Commit | Line | Data |
---|---|---|
d27a4ddd JK |
1 | Title : Kernel Probes (Kprobes) |
2 | Authors : Jim Keniston <jkenisto@us.ibm.com> | |
3 | : Prasanna S Panchamukhi <prasanna@in.ibm.com> | |
4 | ||
5 | CONTENTS | |
6 | ||
7 | 1. Concepts: Kprobes, Jprobes, Return Probes | |
8 | 2. Architectures Supported | |
9 | 3. Configuring Kprobes | |
10 | 4. API Reference | |
11 | 5. Kprobes Features and Limitations | |
12 | 6. Probe Overhead | |
13 | 7. TODO | |
14 | 8. Kprobes Example | |
15 | 9. Jprobes Example | |
16 | 10. Kretprobes Example | |
bf8f6e5b | 17 | Appendix A: The kprobes debugfs interface |
d27a4ddd JK |
18 | |
19 | 1. Concepts: Kprobes, Jprobes, Return Probes | |
20 | ||
21 | Kprobes enables you to dynamically break into any kernel routine and | |
22 | collect debugging and performance information non-disruptively. You | |
23 | can trap at almost any kernel code address, specifying a handler | |
24 | routine to be invoked when the breakpoint is hit. | |
25 | ||
26 | There are currently three types of probes: kprobes, jprobes, and | |
27 | kretprobes (also called return probes). A kprobe can be inserted | |
28 | on virtually any instruction in the kernel. A jprobe is inserted at | |
29 | the entry to a kernel function, and provides convenient access to the | |
30 | function's arguments. A return probe fires when a specified function | |
31 | returns. | |
32 | ||
33 | In the typical case, Kprobes-based instrumentation is packaged as | |
34 | a kernel module. The module's init function installs ("registers") | |
35 | one or more probes, and the exit function unregisters them. A | |
36 | registration function such as register_kprobe() specifies where | |
37 | the probe is to be inserted and what handler is to be called when | |
38 | the probe is hit. | |
39 | ||
40 | The next three subsections explain how the different types of | |
41 | probes work. They explain certain things that you'll need to | |
42 | know in order to make the best use of Kprobes -- e.g., the | |
43 | difference between a pre_handler and a post_handler, and how | |
44 | to use the maxactive and nmissed fields of a kretprobe. But | |
45 | if you're in a hurry to start using Kprobes, you can skip ahead | |
46 | to section 2. | |
47 | ||
48 | 1.1 How Does a Kprobe Work? | |
49 | ||
50 | When a kprobe is registered, Kprobes makes a copy of the probed | |
51 | instruction and replaces the first byte(s) of the probed instruction | |
52 | with a breakpoint instruction (e.g., int3 on i386 and x86_64). | |
53 | ||
54 | When a CPU hits the breakpoint instruction, a trap occurs, the CPU's | |
55 | registers are saved, and control passes to Kprobes via the | |
56 | notifier_call_chain mechanism. Kprobes executes the "pre_handler" | |
57 | associated with the kprobe, passing the handler the addresses of the | |
58 | kprobe struct and the saved registers. | |
59 | ||
60 | Next, Kprobes single-steps its copy of the probed instruction. | |
61 | (It would be simpler to single-step the actual instruction in place, | |
62 | but then Kprobes would have to temporarily remove the breakpoint | |
63 | instruction. This would open a small time window when another CPU | |
64 | could sail right past the probepoint.) | |
65 | ||
66 | After the instruction is single-stepped, Kprobes executes the | |
67 | "post_handler," if any, that is associated with the kprobe. | |
68 | Execution then continues with the instruction following the probepoint. | |
69 | ||
70 | 1.2 How Does a Jprobe Work? | |
71 | ||
72 | A jprobe is implemented using a kprobe that is placed on a function's | |
73 | entry point. It employs a simple mirroring principle to allow | |
74 | seamless access to the probed function's arguments. The jprobe | |
