[PATCHv6 bpf-next,perf/core 00/16] uprobe, bpf: Add session support

[PATCHv6 bpf-next,perf/core 00/16] uprobe, bpf: Add session support

[Jiri Olsa]

hi,
this patchset is adding support for session uprobe attachment and
using it through bpf link for bpf programs.

The session means that the uprobe consumer is executed on entry
and return of probed function with additional control:
  - entry callback can control execution of the return callback
  - entry and return callbacks can share data/cookie

On more details please see patch #2.

The patchset is based on Peter's perf/core [1] tree merged in bpf-next/master.

There's an proposal from Andrii how to get this merged in [2]:

> I think uprobe parts should stay in tip/perf/core (if that's where all
> uprobe code goes in), as we have a bunch of ongoing work that all will
> conflict a bit with each other, if it lands across multiple trees.
> 
> So that means that patches #1 and #2 ideally land in tip/perf/core.
> But you have a lot of BPF-specific things that would be inconvenient
> to route through tip, so I'd say those should go through bpf-next.
> 
> What we can do, if Ingo and Peter are OK with that, is to create a
> stable (non-rebaseable) branch off of your first two patches (applied
> in tip/perf/core), which we'll merge into bpf-next/master and land the
> rest of your patch set there. We've done that with recent struct fd
> changes, and there were few other similar cases in the past, and that
> all worked well.
> 
> Peter, Ingo, are you guys OK with that approach?


v6 changes:
  - added acks [Andrii, Oleg]
  - added missing Fixes tags
  - added fix to force uprobe bpf program to always return 0 [Andrii]
  - separated kprobe session verifier check for return value check
    and squashed similar uprobe session fix in patch 5 [Andrii]
  - move session return handler check for cookie data to handle_uretprobe_chain [Andrii]
  - added threads the session test to speed it up
  - several smaller fixes [Andrii]

thanks,
jirka


[1] git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core
[2] https://lore.kernel.org/bpf/CAEf4BzY8tGCstcD4BVBLPd0V92p--b_vUmQyWydObRJHZPgCLA@mail.gmail.com/
---
Jiri Olsa (16):
      uprobe: Add data pointer to consumer handlers
      uprobe: Add support for session consumer
      bpf: Allow return values 0 and 1 for kprobe session
      bpf: Force uprobe bpf program to always return 0
      bpf: Add support for uprobe multi session attach
      bpf: Add support for uprobe multi session context
      libbpf: Add support for uprobe multi session attach
      selftests/bpf: Add uprobe session test
      selftests/bpf: Add uprobe session cookie test
      selftests/bpf: Add uprobe session recursive test
      selftests/bpf: Add uprobe session verifier test for return value
      selftests/bpf: Add kprobe session verifier test for return value
      selftests/bpf: Add uprobe session single consumer test
      selftests/bpf: Scale down uprobe multi consumer test
      selftests/bpf: Add uprobe sessions to consumer test
      selftests/bpf: Add threads to consumer test

 include/linux/uprobes.h                                            |  25 ++++++-
 include/uapi/linux/bpf.h                                           |   1 +
 kernel/bpf/syscall.c                                               |   9 ++-
 kernel/bpf/verifier.c                                              |  10 +++
 kernel/events/uprobes.c                                            | 148 +++++++++++++++++++++++++++++--------
 kernel/trace/bpf_trace.c                                           |  63 +++++++++++-----
 kernel/trace/trace_uprobe.c                                        |  12 ++-
 tools/include/uapi/linux/bpf.h                                     |   1 +
 tools/lib/bpf/bpf.c                                                |   1 +
 tools/lib/bpf/libbpf.c                                             |  19 ++++-
 tools/lib/bpf/libbpf.h                                             |   4 +-
 tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c              |   2 +-
 tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c         |   2 +
 tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c         | 336 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++----------
 tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c          |  31 ++++++++
 tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c         |   6 +-
 tools/testing/selftests/bpf/progs/uprobe_multi_session.c           |  71 ++++++++++++++++++
 tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c    |  48 ++++++++++++
 tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c |  44 +++++++++++
 tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c    |  44 +++++++++++
 tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c          |  31 ++++++++
 21 files changed, 808 insertions(+), 100 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session.c
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c

[PATCHv6 perf/core 01/16] uprobe: Add data pointer to consumer handlers

[Jiri Olsa]

Adding data pointer to both entry and exit consumer handlers and all
its users. The functionality itself is coming in following change.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/linux/uprobes.h                              |  4 ++--
 kernel/events/uprobes.c                              |  4 ++--
 kernel/trace/bpf_trace.c                             |  6 ++++--
 kernel/trace/trace_uprobe.c                          | 12 ++++++++----
 .../testing/selftests/bpf/bpf_testmod/bpf_testmod.c  |  2 +-
 5 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
index 2b294bf1881f..bb265a632b91 100644
--- a/include/linux/uprobes.h
+++ b/include/linux/uprobes.h
@@ -37,10 +37,10 @@ struct uprobe_consumer {
 	 * for the current process. If filter() is omitted or returns true,
 	 * UPROBE_HANDLER_REMOVE is effectively ignored.
 	 */
-	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs);
+	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs, __u64 *data);
 	int (*ret_handler)(struct uprobe_consumer *self,
 				unsigned long func,
-				struct pt_regs *regs);
+				struct pt_regs *regs, __u64 *data);
 	bool (*filter)(struct uprobe_consumer *self, struct mm_struct *mm);
 
 	struct list_head cons_node;
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 2a0059464383..6b44c386a5df 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -2090,7 +2090,7 @@ static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
 		int rc = 0;
 
 		if (uc->handler) {
-			rc = uc->handler(uc, regs);
+			rc = uc->handler(uc, regs, NULL);
 			WARN(rc & ~UPROBE_HANDLER_MASK,
 				"bad rc=0x%x from %ps()\n", rc, uc->handler);
 		}
@@ -2128,7 +2128,7 @@ handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs)
 	rcu_read_lock_trace();
 	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
 		if (uc->ret_handler)
-			uc->ret_handler(uc, ri->func, regs);
+			uc->ret_handler(uc, ri->func, regs, NULL);
 	}
 	rcu_read_unlock_trace();
 }
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index a582cd25ca87..fdab7ecd8dfa 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -3244,7 +3244,8 @@ uprobe_multi_link_filter(struct uprobe_consumer *con, struct mm_struct *mm)
 }
 
 static int
-uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
+uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs,
+			  __u64 *data)
 {
 	struct bpf_uprobe *uprobe;
 
@@ -3253,7 +3254,8 @@ uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
 }
 
 static int
-uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs)
+uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs,
+			      __u64 *data)
 {
 	struct bpf_uprobe *uprobe;
 
diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
index c40531d2cbad..5895eabe3581 100644
--- a/kernel/trace/trace_uprobe.c
+++ b/kernel/trace/trace_uprobe.c
@@ -89,9 +89,11 @@ static struct trace_uprobe *to_trace_uprobe(struct dyn_event *ev)
 static int register_uprobe_event(struct trace_uprobe *tu);
 static int unregister_uprobe_event(struct trace_uprobe *tu);
 
-static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs);
+static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
+			     __u64 *data);
 static int uretprobe_dispatcher(struct uprobe_consumer *con,
-				unsigned long func, struct pt_regs *regs);
+				unsigned long func, struct pt_regs *regs,
+				__u64 *data);
 
 #ifdef CONFIG_STACK_GROWSUP
 static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
@@ -1517,7 +1519,8 @@ trace_uprobe_register(struct trace_event_call *event, enum trace_reg type,
 	}
 }
 
-static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
+static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
+			     __u64 *data)
 {
 	struct trace_uprobe *tu;
 	struct uprobe_dispatch_data udd;
@@ -1548,7 +1551,8 @@ static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
 }
 
 static int uretprobe_dispatcher(struct uprobe_consumer *con,
-				unsigned long func, struct pt_regs *regs)
+				unsigned long func, struct pt_regs *regs,
+				__u64 *data)
 {
 	struct trace_uprobe *tu;
 	struct uprobe_dispatch_data udd;
diff --git a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
index 8835761d9a12..12005e3dc3e4 100644
--- a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
+++ b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
@@ -461,7 +461,7 @@ static struct bin_attribute bin_attr_bpf_testmod_file __ro_after_init = {
 
 static int
 uprobe_ret_handler(struct uprobe_consumer *self, unsigned long func,
-		   struct pt_regs *regs)
+		   struct pt_regs *regs, __u64 *data)
 
 {
 	regs->ax  = 0x12345678deadbeef;
-- 
2.46.2

[PATCHv6 perf/core 02/16] uprobe: Add support for session consumer

[Jiri Olsa]

This change allows the uprobe consumer to behave as session which
means that 'handler' and 'ret_handler' callbacks are connected in
a way that allows to:

  - control execution of 'ret_handler' from 'handler' callback
  - share data between 'handler' and 'ret_handler' callbacks

The session concept fits to our common use case where we do filtering
on entry uprobe and based on the result we decide to run the return
uprobe (or not).

It's also convenient to share the data between session callbacks.

To achive this we are adding new return value the uprobe consumer
can return from 'handler' callback:

  UPROBE_HANDLER_IGNORE
  - Ignore 'ret_handler' callback for this consumer.

And store cookie and pass it to 'ret_handler' when consumer has both
'handler' and 'ret_handler' callbacks defined.

We store shared data in the return_consumer object array as part of
the return_instance object. This way the handle_uretprobe_chain can
find related return_consumer and its shared data.

We also store entry handler return value, for cases when there are
multiple consumers on single uprobe and some of them are ignored and
some of them not, in which case the return probe gets installed and
we need to have a way to find out which consumer needs to be ignored.

The tricky part is when consumer is registered 'after' the uprobe
entry handler is hit. In such case this consumer's 'ret_handler' gets
executed as well, but it won't have the proper data pointer set,
so we can filter it out.

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Suggested-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/linux/uprobes.h |  21 +++++-
 kernel/events/uprobes.c | 148 ++++++++++++++++++++++++++++++++--------
 2 files changed, 139 insertions(+), 30 deletions(-)

diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
index bb265a632b91..dbaf04189548 100644
--- a/include/linux/uprobes.h
+++ b/include/linux/uprobes.h
@@ -23,8 +23,17 @@ struct inode;
 struct notifier_block;
 struct page;
 
+/*
+ * Allowed return values from uprobe consumer's handler callback
+ * with following meaning:
+ *
+ * UPROBE_HANDLER_REMOVE
+ * - Remove the uprobe breakpoint from current->mm.
+ * UPROBE_HANDLER_IGNORE
+ * - Ignore ret_handler callback for this consumer.
+ */
 #define UPROBE_HANDLER_REMOVE		1
-#define UPROBE_HANDLER_MASK		1
+#define UPROBE_HANDLER_IGNORE		2
 
 #define MAX_URETPROBE_DEPTH		64
 
@@ -44,6 +53,8 @@ struct uprobe_consumer {
 	bool (*filter)(struct uprobe_consumer *self, struct mm_struct *mm);
 
 	struct list_head cons_node;
+
+	__u64 id;	/* set when uprobe_consumer is registered */
 };
 
 #ifdef CONFIG_UPROBES
@@ -83,14 +94,22 @@ struct uprobe_task {
 	unsigned int			depth;
 };
 
+struct return_consumer {
+	__u64	cookie;
+	__u64	id;
+};
+
 struct return_instance {
 	struct uprobe		*uprobe;
 	unsigned long		func;
 	unsigned long		stack;		/* stack pointer */
 	unsigned long		orig_ret_vaddr; /* original return address */
 	bool			chained;	/* true, if instance is nested */
+	int			consumers_cnt;
 
 	struct return_instance	*next;		/* keep as stack */
+
+	struct return_consumer	consumers[] __counted_by(consumers_cnt);
 };
 
 enum rp_check {
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 6b44c386a5df..4ef4b51776eb 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -64,7 +64,7 @@ struct uprobe {
 	struct rcu_head		rcu;
 	loff_t			offset;
 	loff_t			ref_ctr_offset;
-	unsigned long		flags;
+	unsigned long		flags;		/* "unsigned long" so bitops work */
 
 	/*
 	 * The generic code assumes that it has two members of unknown type
@@ -823,8 +823,11 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset,
 
 static void consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc)
 {
+	static atomic64_t id;
+
 	down_write(&uprobe->consumer_rwsem);
 	list_add_rcu(&uc->cons_node, &uprobe->consumers);
+	uc->id = (__u64) atomic64_inc_return(&id);
 	up_write(&uprobe->consumer_rwsem);
 }
 
@@ -1761,6 +1764,34 @@ static struct uprobe_task *get_utask(void)
 	return current->utask;
 }
 
+static size_t ri_size(int consumers_cnt)
+{
+	struct return_instance *ri;
+
+	return sizeof(*ri) + sizeof(ri->consumers[0]) * consumers_cnt;
+}
+
+#define DEF_CNT 4
+
+static struct return_instance *alloc_return_instance(void)
+{
+	struct return_instance *ri;
+
+	ri = kzalloc(ri_size(DEF_CNT), GFP_KERNEL);
+	if (!ri)
+		return ZERO_SIZE_PTR;
+
+	ri->consumers_cnt = DEF_CNT;
+	return ri;
+}
+
+static struct return_instance *dup_return_instance(struct return_instance *old)
+{
+	size_t size = ri_size(old->consumers_cnt);
+
+	return kmemdup(old, size, GFP_KERNEL);
+}
+
 static int dup_utask(struct task_struct *t, struct uprobe_task *o_utask)
 {
 	struct uprobe_task *n_utask;
@@ -1773,11 +1804,10 @@ static int dup_utask(struct task_struct *t, struct uprobe_task *o_utask)
 
 	p = &n_utask->return_instances;
 	for (o = o_utask->return_instances; o; o = o->next) {
-		n = kmalloc(sizeof(struct return_instance), GFP_KERNEL);
+		n = dup_return_instance(o);
 		if (!n)
 			return -ENOMEM;
 
-		*n = *o;
 		/*
 		 * uprobe's refcnt has to be positive at this point, kept by
 		 * utask->return_instances items; return_instances can't be
@@ -1870,35 +1900,31 @@ static void cleanup_return_instances(struct uprobe_task *utask, bool chained,
 	utask->return_instances = ri;
 }
 
-static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
+static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs,
+			      struct return_instance *ri)
 {
 	struct uprobe_task *utask = current->utask;
 	unsigned long orig_ret_vaddr, trampoline_vaddr;
-	struct return_instance *ri;
 	bool chained;
 
 	if (!get_xol_area())
-		return;
+		goto free;
 
 	if (utask->depth >= MAX_URETPROBE_DEPTH) {
 		printk_ratelimited(KERN_INFO "uprobe: omit uretprobe due to"
 				" nestedness limit pid/tgid=%d/%d\n",
 				current->pid, current->tgid);
-		return;
+		goto free;
 	}
 
 	/* we need to bump refcount to store uprobe in utask */
 	if (!try_get_uprobe(uprobe))
-		return;
-
-	ri = kmalloc(sizeof(struct return_instance), GFP_KERNEL);
-	if (!ri)
-		goto fail;
+		goto free;
 
 	trampoline_vaddr = uprobe_get_trampoline_vaddr();
 	orig_ret_vaddr = arch_uretprobe_hijack_return_addr(trampoline_vaddr, regs);
 	if (orig_ret_vaddr == -1)
-		goto fail;
+		goto put;
 
 	/* drop the entries invalidated by longjmp() */
 	chained = (orig_ret_vaddr == trampoline_vaddr);
@@ -1916,7 +1942,7 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
 			 * attack from user-space.
 			 */
 			uprobe_warn(current, "handle tail call");
-			goto fail;
+			goto put;
 		}
 		orig_ret_vaddr = utask->return_instances->orig_ret_vaddr;
 	}
@@ -1931,9 +1957,10 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
 	utask->return_instances = ri;
 
 	return;
-fail:
-	kfree(ri);
+put:
 	put_uprobe(uprobe);
+free:
+	kfree(ri);
 }
 
 /* Prepare to single-step probed instruction out of line. */
@@ -2077,34 +2104,90 @@ static struct uprobe *find_active_uprobe_rcu(unsigned long bp_vaddr, int *is_swb
 	return uprobe;
 }
 
