Based on kernel version 4.9. Page generated on 2016-12-21 14:34 EST.
1 Title : Kernel Probes (Kprobes) 2 Authors : Jim Keniston <firstname.lastname@example.org> 3 : Prasanna S Panchamukhi <email@example.com> 4 : Masami Hiramatsu <firstname.lastname@example.org> 5 6 CONTENTS 7 8 1. Concepts: Kprobes, Jprobes, Return Probes 9 2. Architectures Supported 10 3. Configuring Kprobes 11 4. API Reference 12 5. Kprobes Features and Limitations 13 6. Probe Overhead 14 7. TODO 15 8. Kprobes Example 16 9. Jprobes Example 17 10. Kretprobes Example 18 Appendix A: The kprobes debugfs interface 19 Appendix B: The kprobes sysctl interface 20 21 1. Concepts: Kprobes, Jprobes, Return Probes 22 23 Kprobes enables you to dynamically break into any kernel routine and 24 collect debugging and performance information non-disruptively. You 25 can trap at almost any kernel code address(*), specifying a handler 26 routine to be invoked when the breakpoint is hit. 27 (*: some parts of the kernel code can not be trapped, see 1.5 Blacklist) 28 29 There are currently three types of probes: kprobes, jprobes, and 30 kretprobes (also called return probes). A kprobe can be inserted 31 on virtually any instruction in the kernel. A jprobe is inserted at 32 the entry to a kernel function, and provides convenient access to the 33 function's arguments. A return probe fires when a specified function 34 returns. 35 36 In the typical case, Kprobes-based instrumentation is packaged as 37 a kernel module. The module's init function installs ("registers") 38 one or more probes, and the exit function unregisters them. A 39 registration function such as register_kprobe() specifies where 40 the probe is to be inserted and what handler is to be called when 41 the probe is hit. 42 43 There are also register_/unregister_*probes() functions for batch 44 registration/unregistration of a group of *probes. These functions 45 can speed up unregistration process when you have to unregister 46 a lot of probes at once. 47 48 The next four subsections explain how the different types of 49 probes work and how jump optimization works. They explain certain 50 things that you'll need to know in order to make the best use of 51 Kprobes -- e.g., the difference between a pre_handler and 52 a post_handler, and how to use the maxactive and nmissed fields of 53 a kretprobe. But if you're in a hurry to start using Kprobes, you 54 can skip ahead to section 2. 55 56 1.1 How Does a Kprobe Work? 57 58 When a kprobe is registered, Kprobes makes a copy of the probed 59 instruction and replaces the first byte(s) of the probed instruction 60 with a breakpoint instruction (e.g., int3 on i386 and x86_64). 61 62 When a CPU hits the breakpoint instruction, a trap occurs, the CPU's 63 registers are saved, and control passes to Kprobes via the 64 notifier_call_chain mechanism. Kprobes executes the "pre_handler" 65 associated with the kprobe, passing the handler the addresses of the 66 kprobe struct and the saved registers. 67 68 Next, Kprobes single-steps its copy of the probed instruction. 69 (It would be simpler to single-step the actual instruction in place, 70 but then Kprobes would have to temporarily remove the breakpoint 71 instruction. This would open a small time window when another CPU 72 could sail right past the probepoint.) 73 74 After the instruction is single-stepped, Kprobes executes the 75 "post_handler," if any, that is associated with the kprobe. 76 Execution then continues with the instruction following the probepoint. 77 78 1.2 How Does a Jprobe Work? 79 80 A jprobe is implemented using a kprobe that is placed on a function's 81 entry point. It employs a simple mirroring principle to allow 82 seamless access to the probed function's arguments. The jprobe 83 handler routine should have the same signature (arg list and return 84 type) as the function being probed, and must always end by calling 85 the Kprobes function jprobe_return(). 86 87 Here's how it works. When the probe is hit, Kprobes makes a copy of 88 the saved registers and a generous portion of the stack (see below). 89 Kprobes then points the saved instruction pointer at the jprobe's 90 handler routine, and returns from the trap. As a result, control 91 passes to the handler, which is presented with the same register and 92 stack contents as the probed function. When it is done, the handler 93 calls jprobe_return(), which traps again to restore the original stack 94 contents and processor state and switch to the probed function. 95 96 By convention, the callee owns its arguments, so gcc may produce code 97 that unexpectedly modifies that portion of the stack. This is why 98 Kprobes saves a copy of the stack and restores it after the jprobe 99 handler has run. Up to MAX_STACK_SIZE bytes are copied -- e.g., 100 64 bytes on i386. 