75 | handler routine should have the same signature (arg list and return | |
76 | type) as the function being probed, and must always end by calling | |
77 | the Kprobes function jprobe_return(). | |
78 | ||
79 | Here's how it works. When the probe is hit, Kprobes makes a copy of | |
80 | the saved registers and a generous portion of the stack (see below). | |
81 | Kprobes then points the saved instruction pointer at the jprobe's | |
82 | handler routine, and returns from the trap. As a result, control | |
83 | passes to the handler, which is presented with the same register and | |
84 | stack contents as the probed function. When it is done, the handler | |
85 | calls jprobe_return(), which traps again to restore the original stack | |
86 | contents and processor state and switch to the probed function. | |
87 | ||
88 | By convention, the callee owns its arguments, so gcc may produce code | |
89 | that unexpectedly modifies that portion of the stack. This is why | |
90 | Kprobes saves a copy of the stack and restores it after the jprobe | |
91 | handler has run. Up to MAX_STACK_SIZE bytes are copied -- e.g., | |
92 | 64 bytes on i386. | |
93 | ||
94 | Note that the probed function's args may be passed on the stack | |
95 | or in registers (e.g., for x86_64 or for an i386 fastcall function). | |
96 | The jprobe will work in either case, so long as the handler's | |
97 | prototype matches that of the probed function. | |
98 | ||
99 | 1.3 How Does a Return Probe Work? | |
100 | ||
101 | When you call register_kretprobe(), Kprobes establishes a kprobe at | |
102 | the entry to the function. When the probed function is called and this | |
103 | probe is hit, Kprobes saves a copy of the return address, and replaces | |
104 | the return address with the address of a "trampoline." The trampoline | |
105 | is an arbitrary piece of code -- typically just a nop instruction. | |
106 | At boot time, Kprobes registers a kprobe at the trampoline. | |
107 | ||
108 | When the probed function executes its return instruction, control | |
109 | passes to the trampoline and that probe is hit. Kprobes' trampoline | |
110 | handler calls the user-specified handler associated with the kretprobe, | |
111 | then sets the saved instruction pointer to the saved return address, | |
112 | and that's where execution resumes upon return from the trap. | |
113 | ||
114 | While the probed function is executing, its return address is | |
115 | stored in an object of type kretprobe_instance. Before calling | |
116 | register_kretprobe(), the user sets the maxactive field of the | |
117 | kretprobe struct to specify how many instances of the specified | |
118 | function can be probed simultaneously. register_kretprobe() | |
119 | pre-allocates the indicated number of kretprobe_instance objects. | |
120 | ||
121 | For example, if the function is non-recursive and is called with a | |
122 | spinlock held, maxactive = 1 should be enough. If the function is | |
123 | non-recursive and can never relinquish the CPU (e.g., via a semaphore | |
124 | or preemption), NR_CPUS should be enough. If maxactive <= 0, it is | |
125 | set to a default value. If CONFIG_PREEMPT is enabled, the default | |
126 | is max(10, 2*NR_CPUS). Otherwise, the default is NR_CPUS. | |
127 | ||
128 | It's not a disaster if you set maxactive too low; you'll just miss | |
129 | some probes. In the kretprobe struct, the nmissed field is set to | |
130 | zero when the return probe is registered, and is incremented every | |
131 | time the probed function is entered but there is no kretprobe_instance | |
132 | object available for establishing the return probe. | |
133 | ||
134 | 2. Architectures Supported | |
135 | ||
136 | Kprobes, jprobes, and return probes are implemented on the following | |
