blob: 6653017680ddcd0d813b074245de53c3b59310fb [file] [log] [blame]
Jim Kenistond27a4dd2005-08-04 12:53:35 -07001Title : Kernel Probes (Kprobes)
2Authors : Jim Keniston <jkenisto@us.ibm.com>
Masami Hiramatsub26486b2010-02-25 08:35:04 -05003 : Prasanna S Panchamukhi <prasanna.panchamukhi@gmail.com>
4 : Masami Hiramatsu <mhiramat@redhat.com>
Jim Kenistond27a4dd2005-08-04 12:53:35 -07005
6CONTENTS
7
81. Concepts: Kprobes, Jprobes, Return Probes
92. Architectures Supported
103. Configuring Kprobes
114. API Reference
125. Kprobes Features and Limitations
136. Probe Overhead
147. TODO
158. Kprobes Example
169. Jprobes Example
1710. Kretprobes Example
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -070018Appendix A: The kprobes debugfs interface
Masami Hiramatsub26486b2010-02-25 08:35:04 -050019Appendix B: The kprobes sysctl interface
Jim Kenistond27a4dd2005-08-04 12:53:35 -070020
211. Concepts: Kprobes, Jprobes, Return Probes
22
23Kprobes enables you to dynamically break into any kernel routine and
24collect debugging and performance information non-disruptively. You
25can trap at almost any kernel code address, specifying a handler
26routine to be invoked when the breakpoint is hit.
27
28There are currently three types of probes: kprobes, jprobes, and
29kretprobes (also called return probes). A kprobe can be inserted
30on virtually any instruction in the kernel. A jprobe is inserted at
31the entry to a kernel function, and provides convenient access to the
32function's arguments. A return probe fires when a specified function
33returns.
34
35In the typical case, Kprobes-based instrumentation is packaged as
36a kernel module. The module's init function installs ("registers")
37one or more probes, and the exit function unregisters them. A
38registration function such as register_kprobe() specifies where
39the probe is to be inserted and what handler is to be called when
40the probe is hit.
41
Masami Hiramatsu3b0cb4c2008-04-28 02:14:30 -070042There are also register_/unregister_*probes() functions for batch
43registration/unregistration of a group of *probes. These functions
44can speed up unregistration process when you have to unregister
45a lot of probes at once.
46
Masami Hiramatsub26486b2010-02-25 08:35:04 -050047The next four subsections explain how the different types of
48probes work and how jump optimization works. They explain certain
49things that you'll need to know in order to make the best use of
50Kprobes -- e.g., the difference between a pre_handler and
51a post_handler, and how to use the maxactive and nmissed fields of
52a kretprobe. But if you're in a hurry to start using Kprobes, you
53can skip ahead to section 2.
Jim Kenistond27a4dd2005-08-04 12:53:35 -070054
551.1 How Does a Kprobe Work?
56
57When a kprobe is registered, Kprobes makes a copy of the probed
58instruction and replaces the first byte(s) of the probed instruction
59with a breakpoint instruction (e.g., int3 on i386 and x86_64).
60
61When a CPU hits the breakpoint instruction, a trap occurs, the CPU's
62registers are saved, and control passes to Kprobes via the
63notifier_call_chain mechanism. Kprobes executes the "pre_handler"
64associated with the kprobe, passing the handler the addresses of the
65kprobe struct and the saved registers.
66
67Next, Kprobes single-steps its copy of the probed instruction.
68(It would be simpler to single-step the actual instruction in place,
69but then Kprobes would have to temporarily remove the breakpoint
70instruction. This would open a small time window when another CPU
71could sail right past the probepoint.)
72
73After the instruction is single-stepped, Kprobes executes the
74"post_handler," if any, that is associated with the kprobe.
75Execution then continues with the instruction following the probepoint.
76
771.2 How Does a Jprobe Work?
78
79A jprobe is implemented using a kprobe that is placed on a function's
80entry point. It employs a simple mirroring principle to allow
81seamless access to the probed function's arguments. The jprobe
82handler routine should have the same signature (arg list and return
83type) as the function being probed, and must always end by calling
84the Kprobes function jprobe_return().
85
86Here's how it works. When the probe is hit, Kprobes makes a copy of
87the saved registers and a generous portion of the stack (see below).
