1 /* Copyright (c) 2011-2015 PLUMgrid, http://plumgrid.com
3 * This program is free software; you can redistribute it and/or
4 * modify it under the terms of version 2 of the GNU General Public
5 * License as published by the Free Software Foundation.
7 #include <linux/kernel.h>
8 #include <linux/types.h>
9 #include <linux/slab.h>
10 #include <linux/bpf.h>
11 #include <linux/filter.h>
12 #include <linux/uaccess.h>
13 #include <linux/ctype.h>
17 * trace_call_bpf - invoke BPF program
19 * @ctx: opaque context pointer
21 * kprobe handlers execute BPF programs via this helper.
22 * Can be used from static tracepoints in the future.
24 * Return: BPF programs always return an integer which is interpreted by
26 * 0 - return from kprobe (event is filtered out)
27 * 1 - store kprobe event into ring buffer
28 * Other values are reserved and currently alias to 1
30 unsigned int trace_call_bpf(struct bpf_prog
*prog
, void *ctx
)
34 if (in_nmi()) /* not supported yet */
39 if (unlikely(__this_cpu_inc_return(bpf_prog_active
) != 1)) {
41 * since some bpf program is already running on this cpu,
42 * don't call into another bpf program (same or different)
43 * and don't send kprobe event into ring-buffer,
51 ret
= BPF_PROG_RUN(prog
, ctx
);
55 __this_cpu_dec(bpf_prog_active
);
60 EXPORT_SYMBOL_GPL(trace_call_bpf
);
62 static u64
bpf_probe_read(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
)
64 void *dst
= (void *) (long) r1
;
66 void *unsafe_ptr
= (void *) (long) r3
;
68 return probe_kernel_read(dst
, unsafe_ptr
, size
);
71 static const struct bpf_func_proto bpf_probe_read_proto
= {
72 .func
= bpf_probe_read
,
74 .ret_type
= RET_INTEGER
,
75 .arg1_type
= ARG_PTR_TO_STACK
,
76 .arg2_type
= ARG_CONST_STACK_SIZE
,
77 .arg3_type
= ARG_ANYTHING
,
81 * limited trace_printk()
82 * only %d %u %x %ld %lu %lx %lld %llu %llx %p %s conversion specifiers allowed
84 static u64
bpf_trace_printk(u64 r1
, u64 fmt_size
, u64 r3
, u64 r4
, u64 r5
)
86 char *fmt
= (char *) (long) r1
;
87 bool str_seen
= false;
95 * bpf_check()->check_func_arg()->check_stack_boundary()
96 * guarantees that fmt points to bpf program stack,
97 * fmt_size bytes of it were initialized and fmt_size > 0
99 if (fmt
[--fmt_size
] != 0)
102 /* check format string for allowed specifiers */
103 for (i
= 0; i
< fmt_size
; i
++) {
104 if ((!isprint(fmt
[i
]) && !isspace(fmt
[i
])) || !isascii(fmt
[i
]))
113 /* fmt[i] != 0 && fmt[last] == 0, so we can access fmt[i + 1] */
118 } else if (fmt
[i
] == 'p' || fmt
[i
] == 's') {
121 if (!isspace(fmt
[i
]) && !ispunct(fmt
[i
]) && fmt
[i
] != 0)
124 if (fmt
[i
- 1] == 's') {
126 /* allow only one '%s' per fmt string */
145 strncpy_from_unsafe(buf
,
146 (void *) (long) unsafe_addr
,
157 if (fmt
[i
] != 'd' && fmt
[i
] != 'u' && fmt
[i
] != 'x')
162 return __trace_printk(1/* fake ip will not be printed */, fmt
,
163 mod
[0] == 2 ? r3
: mod
[0] == 1 ? (long) r3
: (u32
) r3
,
164 mod
[1] == 2 ? r4
: mod
[1] == 1 ? (long) r4
: (u32
) r4
,
165 mod
[2] == 2 ? r5
: mod
[2] == 1 ? (long) r5
: (u32
) r5
);
168 static const struct bpf_func_proto bpf_trace_printk_proto
= {
169 .func
= bpf_trace_printk
,
171 .ret_type
= RET_INTEGER
,
172 .arg1_type
= ARG_PTR_TO_STACK
,
173 .arg2_type
= ARG_CONST_STACK_SIZE
,
176 const struct bpf_func_proto
*bpf_get_trace_printk_proto(void)
179 * this program might be calling bpf_trace_printk,
180 * so allocate per-cpu printk buffers
182 trace_printk_init_buffers();
184 return &bpf_trace_printk_proto
;
187 static u64
bpf_perf_event_read(u64 r1
, u64 index
, u64 r3
, u64 r4
, u64 r5
)
189 struct bpf_map
*map
= (struct bpf_map
*) (unsigned long) r1
;
190 struct bpf_array
*array
= container_of(map
, struct bpf_array
, map
);