+static struct return_instance*
+push_consumer(struct return_instance *ri, int idx, __u64 id, __u64 cookie)
+{
+	if (unlikely(ri == ZERO_SIZE_PTR))
+		return ri;
+
+	if (unlikely(idx >= ri->consumers_cnt)) {
+		struct return_instance *old_ri = ri;
+
+		ri->consumers_cnt += DEF_CNT;
+		ri = krealloc(old_ri, ri_size(old_ri->consumers_cnt), GFP_KERNEL);
+		if (!ri) {
+			kfree(old_ri);
+			return ZERO_SIZE_PTR;
+		}
+	}
+
+	ri->consumers[idx].id = id;
+	ri->consumers[idx].cookie = cookie;
+	return ri;
+}
+
+static struct return_consumer *
+return_consumer_find(struct return_instance *ri, int *iter, int id)
+{
+	struct return_consumer *ric;
+	int idx = *iter;
+
+	for (ric = &ri->consumers[idx]; idx < ri->consumers_cnt; idx++, ric++) {
+		if (ric->id == id) {
+			*iter = idx + 1;
+			return ric;
+		}
+	}
+	return NULL;
+}
+
+static bool ignore_ret_handler(int rc)
+{
+	return rc == UPROBE_HANDLER_REMOVE || rc == UPROBE_HANDLER_IGNORE;
+}
+
 static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
 {
 	struct uprobe_consumer *uc;
-	int remove = UPROBE_HANDLER_REMOVE;
-	bool need_prep = false; /* prepare return uprobe, when needed */
-	bool has_consumers = false;
+	bool has_consumers = false, remove = true;
+	struct return_instance *ri = NULL;
+	int push_idx = 0;
 
 	current->utask->auprobe = &uprobe->arch;
 
 	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
+		bool session = uc->handler && uc->ret_handler;
+		__u64 cookie = 0;
 		int rc = 0;
 
 		if (uc->handler) {
-			rc = uc->handler(uc, regs, NULL);
-			WARN(rc & ~UPROBE_HANDLER_MASK,
+			rc = uc->handler(uc, regs, &cookie);
+			WARN(rc < 0 || rc > 2,
 				"bad rc=0x%x from %ps()\n", rc, uc->handler);
 		}
 
-		if (uc->ret_handler)
-			need_prep = true;
-
-		remove &= rc;
+		remove &= rc == UPROBE_HANDLER_REMOVE;
 		has_consumers = true;
+
+		if (!uc->ret_handler || ignore_ret_handler(rc))
+			continue;
+
+		if (!ri)
+			ri = alloc_return_instance();
+
+		if (session)
+			ri = push_consumer(ri, push_idx++, uc->id, cookie);
 	}
 	current->utask->auprobe = NULL;
 
-	if (need_prep && !remove)
-		prepare_uretprobe(uprobe, regs); /* put bp at return */
+	if (!ZERO_OR_NULL_PTR(ri)) {
+		/*
+		 * The push_idx value has the final number of return consumers,
+		 * and ri->consumers_cnt has number of allocated consumers.
+		 */
+		ri->consumers_cnt = push_idx;
+		prepare_uretprobe(uprobe, regs, ri);
+	}
 
 	if (remove && has_consumers) {
 		down_read(&uprobe->register_rwsem);
@@ -2123,12 +2206,19 @@ static void
 handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs)
 {
 	struct uprobe *uprobe = ri->uprobe;
+	struct return_consumer *ric;
 	struct uprobe_consumer *uc;
+	int ric_idx = 0;
 
 	rcu_read_lock_trace();
 	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
-		if (uc->ret_handler)
-			uc->ret_handler(uc, ri->func, regs, NULL);
+		bool session = uc->handler && uc->ret_handler;
+
+		if (uc->ret_handler) {
+			ric = return_consumer_find(ri, &ric_idx, uc->id);
+			if (!session || ric)
+				uc->ret_handler(uc, ri->func, regs, ric ? &ric->cookie : NULL);
+		}
 	}
 	rcu_read_unlock_trace();
 }
-- 
2.46.2

[PATCHv6 bpf-next 03/16] bpf: Allow return values 0 and 1 for kprobe session

[Jiri Olsa]

The kprobe session program can return only 0 or 1,
instruct verifier to check for that.

Fixes: 535a3692ba72 ("bpf: Add support for kprobe session attach")
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 kernel/bpf/verifier.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 7d9b38ffd220..787008872a14 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -15910,6 +15910,15 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char
 			return -ENOTSUPP;
 		}
 		break;
+	case BPF_PROG_TYPE_KPROBE:
+		switch (env->prog->expected_attach_type) {
+		case BPF_TRACE_KPROBE_SESSION:
+			range = retval_range(0, 1);
+			break;
+		default:
+			return 0;
+		}
+		break;
 	case BPF_PROG_TYPE_SK_LOOKUP:
 		range = retval_range(SK_DROP, SK_PASS);
 		break;
-- 
2.46.2

[PATCHv6 bpf-next 04/16] bpf: Force uprobe bpf program to always return 0

[Jiri Olsa]

As suggested by Andrii make uprobe multi bpf programs to always return 0,
so they can't force uprobe removal.

Keeping the int return type for uprobe_prog_run, because it will be used
in following session changes.

Fixes: 89ae89f53d20 ("bpf: Add multi uprobe link")
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 kernel/trace/bpf_trace.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index fdab7ecd8dfa..3c1e5a561df4 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -3209,7 +3209,6 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 	struct bpf_prog *prog = link->link.prog;
 	bool sleepable = prog->sleepable;
 	struct bpf_run_ctx *old_run_ctx;
-	int err = 0;
 
 	if (link->task && !same_thread_group(current, link->task))
 		return 0;
@@ -3222,7 +3221,7 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 	migrate_disable();
 
 	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
-	err = bpf_prog_run(link->link.prog, regs);
+	bpf_prog_run(link->link.prog, regs);
 	bpf_reset_run_ctx(old_run_ctx);
 
 	migrate_enable();
@@ -3231,7 +3230,7 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 		rcu_read_unlock_trace();
 	else
 		rcu_read_unlock();
-	return err;
+	return 0;
 }
 
 static bool
-- 
2.46.2

[PATCHv6 bpf-next 05/16] bpf: Add support for uprobe multi session attach

[Jiri Olsa]

Adding support to attach BPF program for entry and return probe
of the same function. This is common use case which at the moment
requires to create two uprobe multi links.

Adding new BPF_TRACE_UPROBE_SESSION attach type that instructs
kernel to attach single link program to both entry and exit probe.

It's possible to control execution of the BPF program on return
probe simply by returning zero or non zero from the entry BPF
program execution to execute or not the BPF program on return
probe respectively.

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/uapi/linux/bpf.h       |  1 +
 kernel/bpf/syscall.c           |  9 +++++++--
 kernel/bpf/verifier.c          |  1 +
 kernel/trace/bpf_trace.c       | 36 +++++++++++++++++++++++++---------
 tools/include/uapi/linux/bpf.h |  1 +
 tools/lib/bpf/libbpf.c         |  1 +
 6 files changed, 38 insertions(+), 11 deletions(-)

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 8ab4d8184b9d..77d0bc5fa986 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -1116,6 +1116,7 @@ enum bpf_attach_type {
 	BPF_NETKIT_PRIMARY,
 	BPF_NETKIT_PEER,
 	BPF_TRACE_KPROBE_SESSION,
+	BPF_TRACE_UPROBE_SESSION,
 	__MAX_BPF_ATTACH_TYPE
 };
 
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index a8f1808a1ca5..0cf7617e6cb6 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -3983,10 +3983,14 @@ static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog,
 		if (prog->expected_attach_type == BPF_TRACE_UPROBE_MULTI &&
 		    attach_type != BPF_TRACE_UPROBE_MULTI)
 			return -EINVAL;
+		if (prog->expected_attach_type == BPF_TRACE_UPROBE_SESSION &&
+		    attach_type != BPF_TRACE_UPROBE_SESSION)
+			return -EINVAL;
 		if (attach_type != BPF_PERF_EVENT &&
 		    attach_type != BPF_TRACE_KPROBE_MULTI &&
 		    attach_type != BPF_TRACE_KPROBE_SESSION &&
-		    attach_type != BPF_TRACE_UPROBE_MULTI)
+		    attach_type != BPF_TRACE_UPROBE_MULTI &&
+		    attach_type != BPF_TRACE_UPROBE_SESSION)
 			return -EINVAL;
 		return 0;
 	case BPF_PROG_TYPE_SCHED_CLS:
@@ -5239,7 +5243,8 @@ static int link_create(union bpf_attr *attr, bpfptr_t uattr)
 		else if (attr->link_create.attach_type == BPF_TRACE_KPROBE_MULTI ||
 			 attr->link_create.attach_type == BPF_TRACE_KPROBE_SESSION)
 			ret = bpf_kprobe_multi_link_attach(attr, prog);
-		else if (attr->link_create.attach_type == BPF_TRACE_UPROBE_MULTI)
+		else if (attr->link_create.attach_type == BPF_TRACE_UPROBE_MULTI ||
+			 attr->link_create.attach_type == BPF_TRACE_UPROBE_SESSION)
 			ret = bpf_uprobe_multi_link_attach(attr, prog);
 		break;
 	default:
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 787008872a14..c4d7b7369259 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -15913,6 +15913,7 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char
 	case BPF_PROG_TYPE_KPROBE:
 		switch (env->prog->expected_attach_type) {
 		case BPF_TRACE_KPROBE_SESSION:
+		case BPF_TRACE_UPROBE_SESSION:
 			range = retval_range(0, 1);
 			break;
 		default:
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 3c1e5a561df4..5f10994376d0 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -1557,6 +1557,17 @@ static inline bool is_kprobe_session(const struct bpf_prog *prog)
 	return prog->expected_attach_type == BPF_TRACE_KPROBE_SESSION;
 }
 
+static inline bool is_uprobe_multi(const struct bpf_prog *prog)
+{
+	return prog->expected_attach_type == BPF_TRACE_UPROBE_MULTI ||
+	       prog->expected_attach_type == BPF_TRACE_UPROBE_SESSION;
+}
+
+static inline bool is_uprobe_session(const struct bpf_prog *prog)
+{
+	return prog->expected_attach_type == BPF_TRACE_UPROBE_SESSION;
+}
+
 static const struct bpf_func_proto *
 kprobe_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 {
@@ -1574,13 +1585,13 @@ kprobe_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_get_func_ip:
 		if (is_kprobe_multi(prog))
 			return &bpf_get_func_ip_proto_kprobe_multi;
-		if (prog->expected_attach_type == BPF_TRACE_UPROBE_MULTI)
+		if (is_uprobe_multi(prog))
 			return &bpf_get_func_ip_proto_uprobe_multi;
 		return &bpf_get_func_ip_proto_kprobe;
 	case BPF_FUNC_get_attach_cookie:
 		if (is_kprobe_multi(prog))
 			return &bpf_get_attach_cookie_proto_kmulti;
-		if (prog->expected_attach_type == BPF_TRACE_UPROBE_MULTI)
+		if (is_uprobe_multi(prog))
 			return &bpf_get_attach_cookie_proto_umulti;
 		return &bpf_get_attach_cookie_proto_trace;
 	default:
@@ -3074,6 +3085,7 @@ struct bpf_uprobe {
 	u64 cookie;
 	struct uprobe *uprobe;
 	struct uprobe_consumer consumer;
+	bool session;
 };
 
 struct bpf_uprobe_multi_link {
@@ -3247,9 +3259,13 @@ uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs,
 			  __u64 *data)
 {
 	struct bpf_uprobe *uprobe;
+	int ret;
 
 	uprobe = container_of(con, struct bpf_uprobe, consumer);
-	return uprobe_prog_run(uprobe, instruction_pointer(regs), regs);
+	ret = uprobe_prog_run(uprobe, instruction_pointer(regs), regs);
+	if (uprobe->session)
+		return ret ? UPROBE_HANDLER_IGNORE : 0;
+	return 0;
 }
 
 static int
@@ -3259,7 +3275,8 @@ uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, s
 	struct bpf_uprobe *uprobe;
 
 	uprobe = container_of(con, struct bpf_uprobe, consumer);
-	return uprobe_prog_run(uprobe, func, regs);
+	uprobe_prog_run(uprobe, func, regs);
+	return 0;
 }
 
 static u64 bpf_uprobe_multi_entry_ip(struct bpf_run_ctx *ctx)
@@ -3298,7 +3315,7 @@ int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
 	if (sizeof(u64) != sizeof(void *))
 		return -EOPNOTSUPP;
 
-	if (prog->expected_attach_type != BPF_TRACE_UPROBE_MULTI)
+	if (!is_uprobe_multi(prog))
 		return -EINVAL;
 
 	flags = attr->link_create.uprobe_multi.flags;
@@ -3374,11 +3391,12 @@ int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
 
 		uprobes[i].link = link;
 
-		if (flags & BPF_F_UPROBE_MULTI_RETURN)
-			uprobes[i].consumer.ret_handler = uprobe_multi_link_ret_handler;
-		else
+		if (!(flags & BPF_F_UPROBE_MULTI_RETURN))
 			uprobes[i].consumer.handler = uprobe_multi_link_handler;
-
+		if (flags & BPF_F_UPROBE_MULTI_RETURN || is_uprobe_session(prog))
+			uprobes[i].consumer.ret_handler = uprobe_multi_link_ret_handler;
+		if (is_uprobe_session(prog))
+			uprobes[i].session = true;
 		if (pid)
 			uprobes[i].consumer.filter = uprobe_multi_link_filter;
 	}
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 7610883c8191..09bdb1867d4a 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -1116,6 +1116,7 @@ enum bpf_attach_type {
 	BPF_NETKIT_PRIMARY,
 	BPF_NETKIT_PEER,
 	BPF_TRACE_KPROBE_SESSION,
+	BPF_TRACE_UPROBE_SESSION,
 	__MAX_BPF_ATTACH_TYPE
 };
 
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 05ad264ff09b..1caa9d20de5d 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -133,6 +133,7 @@ static const char * const attach_type_name[] = {
 	[BPF_NETKIT_PRIMARY]		= "netkit_primary",
 	[BPF_NETKIT_PEER]		= "netkit_peer",
 	[BPF_TRACE_KPROBE_SESSION]	= "trace_kprobe_session",
+	[BPF_TRACE_UPROBE_SESSION]	= "trace_uprobe_session",
 };
 
 static const char * const link_type_name[] = {
-- 
2.46.2

[PATCHv6 bpf-next 06/16] bpf: Add support for uprobe multi session context

[Jiri Olsa]

Placing bpf_session_run_ctx layer in between bpf_run_ctx and
bpf_uprobe_multi_run_ctx, so the session data can be retrieved
from uprobe_multi link.

Plus granting session kfuncs access to uprobe session programs.

