu->queue); ptr_ring_cleanup(rcpu->queue, NULL); @@ -222,6 +228,62 @@ static void put_cpu_map_entry(struct bpf_cpu_map_entry *rcpu) } } +static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu, + void **frames, int n, + struct xdp_cpumap_stats *stats) +{ + struct xdp_rxq_info rxq; + struct xdp_buff xdp; + int i, nframes = 0; + + if (!rcpu->prog) + return n; + + rcu_read_lock(); + + xdp_set_return_frame_no_direct(); + xdp.rxq = &rxq; + + for (i = 0; i < n; i++) { + struct xdp_frame *xdpf = frames[i]; + u32 act; + int err; + + rxq.dev = xdpf->dev_rx; + rxq.mem = xdpf->mem; + /* TODO: report queue_index to xdp_rxq_info */ + + xdp_convert_frame_to_buff(xdpf, &xdp); + + act = bpf_prog_run_xdp(rcpu->prog, &xdp); + switch (act) { + case XDP_PASS: + err = xdp_update_frame_from_buff(&xdp, xdpf); + if (err < 0) { + xdp_return_frame(xdpf); + stats->drop++; + } else { + frames[nframes++] = xdpf; + stats->pass++; + } + break; + default: + bpf_warn_invalid_xdp_action(act); + /* fallthrough */ + case XDP_DROP: + xdp_return_frame(xdpf); + stats->drop++; + break; + } + } + + xdp_clear_return_frame_no_direct(); + + rcu_read_unlock(); + + return nframes; +} + #define CPUMAP_BATCH 8 static int cpu_map_kthread_run(void *data) @@ -237,11 +299,12 @@ static int cpu_map_kthread_run(void *data) * kthread_stop signal until queue is empty. */ while (!kthread_should_stop() || !__ptr_ring_empty(rcpu->queue)) { + struct xdp_cpumap_stats stats = {}; /* zero stats */ + gfp_t gfp = __GFP_ZERO | GFP_ATOMIC; unsigned int drops = 0, sched = 0; void *frames[CPUMAP_BATCH]; void *skbs[CPUMAP_BATCH]; - gfp_t gfp = __GFP_ZERO | GFP_ATOMIC; - int i, n, m; + int i, n, m, nframes; /* Release CPU reschedule checks */ if (__ptr_ring_empty(rcpu->queue)) { @@ -264,8 +327,8 @@ static int cpu_map_kthread_run(void *data) * kthread CPU pinned. Lockless access to ptr_ring * consume side valid as no-resize allowed of queue. */ - n = __ptr_ring_consume_batched(rcpu->queue, frames, CPUMAP_BATCH); - + n = __ptr_ring_consume_batched(rcpu->queue, frames, + CPUMAP_BATCH); for (i = 0; i < n; i++) { void *f = frames[i]; struct page *page = virt_to_page(f); @@ -277,15 +340,19 @@ static int cpu_map_kthread_run(void *data) prefetchw(page); } - m = kmem_cache_alloc_bulk(skbuff_head_cache, gfp, n, skbs); - if (unlikely(m == 0)) { - for (i = 0; i < n; i++) - skbs[i] = NULL; /* effect: xdp_return_frame */ - drops = n; + /* Support running another XDP prog on this CPU */ + nframes = cpu_map_bpf_prog_run_xdp(rcpu, frames, n, &stats); + if (nframes) { + m = kmem_cache_alloc_bulk(skbuff_head_cache, gfp, nframes, skbs); + if (unlikely(m == 0)) { + for (i = 0; i < nframes; i++) + skbs[i] = NULL; /* effect: xdp_return_frame */ + drops += nframes; + } } local_bh_disable(); - for (i = 0; i < n; i++) { + for (i = 0; i < nframes; i++) { struct xdp_frame *xdpf = frames[i]; struct sk_buff *skb = skbs[i]; int ret; @@ -302,7 +369,7 @@ static int cpu_map_kthread_run(void *data) drops++; } /* Feedback loop via