137 | architectures: | |
138 | ||
139 | - i386 | |
8861da31 | 140 | - x86_64 (AMD-64, EM64T) |
d27a4ddd | 141 | - ppc64 |
8861da31 | 142 | - ia64 (Does not support probes on instruction slot1.) |
d27a4ddd JK |
143 | - sparc64 (Return probes not yet implemented.) |
144 | ||
145 | 3. Configuring Kprobes | |
146 | ||
147 | When configuring the kernel using make menuconfig/xconfig/oldconfig, | |
8861da31 JK |
148 | ensure that CONFIG_KPROBES is set to "y". Under "Instrumentation |
149 | Support", look for "Kprobes". | |
150 | ||
151 | So that you can load and unload Kprobes-based instrumentation modules, | |
152 | make sure "Loadable module support" (CONFIG_MODULES) and "Module | |
153 | unloading" (CONFIG_MODULE_UNLOAD) are set to "y". | |
d27a4ddd | 154 | |
09b18203 AM |
155 | Also make sure that CONFIG_KALLSYMS and perhaps even CONFIG_KALLSYMS_ALL |
156 | are set to "y", since kallsyms_lookup_name() is used by the in-kernel | |
157 | kprobe address resolution code. | |
d27a4ddd JK |
158 | |
159 | If you need to insert a probe in the middle of a function, you may find | |
160 | it useful to "Compile the kernel with debug info" (CONFIG_DEBUG_INFO), | |
161 | so you can use "objdump -d -l vmlinux" to see the source-to-object | |
162 | code mapping. | |
163 | ||
164 | 4. API Reference | |
165 | ||
166 | The Kprobes API includes a "register" function and an "unregister" | |
167 | function for each type of probe. Here are terse, mini-man-page | |
168 | specifications for these functions and the associated probe handlers | |
169 | that you'll write. See the latter half of this document for examples. | |
170 | ||
171 | 4.1 register_kprobe | |
172 | ||
173 | #include <linux/kprobes.h> | |
174 | int register_kprobe(struct kprobe *kp); | |
175 | ||
176 | Sets a breakpoint at the address kp->addr. When the breakpoint is | |
177 | hit, Kprobes calls kp->pre_handler. After the probed instruction | |
178 | is single-stepped, Kprobe calls kp->post_handler. If a fault | |
179 | occurs during execution of kp->pre_handler or kp->post_handler, | |
180 | or during single-stepping of the probed instruction, Kprobes calls | |
181 | kp->fault_handler. Any or all handlers can be NULL. | |
182 | ||
09b18203 AM |
183 | NOTE: |
184 | 1. With the introduction of the "symbol_name" field to struct kprobe, | |
185 | the probepoint address resolution will now be taken care of by the kernel. | |
186 | The following will now work: | |
187 | ||
188 | kp.symbol_name = "symbol_name"; | |
189 | ||
190 | (64-bit powerpc intricacies such as function descriptors are handled | |
191 | transparently) | |
192 | ||
193 | 2. Use the "offset" field of struct kprobe if the offset into the symbol | |
194 | to install a probepoint is known. This field is used to calculate the | |
195 | probepoint. | |
196 | ||
197 | 3. Specify either the kprobe "symbol_name" OR the "addr". If both are | |
198 | specified, kprobe registration will fail with -EINVAL. | |
199 | ||
200 | 4. With CISC architectures (such as i386 and x86_64), the kprobes code | |
201 | does not validate if the kprobe.addr is at an instruction boundary. | |
202 | Use "offset" with caution. | |
203 | ||
d27a4ddd JK |
204 | register_kprobe() returns 0 on success, or a negative errno otherwise. |
205 | ||
206 | User's pre-handler (kp->pre_handler): | |
207 | #include <linux/kprobes.h> | |
208 | #include <linux/ptrace.h> | |
209 | int pre_handler(struct kprobe *p, struct pt_regs *regs); | |
210 | ||
211 | Called with p pointing to the kprobe associated with the breakpoint, | |
212 | and regs pointing to the struct containing the registers saved when | |
213 | the breakpoint was hit. Return 0 here unless you're a Kprobes geek. | |
214 | ||
215 | User's post-handler (kp->post_handler): | |