88Kprobes then points the saved instruction pointer at the jprobe's
89handler routine, and returns from the trap. As a result, control
90passes to the handler, which is presented with the same register and
91stack contents as the probed function. When it is done, the handler
92calls jprobe_return(), which traps again to restore the original stack
93contents and processor state and switch to the probed function.
94
95By convention, the callee owns its arguments, so gcc may produce code
96that unexpectedly modifies that portion of the stack. This is why
97Kprobes saves a copy of the stack and restores it after the jprobe
98handler has run. Up to MAX_STACK_SIZE bytes are copied -- e.g.,
9964 bytes on i386.
100
101Note that the probed function's args may be passed on the stack
Harvey Harrisonb5606c22008-02-13 15:03:16 -0800102or in registers. The jprobe will work in either case, so long as the
103handler's prototype matches that of the probed function.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700104
Abhishek Sagarf47cd9b2008-02-06 01:38:22 -08001051.3 Return Probes
106
1071.3.1 How Does a Return Probe Work?
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700108
109When you call register_kretprobe(), Kprobes establishes a kprobe at
110the entry to the function. When the probed function is called and this
111probe is hit, Kprobes saves a copy of the return address, and replaces
112the return address with the address of a "trampoline." The trampoline
113is an arbitrary piece of code -- typically just a nop instruction.
114At boot time, Kprobes registers a kprobe at the trampoline.
115
116When the probed function executes its return instruction, control
117passes to the trampoline and that probe is hit. Kprobes' trampoline
Abhishek Sagarf47cd9b2008-02-06 01:38:22 -0800118handler calls the user-specified return handler associated with the
119kretprobe, then sets the saved instruction pointer to the saved return
120address, and that's where execution resumes upon return from the trap.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700121
122While the probed function is executing, its return address is
123stored in an object of type kretprobe_instance. Before calling
124register_kretprobe(), the user sets the maxactive field of the
125kretprobe struct to specify how many instances of the specified
126function can be probed simultaneously. register_kretprobe()
127pre-allocates the indicated number of kretprobe_instance objects.
128
129For example, if the function is non-recursive and is called with a
130spinlock held, maxactive = 1 should be enough. If the function is
131non-recursive and can never relinquish the CPU (e.g., via a semaphore
132or preemption), NR_CPUS should be enough. If maxactive <= 0, it is
133set to a default value. If CONFIG_PREEMPT is enabled, the default
134is max(10, 2*NR_CPUS). Otherwise, the default is NR_CPUS.
135
136It's not a disaster if you set maxactive too low; you'll just miss
137some probes. In the kretprobe struct, the nmissed field is set to
138zero when the return probe is registered, and is incremented every
139time the probed function is entered but there is no kretprobe_instance
140object available for establishing the return probe.
141
Abhishek Sagarf47cd9b2008-02-06 01:38:22 -08001421.3.2 Kretprobe entry-handler
143
144Kretprobes also provides an optional user-specified handler which runs
145on function entry. This handler is specified by setting the entry_handler
146field of the kretprobe struct. Whenever the kprobe placed by kretprobe at the
147function entry is hit, the user-defined entry_handler, if any, is invoked.
148If the entry_handler returns 0 (success) then a corresponding return handler
149is guaranteed to be called upon function return. If the entry_handler
150returns a non-zero error then Kprobes leaves the return address as is, and
151the kretprobe has no further effect for that particular function instance.
152
153Multiple entry and return handler invocations are matched using the unique
154kretprobe_instance object associated with them. Additionally, a user
155may also specify per return-instance private data to be part of each
156kretprobe_instance object. This is especially useful when sharing private
157data between corresponding user entry and return handlers. The size of each
158private data object can be specified at kretprobe registration time by
159setting the data_size field of the kretprobe struct. This data can be
160accessed through the data field of each kretprobe_instance object.
161
162In case probed function is entered but there is no kretprobe_instance
163object available, then in addition to incrementing the nmissed count,
164the user entry_handler invocation is also skipped.
165
Masami Hiramatsub26486b2010-02-25 08:35:04 -05001661.4 How Does Jump Optimization Work?
