[bpf-next,v3] net: Add a warning if NAPI cb missed xdp_do_flush().

Commit Message

Sebastian Andrzej Siewior Oct. 7, 2023, 3:43 p.m. UTC

A few drivers were missing a xdp_do_flush() invocation after
XDP_REDIRECT.

Add three helper functions each for one of the per-CPU lists. Return
true if the per-CPU list is non-empty and flush the list.
Add xdp_do_check_flushed() which invokes each helper functions and
creats a warning if one of the functions had a non-empty list.
Hide everything behind CONFIG_DEBUG_NET.

Suggested-by: Jesper Dangaard Brouer <[e-mail]>
Reviewed-by: Toke Høiland-Jørgensen <[e-mail]>
Acked-by: Jakub Kicinski <[e-mail]>
Signed-off-by: Sebastian Andrzej Siewior <[e-mail]>
---
v2…v3:
  - Collected Reviewed/Acked from the list.
  - Added an include dev.h to filter.c, the robot pointed out a missing
    prototype.

v1…v2:
  - Moved xdp_do_check_flushed() to net/core/dev.h.
  - Stripped __ from function names.
  - Removed empty lines within an ifdef block.
  - xdp_do_check_flushed() is now behind CONFIG_DEBUG_NET &&
    CONFIG_BPF_SYSCALL. dev_check_flush and cpu_map_check_flush are now
    only behind CONFIG_DEBUG_NET. They have no empty inline function for
    the !CONFIG_DEBUG_NET case since they are only called in
    CONFIG_DEBUG_NET case.

 include/linux/bpf.h    |  3 +++
 include/net/xdp_sock.h |  9 +++++++++
 kernel/bpf/cpumap.c    | 10 ++++++++++
 kernel/bpf/devmap.c    | 10 ++++++++++
 net/core/dev.c         |  2 ++
 net/core/dev.h         |  6 ++++++
 net/core/filter.c      | 16 ++++++++++++++++
 net/xdp/xsk.c          | 10 ++++++++++
 8 files changed, 66 insertions(+)

Patch

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index a34ac7f00c86c..584adabd411fc 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -2478,6 +2478,9 @@  void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
 		     enum bpf_dynptr_type type, u32 offset, u32 size);
 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
+
+bool dev_check_flush(void);
+bool cpu_map_check_flush(void);
 #else /* !CONFIG_BPF_SYSCALL */
 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
 {
diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index 69b472604b86f..7dd0df2f6f8e6 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -109,4 +109,13 @@  static inline void __xsk_map_flush(void)
 
 #endif /* CONFIG_XDP_SOCKETS */
 
+#if defined(CONFIG_XDP_SOCKETS) && defined(CONFIG_DEBUG_NET)
+bool xsk_map_check_flush(void);
+#else
+static inline bool xsk_map_check_flush(void)
+{
+	return false;
+}
+#endif
+
 #endif /* _LINUX_XDP_SOCK_H */
diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index e42a1bdb7f536..8a0bb80fe48a3 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -764,6 +764,16 @@  void __cpu_map_flush(void)
 	}
 }
 
+#ifdef CONFIG_DEBUG_NET
+bool cpu_map_check_flush(void)
+{
+	if (list_empty(this_cpu_ptr(&cpu_map_flush_list)))
+		return false;
+	__cpu_map_flush();
+	return true;
+}
+#endif
+
 static int __init cpu_map_init(void)
 {
 	int cpu;
diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index 4d42f6ed6c11a..a936c704d4e77 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c
@@ -418,6 +418,16 @@  void __dev_flush(void)
 	}
 }
 
+#ifdef CONFIG_DEBUG_NET
+bool dev_check_flush(void)
+{
+	if (list_empty(this_cpu_ptr(&dev_flush_list)))
+		return false;
+	__dev_flush();
+	return true;
+}
+#endif
+
 /* Elements are kept alive by RCU; either by rcu_read_lock() (from syscall) or
  * by local_bh_disable() (from XDP calls inside NAPI). The
  * rcu_read_lock_bh_held() below makes lockdep accept both.
diff --git a/net/core/dev.c b/net/core/dev.c
index 606a366cc2095..9273b12ecf6fa 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6526,6 +6526,8 @@  static int __napi_poll(struct napi_struct *n, bool *repoll)
 	if (test_bit(NAPI_STATE_SCHED, &n->state)) {
 		work = n->poll(n, weight);
 		trace_napi_poll(n, work, weight);
+
+		xdp_do_check_flushed(n);
 	}
 
 	if (unlikely(work > weight))
diff --git a/net/core/dev.h b/net/core/dev.h
index e075e198092cc..f66125857af77 100644
--- a/net/core/dev.h
+++ b/net/core/dev.h
@@ -136,4 +136,10 @@  static inline void netif_set_gro_ipv4_max_size(struct net_device *dev,
 }
 
 int rps_cpumask_housekeeping(struct cpumask *mask);
+
+#if defined(CONFIG_DEBUG_NET) && defined(CONFIG_BPF_SYSCALL)
+void xdp_do_check_flushed(struct napi_struct *napi);
+#else
+static inline void xdp_do_check_flushed(struct napi_struct *napi) { }
+#endif
 #endif
diff --git a/net/core/filter.c b/net/core/filter.c
index a094694899c99..af2d34d5e1815 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -82,6 +82,8 @@ 
 #include <net/mptcp.h>
 #include <net/netfilter/nf_conntrack_bpf.h>
 
+#include "dev.h"
+
 static const struct bpf_func_proto *
 bpf_sk_base_func_proto(enum bpf_func_id func_id);
 
@@ -4207,6 +4209,20 @@  void xdp_do_flush(void)
 }
 EXPORT_SYMBOL_GPL(xdp_do_flush);
 
+#if defined(CONFIG_DEBUG_NET) && defined(CONFIG_BPF_SYSCALL)
+void xdp_do_check_flushed(struct napi_struct *napi)
+{
+	bool ret;
+
+	ret = dev_check_flush();
+	ret |= cpu_map_check_flush();
+	ret |= xsk_map_check_flush();
+
+	WARN_ONCE(ret, "Missing xdp_do_flush() invocation after NAPI by %ps\n",
+		  napi->poll);
+}
+#endif
+
 void bpf_clear_redirect_map(struct bpf_map *map)
 {
 	struct bpf_redirect_info *ri;
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index f5e96e0d6e01d..ba070fd37d244 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -391,6 +391,16 @@  void __xsk_map_flush(void)
 	}
 }
 
+#ifdef CONFIG_DEBUG_NET
+bool xsk_map_check_flush(void)
+{
+	if (list_empty(this_cpu_ptr(&xskmap_flush_list)))
+		return false;
+	__xsk_map_flush();
+	return true;
+}
+#endif
+
 void xsk_tx_completed(struct xsk_buff_pool *pool, u32 nb_entries)
 {
 	xskq_prod_submit_n(pool->cq, nb_entries);