216 | #include <linux/kprobes.h> | |
217 | #include <linux/ptrace.h> | |
218 | void post_handler(struct kprobe *p, struct pt_regs *regs, | |
219 | unsigned long flags); | |
220 | ||
221 | p and regs are as described for the pre_handler. flags always seems | |
222 | to be zero. | |
223 | ||
224 | User's fault-handler (kp->fault_handler): | |
225 | #include <linux/kprobes.h> | |
226 | #include <linux/ptrace.h> | |
227 | int fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr); | |
228 | ||
229 | p and regs are as described for the pre_handler. trapnr is the | |
230 | architecture-specific trap number associated with the fault (e.g., | |
231 | on i386, 13 for a general protection fault or 14 for a page fault). | |
232 | Returns 1 if it successfully handled the exception. | |
233 | ||
234 | 4.2 register_jprobe | |
235 | ||
236 | #include <linux/kprobes.h> | |
237 | int register_jprobe(struct jprobe *jp) | |
238 | ||
239 | Sets a breakpoint at the address jp->kp.addr, which must be the address | |
240 | of the first instruction of a function. When the breakpoint is hit, | |
241 | Kprobes runs the handler whose address is jp->entry. | |
242 | ||
243 | The handler should have the same arg list and return type as the probed | |
244 | function; and just before it returns, it must call jprobe_return(). | |
245 | (The handler never actually returns, since jprobe_return() returns | |
246 | control to Kprobes.) If the probed function is declared asmlinkage, | |
247 | fastcall, or anything else that affects how args are passed, the | |
248 | handler's declaration must match. | |
249 | ||
250 | register_jprobe() returns 0 on success, or a negative errno otherwise. | |
251 | ||
252 | 4.3 register_kretprobe | |
253 | ||
254 | #include <linux/kprobes.h> | |
255 | int register_kretprobe(struct kretprobe *rp); | |
256 | ||
257 | Establishes a return probe for the function whose address is | |
258 | rp->kp.addr. When that function returns, Kprobes calls rp->handler. | |
259 | You must set rp->maxactive appropriately before you call | |
260 | register_kretprobe(); see "How Does a Return Probe Work?" for details. | |
261 | ||
262 | register_kretprobe() returns 0 on success, or a negative errno | |
263 | otherwise. | |
264 | ||
265 | User's return-probe handler (rp->handler): | |
266 | #include <linux/kprobes.h> | |
267 | #include <linux/ptrace.h> | |
268 | int kretprobe_handler(struct kretprobe_instance *ri, struct pt_regs *regs); | |
269 | ||
270 | regs is as described for kprobe.pre_handler. ri points to the | |
271 | kretprobe_instance object, of which the following fields may be | |
272 | of interest: | |
273 | - ret_addr: the return address | |
274 | - rp: points to the corresponding kretprobe object | |
275 | - task: points to the corresponding task struct | |
09b18203 AM |
276 | |
277 | The regs_return_value(regs) macro provides a simple abstraction to | |
278 | extract the return value from the appropriate register as defined by | |
279 | the architecture's ABI. | |
280 | ||
d27a4ddd JK |
281 | The handler's return value is currently ignored. |
282 | ||
283 | 4.4 unregister_*probe | |
284 | ||
285 | #include <linux/kprobes.h> | |
286 | void unregister_kprobe(struct kprobe *kp); | |
287 | void unregister_jprobe(struct jprobe *jp); | |
288 | void unregister_kretprobe(struct kretprobe *rp); | |
289 | ||
290 | Removes the specified probe. The unregister function can be called | |
291 | at any time after the probe has been registered. | |
292 | ||
293 | 5. Kprobes Features and Limitations | |
294 | ||
8861da31 JK |
295 | Kprobes allows multiple probes at the same address. Currently, |
296 | however, there cannot be multiple jprobes on the same function at | |
297 | the same time. | |
d27a4ddd JK |
298 | |