101 102 Note that the probed function's args may be passed on the stack 103 or in registers. The jprobe will work in either case, so long as the 104 handler's prototype matches that of the probed function. 105 106 Note that in some architectures (e.g.: arm64 and sparc64) the stack 107 copy is not done, as the actual location of stacked parameters may be 108 outside of a reasonable MAX_STACK_SIZE value and because that location 109 cannot be determined by the jprobes code. In this case the jprobes 110 user must be careful to make certain the calling signature of the 111 function does not cause parameters to be passed on the stack (e.g.: 112 more than eight function arguments, an argument of more than sixteen 113 bytes, or more than 64 bytes of argument data, depending on 114 architecture). 115 116 1.3 Return Probes 117 118 1.3.1 How Does a Return Probe Work? 119 120 When you call register_kretprobe(), Kprobes establishes a kprobe at 121 the entry to the function. When the probed function is called and this 122 probe is hit, Kprobes saves a copy of the return address, and replaces 123 the return address with the address of a "trampoline." The trampoline 124 is an arbitrary piece of code -- typically just a nop instruction. 125 At boot time, Kprobes registers a kprobe at the trampoline. 126 127 When the probed function executes its return instruction, control 128 passes to the trampoline and that probe is hit. Kprobes' trampoline 129 handler calls the user-specified return handler associated with the 130 kretprobe, then sets the saved instruction pointer to the saved return 131 address, and that's where execution resumes upon return from the trap. 132 133 While the probed function is executing, its return address is 134 stored in an object of type kretprobe_instance. Before calling 135 register_kretprobe(), the user sets the maxactive field of the 136 kretprobe struct to specify how many instances of the specified 137 function can be probed simultaneously. register_kretprobe() 138 pre-allocates the indicated number of kretprobe_instance objects. 139 140 For example, if the function is non-recursive and is called with a 141 spinlock held, maxactive = 1 should be enough. If the function is 142 non-recursive and can never relinquish the CPU (e.g., via a semaphore 143 or preemption), NR_CPUS should be enough. If maxactive <= 0, it is 144 set to a default value. If CONFIG_PREEMPT is enabled, the default 145 is max(10, 2*NR_CPUS). Otherwise, the default is NR_CPUS. 146 147 It's not a disaster if you set maxactive too low; you'll just miss 148 some probes. In the kretprobe struct, the nmissed field is set to 149 zero when the return probe is registered, and is incremented every 150 time the probed function is entered but there is no kretprobe_instance 151 object available for establishing the return probe. 152 153 1.3.2 Kretprobe entry-handler 154 155 Kretprobes also provides an optional user-specified handler which runs 156 on function entry. This handler is specified by setting the entry_handler 157 field of the kretprobe struct. Whenever the kprobe placed by kretprobe at the 158 function entry is hit, the user-defined entry_handler, if any, is invoked. 159 If the entry_handler returns 0 (success) then a corresponding return handler 160 is guaranteed to be called upon function return. If the entry_handler 161 returns a non-zero error then Kprobes leaves the return address as is, and 162 the kretprobe has no further effect for that particular function instance. 163 164 Multiple entry and return handler invocations are matched using the unique 165 kretprobe_instance object associated with them. Additionally, a user 166 may also specify per return-instance private data to be part of each 167 kretprobe_instance object. This is especially useful when sharing private 168 data between corresponding user entry and return handlers. The size of each 169 private data object can be specified at kretprobe registration time by 170 setting the data_size field of the kretprobe struct. This data can be 171 accessed through the data field of each kretprobe_instance object. 172 173 In case probed function is entered but there is no kretprobe_instance 174 object available, then in addition to incrementing the nmissed count, 175 the user entry_handler invocation is also skipped. 176 177 1.4 How Does Jump Optimization Work? 178 179 If your kernel is built with CONFIG_OPTPROBES=y (currently this flag 180 is automatically set 'y' on x86/x86-64, non-preemptive kernel) and 181 the "debug.kprobes_optimization" kernel parameter is set to 1 (see 182 sysctl(8)), Kprobes tries to reduce probe-hit overhead by using a jump 183 instruction instead of a breakpoint instruction at each probepoint. 