167
Masami Hiramatsu5cc718b2010-03-15 13:00:54 -0400168If your kernel is built with CONFIG_OPTPROBES=y (currently this flag
169is automatically set 'y' on x86/x86-64, non-preemptive kernel) and
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500170the "debug.kprobes_optimization" kernel parameter is set to 1 (see
171sysctl(8)), Kprobes tries to reduce probe-hit overhead by using a jump
172instruction instead of a breakpoint instruction at each probepoint.
173
1741.4.1 Init a Kprobe
175
176When a probe is registered, before attempting this optimization,
177Kprobes inserts an ordinary, breakpoint-based kprobe at the specified
178address. So, even if it's not possible to optimize this particular
179probepoint, there'll be a probe there.
180
1811.4.2 Safety Check
182
183Before optimizing a probe, Kprobes performs the following safety checks:
184
185- Kprobes verifies that the region that will be replaced by the jump
186instruction (the "optimized region") lies entirely within one function.
187(A jump instruction is multiple bytes, and so may overlay multiple
188instructions.)
189
190- Kprobes analyzes the entire function and verifies that there is no
191jump into the optimized region. Specifically:
192 - the function contains no indirect jump;
193 - the function contains no instruction that causes an exception (since
194 the fixup code triggered by the exception could jump back into the
195 optimized region -- Kprobes checks the exception tables to verify this);
196 and
197 - there is no near jump to the optimized region (other than to the first
198 byte).
199
200- For each instruction in the optimized region, Kprobes verifies that
201the instruction can be executed out of line.
202
2031.4.3 Preparing Detour Buffer
204
205Next, Kprobes prepares a "detour" buffer, which contains the following
206instruction sequence:
207- code to push the CPU's registers (emulating a breakpoint trap)
208- a call to the trampoline code which calls user's probe handlers.
209- code to restore registers
210- the instructions from the optimized region
211- a jump back to the original execution path.
212
2131.4.4 Pre-optimization
214
215After preparing the detour buffer, Kprobes verifies that none of the
216following situations exist:
217- The probe has either a break_handler (i.e., it's a jprobe) or a
218post_handler.
219- Other instructions in the optimized region are probed.
220- The probe is disabled.
221In any of the above cases, Kprobes won't start optimizing the probe.
222Since these are temporary situations, Kprobes tries to start
223optimizing it again if the situation is changed.
224
225If the kprobe can be optimized, Kprobes enqueues the kprobe to an
226optimizing list, and kicks the kprobe-optimizer workqueue to optimize
227it. If the to-be-optimized probepoint is hit before being optimized,
228Kprobes returns control to the original instruction path by setting
229the CPU's instruction pointer to the copied code in the detour buffer
230-- thus at least avoiding the single-step.
231
2321.4.5 Optimization
233
234The Kprobe-optimizer doesn't insert the jump instruction immediately;
235rather, it calls synchronize_sched() for safety first, because it's
236possible for a CPU to be interrupted in the middle of executing the
237optimized region(*). As you know, synchronize_sched() can ensure
238that all interruptions that were active when synchronize_sched()
239was called are done, but only if CONFIG_PREEMPT=n. So, this version
240of kprobe optimization supports only kernels with CONFIG_PREEMPT=n.(**)
241
242After that, the Kprobe-optimizer calls stop_machine() to replace
243the optimized region with a jump instruction to the detour buffer,
244using text_poke_smp().
245
2461.4.6 Unoptimization
247
248When an optimized kprobe is unregistered, disabled, or blocked by
249another kprobe, it will be unoptimized. If this happens before
250the optimization is complete, the kprobe is just dequeued from the
251optimized list. If the optimization has been done, the jump is
252replaced with the original code (except for an int3 breakpoint in
253the first byte) by using text_poke_smp().
254
255(*)Please imagine that the 2nd instruction is interrupted and then
256the optimizer replaces the 2nd instruction with the jump *address*
257while the interrupt handler is running. When the interrupt
258returns to original address, there is no valid instruction,
259and it causes an unexpected result.
260
261(**)This optimization-safety checking may be replaced with the
262stop-machine method that ksplice uses for supporting a CONFIG_PREEMPT=y
263kernel.
264
265NOTE for geeks:
266The jump optimization changes the kprobe's pre_handler behavior.
267Without optimization, the pre_handler can change the kernel's execution
268path by changing regs->ip and returning 1. However, when the probe
269is optimized, that modification is ignored. Thus, if you want to
270tweak the kernel's execution path, you need to suppress optimization,
271using one of the following techniques:
272- Specify an empty function for the kprobe's post_handler or break_handler.