299 | In general, you can install a probe anywhere in the kernel. | |
300 | In particular, you can probe interrupt handlers. Known exceptions | |
301 | are discussed in this section. | |
302 | ||
8861da31 JK |
303 | The register_*probe functions will return -EINVAL if you attempt |
304 | to install a probe in the code that implements Kprobes (mostly | |
305 | kernel/kprobes.c and arch/*/kernel/kprobes.c, but also functions such | |
306 | as do_page_fault and notifier_call_chain). | |
d27a4ddd JK |
307 | |
308 | If you install a probe in an inline-able function, Kprobes makes | |
309 | no attempt to chase down all inline instances of the function and | |
310 | install probes there. gcc may inline a function without being asked, | |
311 | so keep this in mind if you're not seeing the probe hits you expect. | |
312 | ||
313 | A probe handler can modify the environment of the probed function | |
314 | -- e.g., by modifying kernel data structures, or by modifying the | |
315 | contents of the pt_regs struct (which are restored to the registers | |
316 | upon return from the breakpoint). So Kprobes can be used, for example, | |
317 | to install a bug fix or to inject faults for testing. Kprobes, of | |
318 | course, has no way to distinguish the deliberately injected faults | |
319 | from the accidental ones. Don't drink and probe. | |
320 | ||
321 | Kprobes makes no attempt to prevent probe handlers from stepping on | |
322 | each other -- e.g., probing printk() and then calling printk() from a | |
8861da31 JK |
323 | probe handler. If a probe handler hits a probe, that second probe's |
324 | handlers won't be run in that instance, and the kprobe.nmissed member | |
325 | of the second probe will be incremented. | |
326 | ||
327 | As of Linux v2.6.15-rc1, multiple handlers (or multiple instances of | |
328 | the same handler) may run concurrently on different CPUs. | |
329 | ||
330 | Kprobes does not use mutexes or allocate memory except during | |
d27a4ddd JK |
331 | registration and unregistration. |
332 | ||
333 | Probe handlers are run with preemption disabled. Depending on the | |
334 | architecture, handlers may also run with interrupts disabled. In any | |
335 | case, your handler should not yield the CPU (e.g., by attempting to | |
336 | acquire a semaphore). | |
337 | ||
338 | Since a return probe is implemented by replacing the return | |
339 | address with the trampoline's address, stack backtraces and calls | |
340 | to __builtin_return_address() will typically yield the trampoline's | |
341 | address instead of the real return address for kretprobed functions. | |
342 | (As far as we can tell, __builtin_return_address() is used only | |
343 | for instrumentation and error reporting.) | |
344 | ||
8861da31 JK |
345 | If the number of times a function is called does not match the number |
346 | of times it returns, registering a return probe on that function may | |
bf8f6e5b AM |
347 | produce undesirable results. In such a case, a line: |
348 | kretprobe BUG!: Processing kretprobe d000000000041aa8 @ c00000000004f48c | |
349 | gets printed. With this information, one will be able to correlate the | |
350 | exact instance of the kretprobe that caused the problem. We have the | |
351 | do_exit() case covered. do_execve() and do_fork() are not an issue. | |
352 | We're unaware of other specific cases where this could be a problem. | |
8861da31 JK |
353 | |
354 | If, upon entry to or exit from a function, the CPU is running on | |
355 | a stack other than that of the current task, registering a return | |
356 | probe on that function may produce undesirable results. For this | |
357 | reason, Kprobes doesn't support return probes (or kprobes or jprobes) | |
358 | on the x86_64 version of __switch_to(); the registration functions | |