184 185 1.4.1 Init a Kprobe 186 187 When a probe is registered, before attempting this optimization, 188 Kprobes inserts an ordinary, breakpoint-based kprobe at the specified 189 address. So, even if it's not possible to optimize this particular 190 probepoint, there'll be a probe there. 191 192 1.4.2 Safety Check 193 194 Before optimizing a probe, Kprobes performs the following safety checks: 195 196 - Kprobes verifies that the region that will be replaced by the jump 197 instruction (the "optimized region") lies entirely within one function. 198 (A jump instruction is multiple bytes, and so may overlay multiple 199 instructions.) 200 201 - Kprobes analyzes the entire function and verifies that there is no 202 jump into the optimized region. Specifically: 203 - the function contains no indirect jump; 204 - the function contains no instruction that causes an exception (since 205 the fixup code triggered by the exception could jump back into the 206 optimized region -- Kprobes checks the exception tables to verify this); 207 and 208 - there is no near jump to the optimized region (other than to the first 209 byte). 210 211 - For each instruction in the optimized region, Kprobes verifies that 212 the instruction can be executed out of line. 213 214 1.4.3 Preparing Detour Buffer 215 216 Next, Kprobes prepares a "detour" buffer, which contains the following 217 instruction sequence: 218 - code to push the CPU's registers (emulating a breakpoint trap) 219 - a call to the trampoline code which calls user's probe handlers. 220 - code to restore registers 221 - the instructions from the optimized region 222 - a jump back to the original execution path. 223 224 1.4.4 Pre-optimization 225 226 After preparing the detour buffer, Kprobes verifies that none of the 227 following situations exist: 228 - The probe has either a break_handler (i.e., it's a jprobe) or a 229 post_handler. 230 - Other instructions in the optimized region are probed. 231 - The probe is disabled. 232 In any of the above cases, Kprobes won't start optimizing the probe. 233 Since these are temporary situations, Kprobes tries to start 234 optimizing it again if the situation is changed. 235 236 If the kprobe can be optimized, Kprobes enqueues the kprobe to an 237 optimizing list, and kicks the kprobe-optimizer workqueue to optimize 238 it. If the to-be-optimized probepoint is hit before being optimized, 239 Kprobes returns control to the original instruction path by setting 240 the CPU's instruction pointer to the copied code in the detour buffer 241 -- thus at least avoiding the single-step. 242 243 1.4.5 Optimization 244 245 The Kprobe-optimizer doesn't insert the jump instruction immediately; 246 rather, it calls synchronize_sched() for safety first, because it's 247 possible for a CPU to be interrupted in the middle of executing the 248 optimized region(*). As you know, synchronize_sched() can ensure 249 that all interruptions that were active when synchronize_sched() 250 was called are done, but only if CONFIG_PREEMPT=n. So, this version 251 of kprobe optimization supports only kernels with CONFIG_PREEMPT=n.(**) 252 253 After that, the Kprobe-optimizer calls stop_machine() to replace 254 the optimized region with a jump instruction to the detour buffer, 255 using text_poke_smp(). 256 257 1.4.6 Unoptimization 258 259 When an optimized kprobe is unregistered, disabled, or blocked by 260 another kprobe, it will be unoptimized. If this happens before 261 the optimization is complete, the kprobe is just dequeued from the 262 optimized list. If the optimization has been done, the jump is 263 replaced with the original code (except for an int3 breakpoint in 264 the first byte) by using text_poke_smp(). 265 266 (*)Please imagine that the 2nd instruction is interrupted and then 267 the optimizer replaces the 2nd instruction with the jump *address* 268 while the interrupt handler is running. When the interrupt 269 returns to original address, there is no valid instruction, 270 and it causes an unexpected result. 271 272 (**)This optimization-safety checking may be replaced with the 273 stop-machine method that ksplice uses for supporting a CONFIG_PREEMPT=y 274 kernel. 275 276 NOTE for geeks: 277 The jump optimization changes the kprobe's pre_handler behavior. 278 Without optimization, the pre_handler can change the kernel's execution 279 path by changing regs->ip and returning 1. However, when the probe 280 is optimized, that modification is ignored. Thus, if you want to 281 tweak the kernel's execution path, you need to suppress optimization, 282 using one of the following techniques: 283 - Specify an empty function for the kprobe's post_handler or break_handler. 