Merge branch 'bpf_get_task_stack'

Song Liu says:

====================
This set introduces a new helper bpf_get_task_stack(). The primary use case
is to dump all /proc/*/stack to seq_file via bpf_iter__task.

A few different approaches have been explored and compared:

  1. A simple wrapper around stack_trace_save_tsk(), as v1 [1].

     This approach introduces new syntax, which is different to existing
     helper bpf_get_stack(). Therefore, this is not ideal.

  2. Extend get_perf_callchain() to support "task" as argument.

     This approach reuses most of bpf_get_stack(). However, extending
     get_perf_callchain() requires non-trivial changes to architecture
     specific code. Which is error prone.

  3. Current (v2) approach, leverages most of existing bpf_get_stack(), and
     uses stack_trace_save_tsk() to handle architecture specific logic.

[1] https://lore.kernel.org/netdev/20200623070802.2310018-1-songliubraving@fb.com/

Changes v4 => v5:
1. Rebase and work around git-am issue. (Alexei)
2. Update commit log for 4/4. (Yonghong)

Changes v3 => v4:
1. Simplify the selftests with bpf_iter.h. (Yonghong)
2. Add example output to commit log of 4/4. (Yonghong)

Changes v2 => v3:
1. Rebase on top of bpf-next. (Yonghong)
2. Sanitize get_callchain_entry(). (Peter)
3. Use has_callchain_buf for bpf_get_task_stack. (Andrii)
4. Other small clean up. (Yonghong, Andrii).

Changes v1 => v2:
1. Reuse most of bpf_get_stack() logic. (Andrii)
2. Fix unsigned long vs. u64 mismatch for 32-bit systems. (Yonghong)
3. Add %pB support in bpf_trace_printk(). (Daniel)
4. Fix buffer size to bytes.
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

include/linux/bpf.h

@@ -1627,6 +1627,7 @@ extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
 extern const struct bpf_func_proto bpf_get_current_comm_proto;
 extern const struct bpf_func_proto bpf_get_stackid_proto;
 extern const struct bpf_func_proto bpf_get_stack_proto;
+extern const struct bpf_func_proto bpf_get_task_stack_proto;
 extern const struct bpf_func_proto bpf_sock_map_update_proto;
 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;

include/linux/perf_event.h

@@ -1244,6 +1244,8 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user,
 extern struct perf_callchain_entry *perf_callchain(struct perf_event *event, struct pt_regs *regs);
 extern int get_callchain_buffers(int max_stack);
 extern void put_callchain_buffers(void);
+extern struct perf_callchain_entry *get_callchain_entry(int *rctx);
+extern void put_callchain_entry(int rctx);
 
 extern int sysctl_perf_event_max_stack;
 extern int sysctl_perf_event_max_contexts_per_stack;

include/uapi/linux/bpf.h

@@ -3285,6 +3285,39 @@ union bpf_attr {
  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
  *	Return
  *		*sk* if casting is valid, or NULL otherwise.
+ *
+ * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
+ *	Description
+ *		Return a user or a kernel stack in bpf program provided buffer.
+ *		To achieve this, the helper needs *task*, which is a valid
+ *		pointer to struct task_struct. To store the stacktrace, the
+ *		bpf program provides *buf* with	a nonnegative *size*.
+ *
+ *		The last argument, *flags*, holds the number of stack frames to
+ *		skip (from 0 to 255), masked with
+ *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
+ *		the following flags:
+ *
+ *		**BPF_F_USER_STACK**
+ *			Collect a user space stack instead of a kernel stack.
+ *		**BPF_F_USER_BUILD_ID**
+ *			Collect buildid+offset instead of ips for user stack,
+ *			only valid if **BPF_F_USER_STACK** is also specified.
+ *
+ *		**bpf_get_task_stack**\ () can collect up to
+ *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
+ *		to sufficient large buffer size. Note that
+ *		this limit can be controlled with the **sysctl** program, and
+ *		that it should be manually increased in order to profile long
+ *		user stacks (such as stacks for Java programs). To do so, use:
+ *
+ *		::
+ *
+ *			# sysctl kernel.perf_event_max_stack=<new value>
+ *	Return
+ *		A non-negative value equal to or less than *size* on success,
+ *		or a negative error in case of failure.
+ *
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -3427,7 +3460,9 @@ union bpf_attr {
 	FN(skc_to_tcp_sock),		\
 	FN(skc_to_tcp_timewait_sock),	\
 	FN(skc_to_tcp_request_sock),	\
-	FN(skc_to_udp6_sock),
+	FN(skc_to_udp6_sock),		\
+	FN(get_task_stack),		\
+	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  * function eBPF program intends to call