359 | return -EINVAL. | |
d27a4ddd JK |
360 | |
361 | 6. Probe Overhead | |
362 | ||
363 | On a typical CPU in use in 2005, a kprobe hit takes 0.5 to 1.0 | |
364 | microseconds to process. Specifically, a benchmark that hits the same | |
365 | probepoint repeatedly, firing a simple handler each time, reports 1-2 | |
366 | million hits per second, depending on the architecture. A jprobe or | |
367 | return-probe hit typically takes 50-75% longer than a kprobe hit. | |
368 | When you have a return probe set on a function, adding a kprobe at | |
369 | the entry to that function adds essentially no overhead. | |
370 | ||
371 | Here are sample overhead figures (in usec) for different architectures. | |
372 | k = kprobe; j = jprobe; r = return probe; kr = kprobe + return probe | |
373 | on same function; jr = jprobe + return probe on same function | |
374 | ||
375 | i386: Intel Pentium M, 1495 MHz, 2957.31 bogomips | |
376 | k = 0.57 usec; j = 1.00; r = 0.92; kr = 0.99; jr = 1.40 | |
377 | ||
378 | x86_64: AMD Opteron 246, 1994 MHz, 3971.48 bogomips | |
379 | k = 0.49 usec; j = 0.76; r = 0.80; kr = 0.82; jr = 1.07 | |
380 | ||
381 | ppc64: POWER5 (gr), 1656 MHz (SMT disabled, 1 virtual CPU per physical CPU) | |
382 | k = 0.77 usec; j = 1.31; r = 1.26; kr = 1.45; jr = 1.99 | |
383 | ||
384 | 7. TODO | |
385 | ||
8861da31 JK |
386 | a. SystemTap (http://sourceware.org/systemtap): Provides a simplified |
387 | programming interface for probe-based instrumentation. Try it out. | |
388 | b. Kernel return probes for sparc64. | |
389 | c. Support for other architectures. | |
390 | d. User-space probes. | |
391 | e. Watchpoint probes (which fire on data references). | |
d27a4ddd JK |
392 | |
393 | 8. Kprobes Example | |
394 | ||
395 | Here's a sample kernel module showing the use of kprobes to dump a | |
396 | stack trace and selected i386 registers when do_fork() is called. | |
397 | ----- cut here ----- | |
398 | /*kprobe_example.c*/ | |
399 | #include <linux/kernel.h> | |
400 | #include <linux/module.h> | |
401 | #include <linux/kprobes.h> | |
d27a4ddd JK |
402 | #include <linux/sched.h> |
403 | ||
404 | /*For each probe you need to allocate a kprobe structure*/ | |
405 | static struct kprobe kp; | |
406 | ||
407 | /*kprobe pre_handler: called just before the probed instruction is executed*/ | |
408 | int handler_pre(struct kprobe *p, struct pt_regs *regs) | |
409 | { | |
410 | printk("pre_handler: p->addr=0x%p, eip=%lx, eflags=0x%lx\n", | |
411 | p->addr, regs->eip, regs->eflags); | |
412 | dump_stack(); | |
413 | return 0; | |
414 | } | |
415 | ||
416 | /*kprobe post_handler: called after the probed instruction is executed*/ | |
417 | void handler_post(struct kprobe *p, struct pt_regs *regs, unsigned long flags) | |
418 | { | |
419 | printk("post_handler: p->addr=0x%p, eflags=0x%lx\n", | |
420 | p->addr, regs->eflags); | |
421 | } | |
422 | ||
423 | /* fault_handler: this is called if an exception is generated for any | |
424 | * instruction within the pre- or post-handler, or when Kprobes | |
425 | * single-steps the probed instruction. | |
426 | */ | |
427 | int handler_fault(struct kprobe *p, struct pt_regs *regs, int trapnr) | |
428 | { | |
429 | printk("fault_handler: p->addr=0x%p, trap #%dn", | |
430 | p->addr, trapnr); | |
431 | /* Return 0 because we don't handle the fault. */ | |
432 | return 0; | |
433 | } | |
434 | ||
09b18203 | 435 | static int __init kprobe_init(void) |
d27a4ddd JK |
436 | { |
437 | int ret; | |
438 | kp.pre_handler = handler_pre; | |
439 | kp.post_handler = handler_post; | |
440 | kp.fault_handler = handler_fault; | |
09b18203 AM |
441 | kp.symbol_name = "do_fork"; |
442 | ||
565762f3 AD |
443 | ret = register_kprobe(&kp); |
444 | if (ret < 0) { | |