284 or 285 - Execute 'sysctl -w debug.kprobes_optimization=n' 286 287 1.5 Blacklist 288 289 Kprobes can probe most of the kernel except itself. This means 290 that there are some functions where kprobes cannot probe. Probing 291 (trapping) such functions can cause a recursive trap (e.g. double 292 fault) or the nested probe handler may never be called. 293 Kprobes manages such functions as a blacklist. 294 If you want to add a function into the blacklist, you just need 295 to (1) include linux/kprobes.h and (2) use NOKPROBE_SYMBOL() macro 296 to specify a blacklisted function. 297 Kprobes checks the given probe address against the blacklist and 298 rejects registering it, if the given address is in the blacklist. 299 300 2. Architectures Supported 301 302 Kprobes, jprobes, and return probes are implemented on the following 303 architectures: 304 305 - i386 (Supports jump optimization) 306 - x86_64 (AMD-64, EM64T) (Supports jump optimization) 307 - ppc64 308 - ia64 (Does not support probes on instruction slot1.) 309 - sparc64 (Return probes not yet implemented.) 310 - arm 311 - ppc 312 - mips 313 - s390 314 315 3. Configuring Kprobes 316 317 When configuring the kernel using make menuconfig/xconfig/oldconfig, 318 ensure that CONFIG_KPROBES is set to "y". Under "General setup", look 319 for "Kprobes". 320 321 So that you can load and unload Kprobes-based instrumentation modules, 322 make sure "Loadable module support" (CONFIG_MODULES) and "Module 323 unloading" (CONFIG_MODULE_UNLOAD) are set to "y". 324 325 Also make sure that CONFIG_KALLSYMS and perhaps even CONFIG_KALLSYMS_ALL 326 are set to "y", since kallsyms_lookup_name() is used by the in-kernel 327 kprobe address resolution code. 328 329 If you need to insert a probe in the middle of a function, you may find 330 it useful to "Compile the kernel with debug info" (CONFIG_DEBUG_INFO), 331 so you can use "objdump -d -l vmlinux" to see the source-to-object 332 code mapping. 333 334 4. API Reference 335 336 The Kprobes API includes a "register" function and an "unregister" 337 function for each type of probe. The API also includes "register_*probes" 338 and "unregister_*probes" functions for (un)registering arrays of probes. 339 Here are terse, mini-man-page specifications for these functions and 340 the associated probe handlers that you'll write. See the files in the 341 samples/kprobes/ sub-directory for examples. 342 343 4.1 register_kprobe 344 345 #include <linux/kprobes.h> 346 int register_kprobe(struct kprobe *kp); 347 348 Sets a breakpoint at the address kp->addr. When the breakpoint is 349 hit, Kprobes calls kp->pre_handler. After the probed instruction 350 is single-stepped, Kprobe calls kp->post_handler. If a fault 351 occurs during execution of kp->pre_handler or kp->post_handler, 352 or during single-stepping of the probed instruction, Kprobes calls 353 kp->fault_handler. Any or all handlers can be NULL. If kp->flags 354 is set KPROBE_FLAG_DISABLED, that kp will be registered but disabled, 355 so, its handlers aren't hit until calling enable_kprobe(kp). 356 357 NOTE: 358 1. With the introduction of the "symbol_name" field to struct kprobe, 359 the probepoint address resolution will now be taken care of by the kernel. 360 The following will now work: 361 362 kp.symbol_name = "symbol_name"; 363 364 (64-bit powerpc intricacies such as function descriptors are handled 365 transparently) 366 367 2. Use the "offset" field of struct kprobe if the offset into the symbol 368 to install a probepoint is known. This field is used to calculate the 369 probepoint. 370 371 3. Specify either the kprobe "symbol_name" OR the "addr". If both are 372 specified, kprobe registration will fail with -EINVAL. 373 374 4. With CISC architectures (such as i386 and x86_64), the kprobes code 375 does not validate if the kprobe.addr is at an instruction boundary. 376 Use "offset" with caution. 377 378 register_kprobe() returns 0 on success, or a negative errno otherwise. 