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 kernel/trace/bpf_trace.c | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 5f10994376d0..01868039d7bc 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -3098,7 +3098,7 @@ struct bpf_uprobe_multi_link {
 };
 
 struct bpf_uprobe_multi_run_ctx {
-	struct bpf_run_ctx run_ctx;
+	struct bpf_session_run_ctx session_ctx;
 	unsigned long entry_ip;
 	struct bpf_uprobe *uprobe;
 };
@@ -3211,16 +3211,22 @@ static const struct bpf_link_ops bpf_uprobe_multi_link_lops = {
 
 static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 			   unsigned long entry_ip,
-			   struct pt_regs *regs)
+			   struct pt_regs *regs,
+			   bool is_return, void *data)
 {
 	struct bpf_uprobe_multi_link *link = uprobe->link;
 	struct bpf_uprobe_multi_run_ctx run_ctx = {
+		.session_ctx = {
+			.is_return = is_return,
+			.data = data,
+		},
 		.entry_ip = entry_ip,
 		.uprobe = uprobe,
 	};
 	struct bpf_prog *prog = link->link.prog;
 	bool sleepable = prog->sleepable;
 	struct bpf_run_ctx *old_run_ctx;
+	int err;
 
 	if (link->task && !same_thread_group(current, link->task))
 		return 0;
@@ -3232,8 +3238,8 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 
 	migrate_disable();
 
-	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
-	bpf_prog_run(link->link.prog, regs);
+	old_run_ctx = bpf_set_run_ctx(&run_ctx.session_ctx.run_ctx);
+	err = bpf_prog_run(link->link.prog, regs);
 	bpf_reset_run_ctx(old_run_ctx);
 
 	migrate_enable();
@@ -3242,7 +3248,7 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
 		rcu_read_unlock_trace();
 	else
 		rcu_read_unlock();
-	return 0;
+	return err;
 }
 
 static bool
@@ -3262,7 +3268,7 @@ uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs,
 	int ret;
 
 	uprobe = container_of(con, struct bpf_uprobe, consumer);
-	ret = uprobe_prog_run(uprobe, instruction_pointer(regs), regs);
+	ret = uprobe_prog_run(uprobe, instruction_pointer(regs), regs, false, data);
 	if (uprobe->session)
 		return ret ? UPROBE_HANDLER_IGNORE : 0;
 	return 0;
@@ -3275,7 +3281,7 @@ uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, s
 	struct bpf_uprobe *uprobe;
 
 	uprobe = container_of(con, struct bpf_uprobe, consumer);
-	uprobe_prog_run(uprobe, func, regs);
+	uprobe_prog_run(uprobe, func, regs, true, data);
 	return 0;
 }
 
@@ -3283,7 +3289,8 @@ static u64 bpf_uprobe_multi_entry_ip(struct bpf_run_ctx *ctx)
 {
 	struct bpf_uprobe_multi_run_ctx *run_ctx;
 
-	run_ctx = container_of(current->bpf_ctx, struct bpf_uprobe_multi_run_ctx, run_ctx);
+	run_ctx = container_of(current->bpf_ctx, struct bpf_uprobe_multi_run_ctx,
+			       session_ctx.run_ctx);
 	return run_ctx->entry_ip;
 }
 
@@ -3291,7 +3298,8 @@ static u64 bpf_uprobe_multi_cookie(struct bpf_run_ctx *ctx)
 {
 	struct bpf_uprobe_multi_run_ctx *run_ctx;
 
-	run_ctx = container_of(current->bpf_ctx, struct bpf_uprobe_multi_run_ctx, run_ctx);
+	run_ctx = container_of(current->bpf_ctx, struct bpf_uprobe_multi_run_ctx,
+			       session_ctx.run_ctx);
 	return run_ctx->uprobe->cookie;
 }
 
@@ -3485,7 +3493,7 @@ static int bpf_kprobe_multi_filter(const struct bpf_prog *prog, u32 kfunc_id)
 	if (!btf_id_set8_contains(&kprobe_multi_kfunc_set_ids, kfunc_id))
 		return 0;
 
-	if (!is_kprobe_session(prog))
+	if (!is_kprobe_session(prog) && !is_uprobe_session(prog))
 		return -EACCES;
 
 	return 0;
-- 
2.46.2

[PATCHv6 bpf-next 07/16] libbpf: Add support for uprobe multi session attach

[Jiri Olsa]

Adding support to attach program in uprobe session mode
with bpf_program__attach_uprobe_multi function.

Adding session bool to bpf_uprobe_multi_opts struct that allows
to load and attach the bpf program via uprobe session.
the attachment to create uprobe multi session.

Also adding new program loader section that allows:
  SEC("uprobe.session/bpf_fentry_test*")

and loads/attaches uprobe program as uprobe session.

Adding sleepable hook (uprobe.session.s) as well.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/lib/bpf/bpf.c    |  1 +
 tools/lib/bpf/libbpf.c | 18 ++++++++++++++++--
 tools/lib/bpf/libbpf.h |  4 +++-
 3 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
index 2a4c71501a17..becdfa701c75 100644
--- a/tools/lib/bpf/bpf.c
+++ b/tools/lib/bpf/bpf.c
@@ -776,6 +776,7 @@ int bpf_link_create(int prog_fd, int target_fd,
 			return libbpf_err(-EINVAL);
 		break;
 	case BPF_TRACE_UPROBE_MULTI:
+	case BPF_TRACE_UPROBE_SESSION:
 		attr.link_create.uprobe_multi.flags = OPTS_GET(opts, uprobe_multi.flags, 0);
 		attr.link_create.uprobe_multi.cnt = OPTS_GET(opts, uprobe_multi.cnt, 0);
 		attr.link_create.uprobe_multi.path = ptr_to_u64(OPTS_GET(opts, uprobe_multi.path, 0));
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 1caa9d20de5d..cbaa8b8a824f 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -9410,8 +9410,10 @@ static const struct bpf_sec_def section_defs[] = {
 	SEC_DEF("kprobe.session+",	KPROBE,	BPF_TRACE_KPROBE_SESSION, SEC_NONE, attach_kprobe_session),
 	SEC_DEF("uprobe.multi+",	KPROBE,	BPF_TRACE_UPROBE_MULTI, SEC_NONE, attach_uprobe_multi),
 	SEC_DEF("uretprobe.multi+",	KPROBE,	BPF_TRACE_UPROBE_MULTI, SEC_NONE, attach_uprobe_multi),
+	SEC_DEF("uprobe.session+",	KPROBE,	BPF_TRACE_UPROBE_SESSION, SEC_NONE, attach_uprobe_multi),
 	SEC_DEF("uprobe.multi.s+",	KPROBE,	BPF_TRACE_UPROBE_MULTI, SEC_SLEEPABLE, attach_uprobe_multi),
 	SEC_DEF("uretprobe.multi.s+",	KPROBE,	BPF_TRACE_UPROBE_MULTI, SEC_SLEEPABLE, attach_uprobe_multi),
+	SEC_DEF("uprobe.session.s+",	KPROBE,	BPF_TRACE_UPROBE_SESSION, SEC_SLEEPABLE, attach_uprobe_multi),
 	SEC_DEF("ksyscall+",		KPROBE,	0, SEC_NONE, attach_ksyscall),
 	SEC_DEF("kretsyscall+",		KPROBE, 0, SEC_NONE, attach_ksyscall),
 	SEC_DEF("usdt+",		KPROBE,	0, SEC_USDT, attach_usdt),
@@ -11733,7 +11735,9 @@ static int attach_uprobe_multi(const struct bpf_program *prog, long cookie, stru
 		ret = 0;
 		break;
 	case 3:
+		opts.session = str_has_pfx(probe_type, "uprobe.session");
 		opts.retprobe = str_has_pfx(probe_type, "uretprobe.multi");
+
 		*link = bpf_program__attach_uprobe_multi(prog, -1, binary_path, func_name, &opts);
 		ret = libbpf_get_error(*link);
 		break;
@@ -11982,10 +11986,12 @@ bpf_program__attach_uprobe_multi(const struct bpf_program *prog,
 	const unsigned long *ref_ctr_offsets = NULL, *offsets = NULL;
 	LIBBPF_OPTS(bpf_link_create_opts, lopts);
 	unsigned long *resolved_offsets = NULL;
+	enum bpf_attach_type attach_type;
 	int err = 0, link_fd, prog_fd;
 	struct bpf_link *link = NULL;
 	char errmsg[STRERR_BUFSIZE];
 	char full_path[PATH_MAX];
+	bool retprobe, session;
 	const __u64 *cookies;
 	const char **syms;
 	size_t cnt;
@@ -12056,12 +12062,20 @@ bpf_program__attach_uprobe_multi(const struct bpf_program *prog,
 		offsets = resolved_offsets;
 	}
 
+	retprobe = OPTS_GET(opts, retprobe, false);
+	session  = OPTS_GET(opts, session, false);
+
+	if (retprobe && session)
+		return libbpf_err_ptr(-EINVAL);
+
+	attach_type = session ? BPF_TRACE_UPROBE_SESSION : BPF_TRACE_UPROBE_MULTI;
+
 	lopts.uprobe_multi.path = path;
 	lopts.uprobe_multi.offsets = offsets;
 	lopts.uprobe_multi.ref_ctr_offsets = ref_ctr_offsets;
 	lopts.uprobe_multi.cookies = cookies;
 	lopts.uprobe_multi.cnt = cnt;
-	lopts.uprobe_multi.flags = OPTS_GET(opts, retprobe, false) ? BPF_F_UPROBE_MULTI_RETURN : 0;
+	lopts.uprobe_multi.flags = retprobe ? BPF_F_UPROBE_MULTI_RETURN : 0;
 
 	if (pid == 0)
 		pid = getpid();
@@ -12075,7 +12089,7 @@ bpf_program__attach_uprobe_multi(const struct bpf_program *prog,
 	}
 	link->detach = &bpf_link__detach_fd;
 
-	link_fd = bpf_link_create(prog_fd, 0, BPF_TRACE_UPROBE_MULTI, &lopts);
+	link_fd = bpf_link_create(prog_fd, 0, attach_type, &lopts);
 	if (link_fd < 0) {
 		err = -errno;
 		pr_warn("prog '%s': failed to attach multi-uprobe: %s\n",
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 91484303849c..b2ce3a72b11d 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -577,10 +577,12 @@ struct bpf_uprobe_multi_opts {
 	size_t cnt;
 	/* create return uprobes */
 	bool retprobe;
+	/* create session kprobes */
+	bool session;
 	size_t :0;
 };
 
-#define bpf_uprobe_multi_opts__last_field retprobe
+#define bpf_uprobe_multi_opts__last_field session
 
 /**
  * @brief **bpf_program__attach_uprobe_multi()** attaches a BPF program
-- 
2.46.2

[PATCHv6 bpf-next 08/16] selftests/bpf: Add uprobe session test

[Jiri Olsa]

Adding uprobe session test and testing that the entry program
return value controls execution of the return probe program.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 47 ++++++++++++
 .../bpf/progs/uprobe_multi_session.c          | 71 +++++++++++++++++++
 2 files changed, 118 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session.c

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index 2c39902b8a09..b10d2dadb462 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -8,6 +8,7 @@
 #include "uprobe_multi_usdt.skel.h"
 #include "uprobe_multi_consumers.skel.h"
 #include "uprobe_multi_pid_filter.skel.h"
+#include "uprobe_multi_session.skel.h"
 #include "bpf/libbpf_internal.h"
 #include "testing_helpers.h"
 #include "../sdt.h"
@@ -1015,6 +1016,50 @@ static void test_pid_filter_process(bool clone_vm)
 	uprobe_multi_pid_filter__destroy(skel);
 }
 
+static void test_session_skel_api(void)
+{
+	struct uprobe_multi_session *skel = NULL;
+	LIBBPF_OPTS(bpf_kprobe_multi_opts, opts);
+	struct bpf_link *link = NULL;
+	int err;
+
+	skel = uprobe_multi_session__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "uprobe_multi_session__open_and_load"))
+		goto cleanup;
+
+	skel->bss->pid = getpid();
+	skel->bss->user_ptr = test_data;
+
+	err = uprobe_multi_session__attach(skel);
+	if (!ASSERT_OK(err, "uprobe_multi_session__attach"))
+		goto cleanup;
+
+	/* trigger all probes */
+	skel->bss->uprobe_multi_func_1_addr = (__u64) uprobe_multi_func_1;
+	skel->bss->uprobe_multi_func_2_addr = (__u64) uprobe_multi_func_2;
+	skel->bss->uprobe_multi_func_3_addr = (__u64) uprobe_multi_func_3;
+
+	uprobe_multi_func_1();
+	uprobe_multi_func_2();
+	uprobe_multi_func_3();
+
+	/*
+	 * We expect 2 for uprobe_multi_func_2 because it runs both entry/return probe,
+	 * uprobe_multi_func_[13] run just the entry probe. All expected numbers are
+	 * doubled, because we run extra test for sleepable session.
+	 */
+	ASSERT_EQ(skel->bss->uprobe_session_result[0], 2, "uprobe_multi_func_1_result");
+	ASSERT_EQ(skel->bss->uprobe_session_result[1], 4, "uprobe_multi_func_2_result");
+	ASSERT_EQ(skel->bss->uprobe_session_result[2], 2, "uprobe_multi_func_3_result");
+
+	/* We expect increase in 3 entry and 1 return session calls -> 4 */
+	ASSERT_EQ(skel->bss->uprobe_multi_sleep_result, 4, "uprobe_multi_sleep_result");
+
+cleanup:
+	bpf_link__destroy(link);
+	uprobe_multi_session__destroy(skel);
+}
+
 static void test_bench_attach_uprobe(void)
 {
 	long attach_start_ns = 0, attach_end_ns = 0;
@@ -1111,4 +1156,6 @@ void test_uprobe_multi_test(void)
 		test_pid_filter_process(false);
 	if (test__start_subtest("filter_clone_vm"))
 		test_pid_filter_process(true);
+	if (test__start_subtest("session"))
+		test_session_skel_api();
 }
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session.c
new file mode 100644
index 000000000000..30bff90b68dc
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session.c
@@ -0,0 +1,71 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <stdbool.h>
+#include "bpf_kfuncs.h"
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+__u64 uprobe_multi_func_1_addr = 0;
+__u64 uprobe_multi_func_2_addr = 0;
+__u64 uprobe_multi_func_3_addr = 0;
+
+__u64 uprobe_session_result[3] = {};
+__u64 uprobe_multi_sleep_result = 0;
+
+void *user_ptr = 0;
+int pid = 0;
+
+static int uprobe_multi_check(void *ctx, bool is_return)
+{
+	const __u64 funcs[] = {
+		uprobe_multi_func_1_addr,
+		uprobe_multi_func_2_addr,
+		uprobe_multi_func_3_addr,
+	};
+	unsigned int i;
+	__u64 addr;
+
+	if (bpf_get_current_pid_tgid() >> 32 != pid)
+		return 1;
+
+	addr = bpf_get_func_ip(ctx);
+
+	for (i = 0; i < ARRAY_SIZE(funcs); i++) {
+		if (funcs[i] == addr) {
+			uprobe_session_result[i]++;
+			break;
+		}
+	}
+
+	/* only uprobe_multi_func_2 executes return probe */
+	if ((addr == uprobe_multi_func_1_addr) ||
+	    (addr == uprobe_multi_func_3_addr))
+		return 1;
+
+	return 0;
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_*")
+int uprobe(struct pt_regs *ctx)
+{
+	return uprobe_multi_check(ctx, bpf_session_is_return());
+}
+
+static __always_inline bool verify_sleepable_user_copy(void)
+{
+	char data[9];
+
+	bpf_copy_from_user(data, sizeof(data), user_ptr);
+	return bpf_strncmp(data, sizeof(data), "test_data") == 0;
+}
+
+SEC("uprobe.session.s//proc/self/exe:uprobe_multi_func_*")
+int uprobe_sleepable(struct pt_regs *ctx)
+{
+	if (verify_sleepable_user_copy())
+		uprobe_multi_sleep_result++;
+	return uprobe_multi_check(ctx, bpf_session_is_return());
+}
-- 
2.46.2

[PATCHv6 bpf-next 09/16] selftests/bpf: Add uprobe session cookie test

[Jiri Olsa]

Adding uprobe session test that verifies the cookie value
get properly propagated from entry to return program.