d27a4ddd | 445 | printk("register_kprobe failed, returned %d\n", ret); |
565762f3 | 446 | return ret; |
d27a4ddd JK |
447 | } |
448 | printk("kprobe registered\n"); | |
449 | return 0; | |
450 | } | |
451 | ||
09b18203 | 452 | static void __exit kprobe_exit(void) |
d27a4ddd JK |
453 | { |
454 | unregister_kprobe(&kp); | |
455 | printk("kprobe unregistered\n"); | |
456 | } | |
457 | ||
09b18203 AM |
458 | module_init(kprobe_init) |
459 | module_exit(kprobe_exit) | |
d27a4ddd JK |
460 | MODULE_LICENSE("GPL"); |
461 | ----- cut here ----- | |
462 | ||
463 | You can build the kernel module, kprobe-example.ko, using the following | |
464 | Makefile: | |
465 | ----- cut here ----- | |
466 | obj-m := kprobe-example.o | |
467 | KDIR := /lib/modules/$(shell uname -r)/build | |
468 | PWD := $(shell pwd) | |
469 | default: | |
470 | $(MAKE) -C $(KDIR) SUBDIRS=$(PWD) modules | |
471 | clean: | |
472 | rm -f *.mod.c *.ko *.o | |
473 | ----- cut here ----- | |
474 | ||
475 | $ make | |
476 | $ su - | |
477 | ... | |
478 | # insmod kprobe-example.ko | |
479 | ||
480 | You will see the trace data in /var/log/messages and on the console | |
481 | whenever do_fork() is invoked to create a new process. | |
482 | ||
483 | 9. Jprobes Example | |
484 | ||
485 | Here's a sample kernel module showing the use of jprobes to dump | |
486 | the arguments of do_fork(). | |
487 | ----- cut here ----- | |
488 | /*jprobe-example.c */ | |
489 | #include <linux/kernel.h> | |
490 | #include <linux/module.h> | |
491 | #include <linux/fs.h> | |
492 | #include <linux/uio.h> | |
493 | #include <linux/kprobes.h> | |
d27a4ddd JK |
494 | |
495 | /* | |
496 | * Jumper probe for do_fork. | |
497 | * Mirror principle enables access to arguments of the probed routine | |
498 | * from the probe handler. | |
499 | */ | |
500 | ||
501 | /* Proxy routine having the same arguments as actual do_fork() routine */ | |
502 | long jdo_fork(unsigned long clone_flags, unsigned long stack_start, | |
503 | struct pt_regs *regs, unsigned long stack_size, | |
504 | int __user * parent_tidptr, int __user * child_tidptr) | |
505 | { | |
506 | printk("jprobe: clone_flags=0x%lx, stack_size=0x%lx, regs=0x%p\n", | |
507 | clone_flags, stack_size, regs); | |
508 | /* Always end with a call to jprobe_return(). */ | |
509 | jprobe_return(); | |
510 | /*NOTREACHED*/ | |
511 | return 0; | |
512 | } | |
513 | ||
514 | static struct jprobe my_jprobe = { | |
9e367d85 | 515 | .entry = jdo_fork |
d27a4ddd JK |
516 | }; |
517 | ||
09b18203 | 518 | static int __init jprobe_init(void) |
d27a4ddd JK |
519 | { |
520 | int ret; | |
09b18203 | 521 | my_jprobe.kp.symbol_name = "do_fork"; |
d27a4ddd JK |
522 | |
523 | if ((ret = register_jprobe(&my_jprobe)) <0) { | |
524 | printk("register_jprobe failed, returned %d\n", ret); | |
525 | return -1; | |
526 | } | |
527 | printk("Planted jprobe at %p, handler addr %p\n", | |
528 | my_jprobe.kp.addr, my_jprobe.entry); | |
529 | return 0; | |
530 | } | |
531 | ||
09b18203 | 532 | static void __exit jprobe_exit(void) |
d27a4ddd JK |
533 | { |
534 | unregister_jprobe(&my_jprobe); | |
535 | printk("jprobe unregistered\n"); | |
536 | } | |
537 | ||
09b18203 AM |
538 | module_init(jprobe_init) |
539 | module_exit(jprobe_exit) | |
d27a4ddd JK |
540 | MODULE_LICENSE("GPL"); |
541 | ----- cut here ----- | |
542 | ||
543 | Build and insert the kernel module as shown in the above kprobe | |
544 | example. You will see the trace data in /var/log/messages and on | |
545 | the console whenever do_fork() is invoked to create a new process. | |
546 | (Some messages may be suppressed if syslogd is configured to | |
547 | eliminate duplicate messages.) | |
548 | ||
549 | 10. Kretprobes Example | |
550 | ||
551 | Here's a sample kernel module showing the use of return probes to | |
