Skip to content

Commit d72e296

Browse files
anakryikoAlexei Starovoitov
anakryiko
authored and
Alexei Starovoitov
committed
libbpf: Add BPF-side of USDT support
Add BPF-side implementation of libbpf-provided USDT support. This consists of single header library, usdt.bpf.h, which is meant to be used from user's BPF-side source code. This header is added to the list of installed libbpf header, along bpf_helpers.h and others. BPF-side implementation consists of two BPF maps: - spec map, which contains "a USDT spec" which encodes information necessary to be able to fetch USDT arguments and other information (argument count, user-provided cookie value, etc) at runtime; - IP-to-spec-ID map, which is only used on kernels that don't support BPF cookie feature. It allows to lookup spec ID based on the place in user application that triggers USDT program. These maps have default sizes, 256 and 1024, which are chosen conservatively to not waste a lot of space, but handling a lot of common cases. But there could be cases when user application needs to either trace a lot of different USDTs, or USDTs are heavily inlined and their arguments are located in a lot of differing locations. For such cases it might be necessary to size those maps up, which libbpf allows to do by overriding BPF_USDT_MAX_SPEC_CNT and BPF_USDT_MAX_IP_CNT macros. It is an important aspect to keep in mind. Single USDT (user-space equivalent of kernel tracepoint) can have multiple USDT "call sites". That is, single logical USDT is triggered from multiple places in user application. This can happen due to function inlining. Each such inlined instance of USDT invocation can have its own unique USDT argument specification (instructions about the location of the value of each of USDT arguments). So while USDT looks very similar to usual uprobe or kernel tracepoint, under the hood it's actually a collection of uprobes, each potentially needing different spec to know how to fetch arguments. User-visible API consists of three helper functions: - bpf_usdt_arg_cnt(), which returns number of arguments of current USDT; - bpf_usdt_arg(), which reads value of specified USDT argument (by it's zero-indexed position) and returns it as 64-bit value; - bpf_usdt_cookie(), which functions like BPF cookie for USDT programs; this is necessary as libbpf doesn't allow specifying actual BPF cookie and utilizes it internally for USDT support implementation. Each bpf_usdt_xxx() APIs expect struct pt_regs * context, passed into BPF program. On kernels that don't support BPF cookie it is used to fetch absolute IP address of the underlying uprobe. usdt.bpf.h also provides BPF_USDT() macro, which functions like BPF_PROG() and BPF_KPROBE() and allows much more user-friendly way to get access to USDT arguments, if USDT definition is static and known to the user. It is expected that majority of use cases won't have to use bpf_usdt_arg_cnt() and bpf_usdt_arg() directly and BPF_USDT() will cover all their needs. Last, usdt.bpf.h is utilizing BPF CO-RE for one single purpose: to detect kernel support for BPF cookie. If BPF CO-RE dependency is undesirable, user application can redefine BPF_USDT_HAS_BPF_COOKIE to either a boolean constant (or equivalently zero and non-zero), or even point it to its own .rodata variable that can be specified from user's application user-space code. It is important that BPF_USDT_HAS_BPF_COOKIE is known to BPF verifier as static value (thus .rodata and not just .data), as otherwise BPF code will still contain bpf_get_attach_cookie() BPF helper call and will fail validation at runtime, if not dead-code eliminated. Signed-off-by: Andrii Nakryiko <[email protected]> Signed-off-by: Alexei Starovoitov <[email protected]> Reviewed-by: Alan Maguire <[email protected]> Link: https://lore.kernel.org/bpf/[email protected]
1 parent 5681893 commit d72e296

File tree

2 files changed

+257
-1
lines changed

2 files changed

+257
-1
lines changed

‎tools/lib/bpf/Makefile

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -239,7 +239,7 @@ install_lib: all_cmd
239239

240240
SRC_HDRS := bpf.h libbpf.h btf.h libbpf_common.h libbpf_legacy.h xsk.h \
241241
bpf_helpers.h bpf_tracing.h bpf_endian.h bpf_core_read.h \
242-
skel_internal.h libbpf_version.h
242+
skel_internal.h libbpf_version.h usdt.bpf.h
243243
GEN_HDRS := $(BPF_GENERATED)
244244