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 31 ++++++++++++
 .../bpf/progs/uprobe_multi_session_cookie.c   | 48 +++++++++++++++++++
 2 files changed, 79 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index b10d2dadb462..cc9030e86821 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -9,6 +9,7 @@
 #include "uprobe_multi_consumers.skel.h"
 #include "uprobe_multi_pid_filter.skel.h"
 #include "uprobe_multi_session.skel.h"
+#include "uprobe_multi_session_cookie.skel.h"
 #include "bpf/libbpf_internal.h"
 #include "testing_helpers.h"
 #include "../sdt.h"
@@ -1060,6 +1061,34 @@ static void test_session_skel_api(void)
 	uprobe_multi_session__destroy(skel);
 }
 
+static void test_session_cookie_skel_api(void)
+{
+	struct uprobe_multi_session_cookie *skel = NULL;
+	int err;
+
+	skel = uprobe_multi_session_cookie__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "uprobe_multi_session_cookie__open_and_load"))
+		goto cleanup;
+
+	skel->bss->pid = getpid();
+
+	err = uprobe_multi_session_cookie__attach(skel);
+	if (!ASSERT_OK(err, "uprobe_multi_session_cookie__attach"))
+		goto cleanup;
+
+	/* trigger all probes */
+	uprobe_multi_func_1();
+	uprobe_multi_func_2();
+	uprobe_multi_func_3();
+
+	ASSERT_EQ(skel->bss->test_uprobe_1_result, 1, "test_uprobe_1_result");
+	ASSERT_EQ(skel->bss->test_uprobe_2_result, 2, "test_uprobe_2_result");
+	ASSERT_EQ(skel->bss->test_uprobe_3_result, 3, "test_uprobe_3_result");
+
+cleanup:
+	uprobe_multi_session_cookie__destroy(skel);
+}
+
 static void test_bench_attach_uprobe(void)
 {
 	long attach_start_ns = 0, attach_end_ns = 0;
@@ -1158,4 +1187,6 @@ void test_uprobe_multi_test(void)
 		test_pid_filter_process(true);
 	if (test__start_subtest("session"))
 		test_session_skel_api();
+	if (test__start_subtest("session_cookie"))
+		test_session_cookie_skel_api();
 }
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c
new file mode 100644
index 000000000000..5befdf944dc6
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session_cookie.c
@@ -0,0 +1,48 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <stdbool.h>
+#include "bpf_kfuncs.h"
+
+char _license[] SEC("license") = "GPL";
+
+int pid = 0;
+
+__u64 test_uprobe_1_result = 0;
+__u64 test_uprobe_2_result = 0;
+__u64 test_uprobe_3_result = 0;
+
+static int check_cookie(__u64 val, __u64 *result)
+{
+	__u64 *cookie;
+
+	if (bpf_get_current_pid_tgid() >> 32 != pid)
+		return 1;
+
+	cookie = bpf_session_cookie();
+
+	if (bpf_session_is_return())
+		*result = *cookie == val ? val : 0;
+	else
+		*cookie = val;
+	return 0;
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
+int uprobe_1(struct pt_regs *ctx)
+{
+	return check_cookie(1, &test_uprobe_1_result);
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_2")
+int uprobe_2(struct pt_regs *ctx)
+{
+	return check_cookie(2, &test_uprobe_2_result);
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_3")
+int uprobe_3(struct pt_regs *ctx)
+{
+	return check_cookie(3, &test_uprobe_3_result);
+}
-- 
2.46.2

[PATCHv6 bpf-next 10/16] selftests/bpf: Add uprobe session recursive test

[Jiri Olsa]

Adding uprobe session test that verifies the cookie value is stored
properly when single uprobe-ed function is executed recursively.

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 57 +++++++++++++++++++
 .../progs/uprobe_multi_session_recursive.c    | 44 ++++++++++++++
 2 files changed, 101 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index cc9030e86821..284cd7fce576 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -10,6 +10,7 @@
 #include "uprobe_multi_pid_filter.skel.h"
 #include "uprobe_multi_session.skel.h"
 #include "uprobe_multi_session_cookie.skel.h"
+#include "uprobe_multi_session_recursive.skel.h"
 #include "bpf/libbpf_internal.h"
 #include "testing_helpers.h"
 #include "../sdt.h"
@@ -36,6 +37,12 @@ noinline void usdt_trigger(void)
 	STAP_PROBE(test, pid_filter_usdt);
 }
 
+noinline void uprobe_session_recursive(int i)
+{
+	if (i)
+		uprobe_session_recursive(i - 1);
+}
+
 struct child {
 	int go[2];
 	int c2p[2]; /* child -> parent channel */
@@ -1089,6 +1096,54 @@ static void test_session_cookie_skel_api(void)
 	uprobe_multi_session_cookie__destroy(skel);
 }
 
+static void test_session_recursive_skel_api(void)
+{
+	struct uprobe_multi_session_recursive *skel = NULL;
+	int i, err;
+
+	skel = uprobe_multi_session_recursive__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "uprobe_multi_session_recursive__open_and_load"))
+		goto cleanup;
+
+	skel->bss->pid = getpid();
+
+	err = uprobe_multi_session_recursive__attach(skel);
+	if (!ASSERT_OK(err, "uprobe_multi_session_recursive__attach"))
+		goto cleanup;
+
+	for (i = 0; i < ARRAY_SIZE(skel->bss->test_uprobe_cookie_entry); i++)
+		skel->bss->test_uprobe_cookie_entry[i] = i + 1;
+
+	uprobe_session_recursive(5);
+
+	/*
+	 *                                         entry uprobe:
+	 * uprobe_session_recursive(5) {             *cookie = 1, return 0
+	 *   uprobe_session_recursive(4) {           *cookie = 2, return 1
+	 *     uprobe_session_recursive(3) {         *cookie = 3, return 0
+	 *       uprobe_session_recursive(2) {       *cookie = 4, return 1
+	 *         uprobe_session_recursive(1) {     *cookie = 5, return 0
+	 *           uprobe_session_recursive(0) {   *cookie = 6, return 1
+	 *                                          return uprobe:
+	 *           } i = 0                          not executed
+	 *         } i = 1                            test_uprobe_cookie_return[0] = 5
+	 *       } i = 2                              not executed
+	 *     } i = 3                                test_uprobe_cookie_return[1] = 3
+	 *   } i = 4                                  not executed
+	 * } i = 5                                    test_uprobe_cookie_return[2] = 1
+	 */
+
+	ASSERT_EQ(skel->bss->idx_entry, 6, "idx_entry");
+	ASSERT_EQ(skel->bss->idx_return, 3, "idx_return");
+
+	ASSERT_EQ(skel->bss->test_uprobe_cookie_return[0], 5, "test_uprobe_cookie_return[0]");
+	ASSERT_EQ(skel->bss->test_uprobe_cookie_return[1], 3, "test_uprobe_cookie_return[1]");
+	ASSERT_EQ(skel->bss->test_uprobe_cookie_return[2], 1, "test_uprobe_cookie_return[2]");
+
+cleanup:
+	uprobe_multi_session_recursive__destroy(skel);
+}
+
 static void test_bench_attach_uprobe(void)
 {
 	long attach_start_ns = 0, attach_end_ns = 0;
@@ -1189,4 +1244,6 @@ void test_uprobe_multi_test(void)
 		test_session_skel_api();
 	if (test__start_subtest("session_cookie"))
 		test_session_cookie_skel_api();
+	if (test__start_subtest("session_cookie_recursive"))
+		test_session_recursive_skel_api();
 }
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c
new file mode 100644
index 000000000000..8fbcd69fae22
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session_recursive.c
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <stdbool.h>
+#include "bpf_kfuncs.h"
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+int pid = 0;
+
+int idx_entry = 0;
+int idx_return = 0;
+
+__u64 test_uprobe_cookie_entry[6];
+__u64 test_uprobe_cookie_return[3];
+
+static int check_cookie(void)
+{
+	__u64 *cookie = bpf_session_cookie();
+
+	if (bpf_session_is_return()) {
+		if (idx_return >= ARRAY_SIZE(test_uprobe_cookie_return))
+			return 1;
+		test_uprobe_cookie_return[idx_return++] = *cookie;
+		return 0;
+	}
+
+	if (idx_entry >= ARRAY_SIZE(test_uprobe_cookie_entry))
+		return 1;
+	*cookie = test_uprobe_cookie_entry[idx_entry];
+	return idx_entry++ % 2;
+}
+
+
+SEC("uprobe.session//proc/self/exe:uprobe_session_recursive")
+int uprobe_recursive(struct pt_regs *ctx)
+{
+	if (bpf_get_current_pid_tgid() >> 32 != pid)
+		return 1;
+
+	return check_cookie();
+}
-- 
2.46.2

[PATCHv6 bpf-next 11/16] selftests/bpf: Add uprobe session verifier test for return value

[Jiri Olsa]

Making sure uprobe.session program can return only [0,1] values.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        |  2 ++
 .../bpf/progs/uprobe_multi_verifier.c         | 31 +++++++++++++++++++
 2 files changed, 33 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index 284cd7fce576..e693eeb1a5a5 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -11,6 +11,7 @@
 #include "uprobe_multi_session.skel.h"
 #include "uprobe_multi_session_cookie.skel.h"
 #include "uprobe_multi_session_recursive.skel.h"
+#include "uprobe_multi_verifier.skel.h"
 #include "bpf/libbpf_internal.h"
 #include "testing_helpers.h"
 #include "../sdt.h"
@@ -1246,4 +1247,5 @@ void test_uprobe_multi_test(void)
 		test_session_cookie_skel_api();
 	if (test__start_subtest("session_cookie_recursive"))
 		test_session_recursive_skel_api();
+	RUN_TESTS(uprobe_multi_verifier);
 }
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c b/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c
new file mode 100644
index 000000000000..fe49f2cb5360
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/usdt.bpf.h>
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+
+SEC("uprobe.session")
+__success
+int uprobe_sesison_return_0(struct pt_regs *ctx)
+{
+	return 0;
+}
+
+SEC("uprobe.session")
+__success
+int uprobe_sesison_return_1(struct pt_regs *ctx)
+{
+	return 1;
+}
+
+SEC("uprobe.session")
+__failure
+__msg("At program exit the register R0 has smin=2 smax=2 should have been in [0, 1]")
+int uprobe_sesison_return_2(struct pt_regs *ctx)
+{
+	return 2;
+}
-- 
2.46.2

[PATCHv6 bpf-next 12/16] selftests/bpf: Add kprobe session verifier test for return value

[Jiri Olsa]

Making sure kprobe.session program can return only [0,1] values.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/kprobe_multi_test.c        |  2 ++
 .../bpf/progs/kprobe_multi_verifier.c         | 31 +++++++++++++++++++
 2 files changed, 33 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c

diff --git a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
index 960c9323d1e0..66ab1cae923e 100644
--- a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
@@ -6,6 +6,7 @@
 #include "kprobe_multi_override.skel.h"
 #include "kprobe_multi_session.skel.h"
 #include "kprobe_multi_session_cookie.skel.h"
+#include "kprobe_multi_verifier.skel.h"
 #include "bpf/libbpf_internal.h"
 #include "bpf/hashmap.h"
 
@@ -764,4 +765,5 @@ void test_kprobe_multi_test(void)
 		test_session_skel_api();
 	if (test__start_subtest("session_cookie"))
 		test_session_cookie_skel_api();
+	RUN_TESTS(kprobe_multi_verifier);
 }
diff --git a/tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c b/tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c
new file mode 100644
index 000000000000..288577e81deb
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/usdt.bpf.h>
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+
+SEC("kprobe.session")
+__success
+int kprobe_session_return_0(struct pt_regs *ctx)
+{
+	return 0;
+}
+
+SEC("kprobe.session")
+__success
+int kprobe_session_return_1(struct pt_regs *ctx)
+{
+	return 1;
+}
+
+SEC("kprobe.session")
+__failure
+__msg("At program exit the register R0 has smin=2 smax=2 should have been in [0, 1]")
+int kprobe_session_return_2(struct pt_regs *ctx)
+{
+	return 2;
+}
-- 
2.46.2

[PATCHv6 bpf-next 13/16] selftests/bpf: Add uprobe session single consumer test

[Jiri Olsa]

Testing that the session ret_handler bypass works on single
uprobe with multiple consumers, each with different session
ignore return value.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 33 ++++++++++++++
 .../bpf/progs/uprobe_multi_session_single.c   | 44 +++++++++++++++++++
 2 files changed, 77 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index e693eeb1a5a5..7e0228f8fcfc 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -9,6 +9,7 @@
 #include "uprobe_multi_consumers.skel.h"
 #include "uprobe_multi_pid_filter.skel.h"
 #include "uprobe_multi_session.skel.h"
+#include "uprobe_multi_session_single.skel.h"
 #include "uprobe_multi_session_cookie.skel.h"
 #include "uprobe_multi_session_recursive.skel.h"
 #include "uprobe_multi_verifier.skel.h"
@@ -1069,6 +1070,36 @@ static void test_session_skel_api(void)
 	uprobe_multi_session__destroy(skel);
 }
 
+static void test_session_single_skel_api(void)
+{
+	struct uprobe_multi_session_single *skel = NULL;
+	LIBBPF_OPTS(bpf_kprobe_multi_opts, opts);
+	int err;
+
+	skel = uprobe_multi_session_single__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "uprobe_multi_session_single__open_and_load"))
+		goto cleanup;
+
+	skel->bss->pid = getpid();
+
+	err = uprobe_multi_session_single__attach(skel);
+	if (!ASSERT_OK(err, "uprobe_multi_session_single__attach"))
+		goto cleanup;
+
+	uprobe_multi_func_1();
+
+	/*
+	 * We expect consumer 0 and 2 to trigger just entry handler (value 1)
+	 * and consumer 1 to hit both (value 2).
+	 */
+	ASSERT_EQ(skel->bss->uprobe_session_result[0], 1, "uprobe_session_result_0");
+	ASSERT_EQ(skel->bss->uprobe_session_result[1], 2, "uprobe_session_result_1");
+	ASSERT_EQ(skel->bss->uprobe_session_result[2], 1, "uprobe_session_result_2");
+
+cleanup:
+	uprobe_multi_session_single__destroy(skel);
+}
+
 static void test_session_cookie_skel_api(void)
 {
 	struct uprobe_multi_session_cookie *skel = NULL;
@@ -1243,6 +1274,8 @@ void test_uprobe_multi_test(void)
 		test_pid_filter_process(true);
 	if (test__start_subtest("session"))
 		test_session_skel_api();
+	if (test__start_subtest("session_single"))
+		test_session_single_skel_api();
 	if (test__start_subtest("session_cookie"))
 		test_session_cookie_skel_api();
 	if (test__start_subtest("session_cookie_recursive"))
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
new file mode 100644
index 000000000000..1fa53d3785f6
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <stdbool.h>
+#include "bpf_kfuncs.h"
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+__u64 uprobe_session_result[3] = {};
+int pid = 0;
+
+static int uprobe_multi_check(void *ctx, bool is_return, int idx)
+{
+	if (bpf_get_current_pid_tgid() >> 32 != pid)
+		return 1;
+
+	uprobe_session_result[idx]++;
+
+	/* only consumer 1 executes return probe */
+	if (idx == 0 || idx == 2)
+		return 1;
+
+	return 0;
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
+int uprobe_0(struct pt_regs *ctx)
+{
+	return uprobe_multi_check(ctx, bpf_session_is_return(), 0);
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
+int uprobe_1(struct pt_regs *ctx)
+{
+	return uprobe_multi_check(ctx, bpf_session_is_return(), 1);
+}
+
+SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
+int uprobe_2(struct pt_regs *ctx)
+{
+	return uprobe_multi_check(ctx, bpf_session_is_return(), 2);
+}
-- 
2.46.2

[PATCHv6 bpf-next 14/16] selftests/bpf: Scale down uprobe multi consumer test

[Jiri Olsa]

We have currently 2 uprobes and 2 uretprobes and we are about
to add sessions uprobes in following change, which makes the
test time unsuitable for CI even with threads.