191 struct perf_event
*event
;
194 if (unlikely(index
>= array
->map
.max_entries
))
197 file
= (struct file
*)array
->ptrs
[index
];
201 event
= file
->private_data
;
203 /* make sure event is local and doesn't have pmu::count */
204 if (event
->oncpu
!= smp_processor_id() ||
209 * we don't know if the function is run successfully by the
210 * return value. It can be judged in other places, such as
213 return perf_event_read_local(event
);
216 static const struct bpf_func_proto bpf_perf_event_read_proto
= {
217 .func
= bpf_perf_event_read
,
219 .ret_type
= RET_INTEGER
,
220 .arg1_type
= ARG_CONST_MAP_PTR
,
221 .arg2_type
= ARG_ANYTHING
,
224 static u64
bpf_perf_event_output(u64 r1
, u64 r2
, u64 index
, u64 r4
, u64 size
)
226 struct pt_regs
*regs
= (struct pt_regs
*) (long) r1
;
227 struct bpf_map
*map
= (struct bpf_map
*) (long) r2
;
228 struct bpf_array
*array
= container_of(map
, struct bpf_array
, map
);
229 void *data
= (void *) (long) r4
;
230 struct perf_sample_data sample_data
;
231 struct perf_event
*event
;
233 struct perf_raw_record raw
= {
238 if (unlikely(index
>= array
->map
.max_entries
))
241 file
= (struct file
*)array
->ptrs
[index
];
245 event
= file
->private_data
;
247 if (unlikely(event
->attr
.type
!= PERF_TYPE_SOFTWARE
||
248 event
->attr
.config
!= PERF_COUNT_SW_BPF_OUTPUT
))
251 if (unlikely(event
->oncpu
!= smp_processor_id()))
254 perf_sample_data_init(&sample_data
, 0, 0);
255 sample_data
.raw
= &raw
;
256 perf_event_output(event
, &sample_data
, regs
);
260 static const struct bpf_func_proto bpf_perf_event_output_proto
= {
261 .func
= bpf_perf_event_output
,
263 .ret_type
= RET_INTEGER
,
264 .arg1_type
= ARG_PTR_TO_CTX
,
265 .arg2_type
= ARG_CONST_MAP_PTR
,
266 .arg3_type
= ARG_ANYTHING
,
267 .arg4_type
= ARG_PTR_TO_STACK
,
268 .arg5_type
= ARG_CONST_STACK_SIZE
,
271 static const struct bpf_func_proto
*kprobe_prog_func_proto(enum bpf_func_id func_id
)
274 case BPF_FUNC_map_lookup_elem
:
275 return &bpf_map_lookup_elem_proto
;
276 case BPF_FUNC_map_update_elem
:
277 return &bpf_map_update_elem_proto
;
278 case BPF_FUNC_map_delete_elem
:
279 return &bpf_map_delete_elem_proto
;
280 case BPF_FUNC_probe_read
:
281 return &bpf_probe_read_proto
;
282 case BPF_FUNC_ktime_get_ns
:
283 return &bpf_ktime_get_ns_proto
;
284 case BPF_FUNC_tail_call
:
285 return &bpf_tail_call_proto
;
286 case BPF_FUNC_get_current_pid_tgid
:
287 return &bpf_get_current_pid_tgid_proto
;
288 case BPF_FUNC_get_current_uid_gid
:
289 return &bpf_get_current_uid_gid_proto
;
290 case BPF_FUNC_get_current_comm
:
291 return &bpf_get_current_comm_proto
;
292 case BPF_FUNC_trace_printk
:
293 return bpf_get_trace_printk_proto();
294 case BPF_FUNC_get_smp_processor_id
:
295 return &bpf_get_smp_processor_id_proto
;
296 case BPF_FUNC_perf_event_read
:
297 return &bpf_perf_event_read_proto
;
298 case BPF_FUNC_perf_event_output
:
299 return &bpf_perf_event_output_proto
;
300 case BPF_FUNC_get_stackid
:
301 return &bpf_get_stackid_proto
;
307 /* bpf+kprobe programs can access fields of 'struct pt_regs' */
308 static bool kprobe_prog_is_valid_access(int off
, int size
, enum bpf_access_type type
)
311 if (off
< 0 || off
>= sizeof(struct pt_regs
))
314 /* only read is allowed */
315 if (type
!= BPF_READ
)
318 /* disallow misaligned access */
325 static const struct bpf_verifier_ops kprobe_prog_ops
= {
326 .get_func_proto
= kprobe_prog_func_proto
,
327 .is_valid_access
= kprobe_prog_is_valid_access
,
330 static struct bpf_prog_type_list kprobe_tl
= {
331 .ops
= &kprobe_prog_ops
,
332 .type
= BPF_PROG_TYPE_KPROBE
,
335 static int __init
register_kprobe_prog_ops(void)
337 bpf_register_prog_type(&kprobe_tl
);
340 late_initcall(register_kprobe_prog_ops
);