245245
INSTALL_PFX := $(DESTDIR)$(prefix)/include/bpf

‎tools/lib/bpf/usdt.bpf.h

Lines changed: 256 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,256 @@
1+
/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
2+
/* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */
3+
#ifndef __USDT_BPF_H__
4+
#define __USDT_BPF_H__
5+
6+
#include <linux/errno.h>
7+
#include <bpf/bpf_helpers.h>
8+
#include <bpf/bpf_tracing.h>
9+
#include <bpf/bpf_core_read.h>
10+
11+
/* Below types and maps are internal implementation details of libbpf's USDT
12+
* support and are subjects to change. Also, bpf_usdt_xxx() API helpers should
13+
* be considered an unstable API as well and might be adjusted based on user
14+
* feedback from using libbpf's USDT support in production.
15+
*/
16+
17+
/* User can override BPF_USDT_MAX_SPEC_CNT to change default size of internal
18+
* map that keeps track of USDT argument specifications. This might be
19+
* necessary if there are a lot of USDT attachments.
20+
*/
21+
#ifndef BPF_USDT_MAX_SPEC_CNT
22+
#define BPF_USDT_MAX_SPEC_CNT 256
23+
#endif
24+
/* User can override BPF_USDT_MAX_IP_CNT to change default size of internal
25+
* map that keeps track of IP (memory address) mapping to USDT argument
26+
* specification.
27+
* Note, if kernel supports BPF cookies, this map is not used and could be
28+
* resized all the way to 1 to save a bit of memory.
29+
*/
30+
#ifndef BPF_USDT_MAX_IP_CNT
31+
#define BPF_USDT_MAX_IP_CNT (4 * BPF_USDT_MAX_SPEC_CNT)
32+
#endif
33+
/* We use BPF CO-RE to detect support for BPF cookie from BPF side. This is
34+
* the only dependency on CO-RE, so if it's undesirable, user can override
35+
* BPF_USDT_HAS_BPF_COOKIE to specify whether to BPF cookie is supported or not.
36+
*/
37+
#ifndef BPF_USDT_HAS_BPF_COOKIE
38+
#define BPF_USDT_HAS_BPF_COOKIE \
39+
bpf_core_enum_value_exists(enum bpf_func_id___usdt, BPF_FUNC_get_attach_cookie___usdt)
40+
#endif
41+
42+
enum __bpf_usdt_arg_type {
43+
BPF_USDT_ARG_CONST,
44+
BPF_USDT_ARG_REG,
45+
BPF_USDT_ARG_REG_DEREF,
46+
};
47+
48+
struct __bpf_usdt_arg_spec {
49+
/* u64 scalar interpreted depending on arg_type, see below */
50+
__u64 val_off;
51+
/* arg location case, see bpf_udst_arg() for details */
52+
enum __bpf_usdt_arg_type arg_type;
53+
/* offset of referenced register within struct pt_regs */
54+
short reg_off;
55+
/* whether arg should be interpreted as signed value */
56+
bool arg_signed;
57+
/* number of bits that need to be cleared and, optionally,
58+
* sign-extended to cast arguments that are 1, 2, or 4 bytes
59+
* long into final 8-byte u64/s64 value returned to user
60+
*/
61+
char arg_bitshift;
62+
};
63+
64+
/* should match USDT_MAX_ARG_CNT in usdt.c exactly */
65+
#define BPF_USDT_MAX_ARG_CNT 12
66+
struct __bpf_usdt_spec {
67+
struct __bpf_usdt_arg_spec args[BPF_USDT_MAX_ARG_CNT];
68+
__u64 usdt_cookie;
69+
short arg_cnt;
70+
};
71+
72+
struct {
73+
__uint(type, BPF_MAP_TYPE_ARRAY);
74+
__uint(max_entries, BPF_USDT_MAX_SPEC_CNT);
75+
__type(key, int);
76+
__type(value, struct __bpf_usdt_spec);
77+
} __bpf_usdt_specs SEC(".maps") __weak;
78+
79+
struct {
80+
__uint(type, BPF_MAP_TYPE_HASH);
81+
__uint(max_entries, BPF_USDT_MAX_IP_CNT);
82+
__type(key, long);
83+
__type(value, __u32);
84+
} __bpf_usdt_ip_to_spec_id SEC(".maps") __weak;
85+
86+
/* don't rely on user's BPF code to have latest definition of bpf_func_id */