tracepoint */ - trace_xdp_cpumap_kthread(rcpu->map_id, n, drops, sched); + trace_xdp_cpumap_kthread(rcpu->map_id, n, drops, sched, &stats); local_bh_enable(); /* resched point, may call do_softirq() */ } @@ -312,13 +379,38 @@ static int cpu_map_kthread_run(void *data) return 0; } +bool cpu_map_prog_allowed(struct bpf_map *map) +{ + return map->map_type == BPF_MAP_TYPE_CPUMAP && + map->value_size != offsetofend(struct bpf_cpumap_val, qsize); +} + +static int __cpu_map_load_bpf_program(struct bpf_cpu_map_entry *rcpu, int fd) +{ + struct bpf_prog *prog; + + prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_XDP); + if (IS_ERR(prog)) + return PTR_ERR(prog); + + if (prog->expected_attach_type != BPF_XDP_CPUMAP) { + bpf_prog_put(prog); + return -EINVAL; + } + + rcpu->value.bpf_prog.id = prog->aux->id; + rcpu->prog = prog; + + return 0; +} + static struct bpf_cpu_map_entry * __cpu_map_entry_alloc(struct bpf_cpumap_val *value, u32 cpu, int map_id) { + int numa, err, i, fd = value->bpf_prog.fd; gfp_t gfp = GFP_KERNEL | __GFP_NOWARN; struct bpf_cpu_map_entry *rcpu; struct xdp_bulk_queue *bq; - int numa, err, i; /* Have map->numa_node, but choose node of redirect target CPU */ numa = cpu_to_node(cpu); @@ -360,6 +452,9 @@ __cpu_map_entry_alloc(struct bpf_cpumap_val *value, u32 cpu, int map_id) get_cpu_map_entry(rcpu); /* 1-refcnt for being in cmap->cpu_map[] */ get_cpu_map_entry(rcpu); /* 1-refcnt for kthread */ + if (fd > 0 && __cpu_map_load_bpf_program(rcpu, fd)) + goto free_ptr_ring; + /* Make sure kthread runs on a single CPU */ kthread_bind(rcpu->kthread, cpu); wake_up_process(rcpu->kthread); diff --git a/net/core/dev.c b/net/core/dev.c index e49b0655f498..dc8673c4c471 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -5205,6 +5205,8 @@ static int generic_xdp_install(struct net_device *dev, struct netdev_bpf *xdp) for (i = 0; i < new->aux->used_map_cnt; i++) { if (dev_map_can_have_prog(new->aux->used_maps[i])) return -EINVAL; + if (cpu_map_prog_allowed(new->aux->used_maps[i])) + return -EINVAL; } } @@ -8637,6 +8639,13 @@ int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, return -EINVAL; } + if (prog->expected_attach_type == BPF_XDP_CPUMAP) { + NL_SET_ERR_MSG(extack, + "BPF_XDP_CPUMAP programs can not be attached to a device"); + bpf_prog_put(prog); + return -EINVAL; + } + /* prog->aux->id may be 0 for orphaned device-bound progs */ if (prog->aux->id && prog->aux->id == prog_id) { bpf_prog_put(prog); diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 04f1a3860f0e..683d2e7725bf 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -224,6 +224,7 @@ enum bpf_attach_type { BPF_CGROUP_INET6_GETSOCKNAME, BPF_XDP_DEVMAP, BPF_CGROUP_INET_SOCK_RELEASE, + BPF_XDP_CPUMAP, __MAX_BPF_ATTACH_TYPE }; @@ -3836,6 +3837,10 @@ struct bpf_devmap_val { */ struct bpf_cpumap_val { __u32 qsize; /* queue size to remote target CPU */ + union { + int fd; /* prog fd on map write */ + __u32 id; /* prog id on map read */ + } bpf_prog; }; enum sk_action {