It's enough for the test to have just 1 uprobe and 1 uretprobe.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 57 ++++++++-----------
 .../bpf/progs/uprobe_multi_consumers.c        | 16 +-----
 2 files changed, 25 insertions(+), 48 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index 7e0228f8fcfc..2effe4d693b4 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -761,10 +761,6 @@ get_program(struct uprobe_multi_consumers *skel, int prog)
 		return skel->progs.uprobe_0;
 	case 1:
 		return skel->progs.uprobe_1;
-	case 2:
-		return skel->progs.uprobe_2;
-	case 3:
-		return skel->progs.uprobe_3;
 	default:
 		ASSERT_FAIL("get_program");
 		return NULL;
@@ -779,10 +775,6 @@ get_link(struct uprobe_multi_consumers *skel, int link)
 		return &skel->links.uprobe_0;
 	case 1:
 		return &skel->links.uprobe_1;
-	case 2:
-		return &skel->links.uprobe_2;
-	case 3:
-		return &skel->links.uprobe_3;
 	default:
 		ASSERT_FAIL("get_link");
 		return NULL;
@@ -799,10 +791,10 @@ static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx)
 		return -1;
 
 	/*
-	 * bit/prog: 0,1 uprobe entry
-	 * bit/prog: 2,3 uprobe return
+	 * bit/prog: 0 uprobe entry
+	 * bit/prog: 1 uprobe return
 	 */
-	opts.retprobe = idx == 2 || idx == 3;
+	opts.retprobe = idx == 1;
 
 	*link = bpf_program__attach_uprobe_multi(prog, 0, "/proc/self/exe",
 						"uprobe_consumer_test",
@@ -832,7 +824,7 @@ uprobe_consumer_test(struct uprobe_multi_consumers *skel,
 	int idx;
 
 	/* detach uprobe for each unset programs in 'before' state ... */
-	for (idx = 0; idx < 4; idx++) {
+	for (idx = 0; idx < 1; idx++) {
 		if (test_bit(idx, before) && !test_bit(idx, after))
 			uprobe_detach(skel, idx);
 	}
@@ -855,7 +847,7 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	printf("consumer_test before %lu after %lu\n", before, after);
 
 	/* 'before' is each, we attach uprobe for every set idx */
-	for (idx = 0; idx < 4; idx++) {
+	for (idx = 0; idx < 1; idx++) {
 		if (test_bit(idx, before)) {
 			if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_before"))
 				goto cleanup;
@@ -866,18 +858,18 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	if (!ASSERT_EQ(err, 0, "uprobe_consumer_test"))
 		goto cleanup;
 
-	for (idx = 0; idx < 4; idx++) {
+	for (idx = 0; idx < 1; idx++) {
 		const char *fmt = "BUG";
 		__u64 val = 0;
 
-		if (idx < 2) {
+		if (idx == 0) {
 			/*
 			 * uprobe entry
 			 *   +1 if define in 'before'
 			 */
 			if (test_bit(idx, before))
 				val++;
-			fmt = "prog 0/1: uprobe";
+			fmt = "prog 0: uprobe";
 		} else {
 			/*
 			 * to trigger uretprobe consumer, the uretprobe needs to be installed,
@@ -885,11 +877,11 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 			 *
 			 *   idxs: 2/3 uprobe return in 'installed' mask
 			 */
-			unsigned long had_uretprobes  = before & 0b1100; /* is uretprobe installed */
+			unsigned long had_uretprobes = before & 0b10; /* is uretprobe installed */
 
 			if (had_uretprobes && test_bit(idx, after))
 				val++;
-			fmt = "idx 2/3: uretprobe";
+			fmt = "prog 1: uretprobe";
 		}
 
 		if (!ASSERT_EQ(skel->bss->uprobe_result[idx], val, fmt))
@@ -900,7 +892,7 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	ret = 0;
 
 cleanup:
-	for (idx = 0; idx < 4; idx++)
+	for (idx = 0; idx < 1; idx++)
 		uprobe_detach(skel, idx);
 	return ret;
 }
@@ -918,42 +910,41 @@ static void test_consumers(void)
 	 * The idea of this test is to try all possible combinations of
 	 * uprobes consumers attached on single function.
 	 *
-	 *  - 2 uprobe entry consumer
-	 *  - 2 uprobe exit consumers
+	 *  - 1 uprobe entry consumer
+	 *  - 1 uprobe exit consumer
 	 *
-	 * The test uses 4 uprobes attached on single function, but that
-	 * translates into single uprobe with 4 consumers in kernel.
+	 * The test uses 2 uprobes attached on single function, but that
+	 * translates into single uprobe with 2 consumers in kernel.
 	 *
 	 * The before/after values present the state of attached consumers
 	 * before and after the probed function:
 	 *
-	 *  bit/prog 0,1 : uprobe entry
-	 *  bit/prog 2,3 : uprobe return
+	 *  bit/prog 0 : uprobe entry
+	 *  bit/prog 1 : uprobe return
 	 *
 	 * For example for:
 	 *
-	 *   before = 0b0101
-	 *   after  = 0b0110
+	 *   before = 0b01
+	 *   after  = 0b10
 	 *
 	 * it means that before we call 'uprobe_consumer_test' we attach
 	 * uprobes defined in 'before' value:
 	 *
-	 *   - bit/prog 0: uprobe entry
-	 *   - bit/prog 2: uprobe return
+	 *   - bit/prog 1: uprobe entry
 	 *
 	 * uprobe_consumer_test is called and inside it we attach and detach
 	 * uprobes based on 'after' value:
 	 *
-	 *   - bit/prog 0: stays untouched
-	 *   - bit/prog 2: uprobe return is detached
+	 *   - bit/prog 0: is detached
+	 *   - bit/prog 1: is attached
 	 *
 	 * uprobe_consumer_test returns and we check counters values increased
 	 * by bpf programs on each uprobe to match the expected count based on
 	 * before/after bits.
 	 */
 
-	for (before = 0; before < 16; before++) {
-		for (after = 0; after < 16; after++)
+	for (before = 0; before < 4; before++) {
+		for (after = 0; after < 4; after++)
 			if (consumer_test(skel, before, after))
 				goto out;
 	}
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c b/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
index 7e0fdcbbd242..1c64d7263911 100644
--- a/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
@@ -8,7 +8,7 @@
 
 char _license[] SEC("license") = "GPL";
 
-__u64 uprobe_result[4];
+__u64 uprobe_result[2];
 
 SEC("uprobe.multi")
 int uprobe_0(struct pt_regs *ctx)
@@ -23,17 +23,3 @@ int uprobe_1(struct pt_regs *ctx)
 	uprobe_result[1]++;
 	return 0;
 }
-
-SEC("uprobe.multi")
-int uprobe_2(struct pt_regs *ctx)
-{
-	uprobe_result[2]++;
-	return 0;
-}
-
-SEC("uprobe.multi")
-int uprobe_3(struct pt_regs *ctx)
-{
-	uprobe_result[3]++;
-	return 0;
-}
-- 
2.46.2

[PATCHv6 bpf-next 15/16] selftests/bpf: Add uprobe sessions to consumer test

[Jiri Olsa]

Adding uprobe session consumers to the consumer test,
so we get the session into the test mix.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 63 +++++++++++++++----
 .../bpf/progs/uprobe_multi_consumers.c        | 16 ++++-
 2 files changed, 66 insertions(+), 13 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index 2effe4d693b4..df9314309bc3 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -761,6 +761,10 @@ get_program(struct uprobe_multi_consumers *skel, int prog)
 		return skel->progs.uprobe_0;
 	case 1:
 		return skel->progs.uprobe_1;
+	case 2:
+		return skel->progs.uprobe_2;
+	case 3:
+		return skel->progs.uprobe_3;
 	default:
 		ASSERT_FAIL("get_program");
 		return NULL;
@@ -775,6 +779,10 @@ get_link(struct uprobe_multi_consumers *skel, int link)
 		return &skel->links.uprobe_0;
 	case 1:
 		return &skel->links.uprobe_1;
+	case 2:
+		return &skel->links.uprobe_2;
+	case 3:
+		return &skel->links.uprobe_3;
 	default:
 		ASSERT_FAIL("get_link");
 		return NULL;
@@ -793,8 +801,11 @@ static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx)
 	/*
 	 * bit/prog: 0 uprobe entry
 	 * bit/prog: 1 uprobe return
+	 * bit/prog: 2 uprobe session without return
+	 * bit/prog: 3 uprobe session with return
 	 */
 	opts.retprobe = idx == 1;
+	opts.session  = idx == 2 || idx == 3;
 
 	*link = bpf_program__attach_uprobe_multi(prog, 0, "/proc/self/exe",
 						"uprobe_consumer_test",
@@ -824,7 +835,7 @@ uprobe_consumer_test(struct uprobe_multi_consumers *skel,
 	int idx;
 
 	/* detach uprobe for each unset programs in 'before' state ... */
-	for (idx = 0; idx < 1; idx++) {
+	for (idx = 0; idx < 4; idx++) {
 		if (test_bit(idx, before) && !test_bit(idx, after))
 			uprobe_detach(skel, idx);
 	}
@@ -847,7 +858,7 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	printf("consumer_test before %lu after %lu\n", before, after);
 
 	/* 'before' is each, we attach uprobe for every set idx */
-	for (idx = 0; idx < 1; idx++) {
+	for (idx = 0; idx < 4; idx++) {
 		if (test_bit(idx, before)) {
 			if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_before"))
 				goto cleanup;
@@ -858,11 +869,13 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	if (!ASSERT_EQ(err, 0, "uprobe_consumer_test"))
 		goto cleanup;
 
-	for (idx = 0; idx < 1; idx++) {
+	for (idx = 0; idx < 4; idx++) {
+		unsigned long had_uretprobes;
 		const char *fmt = "BUG";
 		__u64 val = 0;
 
-		if (idx == 0) {
+		switch (idx) {
+		case 0:
 			/*
 			 * uprobe entry
 			 *   +1 if define in 'before'
@@ -870,18 +883,42 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 			if (test_bit(idx, before))
 				val++;
 			fmt = "prog 0: uprobe";
-		} else {
+			break;
+		case 1:
 			/*
 			 * to trigger uretprobe consumer, the uretprobe needs to be installed,
 			 * which means one of the 'return' uprobes was alive when probe was hit:
 			 *
-			 *   idxs: 2/3 uprobe return in 'installed' mask
+			 *   idxs: 1/2 uprobe return in 'installed' mask
 			 */
-			unsigned long had_uretprobes = before & 0b10; /* is uretprobe installed */
+			had_uretprobes = before & 0b0110; /* is uretprobe installed */
 
 			if (had_uretprobes && test_bit(idx, after))
 				val++;
 			fmt = "prog 1: uretprobe";
+			break;
+		case 2:
+			/*
+			 * session with return
+			 *  +1 if defined in 'before'
+			 *  +1 if defined in 'after'
+			 */
+			if (test_bit(idx, before)) {
+				val++;
+				if (test_bit(idx, after))
+					val++;
+			}
+			fmt = "prog 2  : session with return";
+			break;
+		case 3:
+			/*
+			 * session without return
+			 *   +1 if defined in 'before'
+			 */
+			if (test_bit(idx, before))
+				val++;
+			fmt = "prog 3  : session with NO return";
+			break;
 		}
 
 		if (!ASSERT_EQ(skel->bss->uprobe_result[idx], val, fmt))
@@ -892,7 +929,7 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	ret = 0;
 
 cleanup:
-	for (idx = 0; idx < 1; idx++)
+	for (idx = 0; idx < 4; idx++)
 		uprobe_detach(skel, idx);
 	return ret;
 }
@@ -912,9 +949,11 @@ static void test_consumers(void)
 	 *
 	 *  - 1 uprobe entry consumer
 	 *  - 1 uprobe exit consumer
+	 *  - 1 uprobe session with return
+	 *  - 1 uprobe session without return
 	 *
-	 * The test uses 2 uprobes attached on single function, but that
-	 * translates into single uprobe with 2 consumers in kernel.
+	 * The test uses 4 uprobes attached on single function, but that
+	 * translates into single uprobe with 4 consumers in kernel.
 	 *
 	 * The before/after values present the state of attached consumers
 	 * before and after the probed function:
@@ -943,8 +982,8 @@ static void test_consumers(void)
 	 * before/after bits.
 	 */
 
-	for (before = 0; before < 4; before++) {
-		for (after = 0; after < 4; after++)
+	for (before = 0; before < 16; before++) {
+		for (after = 0; after < 16; after++)
 			if (consumer_test(skel, before, after))
 				goto out;
 	}
diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c b/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
index 1c64d7263911..93752bb5690b 100644
--- a/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
+++ b/tools/testing/selftests/bpf/progs/uprobe_multi_consumers.c
@@ -8,7 +8,7 @@
 
 char _license[] SEC("license") = "GPL";
 
-__u64 uprobe_result[2];
+__u64 uprobe_result[4];
 
 SEC("uprobe.multi")
 int uprobe_0(struct pt_regs *ctx)
@@ -23,3 +23,17 @@ int uprobe_1(struct pt_regs *ctx)
 	uprobe_result[1]++;
 	return 0;
 }
+
+SEC("uprobe.session")
+int uprobe_2(struct pt_regs *ctx)
+{
+	uprobe_result[2]++;
+	return 0;
+}
+
+SEC("uprobe.session")
+int uprobe_3(struct pt_regs *ctx)
+{
+	uprobe_result[3]++;
+	return 1;
+}
-- 
2.46.2

[PATCHv6 bpf-next 16/16] selftests/bpf: Add threads to consumer test

[Jiri Olsa]

With recent uprobe fix [1] the sync time after unregistering uprobe is
much longer and prolongs the consumer test which creates and destroys
hundreds of uprobes.

This change adds 16 threads (which fits the test logic) and speeds up
the test.

Before the change:

  # perf stat --null ./test_progs -t uprobe_multi_test/consumers
  #421/9   uprobe_multi_test/consumers:OK
  #421     uprobe_multi_test:OK
  Summary: 1/1 PASSED, 0 SKIPPED, 0 FAILED

   Performance counter stats for './test_progs -t uprobe_multi_test/consumers':

        28.818778973 seconds time elapsed

         0.745518000 seconds user
         0.919186000 seconds sys

After the change:

  # perf stat --null ./test_progs -t uprobe_multi_test/consumers 2>&1
  #421/9   uprobe_multi_test/consumers:OK
  #421     uprobe_multi_test:OK
  Summary: 1/1 PASSED, 0 SKIPPED, 0 FAILED

   Performance counter stats for './test_progs -t uprobe_multi_test/consumers':

         3.504790814 seconds time elapsed

         0.012141000 seconds user
         0.751760000 seconds sys

[1] commit 87195a1ee332 ("uprobes: switch to RCU Tasks Trace flavor for better performance")

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../bpf/prog_tests/uprobe_multi_test.c        | 98 +++++++++++++++----
 1 file changed, 80 insertions(+), 18 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index df9314309bc3..c1ef7355dce3 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -789,7 +789,7 @@ get_link(struct uprobe_multi_consumers *skel, int link)
 	}
 }
 
-static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx)
+static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx, unsigned long offset)
 {
 	struct bpf_program *prog = get_program(skel, idx);
 	struct bpf_link **link = get_link(skel, idx);
@@ -798,6 +798,9 @@ static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx)
 	if (!prog || !link)
 		return -1;
 
+	opts.offsets = &offset;
+	opts.cnt = 1;
+
 	/*
 	 * bit/prog: 0 uprobe entry
 	 * bit/prog: 1 uprobe return
@@ -807,9 +810,7 @@ static int uprobe_attach(struct uprobe_multi_consumers *skel, int idx)
 	opts.retprobe = idx == 1;
 	opts.session  = idx == 2 || idx == 3;
 
-	*link = bpf_program__attach_uprobe_multi(prog, 0, "/proc/self/exe",
-						"uprobe_consumer_test",
-						&opts);
+	*link = bpf_program__attach_uprobe_multi(prog, 0, "/proc/self/exe", NULL, &opts);
 	if (!ASSERT_OK_PTR(*link, "bpf_program__attach_uprobe_multi"))
 		return -1;
 	return 0;
@@ -830,7 +831,8 @@ static bool test_bit(int bit, unsigned long val)
 
 noinline int
 uprobe_consumer_test(struct uprobe_multi_consumers *skel,
-		     unsigned long before, unsigned long after)
+		     unsigned long before, unsigned long after,
+		     unsigned long offset)
 {
 	int idx;
 
@@ -843,15 +845,43 @@ uprobe_consumer_test(struct uprobe_multi_consumers *skel,
 	/* ... and attach all new programs in 'after' state */
 	for (idx = 0; idx < 4; idx++) {
 		if (!test_bit(idx, before) && test_bit(idx, after)) {
-			if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_after"))
+			if (!ASSERT_OK(uprobe_attach(skel, idx, offset), "uprobe_attach_after"))
 				return -1;
 		}
 	}
 	return 0;
 }
 