87+
enum bpf_func_id___usdt {
88+
BPF_FUNC_get_attach_cookie___usdt = 0xBAD, /* value doesn't matter */
89+
};
90+
91+
static __always_inline
92+
int __bpf_usdt_spec_id(struct pt_regs *ctx)
93+
{
94+
if (!BPF_USDT_HAS_BPF_COOKIE) {
95+
long ip = PT_REGS_IP(ctx);
96+
int *spec_id_ptr;
97+
98+
spec_id_ptr = bpf_map_lookup_elem(&__bpf_usdt_ip_to_spec_id, &ip);
99+
return spec_id_ptr ? *spec_id_ptr : -ESRCH;
100+
}
101+
102+
return bpf_get_attach_cookie(ctx);
103+
}
104+
105+
/* Return number of USDT arguments defined for currently traced USDT. */
106+
static inline __noinline
107+
int bpf_usdt_arg_cnt(struct pt_regs *ctx)
108+
{
109+
struct __bpf_usdt_spec *spec;
110+
int spec_id;
111+
112+
spec_id = __bpf_usdt_spec_id(ctx);
113+
if (spec_id < 0)
114+
return -ESRCH;
115+
116+
spec = bpf_map_lookup_elem(&__bpf_usdt_specs, &spec_id);
117+
if (!spec)
118+
return -ESRCH;
119+
120+
return spec->arg_cnt;
121+
}
122+
123+
/* Fetch USDT argument #*arg_num* (zero-indexed) and put its value into *res.
124+
* Returns 0 on success; negative error, otherwise.
125+
* On error *res is guaranteed to be set to zero.
126+
*/
127+
static inline __noinline
128+
int bpf_usdt_arg(struct pt_regs *ctx, __u64 arg_num, long *res)
129+
{
130+
struct __bpf_usdt_spec *spec;
131+
struct __bpf_usdt_arg_spec *arg_spec;
132+
unsigned long val;
133+
int err, spec_id;
134+
135+
*res = 0;
136+
137+
spec_id = __bpf_usdt_spec_id(ctx);
138+
if (spec_id < 0)
139+
return -ESRCH;
140+
141+
spec = bpf_map_lookup_elem(&__bpf_usdt_specs, &spec_id);
142+
if (!spec)
143+
return -ESRCH;
144+
145+
if (arg_num >= BPF_USDT_MAX_ARG_CNT || arg_num >= spec->arg_cnt)
146+
return -ENOENT;
147+
148+
arg_spec = &spec->args[arg_num];
149+
switch (arg_spec->arg_type) {
150+
case BPF_USDT_ARG_CONST:
151+
/* Arg is just a constant ("-4@$-9" in USDT arg spec).
152+
* value is recorded in arg_spec->val_off directly.
153+
*/
154+
val = arg_spec->val_off;
155+
break;
156+
case BPF_USDT_ARG_REG:
157+
/* Arg is in a register (e.g, "8@%rax" in USDT arg spec),
158+
* so we read the contents of that register directly from
159+
* struct pt_regs. To keep things simple user-space parts
160+
* record offsetof(struct pt_regs, <regname>) in arg_spec->reg_off.
161+
*/
162+
err = bpf_probe_read_kernel(&val, sizeof(val), (void *)ctx + arg_spec->reg_off);
163+
if (err)
164+
return err;
165+
break;
166+
case BPF_USDT_ARG_REG_DEREF:
167+
/* Arg is in memory addressed by register, plus some offset
168+
* (e.g., "-4@-1204(%rbp)" in USDT arg spec). Register is
169+
* identified lik with BPF_USDT_ARG_REG case, and the offset
170+
* is in arg_spec->val_off. We first fetch register contents
171+
* from pt_regs, then do another user-space probe read to
172+
* fetch argument value itself.
173+
*/
174+
err = bpf_probe_read_kernel(&val, sizeof(val), (void *)ctx + arg_spec->reg_off);
175+
if (err)
176+
return err;
177+
err = bpf_probe_read_user(&val, sizeof(val), (void *)val + arg_spec->val_off);
178+
if (err)
179+
return err;
180+
break;
181+
default:
182+
return -EINVAL;
183+
}
184+
185+
/* cast arg from 1, 2, or 4 bytes to final 8 byte size clearing
186+
* necessary upper arg_bitshift bits, with sign extension if argument
187+
* is signed
188+
*/
189+
val <<= arg_spec->arg_bitshift;
190+
if (arg_spec->arg_signed)
191+
val = ((long)val) >> arg_spec->arg_bitshift;
192+
else
193+