273 or
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500274- Execute 'sysctl -w debug.kprobes_optimization=n'
275
Jim Kenistond27a4dd2005-08-04 12:53:35 -07002762. Architectures Supported
277
278Kprobes, jprobes, and return probes are implemented on the following
279architectures:
280
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500281- i386 (Supports jump optimization)
282- x86_64 (AMD-64, EM64T) (Supports jump optimization)
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700283- ppc64
Jim Keniston8861da32006-02-14 13:53:06 -0800284- ia64 (Does not support probes on instruction slot1.)
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700285- sparc64 (Return probes not yet implemented.)
Nicolas Pitre5de865b2007-12-03 17:15:52 -0500286- arm
Kumar Galaf8279622008-06-26 02:01:37 -0500287- ppc
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700288
2893. Configuring Kprobes
290
291When configuring the kernel using make menuconfig/xconfig/oldconfig,
Jim Keniston8861da32006-02-14 13:53:06 -0800292ensure that CONFIG_KPROBES is set to "y". Under "Instrumentation
293Support", look for "Kprobes".
294
295So that you can load and unload Kprobes-based instrumentation modules,
296make sure "Loadable module support" (CONFIG_MODULES) and "Module
297unloading" (CONFIG_MODULE_UNLOAD) are set to "y".
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700298
Ananth N Mavinakayanahalli09b18202006-10-02 02:17:32 -0700299Also make sure that CONFIG_KALLSYMS and perhaps even CONFIG_KALLSYMS_ALL
300are set to "y", since kallsyms_lookup_name() is used by the in-kernel
301kprobe address resolution code.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700302
303If you need to insert a probe in the middle of a function, you may find
304it useful to "Compile the kernel with debug info" (CONFIG_DEBUG_INFO),
305so you can use "objdump -d -l vmlinux" to see the source-to-object
306code mapping.
307
3084. API Reference
309
310The Kprobes API includes a "register" function and an "unregister"
Masami Hiramatsu3b0cb4c2008-04-28 02:14:30 -0700311function for each type of probe. The API also includes "register_*probes"
312and "unregister_*probes" functions for (un)registering arrays of probes.
313Here are terse, mini-man-page specifications for these functions and
314the associated probe handlers that you'll write. See the files in the
315samples/kprobes/ sub-directory for examples.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700316
3174.1 register_kprobe
318
319#include <linux/kprobes.h>
320int register_kprobe(struct kprobe *kp);
321
322Sets a breakpoint at the address kp->addr. When the breakpoint is
323hit, Kprobes calls kp->pre_handler. After the probed instruction
324is single-stepped, Kprobe calls kp->post_handler. If a fault
325occurs during execution of kp->pre_handler or kp->post_handler,
326or during single-stepping of the probed instruction, Kprobes calls
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700327kp->fault_handler. Any or all handlers can be NULL. If kp->flags
328is set KPROBE_FLAG_DISABLED, that kp will be registered but disabled,
Francis Galieguea33f3222010-04-23 00:08:02 +0200329so, its handlers aren't hit until calling enable_kprobe(kp).
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700330
Ananth N Mavinakayanahalli09b18202006-10-02 02:17:32 -0700331NOTE:
3321. With the introduction of the "symbol_name" field to struct kprobe,
333the probepoint address resolution will now be taken care of by the kernel.
334The following will now work:
335
336 kp.symbol_name = "symbol_name";
337
338(64-bit powerpc intricacies such as function descriptors are handled
339transparently)
340
3412. Use the "offset" field of struct kprobe if the offset into the symbol
342to install a probepoint is known. This field is used to calculate the
343probepoint.
344
3453. Specify either the kprobe "symbol_name" OR the "addr". If both are
346specified, kprobe registration will fail with -EINVAL.
347
3484. With CISC architectures (such as i386 and x86_64), the kprobes code
349does not validate if the kprobe.addr is at an instruction boundary.
350Use "offset" with caution.
351
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700352register_kprobe() returns 0 on success, or a negative errno otherwise.
353
354User's pre-handler (kp->pre_handler):
355#include <linux/kprobes.h>
356#include <linux/ptrace.h>
357int pre_handler(struct kprobe *p, struct pt_regs *regs);
358
359Called with p pointing to the kprobe associated with the breakpoint,
360and regs pointing to the struct containing the registers saved when
361the breakpoint was hit. Return 0 here unless you're a Kprobes geek.
