Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Martin KaFai Lau says:

====================
pull-request: bpf 2023-08-03

We've added 5 non-merge commits during the last 7 day(s) which contain
a total of 3 files changed, 37 insertions(+), 20 deletions(-).

The main changes are:

1) Disable preemption in perf_event_output helpers code,
   from Jiri Olsa

2) Add length check for SK_DIAG_BPF_STORAGE_REQ_MAP_FD parsing,
   from Lin Ma

3) Multiple warning splat fixes in cpumap from Hou Tao

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  bpf, cpumap: Handle skb as well when clean up ptr_ring
  bpf, cpumap: Make sure kthread is running before map update returns
  bpf: Add length check for SK_DIAG_BPF_STORAGE_REQ_MAP_FD parsing
  bpf: Disable preemption in bpf_event_output
  bpf: Disable preemption in bpf_perf_event_output
====================

Link: https://lore.kernel.org/r/20230803181429.994607-1-martin.lau@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

2 files changed, 33 insertions, 19 deletions

diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index 6ae02be7a48e..286ab3db0fde 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -28,6 +28,7 @@
 #include <linux/sched.h>
 #include <linux/workqueue.h>
 #include <linux/kthread.h>
+#include <linux/completion.h>
 #include <trace/events/xdp.h>
 #include <linux/btf_ids.h>
 
@@ -73,6 +74,7 @@ struct bpf_cpu_map_entry {
 	struct rcu_head rcu;
 
 	struct work_struct kthread_stop_wq;
+	struct completion kthread_running;
 };
 
 struct bpf_cpu_map {
@@ -129,11 +131,17 @@ static void __cpu_map_ring_cleanup(struct ptr_ring *ring)
 	 * invoked cpu_map_kthread_stop(). Catch any broken behaviour
 	 * gracefully and warn once.
 	 */
-	struct xdp_frame *xdpf;
+	void *ptr;
 
-	while ((xdpf = ptr_ring_consume(ring)))
-		if (WARN_ON_ONCE(xdpf))
-			xdp_return_frame(xdpf);
+	while ((ptr = ptr_ring_consume(ring))) {
+		WARN_ON_ONCE(1);
+		if (unlikely(__ptr_test_bit(0, &ptr))) {
+			__ptr_clear_bit(0, &ptr);
+			kfree_skb(ptr);
+			continue;
+		}
+		xdp_return_frame(ptr);
+	}
 }
 
 static void put_cpu_map_entry(struct bpf_cpu_map_entry *rcpu)
@@ -153,7 +161,6 @@ static void put_cpu_map_entry(struct bpf_cpu_map_entry *rcpu)
 static void cpu_map_kthread_stop(struct work_struct *work)
 {
 	struct bpf_cpu_map_entry *rcpu;
-	int err;
 
 	rcpu = container_of(work, struct bpf_cpu_map_entry, kthread_stop_wq);
 
@@ -163,14 +170,7 @@ static void cpu_map_kthread_stop(struct work_struct *work)
 	rcu_barrier();
 
 	/* kthread_stop will wake_up_process and wait for it to complete */
-	err = kthread_stop(rcpu->kthread);
-	if (err) {
-		/* kthread_stop may be called before cpu_map_kthread_run
-		 * is executed, so we need to release the memory related
-		 * to rcpu.
-		 */
-		put_cpu_map_entry(rcpu);
-	}
+	kthread_stop(rcpu->kthread);
 }
 
 static void cpu_map_bpf_prog_run_skb(struct bpf_cpu_map_entry *rcpu,
@@ -298,11 +298,11 @@ static int cpu_map_bpf_prog_run(struct bpf_cpu_map_entry *rcpu, void **frames,
 	return nframes;
 }
 
-
 static int cpu_map_kthread_run(void *data)
 {
 	struct bpf_cpu_map_entry *rcpu = data;
 
+	complete(&rcpu->kthread_running);
 	set_current_state(TASK_INTERRUPTIBLE);
 
 	/* When kthread gives stop order, then rcpu have been disconnected
@@ -467,6 +467,7 @@ __cpu_map_entry_alloc(struct bpf_map *map, struct bpf_cpumap_val *value,
 		goto free_ptr_ring;
 
 	/* Setup kthread */
+	init_completion(&rcpu->kthread_running);
 	rcpu->kthread = kthread_create_on_node(cpu_map_kthread_run, rcpu, numa,
 					       "cpumap/%d/map:%d", cpu,
 					       map->id);
@@ -480,6 +481,12 @@ __cpu_map_entry_alloc(struct bpf_map *map, struct bpf_cpumap_val *value,
 	kthread_bind(rcpu->kthread, cpu);
 	wake_up_process(rcpu->kthread);
 
+	/* Make sure kthread has been running, so kthread_stop() will not
+	 * stop the kthread prematurely and all pending frames or skbs
+	 * will be handled by the kthread before kthread_stop() returns.
+	 */
+	wait_for_completion(&rcpu->kthread_running);
+
 	return rcpu;
 
 free_prog:
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 5f2dcabad202..bd1a42b23f3f 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -661,8 +661,7 @@ static DEFINE_PER_CPU(int, bpf_trace_nest_level);
 BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map,
 	   u64, flags, void *, data, u64, size)
 {
-	struct bpf_trace_sample_data *sds = this_cpu_ptr(&bpf_trace_sds);
-	int nest_level = this_cpu_inc_return(bpf_trace_nest_level);
+	struct bpf_trace_sample_data *sds;
 	struct perf_raw_record raw = {
 		.frag = {
 			.size = size,
@@ -670,7 +669,11 @@ BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map,
 		},
 	};
 	struct perf_sample_data *sd;
-	int err;
+	int nest_level, err;
+
+	preempt_disable();
+	sds = this_cpu_ptr(&bpf_trace_sds);
+	nest_level = this_cpu_inc_return(bpf_trace_nest_level);
 
 	if (WARN_ON_ONCE(nest_level > ARRAY_SIZE(sds->sds))) {
 		err = -EBUSY;
@@ -688,9 +691,9 @@ BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map,
 	perf_sample_save_raw_data(sd, &raw);
 
 	err = __bpf_perf_event_output(regs, map, flags, sd);
-
 out:
 	this_cpu_dec(bpf_trace_nest_level);
+	preempt_enable();
 	return err;
 }
 
@@ -715,7 +718,6 @@ static DEFINE_PER_CPU(struct bpf_trace_sample_data, bpf_misc_sds);
 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
 		     void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy)
 {
-	int nest_level = this_cpu_inc_return(bpf_event_output_nest_level);
 	struct perf_raw_frag frag = {
 		.copy		= ctx_copy,
 		.size		= ctx_size,
@@ -732,8 +734,12 @@ u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
 	};
 	struct perf_sample_data *sd;
 	struct pt_regs *regs;
+	int nest_level;
 	u64 ret;
 
+	preempt_disable();
+	nest_level = this_cpu_inc_return(bpf_event_output_nest_level);
+
 	if (WARN_ON_ONCE(nest_level > ARRAY_SIZE(bpf_misc_sds.sds))) {
 		ret = -EBUSY;
 		goto out;
@@ -748,6 +754,7 @@ u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
 	ret = __bpf_perf_event_output(regs, map, flags, sd);
 out:
 	this_cpu_dec(bpf_event_output_nest_level);
+	preempt_enable();
 	return ret;
 }