552 | report failed calls to sys_open(). | |
553 | ----- cut here ----- | |
554 | /*kretprobe-example.c*/ | |
555 | #include <linux/kernel.h> | |
556 | #include <linux/module.h> | |
557 | #include <linux/kprobes.h> | |
d27a4ddd JK |
558 | |
559 | static const char *probed_func = "sys_open"; | |
560 | ||
561 | /* Return-probe handler: If the probed function fails, log the return value. */ | |
562 | static int ret_handler(struct kretprobe_instance *ri, struct pt_regs *regs) | |
563 | { | |
09b18203 | 564 | int retval = regs_return_value(regs); |
d27a4ddd JK |
565 | if (retval < 0) { |
566 | printk("%s returns %d\n", probed_func, retval); | |
567 | } | |
568 | return 0; | |
569 | } | |
570 | ||
571 | static struct kretprobe my_kretprobe = { | |
572 | .handler = ret_handler, | |
573 | /* Probe up to 20 instances concurrently. */ | |
574 | .maxactive = 20 | |
575 | }; | |
576 | ||
09b18203 | 577 | static int __init kretprobe_init(void) |
d27a4ddd JK |
578 | { |
579 | int ret; | |
09b18203 AM |
580 | my_kretprobe.kp.symbol_name = (char *)probed_func; |
581 | ||
d27a4ddd JK |
582 | if ((ret = register_kretprobe(&my_kretprobe)) < 0) { |
583 | printk("register_kretprobe failed, returned %d\n", ret); | |
584 | return -1; | |
585 | } | |
586 | printk("Planted return probe at %p\n", my_kretprobe.kp.addr); | |
587 | return 0; | |
588 | } | |
589 | ||
09b18203 | 590 | static void __exit kretprobe_exit(void) |
d27a4ddd JK |
591 | { |
592 | unregister_kretprobe(&my_kretprobe); | |
593 | printk("kretprobe unregistered\n"); | |
594 | /* nmissed > 0 suggests that maxactive was set too low. */ | |
595 | printk("Missed probing %d instances of %s\n", | |
596 | my_kretprobe.nmissed, probed_func); | |
597 | } | |
598 | ||
09b18203 AM |
599 | module_init(kretprobe_init) |
600 | module_exit(kretprobe_exit) | |
d27a4ddd JK |
601 | MODULE_LICENSE("GPL"); |
602 | ----- cut here ----- | |
603 | ||
604 | Build and insert the kernel module as shown in the above kprobe | |
605 | example. You will see the trace data in /var/log/messages and on the | |
606 | console whenever sys_open() returns a negative value. (Some messages | |
607 | may be suppressed if syslogd is configured to eliminate duplicate | |
608 | messages.) | |
609 | ||
610 | For additional information on Kprobes, refer to the following URLs: | |
611 | http://www-106.ibm.com/developerworks/library/l-kprobes.html?ca=dgr-lnxw42Kprobe | |
612 | http://www.redhat.com/magazine/005mar05/features/kprobes/ | |
09b18203 AM |
613 | http://www-users.cs.umn.edu/~boutcher/kprobes/ |
614 | http://www.linuxsymposium.org/2006/linuxsymposium_procv2.pdf (pages 101-115) | |
bf8f6e5b AM |
615 | |
616 | ||
617 | Appendix A: The kprobes debugfs interface | |
618 | ||
619 | With recent kernels (> 2.6.20) the list of registered kprobes is visible | |
620 | under the /debug/kprobes/ directory (assuming debugfs is mounted at /debug). | |
621 | ||
622 | /debug/kprobes/list: Lists all registered probes on the system | |
623 | ||
624 | c015d71a k vfs_read+0x0 | |
625 | c011a316 j do_fork+0x0 | |
626 | c03dedc5 r tcp_v4_rcv+0x0 | |
627 | ||
628 | The first column provides the kernel address where the probe is inserted. | |
629 | The second column identifies the type of probe (k - kprobe, r - kretprobe | |
630 | and j - jprobe), while the third column specifies the symbol+offset of | |
631 | the probe. If the probed function belongs to a module, the module name | |
632 | is also specified. | |
633 | ||
634 | /debug/kprobes/enabled: Turn kprobes ON/OFF | |
635 | ||
636 | Provides a knob to globally turn registered kprobes ON or OFF. By default, | |
637 | all kprobes are enabled. By echoing "0" to this file, all registered probes | |
638 | will be disarmed, till such time a "1" is echoed to this file. |