val = val >> arg_spec->arg_bitshift;
194+
*res = val;
195+
return 0;
196+
}
197+
198+
/* Retrieve user-specified cookie value provided during attach as
199+
* bpf_usdt_opts.usdt_cookie. This serves the same purpose as BPF cookie
200+
* returned by bpf_get_attach_cookie(). Libbpf's support for USDT is itself
201+
* utilizaing BPF cookies internally, so user can't use BPF cookie directly
202+
* for USDT programs and has to use bpf_usdt_cookie() API instead.
203+
*/
204+
static inline __noinline
205+
long bpf_usdt_cookie(struct pt_regs *ctx)
206+
{
207+
struct __bpf_usdt_spec *spec;
208+
int spec_id;
209+
210+
spec_id = __bpf_usdt_spec_id(ctx);
211+
if (spec_id < 0)
212+
return 0;
213+
214+
spec = bpf_map_lookup_elem(&__bpf_usdt_specs, &spec_id);
215+
if (!spec)
216+
return 0;
217+
218+
return spec->usdt_cookie;
219+
}
220+
221+
/* we rely on ___bpf_apply() and ___bpf_narg() macros already defined in bpf_tracing.h */
222+
#define ___bpf_usdt_args0() ctx
223+
#define ___bpf_usdt_args1(x) ___bpf_usdt_args0(), ({ long _x; bpf_usdt_arg(ctx, 0, &_x); (void *)_x; })
224+
#define ___bpf_usdt_args2(x, args...) ___bpf_usdt_args1(args), ({ long _x; bpf_usdt_arg(ctx, 1, &_x); (void *)_x; })
225+
#define ___bpf_usdt_args3(x, args...) ___bpf_usdt_args2(args), ({ long _x; bpf_usdt_arg(ctx, 2, &_x); (void *)_x; })
226+
#define ___bpf_usdt_args4(x, args...) ___bpf_usdt_args3(args), ({ long _x; bpf_usdt_arg(ctx, 3, &_x); (void *)_x; })
227+
#define ___bpf_usdt_args5(x, args...) ___bpf_usdt_args4(args), ({ long _x; bpf_usdt_arg(ctx, 4, &_x); (void *)_x; })
228+
#define ___bpf_usdt_args6(x, args...) ___bpf_usdt_args5(args), ({ long _x; bpf_usdt_arg(ctx, 5, &_x); (void *)_x; })
229+
#define ___bpf_usdt_args7(x, args...) ___bpf_usdt_args6(args), ({ long _x; bpf_usdt_arg(ctx, 6, &_x); (void *)_x; })
230+
#define ___bpf_usdt_args8(x, args...) ___bpf_usdt_args7(args), ({ long _x; bpf_usdt_arg(ctx, 7, &_x); (void *)_x; })
231+
#define ___bpf_usdt_args9(x, args...) ___bpf_usdt_args8(args), ({ long _x; bpf_usdt_arg(ctx, 8, &_x); (void *)_x; })
232+
#define ___bpf_usdt_args10(x, args...) ___bpf_usdt_args9(args), ({ long _x; bpf_usdt_arg(ctx, 9, &_x); (void *)_x; })
233+
#define ___bpf_usdt_args11(x, args...) ___bpf_usdt_args10(args), ({ long _x; bpf_usdt_arg(ctx, 10, &_x); (void *)_x; })
234+
#define ___bpf_usdt_args12(x, args...) ___bpf_usdt_args11(args), ({ long _x; bpf_usdt_arg(ctx, 11, &_x); (void *)_x; })
235+
#define ___bpf_usdt_args(args...) ___bpf_apply(___bpf_usdt_args, ___bpf_narg(args))(args)
236+
237+
/*
238+
* BPF_USDT serves the same purpose for USDT handlers as BPF_PROG for
239+
* tp_btf/fentry/fexit BPF programs and BPF_KPROBE for kprobes.
240+
* Original struct pt_regs * context is preserved as 'ctx' argument.
241+
*/
242+
#define BPF_USDT(name, args...) \
243+
name(struct pt_regs *ctx); \
244+
static __attribute__((always_inline)) typeof(name(0)) \
245+
____##name(struct pt_regs *ctx, ##args); \
246+
typeof(name(0)) name(struct pt_regs *ctx) \
247+
{ \
248+
_Pragma("GCC diagnostic push") \
249+
_Pragma("GCC diagnostic ignored \"-Wint-conversion\"") \
250+
return ____##name(___bpf_usdt_args(args)); \
251+
_Pragma("GCC diagnostic pop") \
252+
} \
253+
static __attribute__((always_inline)) typeof(name(0)) \
254+
____##name(struct pt_regs *ctx, ##args)
255+
256+
#endif /* __USDT_BPF_H__ */

0 commit comments

Comments
 (0)