362
363User's post-handler (kp->post_handler):
364#include <linux/kprobes.h>
365#include <linux/ptrace.h>
366void post_handler(struct kprobe *p, struct pt_regs *regs,
367 unsigned long flags);
368
369p and regs are as described for the pre_handler. flags always seems
370to be zero.
371
372User's fault-handler (kp->fault_handler):
373#include <linux/kprobes.h>
374#include <linux/ptrace.h>
375int fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr);
376
377p and regs are as described for the pre_handler. trapnr is the
378architecture-specific trap number associated with the fault (e.g.,
379on i386, 13 for a general protection fault or 14 for a page fault).
380Returns 1 if it successfully handled the exception.
381
3824.2 register_jprobe
383
384#include <linux/kprobes.h>
385int register_jprobe(struct jprobe *jp)
386
387Sets a breakpoint at the address jp->kp.addr, which must be the address
388of the first instruction of a function. When the breakpoint is hit,
389Kprobes runs the handler whose address is jp->entry.
390
391The handler should have the same arg list and return type as the probed
392function; and just before it returns, it must call jprobe_return().
393(The handler never actually returns, since jprobe_return() returns
Harvey Harrisonb5606c22008-02-13 15:03:16 -0800394control to Kprobes.) If the probed function is declared asmlinkage
395or anything else that affects how args are passed, the handler's
396declaration must match.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700397
398register_jprobe() returns 0 on success, or a negative errno otherwise.
399
4004.3 register_kretprobe
401
402#include <linux/kprobes.h>
403int register_kretprobe(struct kretprobe *rp);
404
405Establishes a return probe for the function whose address is
406rp->kp.addr. When that function returns, Kprobes calls rp->handler.
407You must set rp->maxactive appropriately before you call
408register_kretprobe(); see "How Does a Return Probe Work?" for details.
409
410register_kretprobe() returns 0 on success, or a negative errno
411otherwise.
412
413User's return-probe handler (rp->handler):
414#include <linux/kprobes.h>
415#include <linux/ptrace.h>
416int kretprobe_handler(struct kretprobe_instance *ri, struct pt_regs *regs);
417
418regs is as described for kprobe.pre_handler. ri points to the
419kretprobe_instance object, of which the following fields may be
420of interest:
421- ret_addr: the return address
422- rp: points to the corresponding kretprobe object
423- task: points to the corresponding task struct
Abhishek Sagarf47cd9b2008-02-06 01:38:22 -0800424- data: points to per return-instance private data; see "Kretprobe
425 entry-handler" for details.
Ananth N Mavinakayanahalli09b18202006-10-02 02:17:32 -0700426
427The regs_return_value(regs) macro provides a simple abstraction to
428extract the return value from the appropriate register as defined by
429the architecture's ABI.
430
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700431The handler's return value is currently ignored.
432
4334.4 unregister_*probe
434
435#include <linux/kprobes.h>
436void unregister_kprobe(struct kprobe *kp);
437void unregister_jprobe(struct jprobe *jp);
438void unregister_kretprobe(struct kretprobe *rp);
439
440Removes the specified probe. The unregister function can be called
441at any time after the probe has been registered.
442
Masami Hiramatsu3b0cb4c2008-04-28 02:14:30 -0700443NOTE:
444If the functions find an incorrect probe (ex. an unregistered probe),
445they clear the addr field of the probe.
446
4474.5 register_*probes
448
449#include <linux/kprobes.h>
450int register_kprobes(struct kprobe **kps, int num);
451int register_kretprobes(struct kretprobe **rps, int num);
452int register_jprobes(struct jprobe **jps, int num);
453
454Registers each of the num probes in the specified array. If any
455error occurs during registration, all probes in the array, up to
456the bad probe, are safely unregistered before the register_*probes
457function returns.
458- kps/rps/jps: an array of pointers to *probe data structures
459- num: the number of the array entries.
460
461NOTE:
462You have to allocate(or define) an array of pointers and set all
463of the array entries before using these functions.
