Merge branch 'ftrace-bpf-use-single-direct-ops-for-bpf-trampolines'

Jiri Olsa says:

====================
ftrace,bpf: Use single direct ops for bpf trampolines

hi,
while poking the multi-tracing interface I ended up with just one ftrace_ops
object to attach all trampolines.

This change allows to use less direct API calls during the attachment changes
in the future code, so in effect speeding up the attachment.

In current code we get a speed up from using just a single ftrace_ops object.

- with current code:

  Performance counter stats for 'bpftrace -e fentry:vmlinux:ksys_* {} -c true':

     6,364,157,902      cycles:k
       828,728,902      cycles:u
     1,064,803,824      instructions:u                   #    1.28  insn per cycle
    23,797,500,067      instructions:k                   #    3.74  insn per cycle

       4.416004987 seconds time elapsed

       0.164121000 seconds user
       1.289550000 seconds sys

- with the fix:

   Performance counter stats for 'bpftrace -e fentry:vmlinux:ksys_* {} -c true':

     6,535,857,905      cycles:k
       810,809,429      cycles:u
     1,064,594,027      instructions:u                   #    1.31  insn per cycle
    23,962,552,894      instructions:k                   #    3.67  insn per cycle

       1.666961239 seconds time elapsed

       0.157412000 seconds user
       1.283396000 seconds sys

The speedup seems to be related to the fact that with single ftrace_ops object
we don't call ftrace_shutdown anymore (we use ftrace_update_ops instead) and
we skip the synchronize rcu calls (each ~100ms) at the end of that function.

rfc: https://lore.kernel.org/bpf/20250729102813.1531457-1-jolsa@kernel.org/
v1:  https://lore.kernel.org/bpf/20250923215147.1571952-1-jolsa@kernel.org/
v2:  https://lore.kernel.org/bpf/20251113123750.2507435-1-jolsa@kernel.org/
v3:  https://lore.kernel.org/bpf/20251120212402.466524-1-jolsa@kernel.org/
v4:  https://lore.kernel.org/bpf/20251203082402.78816-1-jolsa@kernel.org/
v5:  https://lore.kernel.org/bpf/20251215211402.353056-10-jolsa@kernel.org/

v6 changes:
- rename add_hash_entry_direct to add_ftrace_hash_entry_direct [Steven]
- factor hash_add/hash_sub [Steven]
- add kerneldoc header for update_ftrace_direct_* functions [Steven]
- few assorted smaller fixes [Steven]
- added missing direct_ops wrappers for !CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
  case [Steven]

v5 changes:
- do not export ftrace_hash object [Steven]
- fix update_ftrace_direct_add new_filter_hash leak [ci]

v4 changes:
- rebased on top of bpf-next/master (with jmp attach changes)
  added patch 1 to deal with that
- added extra checks for update_ftrace_direct_del/mod to address
  the ci bot review

v3 changes:
- rebased on top of bpf-next/master
- fixed update_ftrace_direct_del cleanup path
- added missing inline to update_ftrace_direct_* stubs

v2 changes:
- rebased on top fo bpf-next/master plus Song's livepatch fixes [1]
- renamed the API functions [2] [Steven]
- do not export the new api [Steven]
- kept the original direct interface:

  I'm not sure if we want to melt both *_ftrace_direct and the new interface
  into single one. It's bit different in semantic (hence the name change as
  Steven suggested [2]) and I don't think the changes are not that big so
  we could easily keep both APIs.

v1 changes:
- make the change x86 specific, after discussing with Mark options for
  arm64 [Mark]

thanks,
jirka

[1] https://lore.kernel.org/bpf/20251027175023.1521602-1-song@kernel.org/
[2] https://lore.kernel.org/bpf/20250924050415.4aefcb91@batman.local.home/
---
====================

Link: https://patch.msgid.link/20251230145010.103439-1-jolsa@kernel.org
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>