+/*
+ * We generate 16 consumer_testX functions that will have uprobe installed on
+ * and will be called in separate threads. All function pointer are stored in
+ * "consumers" section and each thread will pick one function based on index.
+ */
+
+extern const void *__start_consumers;
+
+#define __CONSUMER_TEST(func) 							\
+noinline int func(struct uprobe_multi_consumers *skel, unsigned long before,	\
+		  unsigned long after, unsigned long offset)			\
+{										\
+	return uprobe_consumer_test(skel, before, after, offset);		\
+}										\
+static void *__ ## func __used __attribute__((section("consumers"))) = (void *) func;
+
+#define CONSUMER_TEST(func) __CONSUMER_TEST(func)
+
+#define C1  CONSUMER_TEST(__PASTE(consumer_test, __COUNTER__))
+#define C4  C1 C1 C1 C1
+#define C16 C4 C4 C4 C4
+
+C16
+
+typedef int (*test_t)(struct uprobe_multi_consumers *, unsigned long,
+		      unsigned long, unsigned long);
+
 static int consumer_test(struct uprobe_multi_consumers *skel,
-			 unsigned long before, unsigned long after)
+			 unsigned long before, unsigned long after,
+			 test_t test, unsigned long offset)
 {
 	int err, idx, ret = -1;
 
@@ -860,12 +890,12 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	/* 'before' is each, we attach uprobe for every set idx */
 	for (idx = 0; idx < 4; idx++) {
 		if (test_bit(idx, before)) {
-			if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_before"))
+			if (!ASSERT_OK(uprobe_attach(skel, idx, offset), "uprobe_attach_before"))
 				goto cleanup;
 		}
 	}
 
-	err = uprobe_consumer_test(skel, before, after);
+	err = test(skel, before, after, offset);
 	if (!ASSERT_EQ(err, 0, "uprobe_consumer_test"))
 		goto cleanup;
 
@@ -934,14 +964,46 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 	return ret;
 }
 
-static void test_consumers(void)
+#define CONSUMER_MAX 16
+
+/*
+ * Each thread runs 1/16 of the load by running test for single
+ * 'before' number (based on thread index) and full scale of
+ * 'after' numbers.
+ */
+static void *consumer_thread(void *arg)
 {
+	unsigned long idx = (unsigned long) arg;
 	struct uprobe_multi_consumers *skel;
-	int before, after;
+	unsigned long offset;
+	const void *func;
+	int after;
 
 	skel = uprobe_multi_consumers__open_and_load();
 	if (!ASSERT_OK_PTR(skel, "uprobe_multi_consumers__open_and_load"))
-		return;
+		return NULL;
+
+	func = *((&__start_consumers) + idx);
+
+	offset = get_uprobe_offset(func);
+	if (!ASSERT_GE(offset, 0, "uprobe_offset"))
+		goto out;
+
+	for (after = 0; after < CONSUMER_MAX; after++)
+		if (consumer_test(skel, idx, after, func, offset))
+			goto out;
+
+out:
+	uprobe_multi_consumers__destroy(skel);
+	return NULL;
+}
+
+
+static void test_consumers(void)
+{
+	pthread_t pt[CONSUMER_MAX];
+	unsigned long idx;
+	int err;
 
 	/*
 	 * The idea of this test is to try all possible combinations of
@@ -982,14 +1044,14 @@ static void test_consumers(void)
 	 * before/after bits.
 	 */
 
-	for (before = 0; before < 16; before++) {
-		for (after = 0; after < 16; after++)
-			if (consumer_test(skel, before, after))
-				goto out;
+	for (idx = 0; idx < CONSUMER_MAX; idx++) {
+		err = pthread_create(&pt[idx], NULL, consumer_thread, (void *) idx);
+		if (!ASSERT_OK(err, "pthread_create"))
+			break;
 	}
 
-out:
-	uprobe_multi_consumers__destroy(skel);
+	while (idx)
+		pthread_join(pt[--idx], NULL);
 }
 
 static struct bpf_program *uprobe_multi_program(struct uprobe_multi_pid_filter *skel, int idx)
-- 
2.46.2

Re: [PATCHv6 bpf-next 03/16] bpf: Allow return values 0 and 1 for kprobe session

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:10 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> The kprobe session program can return only 0 or 1,
> instruct verifier to check for that.
>
> Fixes: 535a3692ba72 ("bpf: Add support for kprobe session attach")
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  kernel/bpf/verifier.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>

Acked-by: Andrii Nakryiko <andrii@kernel.org>

> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 7d9b38ffd220..787008872a14 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -15910,6 +15910,15 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char
>                         return -ENOTSUPP;
>                 }
>                 break;
> +       case BPF_PROG_TYPE_KPROBE:
> +               switch (env->prog->expected_attach_type) {
> +               case BPF_TRACE_KPROBE_SESSION:
> +                       range = retval_range(0, 1);
> +                       break;
> +               default:
> +                       return 0;
> +               }
> +               break;
>         case BPF_PROG_TYPE_SK_LOOKUP:
>                 range = retval_range(SK_DROP, SK_PASS);
>                 break;
> --
> 2.46.2
>

Re: [PATCHv6 bpf-next 04/16] bpf: Force uprobe bpf program to always return 0

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:10 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> As suggested by Andrii make uprobe multi bpf programs to always return 0,
> so they can't force uprobe removal.
>
> Keeping the int return type for uprobe_prog_run, because it will be used
> in following session changes.
>
> Fixes: 89ae89f53d20 ("bpf: Add multi uprobe link")
> Suggested-by: Andrii Nakryiko <andrii@kernel.org>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  kernel/trace/bpf_trace.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
>

Acked-by: Andrii Nakryiko <andrii@kernel.org>

> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index fdab7ecd8dfa..3c1e5a561df4 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -3209,7 +3209,6 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
>         struct bpf_prog *prog = link->link.prog;
>         bool sleepable = prog->sleepable;
>         struct bpf_run_ctx *old_run_ctx;
> -       int err = 0;
>
>         if (link->task && !same_thread_group(current, link->task))
>                 return 0;
> @@ -3222,7 +3221,7 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
>         migrate_disable();
>
>         old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
> -       err = bpf_prog_run(link->link.prog, regs);
> +       bpf_prog_run(link->link.prog, regs);
>         bpf_reset_run_ctx(old_run_ctx);
>
>         migrate_enable();
> @@ -3231,7 +3230,7 @@ static int uprobe_prog_run(struct bpf_uprobe *uprobe,
>                 rcu_read_unlock_trace();
>         else
>                 rcu_read_unlock();
> -       return err;
> +       return 0;
>  }
>
>  static bool
> --
> 2.46.2
>

Re: [PATCHv6 bpf-next 07/16] libbpf: Add support for uprobe multi session attach

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:11 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding support to attach program in uprobe session mode
> with bpf_program__attach_uprobe_multi function.
>
> Adding session bool to bpf_uprobe_multi_opts struct that allows
> to load and attach the bpf program via uprobe session.
> the attachment to create uprobe multi session.
>
> Also adding new program loader section that allows:
>   SEC("uprobe.session/bpf_fentry_test*")
>
> and loads/attaches uprobe program as uprobe session.
>
> Adding sleepable hook (uprobe.session.s) as well.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  tools/lib/bpf/bpf.c    |  1 +
>  tools/lib/bpf/libbpf.c | 18 ++++++++++++++++--
>  tools/lib/bpf/libbpf.h |  4 +++-
>  3 files changed, 20 insertions(+), 3 deletions(-)
>

Acked-by: Andrii Nakryiko <andrii@kernel.org>

[...]

Re: [PATCHv6 bpf-next 08/16] selftests/bpf: Add uprobe session test

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:11 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding uprobe session test and testing that the entry program
> return value controls execution of the return probe program.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        | 47 ++++++++++++
>  .../bpf/progs/uprobe_multi_session.c          | 71 +++++++++++++++++++
>  2 files changed, 118 insertions(+)
>  create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session.c
>

LGTM

Acked-by: Andrii Nakryiko <andrii@kernel.org>

[...]

Re: [PATCHv6 bpf-next 11/16] selftests/bpf: Add uprobe session verifier test for return value

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Making sure uprobe.session program can return only [0,1] values.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        |  2 ++
>  .../bpf/progs/uprobe_multi_verifier.c         | 31 +++++++++++++++++++
>  2 files changed, 33 insertions(+)
>  create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c
>

Nice

Acked-by: Andrii Nakryiko <andrii@kernel.org>

> diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> index 284cd7fce576..e693eeb1a5a5 100644
> --- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> +++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> @@ -11,6 +11,7 @@
>  #include "uprobe_multi_session.skel.h"
>  #include "uprobe_multi_session_cookie.skel.h"
>  #include "uprobe_multi_session_recursive.skel.h"
> +#include "uprobe_multi_verifier.skel.h"
>  #include "bpf/libbpf_internal.h"
>  #include "testing_helpers.h"
>  #include "../sdt.h"
> @@ -1246,4 +1247,5 @@ void test_uprobe_multi_test(void)
>                 test_session_cookie_skel_api();
>         if (test__start_subtest("session_cookie_recursive"))
>                 test_session_recursive_skel_api();
> +       RUN_TESTS(uprobe_multi_verifier);
>  }
> diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c b/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c
> new file mode 100644
> index 000000000000..fe49f2cb5360
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/uprobe_multi_verifier.c
> @@ -0,0 +1,31 @@
> +// SPDX-License-Identifier: GPL-2.0
> +#include "vmlinux.h"
> +#include <bpf/bpf_helpers.h>
> +#include <bpf/bpf_tracing.h>
> +#include <bpf/usdt.bpf.h>
> +#include "bpf_misc.h"
> +
> +char _license[] SEC("license") = "GPL";
> +
> +
> +SEC("uprobe.session")
> +__success
> +int uprobe_sesison_return_0(struct pt_regs *ctx)
> +{
> +       return 0;
> +}
> +
> +SEC("uprobe.session")
> +__success
> +int uprobe_sesison_return_1(struct pt_regs *ctx)
> +{
> +       return 1;
> +}
> +
> +SEC("uprobe.session")
> +__failure
> +__msg("At program exit the register R0 has smin=2 smax=2 should have been in [0, 1]")
> +int uprobe_sesison_return_2(struct pt_regs *ctx)
> +{
> +       return 2;
> +}
> --
> 2.46.2
>

Re: [PATCHv6 bpf-next 12/16] selftests/bpf: Add kprobe session verifier test for return value

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Making sure kprobe.session program can return only [0,1] values.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/kprobe_multi_test.c        |  2 ++
>  .../bpf/progs/kprobe_multi_verifier.c         | 31 +++++++++++++++++++
>  2 files changed, 33 insertions(+)
>  create mode 100644 tools/testing/selftests/bpf/progs/kprobe_multi_verifier.c
>

Acked-by: Andrii Nakryiko <andrii@kernel.org>

[...]

Re: [PATCHv6 bpf-next 13/16] selftests/bpf: Add uprobe session single consumer test

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Testing that the session ret_handler bypass works on single
> uprobe with multiple consumers, each with different session
> ignore return value.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        | 33 ++++++++++++++
>  .../bpf/progs/uprobe_multi_session_single.c   | 44 +++++++++++++++++++
>  2 files changed, 77 insertions(+)
>  create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
>

see the nit, but regardless:

Acked-by: Andrii Nakryiko <andrii@kernel.org>

[...]

> diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
> new file mode 100644
> index 000000000000..1fa53d3785f6
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
> @@ -0,0 +1,44 @@
> +// SPDX-License-Identifier: GPL-2.0
> +#include <linux/bpf.h>
> +#include <bpf/bpf_helpers.h>
> +#include <bpf/bpf_tracing.h>
> +#include <stdbool.h>
> +#include "bpf_kfuncs.h"
> +#include "bpf_misc.h"
> +
> +char _license[] SEC("license") = "GPL";
> +
> +__u64 uprobe_session_result[3] = {};
> +int pid = 0;
> +
> +static int uprobe_multi_check(void *ctx, bool is_return, int idx)

nit: you don't use is_return

> +{
> +       if (bpf_get_current_pid_tgid() >> 32 != pid)
> +               return 1;
> +
> +       uprobe_session_result[idx]++;
> +
> +       /* only consumer 1 executes return probe */
> +       if (idx == 0 || idx == 2)
> +               return 1;
> +
> +       return 0;
> +}
> +
> +SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
> +int uprobe_0(struct pt_regs *ctx)
> +{
> +       return uprobe_multi_check(ctx, bpf_session_is_return(), 0);
> +}
> +
> +SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
> +int uprobe_1(struct pt_regs *ctx)
> +{
> +       return uprobe_multi_check(ctx, bpf_session_is_return(), 1);
> +}
> +
> +SEC("uprobe.session//proc/self/exe:uprobe_multi_func_1")
> +int uprobe_2(struct pt_regs *ctx)
> +{
> +       return uprobe_multi_check(ctx, bpf_session_is_return(), 2);
> +}
> --
> 2.46.2
>

Re: [PATCHv6 bpf-next 14/16] selftests/bpf: Scale down uprobe multi consumer test

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> We have currently 2 uprobes and 2 uretprobes and we are about
> to add sessions uprobes in following change, which makes the
> test time unsuitable for CI even with threads.
>
> It's enough for the test to have just 1 uprobe and 1 uretprobe.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        | 57 ++++++++-----------
>  .../bpf/progs/uprobe_multi_consumers.c        | 16 +-----
>  2 files changed, 25 insertions(+), 48 deletions(-)
>

[...]

>         /* 'before' is each, we attach uprobe for every set idx */
> -       for (idx = 0; idx < 4; idx++) {
> +       for (idx = 0; idx < 1; idx++) {
>                 if (test_bit(idx, before)) {
>                         if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_before"))
>                                 goto cleanup;
> @@ -866,18 +858,18 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
>         if (!ASSERT_EQ(err, 0, "uprobe_consumer_test"))
>                 goto cleanup;
>
> -       for (idx = 0; idx < 4; idx++) {
> +       for (idx = 0; idx < 1; idx++) {

here and everywhere else, either idx <= 1 or idx < 2, no?

>                 const char *fmt = "BUG";
>                 __u64 val = 0;
>
> -               if (idx < 2) {
> +               if (idx == 0) {
>                         /*
>                          * uprobe entry
>                          *   +1 if define in 'before'
>                          */
>                         if (test_bit(idx, before))
>                                 val++;
> -                       fmt = "prog 0/1: uprobe";
> +                       fmt = "prog 0: uprobe";
>                 } else {
>                         /*
>                          * to trigger uretprobe consumer, the uretprobe needs to be installed,

[...]

Re: [PATCHv6 bpf-next 15/16] selftests/bpf: Add uprobe sessions to consumer test

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:13 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding uprobe session consumers to the consumer test,
> so we get the session into the test mix.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        | 63 +++++++++++++++----
>  .../bpf/progs/uprobe_multi_consumers.c        | 16 ++++-
>  2 files changed, 66 insertions(+), 13 deletions(-)
>

you are undoing most of the changes done in the previous patch... it
seems like it would be better to just combine  both patches

> diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> index 2effe4d693b4..df9314309bc3 100644
> --- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> +++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> @@ -761,6 +761,10 @@ get_program(struct uprobe_multi_consumers *skel, int prog)
>                 return skel->progs.uprobe_0;
>         case 1:
>                 return skel->progs.uprobe_1;
> +       case 2:
> +               return skel->progs.uprobe_2;
> +       case 3:
> +               return skel->progs.uprobe_3;
>         default:
>                 ASSERT_FAIL("get_program");
>                 return NULL;

[...]