464
4654.6 unregister_*probes
466
467#include <linux/kprobes.h>
468void unregister_kprobes(struct kprobe **kps, int num);
469void unregister_kretprobes(struct kretprobe **rps, int num);
470void unregister_jprobes(struct jprobe **jps, int num);
471
472Removes each of the num probes in the specified array at once.
473
474NOTE:
475If the functions find some incorrect probes (ex. unregistered
476probes) in the specified array, they clear the addr field of those
477incorrect probes. However, other probes in the array are
478unregistered correctly.
479
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -07004804.7 disable_*probe
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700481
482#include <linux/kprobes.h>
483int disable_kprobe(struct kprobe *kp);
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -0700484int disable_kretprobe(struct kretprobe *rp);
485int disable_jprobe(struct jprobe *jp);
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700486
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -0700487Temporarily disables the specified *probe. You can enable it again by using
488enable_*probe(). You must specify the probe which has been registered.
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700489
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -07004904.8 enable_*probe
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700491
492#include <linux/kprobes.h>
493int enable_kprobe(struct kprobe *kp);
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -0700494int enable_kretprobe(struct kretprobe *rp);
495int enable_jprobe(struct jprobe *jp);
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700496
Masami Hiramatsu8f9b1522009-04-06 19:01:02 -0700497Enables *probe which has been disabled by disable_*probe(). You must specify
498the probe which has been registered.
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700499
Jim Kenistond27a4dd2005-08-04 12:53:35 -07005005. Kprobes Features and Limitations
501
Jim Keniston8861da32006-02-14 13:53:06 -0800502Kprobes allows multiple probes at the same address. Currently,
503however, there cannot be multiple jprobes on the same function at
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500504the same time. Also, a probepoint for which there is a jprobe or
505a post_handler cannot be optimized. So if you install a jprobe,
506or a kprobe with a post_handler, at an optimized probepoint, the
507probepoint will be unoptimized automatically.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700508
509In general, you can install a probe anywhere in the kernel.
510In particular, you can probe interrupt handlers. Known exceptions
511are discussed in this section.
512
Jim Keniston8861da32006-02-14 13:53:06 -0800513The register_*probe functions will return -EINVAL if you attempt
514to install a probe in the code that implements Kprobes (mostly
515kernel/kprobes.c and arch/*/kernel/kprobes.c, but also functions such
516as do_page_fault and notifier_call_chain).
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700517
518If you install a probe in an inline-able function, Kprobes makes
519no attempt to chase down all inline instances of the function and
520install probes there. gcc may inline a function without being asked,
521so keep this in mind if you're not seeing the probe hits you expect.
522
523A probe handler can modify the environment of the probed function
524-- e.g., by modifying kernel data structures, or by modifying the
525contents of the pt_regs struct (which are restored to the registers
526upon return from the breakpoint). So Kprobes can be used, for example,
527to install a bug fix or to inject faults for testing. Kprobes, of
528course, has no way to distinguish the deliberately injected faults
529from the accidental ones. Don't drink and probe.
530
531Kprobes makes no attempt to prevent probe handlers from stepping on
532each other -- e.g., probing printk() and then calling printk() from a
Jim Keniston8861da32006-02-14 13:53:06 -0800533probe handler. If a probe handler hits a probe, that second probe's
534handlers won't be run in that instance, and the kprobe.nmissed member
535of the second probe will be incremented.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700536
Jim Keniston8861da32006-02-14 13:53:06 -0800537As of Linux v2.6.15-rc1, multiple handlers (or multiple instances of
538the same handler) may run concurrently on different CPUs.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700539
Jim Keniston8861da32006-02-14 13:53:06 -0800540Kprobes does not use mutexes or allocate memory except during
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700541registration and unregistration.
542
543Probe handlers are run with preemption disabled. Depending on the
544architecture, handlers may also run with interrupts disabled. In any
545case, your handler should not yield the CPU (e.g., by attempting to
546acquire a semaphore).
547
548Since a return probe is implemented by replacing the return
549address with the trampoline's address, stack backtraces and calls
550to __builtin_return_address() will typically yield the trampoline's
551address instead of the real return address for kretprobed functions.
552(As far as we can tell, __builtin_return_address() is used only
553for instrumentation and error reporting.)