379 380 User's pre-handler (kp->pre_handler): 381 #include <linux/kprobes.h> 382 #include <linux/ptrace.h> 383 int pre_handler(struct kprobe *p, struct pt_regs *regs); 384 385 Called with p pointing to the kprobe associated with the breakpoint, 386 and regs pointing to the struct containing the registers saved when 387 the breakpoint was hit. Return 0 here unless you're a Kprobes geek. 388 389 User's post-handler (kp->post_handler): 390 #include <linux/kprobes.h> 391 #include <linux/ptrace.h> 392 void post_handler(struct kprobe *p, struct pt_regs *regs, 393 unsigned long flags); 394 395 p and regs are as described for the pre_handler. flags always seems 396 to be zero. 397 398 User's fault-handler (kp->fault_handler): 399 #include <linux/kprobes.h> 400 #include <linux/ptrace.h> 401 int fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr); 402 403 p and regs are as described for the pre_handler. trapnr is the 404 architecture-specific trap number associated with the fault (e.g., 405 on i386, 13 for a general protection fault or 14 for a page fault). 406 Returns 1 if it successfully handled the exception. 407 408 4.2 register_jprobe 409 410 #include <linux/kprobes.h> 411 int register_jprobe(struct jprobe *jp) 412 413 Sets a breakpoint at the address jp->kp.addr, which must be the address 414 of the first instruction of a function. When the breakpoint is hit, 415 Kprobes runs the handler whose address is jp->entry. 416 417 The handler should have the same arg list and return type as the probed 418 function; and just before it returns, it must call jprobe_return(). 419 (The handler never actually returns, since jprobe_return() returns 420 control to Kprobes.) If the probed function is declared asmlinkage 421 or anything else that affects how args are passed, the handler's 422 declaration must match. 423 424 register_jprobe() returns 0 on success, or a negative errno otherwise. 425 426 4.3 register_kretprobe 427 428 #include <linux/kprobes.h> 429 int register_kretprobe(struct kretprobe *rp); 430 431 Establishes a return probe for the function whose address is 432 rp->kp.addr. When that function returns, Kprobes calls rp->handler. 433 You must set rp->maxactive appropriately before you call 434 register_kretprobe(); see "How Does a Return Probe Work?" for details. 435 436 register_kretprobe() returns 0 on success, or a negative errno 437 otherwise. 438 439 User's return-probe handler (rp->handler): 440 #include <linux/kprobes.h> 441 #include <linux/ptrace.h> 442 int kretprobe_handler(struct kretprobe_instance *ri, struct pt_regs *regs); 443 444 regs is as described for kprobe.pre_handler. ri points to the 445 kretprobe_instance object, of which the following fields may be 446 of interest: 447 - ret_addr: the return address 448 - rp: points to the corresponding kretprobe object 449 - task: points to the corresponding task struct 450 - data: points to per return-instance private data; see "Kretprobe 451 entry-handler" for details. 452 453 The regs_return_value(regs) macro provides a simple abstraction to 454 extract the return value from the appropriate register as defined by 455 the architecture's ABI. 456 457 The handler's return value is currently ignored. 458 459 4.4 unregister_*probe 460 461 #include <linux/kprobes.h> 462 void unregister_kprobe(struct kprobe *kp); 463 void unregister_jprobe(struct jprobe *jp); 464 void unregister_kretprobe(struct kretprobe *rp); 465 466 Removes the specified probe. The unregister function can be called 467 at any time after the probe has been registered. 468 469 NOTE: 470 If the functions find an incorrect probe (ex. an unregistered probe), 471 they clear the addr field of the probe. 472 473 4.5 register_*probes 474 475 #include <linux/kprobes.h> 476 int register_kprobes(struct kprobe **kps, int num); 477 int register_kretprobes(struct kretprobe **rps, int num); 478 int register_jprobes(struct jprobe **jps, int num); 479 480 Registers each of the num probes in the specified array. If any 481 error occurs during registration, all probes in the array, up to 482 the bad probe, are safely unregistered before the register_*probes 483 function returns. 484 - kps/rps/jps: an array of pointers to *probe data structures 485 - num: the number of the array entries. 486 487 NOTE: 488 You have to allocate(or define) an array of pointers and set all 489 of the array entries before using these functions. 490 491 4.6 unregister_*probes 492 493 #include <linux/kprobes.h> 494 void unregister_kprobes(struct kprobe **kps, int num); 495 void unregister_kretprobes(struct kretprobe **rps, int num); 496 void unregister_jprobes(struct jprobe **jps, int num); 497 498 Removes each of the num probes in the specified array at once. 499 500 NOTE: 501 If the functions find some incorrect probes (ex. unregistered 502 probes) in the specified array, they clear the addr field of those 503 incorrect probes. However, other probes in the array are 504 unregistered correctly. 505 506 4.7 disable_*probe 507 508 #include <linux/kprobes.h> 509 int disable_kprobe(struct kprobe *kp); 510 int disable_kretprobe(struct kretprobe *rp); 511 int disable_jprobe(struct jprobe *jp); 512 513 Temporarily disables the specified *probe. You can enable it again by using 514 enable_*probe(). You must specify the probe which has been registered. 