Re: [PATCHv6 bpf-next 16/16] selftests/bpf: Add threads to consumer test

[Andrii Nakryiko]

On Thu, Oct 10, 2024 at 1:13 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> With recent uprobe fix [1] the sync time after unregistering uprobe is
> much longer and prolongs the consumer test which creates and destroys
> hundreds of uprobes.
>
> This change adds 16 threads (which fits the test logic) and speeds up
> the test.
>
> Before the change:
>
>   # perf stat --null ./test_progs -t uprobe_multi_test/consumers
>   #421/9   uprobe_multi_test/consumers:OK
>   #421     uprobe_multi_test:OK
>   Summary: 1/1 PASSED, 0 SKIPPED, 0 FAILED
>
>    Performance counter stats for './test_progs -t uprobe_multi_test/consumers':
>
>         28.818778973 seconds time elapsed
>
>          0.745518000 seconds user
>          0.919186000 seconds sys
>
> After the change:
>
>   # perf stat --null ./test_progs -t uprobe_multi_test/consumers 2>&1
>   #421/9   uprobe_multi_test/consumers:OK
>   #421     uprobe_multi_test:OK
>   Summary: 1/1 PASSED, 0 SKIPPED, 0 FAILED
>
>    Performance counter stats for './test_progs -t uprobe_multi_test/consumers':
>
>          3.504790814 seconds time elapsed
>
>          0.012141000 seconds user
>          0.751760000 seconds sys
>
> [1] commit 87195a1ee332 ("uprobes: switch to RCU Tasks Trace flavor for better performance")
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  .../bpf/prog_tests/uprobe_multi_test.c        | 98 +++++++++++++++----
>  1 file changed, 80 insertions(+), 18 deletions(-)
>

OK, LGTM

Acked-by: Andrii Nakryiko <andrii@kernel.org>

[...]

Re: [PATCHv6 perf/core 02/16] uprobe: Add support for session consumer

[Oleg Nesterov]

On 10/10, Jiri Olsa wrote:
>
>  include/linux/uprobes.h |  21 +++++-
>  kernel/events/uprobes.c | 148 ++++++++++++++++++++++++++++++++--------
>  2 files changed, 139 insertions(+), 30 deletions(-)

Reviewed-by: Oleg Nesterov <oleg@redhat.com>

Re: [PATCHv6 bpf-next 13/16] selftests/bpf: Add uprobe session single consumer test

[Jiri Olsa]

On Thu, Oct 10, 2024 at 07:25:59PM -0700, Andrii Nakryiko wrote:
> On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
> >
> > Testing that the session ret_handler bypass works on single
> > uprobe with multiple consumers, each with different session
> > ignore return value.
> >
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >  .../bpf/prog_tests/uprobe_multi_test.c        | 33 ++++++++++++++
> >  .../bpf/progs/uprobe_multi_session_single.c   | 44 +++++++++++++++++++
> >  2 files changed, 77 insertions(+)
> >  create mode 100644 tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
> >
> 
> see the nit, but regardless:
> 
> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> 
> [...]
> 
> > diff --git a/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
> > new file mode 100644
> > index 000000000000..1fa53d3785f6
> > --- /dev/null
> > +++ b/tools/testing/selftests/bpf/progs/uprobe_multi_session_single.c
> > @@ -0,0 +1,44 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +#include <linux/bpf.h>
> > +#include <bpf/bpf_helpers.h>
> > +#include <bpf/bpf_tracing.h>
> > +#include <stdbool.h>
> > +#include "bpf_kfuncs.h"
> > +#include "bpf_misc.h"
> > +
> > +char _license[] SEC("license") = "GPL";
> > +
> > +__u64 uprobe_session_result[3] = {};
> > +int pid = 0;
> > +
> > +static int uprobe_multi_check(void *ctx, bool is_return, int idx)
> 
> nit: you don't use is_return

ugh true, thanks

jirka

Re: [PATCHv6 bpf-next 14/16] selftests/bpf: Scale down uprobe multi consumer test

[Jiri Olsa]

On Thu, Oct 10, 2024 at 07:27:47PM -0700, Andrii Nakryiko wrote:
> On Thu, Oct 10, 2024 at 1:12 PM Jiri Olsa <jolsa@kernel.org> wrote:
> >
> > We have currently 2 uprobes and 2 uretprobes and we are about
> > to add sessions uprobes in following change, which makes the
> > test time unsuitable for CI even with threads.
> >
> > It's enough for the test to have just 1 uprobe and 1 uretprobe.
> >
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >  .../bpf/prog_tests/uprobe_multi_test.c        | 57 ++++++++-----------
> >  .../bpf/progs/uprobe_multi_consumers.c        | 16 +-----
> >  2 files changed, 25 insertions(+), 48 deletions(-)
> >
> 
> [...]
> 
> >         /* 'before' is each, we attach uprobe for every set idx */
> > -       for (idx = 0; idx < 4; idx++) {
> > +       for (idx = 0; idx < 1; idx++) {
> >                 if (test_bit(idx, before)) {
> >                         if (!ASSERT_OK(uprobe_attach(skel, idx), "uprobe_attach_before"))
> >                                 goto cleanup;
> > @@ -866,18 +858,18 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
> >         if (!ASSERT_EQ(err, 0, "uprobe_consumer_test"))
> >                 goto cleanup;
> >
> > -       for (idx = 0; idx < 4; idx++) {
> > +       for (idx = 0; idx < 1; idx++) {
> 
> here and everywhere else, either idx <= 1 or idx < 2, no?

right, it's changed in the next patch that adds session support,
I guess I'll combine them as you suggested in the other email

jirka

> 
> >                 const char *fmt = "BUG";
> >                 __u64 val = 0;
> >
> > -               if (idx < 2) {
> > +               if (idx == 0) {
> >                         /*
> >                          * uprobe entry
> >                          *   +1 if define in 'before'
> >                          */
> >                         if (test_bit(idx, before))
> >                                 val++;
> > -                       fmt = "prog 0/1: uprobe";
> > +                       fmt = "prog 0: uprobe";
> >                 } else {
> >                         /*
> >                          * to trigger uretprobe consumer, the uretprobe needs to be installed,
> 
> [...]

Re: [PATCHv6 bpf-next 15/16] selftests/bpf: Add uprobe sessions to consumer test

[Jiri Olsa]

On Thu, Oct 10, 2024 at 07:30:19PM -0700, Andrii Nakryiko wrote:
> On Thu, Oct 10, 2024 at 1:13 PM Jiri Olsa <jolsa@kernel.org> wrote:
> >
> > Adding uprobe session consumers to the consumer test,
> > so we get the session into the test mix.
> >
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >  .../bpf/prog_tests/uprobe_multi_test.c        | 63 +++++++++++++++----
> >  .../bpf/progs/uprobe_multi_consumers.c        | 16 ++++-
> >  2 files changed, 66 insertions(+), 13 deletions(-)
> >
> 
> you are undoing most of the changes done in the previous patch... it
> seems like it would be better to just combine  both patches

ok, the diff looks actualy ok, I'll squash them together

thanks,
jirka

> 
> > diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> > index 2effe4d693b4..df9314309bc3 100644
> > --- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> > +++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
> > @@ -761,6 +761,10 @@ get_program(struct uprobe_multi_consumers *skel, int prog)
> >                 return skel->progs.uprobe_0;
> >         case 1:
> >                 return skel->progs.uprobe_1;
> > +       case 2:
> > +               return skel->progs.uprobe_2;
> > +       case 3:
> > +               return skel->progs.uprobe_3;
> >         default:
> >                 ASSERT_FAIL("get_program");
> >                 return NULL;
> 
> [...]

Re: [PATCHv6 perf/core 01/16] uprobe: Add data pointer to consumer handlers

[Masami Hiramatsu]

On Thu, 10 Oct 2024 22:09:42 +0200
Jiri Olsa <jolsa@kernel.org> wrote:

> Adding data pointer to both entry and exit consumer handlers and all
> its users. The functionality itself is coming in following change.
> 

Looks good to me.

Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Alexei, please merge this series via bpf tree, since most of the patches
in this series are for bpf.

Thank you,

> Acked-by: Oleg Nesterov <oleg@redhat.com>
> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  include/linux/uprobes.h                              |  4 ++--
>  kernel/events/uprobes.c                              |  4 ++--
>  kernel/trace/bpf_trace.c                             |  6 ++++--
>  kernel/trace/trace_uprobe.c                          | 12 ++++++++----
>  .../testing/selftests/bpf/bpf_testmod/bpf_testmod.c  |  2 +-
>  5 files changed, 17 insertions(+), 11 deletions(-)
> 
> diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> index 2b294bf1881f..bb265a632b91 100644
> --- a/include/linux/uprobes.h
> +++ b/include/linux/uprobes.h
> @@ -37,10 +37,10 @@ struct uprobe_consumer {
>  	 * for the current process. If filter() is omitted or returns true,
>  	 * UPROBE_HANDLER_REMOVE is effectively ignored.
>  	 */
> -	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs);
> +	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs, __u64 *data);
>  	int (*ret_handler)(struct uprobe_consumer *self,
>  				unsigned long func,
> -				struct pt_regs *regs);
> +				struct pt_regs *regs, __u64 *data);
>  	bool (*filter)(struct uprobe_consumer *self, struct mm_struct *mm);
>  
>  	struct list_head cons_node;
> diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> index 2a0059464383..6b44c386a5df 100644
> --- a/kernel/events/uprobes.c
> +++ b/kernel/events/uprobes.c
> @@ -2090,7 +2090,7 @@ static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
>  		int rc = 0;
>  
>  		if (uc->handler) {
> -			rc = uc->handler(uc, regs);
> +			rc = uc->handler(uc, regs, NULL);
>  			WARN(rc & ~UPROBE_HANDLER_MASK,
>  				"bad rc=0x%x from %ps()\n", rc, uc->handler);
>  		}
> @@ -2128,7 +2128,7 @@ handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs)
>  	rcu_read_lock_trace();
>  	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
>  		if (uc->ret_handler)
> -			uc->ret_handler(uc, ri->func, regs);
> +			uc->ret_handler(uc, ri->func, regs, NULL);
>  	}
>  	rcu_read_unlock_trace();
>  }
> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index a582cd25ca87..fdab7ecd8dfa 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -3244,7 +3244,8 @@ uprobe_multi_link_filter(struct uprobe_consumer *con, struct mm_struct *mm)
>  }
>  
>  static int
> -uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
> +uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs,
> +			  __u64 *data)
>  {
>  	struct bpf_uprobe *uprobe;
>  
> @@ -3253,7 +3254,8 @@ uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
>  }
>  
>  static int
> -uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs)
> +uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs,
> +			      __u64 *data)
>  {
>  	struct bpf_uprobe *uprobe;
>  
> diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
> index c40531d2cbad..5895eabe3581 100644
> --- a/kernel/trace/trace_uprobe.c
> +++ b/kernel/trace/trace_uprobe.c
> @@ -89,9 +89,11 @@ static struct trace_uprobe *to_trace_uprobe(struct dyn_event *ev)
>  static int register_uprobe_event(struct trace_uprobe *tu);
>  static int unregister_uprobe_event(struct trace_uprobe *tu);
>  
> -static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs);
> +static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
> +			     __u64 *data);
>  static int uretprobe_dispatcher(struct uprobe_consumer *con,
> -				unsigned long func, struct pt_regs *regs);
> +				unsigned long func, struct pt_regs *regs,
> +				__u64 *data);
>  
>  #ifdef CONFIG_STACK_GROWSUP
>  static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
> @@ -1517,7 +1519,8 @@ trace_uprobe_register(struct trace_event_call *event, enum trace_reg type,
>  	}
>  }
>  
> -static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
> +static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
> +			     __u64 *data)
>  {
>  	struct trace_uprobe *tu;
>  	struct uprobe_dispatch_data udd;
> @@ -1548,7 +1551,8 @@ static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
>  }
>  
>  static int uretprobe_dispatcher(struct uprobe_consumer *con,
> -				unsigned long func, struct pt_regs *regs)
> +				unsigned long func, struct pt_regs *regs,
> +				__u64 *data)
>  {
>  	struct trace_uprobe *tu;
>  	struct uprobe_dispatch_data udd;
> diff --git a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> index 8835761d9a12..12005e3dc3e4 100644
> --- a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> +++ b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> @@ -461,7 +461,7 @@ static struct bin_attribute bin_attr_bpf_testmod_file __ro_after_init = {
>  
>  static int
>  uprobe_ret_handler(struct uprobe_consumer *self, unsigned long func,
> -		   struct pt_regs *regs)
> +		   struct pt_regs *regs, __u64 *data)
>  
>  {
>  	regs->ax  = 0x12345678deadbeef;
> -- 
> 2.46.2
> 


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>

Re: [PATCHv6 perf/core 02/16] uprobe: Add support for session consumer

[Masami Hiramatsu]

On Thu, 10 Oct 2024 22:09:43 +0200
Jiri Olsa <jolsa@kernel.org> wrote:

> This change allows the uprobe consumer to behave as session which
> means that 'handler' and 'ret_handler' callbacks are connected in
> a way that allows to:
> 
>   - control execution of 'ret_handler' from 'handler' callback
>   - share data between 'handler' and 'ret_handler' callbacks
> 
> The session concept fits to our common use case where we do filtering
> on entry uprobe and based on the result we decide to run the return
> uprobe (or not).
> 
> It's also convenient to share the data between session callbacks.
> 
> To achive this we are adding new return value the uprobe consumer
> can return from 'handler' callback:
> 
>   UPROBE_HANDLER_IGNORE
>   - Ignore 'ret_handler' callback for this consumer.

This is interesting idea. This means we can cancel exit handler
dynamically.

> 
> And store cookie and pass it to 'ret_handler' when consumer has both
> 'handler' and 'ret_handler' callbacks defined.
> 
> We store shared data in the return_consumer object array as part of
> the return_instance object. This way the handle_uretprobe_chain can
> find related return_consumer and its shared data.
> 
> We also store entry handler return value, for cases when there are
> multiple consumers on single uprobe and some of them are ignored and
> some of them not, in which case the return probe gets installed and
> we need to have a way to find out which consumer needs to be ignored.
> 
> The tricky part is when consumer is registered 'after' the uprobe
> entry handler is hit. In such case this consumer's 'ret_handler' gets
> executed as well, but it won't have the proper data pointer set,
> so we can filter it out.

Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Thank you,

> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> Suggested-by: Oleg Nesterov <oleg@redhat.com>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  include/linux/uprobes.h |  21 +++++-
>  kernel/events/uprobes.c | 148 ++++++++++++++++++++++++++++++++--------
>  2 files changed, 139 insertions(+), 30 deletions(-)
> 
> diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> index bb265a632b91..dbaf04189548 100644
> --- a/include/linux/uprobes.h
> +++ b/include/linux/uprobes.h
> @@ -23,8 +23,17 @@ struct inode;
>  struct notifier_block;
>  struct page;
>  
> +/*
> + * Allowed return values from uprobe consumer's handler callback
> + * with following meaning:
> + *
> + * UPROBE_HANDLER_REMOVE
> + * - Remove the uprobe breakpoint from current->mm.
> + * UPROBE_HANDLER_IGNORE
> + * - Ignore ret_handler callback for this consumer.
> + */
>  #define UPROBE_HANDLER_REMOVE		1
> -#define UPROBE_HANDLER_MASK		1
> +#define UPROBE_HANDLER_IGNORE		2
>  
>  #define MAX_URETPROBE_DEPTH		64
>  
> @@ -44,6 +53,8 @@ struct uprobe_consumer {
>  	bool (*filter)(struct uprobe_consumer *self, struct mm_struct *mm);
>  
>  	struct list_head cons_node;
> +
> +	__u64 id;	/* set when uprobe_consumer is registered */
>  };
>  
>  #ifdef CONFIG_UPROBES
> @@ -83,14 +94,22 @@ struct uprobe_task {
>  	unsigned int			depth;
>  };
>  
> +struct return_consumer {
> +	__u64	cookie;
> +	__u64	id;
> +};
> +
>  struct return_instance {
>  	struct uprobe		*uprobe;
>  	unsigned long		func;
>  	unsigned long		stack;		/* stack pointer */
>  	unsigned long		orig_ret_vaddr; /* original return address */
>  	bool			chained;	/* true, if instance is nested */
> +	int			consumers_cnt;
>  
>  	struct return_instance	*next;		/* keep as stack */
> +
> +	struct return_consumer	consumers[] __counted_by(consumers_cnt);
>  };
>  
>  enum rp_check {
> diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> index 6b44c386a5df..4ef4b51776eb 100644
> --- a/kernel/events/uprobes.c
> +++ b/kernel/events/uprobes.c
> @@ -64,7 +64,7 @@ struct uprobe {
>  	struct rcu_head		rcu;
>  	loff_t			offset;
>  	loff_t			ref_ctr_offset;
> -	unsigned long		flags;
> +	unsigned long		flags;		/* "unsigned long" so bitops work */
>  
>  	/*
>  	 * The generic code assumes that it has two members of unknown type
> @@ -823,8 +823,11 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset,
>  
>  static void consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc)
>  {
> +	static atomic64_t id;
> +
>  	down_write(&uprobe->consumer_rwsem);
>  	list_add_rcu(&uc->cons_node, &uprobe->consumers);
> +	uc->id = (__u64) atomic64_inc_return(&id);
>  	up_write(&uprobe->consumer_rwsem);
>  }
>  
> @@ -1761,6 +1764,34 @@ static struct uprobe_task *get_utask(void)
>  	return current->utask;
>  }
>  
> +static size_t ri_size(int consumers_cnt)
> +{
> +	struct return_instance *ri;
> +
> +	return sizeof(*ri) + sizeof(ri->consumers[0]) * consumers_cnt;
> +}
> +
> +#define DEF_CNT 4
> +
> +static struct return_instance *alloc_return_instance(void)
> +{
> +	struct return_instance *ri;
> +
> +	ri = kzalloc(ri_size(DEF_CNT), GFP_KERNEL);
> +	if (!ri)
> +		return ZERO_SIZE_PTR;
> +
> +	ri->consumers_cnt = DEF_CNT;
> +	return ri;
> +}
> +
> +static struct return_instance *dup_return_instance(struct return_instance *old)
> +{
> +	size_t size = ri_size(old->consumers_cnt);
> +
> +	return kmemdup(old, size, GFP_KERNEL);
> +}
> +
>  static int dup_utask(struct task_struct *t, struct uprobe_task *o_utask)
>  {
>  	struct uprobe_task *n_utask;
> @@ -1773,11 +1804,10 @@ static int dup_utask(struct task_struct *t, struct uprobe_task *o_utask)
>  
>  	p = &n_utask->return_instances;
>  	for (o = o_utask->return_instances; o; o = o->next) {
> -		n = kmalloc(sizeof(struct return_instance), GFP_KERNEL);
> +		n = dup_return_instance(o);
>  		if (!n)
>  			return -ENOMEM;
>  
> -		*n = *o;
>  		/*
>  		 * uprobe's refcnt has to be positive at this point, kept by
>  		 * utask->return_instances items; return_instances can't be
> @@ -1870,35 +1900,31 @@ static void cleanup_return_instances(struct uprobe_task *utask, bool chained,
>  	utask->return_instances = ri;
>  }
>  
> -static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
> +static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs,
> +			      struct return_instance *ri)
>  {
>  	struct uprobe_task *utask = current->utask;
>  	unsigned long orig_ret_vaddr, trampoline_vaddr;
> -	struct return_instance *ri;
>  	bool chained;
>  
>  	if (!get_xol_area())
> -		return;
> +		goto free;
>  
>  	if (utask->depth >= MAX_URETPROBE_DEPTH) {
>  		printk_ratelimited(KERN_INFO "uprobe: omit uretprobe due to"
>  				" nestedness limit pid/tgid=%d/%d\n",
>  				current->pid, current->tgid);
> -		return;
> +		goto free;
>  	}
>  
>  	/* we need to bump refcount to store uprobe in utask */
>  	if (!try_get_uprobe(uprobe))
> -		return;
> -
> -	ri = kmalloc(sizeof(struct return_instance), GFP_KERNEL);
> -	if (!ri)
> -		goto fail;
> +		goto free;
>  
>  	trampoline_vaddr = uprobe_get_trampoline_vaddr();
>  	orig_ret_vaddr = arch_uretprobe_hijack_return_addr(trampoline_vaddr, regs);
>  	if (orig_ret_vaddr == -1)
> -		goto fail;
> +		goto put;
>  
>  	/* drop the entries invalidated by longjmp() */
>  	chained = (orig_ret_vaddr == trampoline_vaddr);
> @@ -1916,7 +1942,7 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
>  			 * attack from user-space.
>  			 */
>  			uprobe_warn(current, "handle tail call");
> -			goto fail;
> +			goto put;
>  		}
>  		orig_ret_vaddr = utask->return_instances->orig_ret_vaddr;
>  	}
> @@ -1931,9 +1957,10 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs)
>  	utask->return_instances = ri;
>  
>  	return;
> -fail:
> -	kfree(ri);
> +put:
>  	put_uprobe(uprobe);
> +free:
> +	kfree(ri);
>  }
>  
>  /* Prepare to single-step probed instruction out of line. */
> @@ -2077,34 +2104,90 @@ static struct uprobe *find_active_uprobe_rcu(unsigned long bp_vaddr, int *is_swb
>  	return uprobe;
>  }
>  
> +static struct return_instance*
> +push_consumer(struct return_instance *ri, int idx, __u64 id, __u64 cookie)
> +{
> +	if (unlikely(ri == ZERO_SIZE_PTR))
> +		return ri;
> +
> +	if (unlikely(idx >= ri->consumers_cnt)) {
> +		struct return_instance *old_ri = ri;
> +
> +		ri->consumers_cnt += DEF_CNT;
> +		ri = krealloc(old_ri, ri_size(old_ri->consumers_cnt), GFP_KERNEL);
> +		if (!ri) {
> +			kfree(old_ri);
> +			return ZERO_SIZE_PTR;
> +		}
> +	}
> +
> +	ri->consumers[idx].id = id;
> +	ri->consumers[idx].cookie = cookie;
> +	return ri;
> +}
> +
> +static struct return_consumer *
> +return_consumer_find(struct return_instance *ri, int *iter, int id)
> +{
> +	struct return_consumer *ric;
> +	int idx = *iter;
> +
> +	for (ric = &ri->consumers[idx]; idx < ri->consumers_cnt; idx++, ric++) {
> +		if (ric->id == id) {
> +			*iter = idx + 1;
> +			return ric;
> +		}
> +	}
> +	return NULL;
> +}
> +
> +static bool ignore_ret_handler(int rc)
> +{
> +	return rc == UPROBE_HANDLER_REMOVE || rc == UPROBE_HANDLER_IGNORE;
> +}
> +
>  static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
>  {
>  	struct uprobe_consumer *uc;
> -	int remove = UPROBE_HANDLER_REMOVE;
> -	bool need_prep = false; /* prepare return uprobe, when needed */
> -	bool has_consumers = false;
> +	bool has_consumers = false, remove = true;
> +	struct return_instance *ri = NULL;
> +	int push_idx = 0;
>  
>  	current->utask->auprobe = &uprobe->arch;
>  
>  	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
> +		bool session = uc->handler && uc->ret_handler;
> +		__u64 cookie = 0;
>  		int rc = 0;
>  
>  		if (uc->handler) {
> -			rc = uc->handler(uc, regs, NULL);
> -			WARN(rc & ~UPROBE_HANDLER_MASK,
> +			rc = uc->handler(uc, regs, &cookie);
> +			WARN(rc < 0 || rc > 2,
>  				"bad rc=0x%x from %ps()\n", rc, uc->handler);
>  		}
>  
> -		if (uc->ret_handler)
> -			need_prep = true;
> -
> -		remove &= rc;
> +		remove &= rc == UPROBE_HANDLER_REMOVE;
>  		has_consumers = true;
> +
> +		if (!uc->ret_handler || ignore_ret_handler(rc))
> +			continue;
> +
> +		if (!ri)
> +			ri = alloc_return_instance();
> +
> +		if (session)
> +			ri = push_consumer(ri, push_idx++, uc->id, cookie);
>  	}
>  	current->utask->auprobe = NULL;
>  
> -	if (need_prep && !remove)
> -		prepare_uretprobe(uprobe, regs); /* put bp at return */
> +	if (!ZERO_OR_NULL_PTR(ri)) {
> +		/*
> +		 * The push_idx value has the final number of return consumers,
> +		 * and ri->consumers_cnt has number of allocated consumers.
> +		 */
> +		ri->consumers_cnt = push_idx;
> +		prepare_uretprobe(uprobe, regs, ri);
> +	}
>  
>  	if (remove && has_consumers) {
>  		down_read(&uprobe->register_rwsem);
> @@ -2123,12 +2206,19 @@ static void
>  handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs)
>  {
>  	struct uprobe *uprobe = ri->uprobe;
> +	struct return_consumer *ric;
>  	struct uprobe_consumer *uc;
> +	int ric_idx = 0;
>  
>  	rcu_read_lock_trace();
>  	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
> -		if (uc->ret_handler)
> -			uc->ret_handler(uc, ri->func, regs, NULL);
> +		bool session = uc->handler && uc->ret_handler;
> +
> +		if (uc->ret_handler) {
> +			ric = return_consumer_find(ri, &ric_idx, uc->id);
> +			if (!session || ric)
> +				uc->ret_handler(uc, ri->func, regs, ric ? &ric->cookie : NULL);
> +		}
>  	}
>  	rcu_read_unlock_trace();
>  }
> -- 
> 2.46.2
> 


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>

Re: [PATCHv6 perf/core 01/16] uprobe: Add data pointer to consumer handlers

[Jiri Olsa]

On Wed, Oct 16, 2024 at 07:24:26AM +0900, Masami Hiramatsu wrote:
> On Thu, 10 Oct 2024 22:09:42 +0200
> Jiri Olsa <jolsa@kernel.org> wrote:
> 
> > Adding data pointer to both entry and exit consumer handlers and all
> > its users. The functionality itself is coming in following change.
> > 
> 
> Looks good to me.
> 
> Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> 
> Alexei, please merge this series via bpf tree, since most of the patches
> in this series are for bpf.


hi,
so the uprobe changes are based on Peter's perf/core and bpf changes
on bpf-next/master.. I merged them locally as a base for this patchset

Andrii has an idea how we could proceed with merging:

	> I think uprobe parts should stay in tip/perf/core (if that's where all
	> uprobe code goes in), as we have a bunch of ongoing work that all will
	> conflict a bit with each other, if it lands across multiple trees.
	>
	> So that means that patches #1 and #2 ideally land in tip/perf/core.
	> But you have a lot of BPF-specific things that would be inconvenient
	> to route through tip, so I'd say those should go through bpf-next.
	>
	> What we can do, if Ingo and Peter are OK with that, is to create a
	> stable (non-rebaseable) branch off of your first two patches (applied
	> in tip/perf/core), which we'll merge into bpf-next/master and land the
	> rest of your patch set there. We've done that with recent struct fd
	> changes, and there were few other similar cases in the past, and that
	> all worked well.
	>
	> Peter, Ingo, are you guys OK with that approach?

jirka

> 
> Thank you,
> 
> > Acked-by: Oleg Nesterov <oleg@redhat.com>
> > Acked-by: Andrii Nakryiko <andrii@kernel.org>
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >  include/linux/uprobes.h                              |  4 ++--
> >  kernel/events/uprobes.c                              |  4 ++--
> >  kernel/trace/bpf_trace.c                             |  6 ++++--
> >  kernel/trace/trace_uprobe.c                          | 12 ++++++++----
> >  .../testing/selftests/bpf/bpf_testmod/bpf_testmod.c  |  2 +-
> >  5 files changed, 17 insertions(+), 11 deletions(-)
> > 
> > diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> > index 2b294bf1881f..bb265a632b91 100644
> > --- a/include/linux/uprobes.h
> > +++ b/include/linux/uprobes.h
> > @@ -37,10 +37,10 @@ struct uprobe_consumer {
> >  	 * for the current process. If filter() is omitted or returns true,
> >  	 * UPROBE_HANDLER_REMOVE is effectively ignored.
> >  	 */
> > -	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs);
> > +	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs, __u64 *data);
> >  	int (*ret_handler)(struct uprobe_consumer *self,
> >  				unsigned long func,
> > -				struct pt_regs *regs);
> > +				struct pt_regs *regs, __u64 *data);
> >  	bool (*filter)(struct uprobe_consumer *self, struct mm_struct *mm);
> >  
> >  	struct list_head cons_node;
> > diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> > index 2a0059464383..6b44c386a5df 100644
> > --- a/kernel/events/uprobes.c
> > +++ b/kernel/events/uprobes.c
> > @@ -2090,7 +2090,7 @@ static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
> >  		int rc = 0;
> >  
> >  		if (uc->handler) {
> > -			rc = uc->handler(uc, regs);
> > +			rc = uc->handler(uc, regs, NULL);
> >  			WARN(rc & ~UPROBE_HANDLER_MASK,
> >  				"bad rc=0x%x from %ps()\n", rc, uc->handler);
> >  		}
> > @@ -2128,7 +2128,7 @@ handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs)
> >  	rcu_read_lock_trace();
> >  	list_for_each_entry_rcu(uc, &uprobe->consumers, cons_node, rcu_read_lock_trace_held()) {
> >  		if (uc->ret_handler)
> > -			uc->ret_handler(uc, ri->func, regs);
> > +			uc->ret_handler(uc, ri->func, regs, NULL);
> >  	}
> >  	rcu_read_unlock_trace();
> >  }
> > diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> > index a582cd25ca87..fdab7ecd8dfa 100644
> > --- a/kernel/trace/bpf_trace.c
> > +++ b/kernel/trace/bpf_trace.c
> > @@ -3244,7 +3244,8 @@ uprobe_multi_link_filter(struct uprobe_consumer *con, struct mm_struct *mm)
> >  }
> >  
> >  static int
> > -uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
> > +uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs,
> > +			  __u64 *data)
> >  {
> >  	struct bpf_uprobe *uprobe;
> >  
> > @@ -3253,7 +3254,8 @@ uprobe_multi_link_handler(struct uprobe_consumer *con, struct pt_regs *regs)
> >  }
> >  
> >  static int
> > -uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs)
> > +uprobe_multi_link_ret_handler(struct uprobe_consumer *con, unsigned long func, struct pt_regs *regs,
> > +			      __u64 *data)
> >  {
> >  	struct bpf_uprobe *uprobe;
> >  
> > diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
> > index c40531d2cbad..5895eabe3581 100644
> > --- a/kernel/trace/trace_uprobe.c
> > +++ b/kernel/trace/trace_uprobe.c
> > @@ -89,9 +89,11 @@ static struct trace_uprobe *to_trace_uprobe(struct dyn_event *ev)
> >  static int register_uprobe_event(struct trace_uprobe *tu);
> >  static int unregister_uprobe_event(struct trace_uprobe *tu);
> >  
> > -static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs);
> > +static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
> > +			     __u64 *data);
> >  static int uretprobe_dispatcher(struct uprobe_consumer *con,
> > -				unsigned long func, struct pt_regs *regs);
> > +				unsigned long func, struct pt_regs *regs,
> > +				__u64 *data);
> >  
> >  #ifdef CONFIG_STACK_GROWSUP
> >  static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
> > @@ -1517,7 +1519,8 @@ trace_uprobe_register(struct trace_event_call *event, enum trace_reg type,
> >  	}
> >  }
> >  
> > -static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
> > +static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs,
> > +			     __u64 *data)
> >  {
> >  	struct trace_uprobe *tu;
> >  	struct uprobe_dispatch_data udd;
> > @@ -1548,7 +1551,8 @@ static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
> >  }
> >  
> >  static int uretprobe_dispatcher(struct uprobe_consumer *con,
> > -				unsigned long func, struct pt_regs *regs)
> > +				unsigned long func, struct pt_regs *regs,
> > +				__u64 *data)
> >  {
> >  	struct trace_uprobe *tu;
> >  	struct uprobe_dispatch_data udd;
> > diff --git a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > index 8835761d9a12..12005e3dc3e4 100644
> > --- a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > +++ b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > @@ -461,7 +461,7 @@ static struct bin_attribute bin_attr_bpf_testmod_file __ro_after_init = {
> >  
> >  static int
> >  uprobe_ret_handler(struct uprobe_consumer *self, unsigned long func,
> > -		   struct pt_regs *regs)
> > +		   struct pt_regs *regs, __u64 *data)
> >  
> >  {
> >  	regs->ax  = 0x12345678deadbeef;
> > -- 
> > 2.46.2
> > 
> 
> 
> -- 
> Masami Hiramatsu (Google) <mhiramat@kernel.org>