554
Jim Keniston8861da32006-02-14 13:53:06 -0800555If the number of times a function is called does not match the number
556of times it returns, registering a return probe on that function may
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700557produce undesirable results. In such a case, a line:
558kretprobe BUG!: Processing kretprobe d000000000041aa8 @ c00000000004f48c
559gets printed. With this information, one will be able to correlate the
560exact instance of the kretprobe that caused the problem. We have the
561do_exit() case covered. do_execve() and do_fork() are not an issue.
562We're unaware of other specific cases where this could be a problem.
Jim Keniston8861da32006-02-14 13:53:06 -0800563
564If, upon entry to or exit from a function, the CPU is running on
565a stack other than that of the current task, registering a return
566probe on that function may produce undesirable results. For this
567reason, Kprobes doesn't support return probes (or kprobes or jprobes)
568on the x86_64 version of __switch_to(); the registration functions
569return -EINVAL.
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700570
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500571On x86/x86-64, since the Jump Optimization of Kprobes modifies
572instructions widely, there are some limitations to optimization. To
573explain it, we introduce some terminology. Imagine a 3-instruction
574sequence consisting of a two 2-byte instructions and one 3-byte
575instruction.
576
577 IA
578 |
579[-2][-1][0][1][2][3][4][5][6][7]
580 [ins1][ins2][ ins3 ]
581 [<- DCR ->]
582 [<- JTPR ->]
583
584ins1: 1st Instruction
585ins2: 2nd Instruction
586ins3: 3rd Instruction
587IA: Insertion Address
588JTPR: Jump Target Prohibition Region
589DCR: Detoured Code Region
590
591The instructions in DCR are copied to the out-of-line buffer
592of the kprobe, because the bytes in DCR are replaced by
593a 5-byte jump instruction. So there are several limitations.
594
595a) The instructions in DCR must be relocatable.
596b) The instructions in DCR must not include a call instruction.
597c) JTPR must not be targeted by any jump or call instruction.
598d) DCR must not straddle the border betweeen functions.
599
600Anyway, these limitations are checked by the in-kernel instruction
601decoder, so you don't need to worry about that.
602
Jim Kenistond27a4dd2005-08-04 12:53:35 -07006036. Probe Overhead
604
605On a typical CPU in use in 2005, a kprobe hit takes 0.5 to 1.0
606microseconds to process. Specifically, a benchmark that hits the same
607probepoint repeatedly, firing a simple handler each time, reports 1-2
608million hits per second, depending on the architecture. A jprobe or
609return-probe hit typically takes 50-75% longer than a kprobe hit.
610When you have a return probe set on a function, adding a kprobe at
611the entry to that function adds essentially no overhead.
612
613Here are sample overhead figures (in usec) for different architectures.
614k = kprobe; j = jprobe; r = return probe; kr = kprobe + return probe
615on same function; jr = jprobe + return probe on same function
616
617i386: Intel Pentium M, 1495 MHz, 2957.31 bogomips
618k = 0.57 usec; j = 1.00; r = 0.92; kr = 0.99; jr = 1.40
619
620x86_64: AMD Opteron 246, 1994 MHz, 3971.48 bogomips
621k = 0.49 usec; j = 0.76; r = 0.80; kr = 0.82; jr = 1.07
622
623ppc64: POWER5 (gr), 1656 MHz (SMT disabled, 1 virtual CPU per physical CPU)
624k = 0.77 usec; j = 1.31; r = 1.26; kr = 1.45; jr = 1.99
625
Masami Hiramatsub26486b2010-02-25 08:35:04 -05006266.1 Optimized Probe Overhead
627
628Typically, an optimized kprobe hit takes 0.07 to 0.1 microseconds to
629process. Here are sample overhead figures (in usec) for x86 architectures.
630k = unoptimized kprobe, b = boosted (single-step skipped), o = optimized kprobe,
631r = unoptimized kretprobe, rb = boosted kretprobe, ro = optimized kretprobe.
632
633i386: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips
634k = 0.80 usec; b = 0.33; o = 0.05; r = 1.10; rb = 0.61; ro = 0.33
635
636x86-64: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips
637k = 0.99 usec; b = 0.43; o = 0.06; r = 1.24; rb = 0.68; ro = 0.30
638
Jim Kenistond27a4dd2005-08-04 12:53:35 -07006397. TODO
640
Jim Keniston8861da32006-02-14 13:53:06 -0800641a. SystemTap (http://sourceware.org/systemtap): Provides a simplified
642programming interface for probe-based instrumentation. Try it out.