kernel/bpf/stackmap.c

@@ -348,6 +348,40 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 	}
 }
 
+static struct perf_callchain_entry *
+get_callchain_entry_for_task(struct task_struct *task, u32 init_nr)
+{
+	struct perf_callchain_entry *entry;
+	int rctx;
+
+	entry = get_callchain_entry(&rctx);
+
+	if (!entry)
+		return NULL;
+
+	entry->nr = init_nr +
+		stack_trace_save_tsk(task, (unsigned long *)(entry->ip + init_nr),
+				     sysctl_perf_event_max_stack - init_nr, 0);
+
+	/* stack_trace_save_tsk() works on unsigned long array, while
+	 * perf_callchain_entry uses u64 array. For 32-bit systems, it is
+	 * necessary to fix this mismatch.
+	 */
+	if (__BITS_PER_LONG != 64) {
+		unsigned long *from = (unsigned long *) entry->ip;
+		u64 *to = entry->ip;
+		int i;
+
+		/* copy data from the end to avoid using extra buffer */
+		for (i = entry->nr - 1; i >= (int)init_nr; i--)
+			to[i] = (u64)(from[i]);
+	}
+
+	put_callchain_entry(rctx);
+
+	return entry;
+}
+
 BPF_CALL_3(bpf_get_stackid, struct pt_regs *, regs, struct bpf_map *, map,
 	   u64, flags)
 {
@@ -448,8 +482,8 @@ const struct bpf_func_proto bpf_get_stackid_proto = {
 	.arg3_type	= ARG_ANYTHING,
 };
 
-BPF_CALL_4(bpf_get_stack, struct pt_regs *, regs, void *, buf, u32, size,
-	   u64, flags)
+static long __bpf_get_stack(struct pt_regs *regs, struct task_struct *task,
+			    void *buf, u32 size, u64 flags)
 {
 	u32 init_nr, trace_nr, copy_len, elem_size, num_elem;
 	bool user_build_id = flags & BPF_F_USER_BUILD_ID;
@@ -471,13 +505,22 @@ BPF_CALL_4(bpf_get_stack, struct pt_regs *, regs, void *, buf, u32, size,
 	if (unlikely(size % elem_size))
 		goto clear;
 
+	/* cannot get valid user stack for task without user_mode regs */
+	if (task && user && !user_mode(regs))
+		goto err_fault;
+
 	num_elem = size / elem_size;
 	if (sysctl_perf_event_max_stack < num_elem)
 		init_nr = 0;
 	else
 		init_nr = sysctl_perf_event_max_stack - num_elem;
-	trace = get_perf_callchain(regs, init_nr, kernel, user,
-				   sysctl_perf_event_max_stack, false, false);
+
+	if (kernel && task)
+		trace = get_callchain_entry_for_task(task, init_nr);
+	else
+		trace = get_perf_callchain(regs, init_nr, kernel, user,
+					   sysctl_perf_event_max_stack,
+					   false, false);
 	if (unlikely(!trace))
 		goto err_fault;
 
@@ -505,6 +548,12 @@ BPF_CALL_4(bpf_get_stack, struct pt_regs *, regs, void *, buf, u32, size,
 	return err;
 }
 
+BPF_CALL_4(bpf_get_stack, struct pt_regs *, regs, void *, buf, u32, size,
+	   u64, flags)
+{
+	return __bpf_get_stack(regs, NULL, buf, size, flags);
+}
+
 const struct bpf_func_proto bpf_get_stack_proto = {
 	.func		= bpf_get_stack,
 	.gpl_only	= true,
@@ -515,6 +564,26 @@ const struct bpf_func_proto bpf_get_stack_proto = {
 	.arg4_type	= ARG_ANYTHING,
 };
 
+BPF_CALL_4(bpf_get_task_stack, struct task_struct *, task, void *, buf,
+	   u32, size, u64, flags)
+{
+	struct pt_regs *regs = task_pt_regs(task);
+
+	return __bpf_get_stack(regs, task, buf, size, flags);
+}
+
+static int bpf_get_task_stack_btf_ids[5];
+const struct bpf_func_proto bpf_get_task_stack_proto = {
+	.func		= bpf_get_task_stack,
+	.gpl_only	= false,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg2_type	= ARG_PTR_TO_UNINIT_MEM,
+	.arg3_type	= ARG_CONST_SIZE_OR_ZERO,
+	.arg4_type	= ARG_ANYTHING,
+	.btf_id		= bpf_get_task_stack_btf_ids,
+};
+
 /* Called from eBPF program */
 static void *stack_map_lookup_elem(struct bpf_map *map, void *key)
 {

kernel/bpf/verifier.c

@@ -4864,7 +4864,9 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn
 	if (err)
 		return err;
 