2 files changed, 34 insertions, 4 deletions

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 9272a237cced..5524f9429e76 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -1329,14 +1329,17 @@ struct bpf_tramp_image {
 };
 
 struct bpf_trampoline {
-	/* hlist for trampoline_table */
-	struct hlist_node hlist;
+	/* hlist for trampoline_key_table */
+	struct hlist_node hlist_key;
+	/* hlist for trampoline_ip_table */
+	struct hlist_node hlist_ip;
 	struct ftrace_ops *fops;
 	/* serializes access to fields of this trampoline */
 	struct mutex mutex;
 	refcount_t refcnt;
 	u32 flags;
 	u64 key;
+	unsigned long ip;
 	struct {
 		struct btf_func_model model;
 		void *addr;
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index a3a8989e3268..705db0a6d995 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -82,6 +82,7 @@ static inline void early_trace_init(void) { }
 
 struct module;
 struct ftrace_hash;
+struct ftrace_func_entry;
 
 #if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_MODULES) && \
 	defined(CONFIG_DYNAMIC_FTRACE)
@@ -359,7 +360,6 @@ enum {
 	FTRACE_OPS_FL_DIRECT			= BIT(17),
 	FTRACE_OPS_FL_SUBOP			= BIT(18),
 	FTRACE_OPS_FL_GRAPH			= BIT(19),
-	FTRACE_OPS_FL_JMP			= BIT(20),
 };
 
 #ifndef CONFIG_DYNAMIC_FTRACE_WITH_ARGS
@@ -403,9 +403,17 @@ enum ftrace_ops_cmd {
  *        Negative on failure. The return value is dependent on the
  *        callback.
  */
-typedef int (*ftrace_ops_func_t)(struct ftrace_ops *op, enum ftrace_ops_cmd cmd);
+typedef int (*ftrace_ops_func_t)(struct ftrace_ops *op, unsigned long ip, enum ftrace_ops_cmd cmd);
 
 #ifdef CONFIG_DYNAMIC_FTRACE
+
+#define FTRACE_HASH_DEFAULT_BITS 10
+
+struct ftrace_hash *alloc_ftrace_hash(int size_bits);
+void free_ftrace_hash(struct ftrace_hash *hash);
+struct ftrace_func_entry *add_ftrace_hash_entry_direct(struct ftrace_hash *hash,
+						       unsigned long ip, unsigned long direct);
+
 /* The hash used to know what functions callbacks trace */
 struct ftrace_ops_hash {
 	struct ftrace_hash __rcu	*notrace_hash;
@@ -535,6 +543,10 @@ int unregister_ftrace_direct(struct ftrace_ops *ops, unsigned long addr,
 int modify_ftrace_direct(struct ftrace_ops *ops, unsigned long addr);
 int modify_ftrace_direct_nolock(struct ftrace_ops *ops, unsigned long addr);
 
+int update_ftrace_direct_add(struct ftrace_ops *ops, struct ftrace_hash *hash);
+int update_ftrace_direct_del(struct ftrace_ops *ops, struct ftrace_hash *hash);
+int update_ftrace_direct_mod(struct ftrace_ops *ops, struct ftrace_hash *hash, bool do_direct_lock);
+
 void ftrace_stub_direct_tramp(void);
 
 #else
@@ -561,6 +573,21 @@ static inline int modify_ftrace_direct_nolock(struct ftrace_ops *ops, unsigned l
 	return -ENODEV;
 }
 
+static inline int update_ftrace_direct_add(struct ftrace_ops *ops, struct ftrace_hash *hash)
+{
+	return -ENODEV;
+}
+
+static inline int update_ftrace_direct_del(struct ftrace_ops *ops, struct ftrace_hash *hash)
+{
+	return -ENODEV;
+}
+
+static inline int update_ftrace_direct_mod(struct ftrace_ops *ops, struct ftrace_hash *hash, bool do_direct_lock)
+{
+	return -ENODEV;
+}
+
 /*
  * This must be implemented by the architecture.
  * It is the way the ftrace direct_ops helper, when called