643b. Kernel return probes for sparc64.
644c. Support for other architectures.
645d. User-space probes.
646e. Watchpoint probes (which fire on data references).
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700647
6488. Kprobes Example
649
Ananth N Mavinakayanahalli804defe2008-03-04 14:28:38 -0800650See samples/kprobes/kprobe_example.c
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700651
6529. Jprobes Example
653
Ananth N Mavinakayanahalli804defe2008-03-04 14:28:38 -0800654See samples/kprobes/jprobe_example.c
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700655
65610. Kretprobes Example
657
Ananth N Mavinakayanahalli804defe2008-03-04 14:28:38 -0800658See samples/kprobes/kretprobe_example.c
Jim Kenistond27a4dd2005-08-04 12:53:35 -0700659
660For additional information on Kprobes, refer to the following URLs:
661http://www-106.ibm.com/developerworks/library/l-kprobes.html?ca=dgr-lnxw42Kprobe
662http://www.redhat.com/magazine/005mar05/features/kprobes/
Ananth N Mavinakayanahalli09b18202006-10-02 02:17:32 -0700663http://www-users.cs.umn.edu/~boutcher/kprobes/
664http://www.linuxsymposium.org/2006/linuxsymposium_procv2.pdf (pages 101-115)
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700665
666
667Appendix A: The kprobes debugfs interface
668
669With recent kernels (> 2.6.20) the list of registered kprobes is visible
GeunSik Lim156f5a72009-06-02 15:01:37 +0900670under the /sys/kernel/debug/kprobes/ directory (assuming debugfs is mounted at //sys/kernel/debug).
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700671
GeunSik Lim156f5a72009-06-02 15:01:37 +0900672/sys/kernel/debug/kprobes/list: Lists all registered probes on the system
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700673
674c015d71a k vfs_read+0x0
675c011a316 j do_fork+0x0
676c03dedc5 r tcp_v4_rcv+0x0
677
678The first column provides the kernel address where the probe is inserted.
679The second column identifies the type of probe (k - kprobe, r - kretprobe
680and j - jprobe), while the third column specifies the symbol+offset of
681the probe. If the probed function belongs to a module, the module name
Masami Hiramatsue8386a02009-01-06 14:41:52 -0800682is also specified. Following columns show probe status. If the probe is on
683a virtual address that is no longer valid (module init sections, module
684virtual addresses that correspond to modules that've been unloaded),
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700685such probes are marked with [GONE]. If the probe is temporarily disabled,
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500686such probes are marked with [DISABLED]. If the probe is optimized, it is
687marked with [OPTIMIZED].
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700688
GeunSik Lim156f5a72009-06-02 15:01:37 +0900689/sys/kernel/debug/kprobes/enabled: Turn kprobes ON/OFF forcibly.
Ananth N Mavinakayanahallibf8f6e52007-05-08 00:34:16 -0700690
Masami Hiramatsude5bd882009-04-06 19:01:02 -0700691Provides a knob to globally and forcibly turn registered kprobes ON or OFF.
692By default, all kprobes are enabled. By echoing "0" to this file, all
693registered probes will be disarmed, till such time a "1" is echoed to this
694file. Note that this knob just disarms and arms all kprobes and doesn't
695change each probe's disabling state. This means that disabled kprobes (marked
696[DISABLED]) will be not enabled if you turn ON all kprobes by this knob.
Masami Hiramatsub26486b2010-02-25 08:35:04 -0500697
698
699Appendix B: The kprobes sysctl interface
700
701/proc/sys/debug/kprobes-optimization: Turn kprobes optimization ON/OFF.
702
703When CONFIG_OPTPROBES=y, this sysctl interface appears and it provides
704a knob to globally and forcibly turn jump optimization (see section
7051.4) ON or OFF. By default, jump optimization is allowed (ON).
706If you echo "0" to this file or set "debug.kprobes_optimization" to
7070 via sysctl, all optimized probes will be unoptimized, and any new
708probes registered after that will not be optimized. Note that this
709knob *changes* the optimized state. This means that optimized probes
710(marked [OPTIMIZED]) will be unoptimized ([OPTIMIZED] tag will be
711removed). If the knob is turned on, they will be optimized again.
712