-	if (func_id == BPF_FUNC_get_stack && !env->prog->has_callchain_buf) {
+	if ((func_id == BPF_FUNC_get_stack ||
+	     func_id == BPF_FUNC_get_task_stack) &&
+	    !env->prog->has_callchain_buf) {
 		const char *err_str;
 
 #ifdef CONFIG_PERF_EVENTS

kernel/events/callchain.c

@@ -149,7 +149,7 @@ void put_callchain_buffers(void)
 	}
 }
 
-static struct perf_callchain_entry *get_callchain_entry(int *rctx)
+struct perf_callchain_entry *get_callchain_entry(int *rctx)
 {
 	int cpu;
 	struct callchain_cpus_entries *entries;
@@ -159,8 +159,10 @@ static struct perf_callchain_entry *get_callchain_entry(int *rctx)
 		return NULL;
 
 	entries = rcu_dereference(callchain_cpus_entries);
-	if (!entries)
+	if (!entries) {
+		put_recursion_context(this_cpu_ptr(callchain_recursion), *rctx);
 		return NULL;
+	}
 
 	cpu = smp_processor_id();
 
@@ -168,7 +170,7 @@ static struct perf_callchain_entry *get_callchain_entry(int *rctx)
 		(*rctx * perf_callchain_entry__sizeof()));
 }
 
-static void
+void
 put_callchain_entry(int rctx)
 {
 	put_recursion_context(this_cpu_ptr(callchain_recursion), rctx);
@@ -183,11 +185,8 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user,
 	int rctx;
 
 	entry = get_callchain_entry(&rctx);
-	if (rctx == -1)
-		return NULL;
-
 	if (!entry)
-		goto exit_put;
+		return NULL;
 
 	ctx.entry     = entry;
 	ctx.max_stack = max_stack;

kernel/trace/bpf_trace.c

@@ -376,7 +376,7 @@ static void bpf_trace_copy_string(char *buf, void *unsafe_ptr, char fmt_ptype,
 
 /*
  * Only limited trace_printk() conversion specifiers allowed:
- * %d %i %u %x %ld %li %lu %lx %lld %lli %llu %llx %p %pks %pus %s
+ * %d %i %u %x %ld %li %lu %lx %lld %lli %llu %llx %p %pB %pks %pus %s
  */
 BPF_CALL_5(bpf_trace_printk, char *, fmt, u32, fmt_size, u64, arg1,
 	   u64, arg2, u64, arg3)
@@ -420,6 +420,11 @@ BPF_CALL_5(bpf_trace_printk, char *, fmt, u32, fmt_size, u64, arg1,
 				goto fmt_str;
 			}
 
+			if (fmt[i + 1] == 'B') {
+				i++;
+				goto fmt_next;
+			}
+
 			/* disallow any further format extensions */
 			if (fmt[i + 1] != 0 &&
 			    !isspace(fmt[i + 1]) &&
@@ -636,7 +641,8 @@ BPF_CALL_5(bpf_seq_printf, struct seq_file *, m, char *, fmt, u32, fmt_size,
 		if (fmt[i] == 'p') {
 			if (fmt[i + 1] == 0 ||
 			    fmt[i + 1] == 'K' ||
-			    fmt[i + 1] == 'x') {
+			    fmt[i + 1] == 'x' ||
+			    fmt[i + 1] == 'B') {
 				/* just kernel pointers */
 				params[fmt_cnt] = args[fmt_cnt];
 				fmt_cnt++;
@@ -1137,6 +1143,8 @@ bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_ringbuf_query_proto;
 	case BPF_FUNC_jiffies64:
 		return &bpf_jiffies64_proto;
+	case BPF_FUNC_get_task_stack:
+		return &bpf_get_task_stack_proto;
 	default:
 		return NULL;
 	}

scripts/bpf_helpers_doc.py

@@ -426,6 +426,7 @@ class PrinterHelpers(Printer):
             'struct tcp_timewait_sock',
             'struct tcp_request_sock',
             'struct udp6_sock',
+            'struct task_struct',
 
             'struct __sk_buff',
             'struct sk_msg_md',
@@ -468,6 +469,7 @@ class PrinterHelpers(Printer):
             'struct tcp_timewait_sock',
             'struct tcp_request_sock',
             'struct udp6_sock',
+            'struct task_struct',
     }
     mapped_types = {
             'u8': '__u8',

tools/include/uapi/linux/bpf.h

@@ -3285,6 +3285,39 @@ union bpf_attr {
  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
  *	Return
  *		*sk* if casting is valid, or NULL otherwise.
+ *
+ * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
+ *	Description
+ *		Return a user or a kernel stack in bpf program provided buffer.
+ *		To achieve this, the helper needs *task*, which is a valid
+ *		pointer to struct task_struct. To store the stacktrace, the
+ *		bpf program provides *buf* with	a nonnegative *size*.
+ *
+ *		The last argument, *flags*, holds the number of stack frames to
+ *		skip (from 0 to 255), masked with
+ *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
+ *		the following flags:
+ *
+ *		**BPF_F_USER_STACK**
+ *			Collect a user space stack instead of a kernel stack.
+ *		**BPF_F_USER_BUILD_ID**
+ *			Collect buildid+offset instead of ips for user stack,
+ *			only valid if **BPF_F_USER_STACK** is also specified.
+ *
+ *		**bpf_get_task_stack**\ () can collect up to
+ *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
+ *		to sufficient large buffer size. Note that
+ *		this limit can be controlled with the **sysctl** program, and
+ *		that it should be manually increased in order to profile long
+ *		user stacks (such as stacks for Java programs). To do so, use:
+ *
+ *		::
+ *
+ *			# sysctl kernel.perf_event_max_stack=<new value>
+ *	Return
+ *		A non-negative value equal to or less than *size* on success,
+ *		or a negative error in case of failure.
+ *
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -3427,7 +3460,9 @@ union bpf_attr {
 	FN(skc_to_tcp_sock),		\
 	FN(skc_to_tcp_timewait_sock),	\
 	FN(skc_to_tcp_request_sock),	\
-	FN(skc_to_udp6_sock),
+	FN(skc_to_udp6_sock),		\
+	FN(get_task_stack),		\
+	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  * function eBPF program intends to call

tools/testing/selftests/bpf/prog_tests/bpf_iter.c

@@ -5,6 +5,7 @@
 #include "bpf_iter_netlink.skel.h"
 #include "bpf_iter_bpf_map.skel.h"
 #include "bpf_iter_task.skel.h"
+#include "bpf_iter_task_stack.skel.h"
 #include "bpf_iter_task_file.skel.h"
 #include "bpf_iter_tcp4.skel.h"
 #include "bpf_iter_tcp6.skel.h"
@@ -110,6 +111,20 @@ static void test_task(void)
 	bpf_iter_task__destroy(skel);
 }
 
+static void test_task_stack(void)
+{
+	struct bpf_iter_task_stack *skel;
+
+	skel = bpf_iter_task_stack__open_and_load();
+	if (CHECK(!skel, "bpf_iter_task_stack__open_and_load",
+		  "skeleton open_and_load failed\n"))
+		return;
+
+	do_dummy_read(skel->progs.dump_task_stack);
+
+	bpf_iter_task_stack__destroy(skel);
+}
+
 static void test_task_file(void)
 {
 	struct bpf_iter_task_file *skel;
@@ -452,6 +467,8 @@ void test_bpf_iter(void)
 		test_bpf_map();
 	if (test__start_subtest("task"))
 		test_task();
+	if (test__start_subtest("task_stack"))
+		test_task_stack();
 	if (test__start_subtest("task_file"))
 		test_task_file();
 	if (test__start_subtest("tcp4"))

tools/testing/selftests/bpf/progs/bpf_iter_task_stack.c

@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include "bpf_iter.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+char _license[] SEC("license") = "GPL";
+
+#define MAX_STACK_TRACE_DEPTH   64
+unsigned long entries[MAX_STACK_TRACE_DEPTH];
+#define SIZE_OF_ULONG (sizeof(unsigned long))
+
+SEC("iter/task")
+int dump_task_stack(struct bpf_iter__task *ctx)
+{
+	struct seq_file *seq = ctx->meta->seq;
+	struct task_struct *task = ctx->task;
+	long i, retlen;
+
+	if (task == (void *)0)
+		return 0;
+
+	retlen = bpf_get_task_stack(task, entries,
+				    MAX_STACK_TRACE_DEPTH * SIZE_OF_ULONG, 0);
+	if (retlen < 0)
+		return 0;
+
+	BPF_SEQ_PRINTF(seq, "pid: %8u num_entries: %8u\n", task->pid,
+		       retlen / SIZE_OF_ULONG);
+	for (i = 0; i < MAX_STACK_TRACE_DEPTH; i++) {
+		if (retlen > i * SIZE_OF_ULONG)
+			BPF_SEQ_PRINTF(seq, "[<0>] %pB\n", (void *)entries[i]);
+	}
+	BPF_SEQ_PRINTF(seq, "\n");
+
+	return 0;
+}