515 516 4.8 enable_*probe 517 518 #include <linux/kprobes.h> 519 int enable_kprobe(struct kprobe *kp); 520 int enable_kretprobe(struct kretprobe *rp); 521 int enable_jprobe(struct jprobe *jp); 522 523 Enables *probe which has been disabled by disable_*probe(). You must specify 524 the probe which has been registered. 525 526 5. Kprobes Features and Limitations 527 528 Kprobes allows multiple probes at the same address. Currently, 529 however, there cannot be multiple jprobes on the same function at 530 the same time. Also, a probepoint for which there is a jprobe or 531 a post_handler cannot be optimized. So if you install a jprobe, 532 or a kprobe with a post_handler, at an optimized probepoint, the 533 probepoint will be unoptimized automatically. 534 535 In general, you can install a probe anywhere in the kernel. 536 In particular, you can probe interrupt handlers. Known exceptions 537 are discussed in this section. 538 539 The register_*probe functions will return -EINVAL if you attempt 540 to install a probe in the code that implements Kprobes (mostly 541 kernel/kprobes.c and arch/*/kernel/kprobes.c, but also functions such 542 as do_page_fault and notifier_call_chain). 543 544 If you install a probe in an inline-able function, Kprobes makes 545 no attempt to chase down all inline instances of the function and 546 install probes there. gcc may inline a function without being asked, 547 so keep this in mind if you're not seeing the probe hits you expect. 548 549 A probe handler can modify the environment of the probed function 550 -- e.g., by modifying kernel data structures, or by modifying the 551 contents of the pt_regs struct (which are restored to the registers 552 upon return from the breakpoint). So Kprobes can be used, for example, 553 to install a bug fix or to inject faults for testing. Kprobes, of 554 course, has no way to distinguish the deliberately injected faults 555 from the accidental ones. Don't drink and probe. 556 557 Kprobes makes no attempt to prevent probe handlers from stepping on 558 each other -- e.g., probing printk() and then calling printk() from a 559 probe handler. If a probe handler hits a probe, that second probe's 560 handlers won't be run in that instance, and the kprobe.nmissed member 561 of the second probe will be incremented. 562 563 As of Linux v2.6.15-rc1, multiple handlers (or multiple instances of 564 the same handler) may run concurrently on different CPUs. 565 566 Kprobes does not use mutexes or allocate memory except during 567 registration and unregistration. 568 569 Probe handlers are run with preemption disabled. Depending on the 570 architecture and optimization state, handlers may also run with 571 interrupts disabled (e.g., kretprobe handlers and optimized kprobe 572 handlers run without interrupt disabled on x86/x86-64). In any case, 573 your handler should not yield the CPU (e.g., by attempting to acquire 574 a semaphore). 575 576 Since a return probe is implemented by replacing the return 577 address with the trampoline's address, stack backtraces and calls 578 to __builtin_return_address() will typically yield the trampoline's 579 address instead of the real return address for kretprobed functions. 580 (As far as we can tell, __builtin_return_address() is used only 581 for instrumentation and error reporting.) 582 583 If the number of times a function is called does not match the number 584 of times it returns, registering a return probe on that function may 585 produce undesirable results. In such a case, a line: 586 kretprobe BUG!: Processing kretprobe d000000000041aa8 @ c00000000004f48c 587 gets printed. With this information, one will be able to correlate the 588 exact instance of the kretprobe that caused the problem. We have the 589 do_exit() case covered. do_execve() and do_fork() are not an issue. 590 We're unaware of other specific cases where this could be a problem. 591 592 If, upon entry to or exit from a function, the CPU is running on 593 a stack other than that of the current task, registering a return 594 probe on that function may produce undesirable results. For this 595 reason, Kprobes doesn't support return probes (or kprobes or jprobes) 596 on the x86_64 version of __switch_to(); the registration functions 597 return -EINVAL. 598 599 On x86/x86-64, since the Jump Optimization of Kprobes modifies 600 instructions widely, there are some limitations to optimization. To 601 explain it, we introduce some terminology. Imagine a 3-instruction 602 sequence consisting of a two 2-byte instructions and one 3-byte 603 instruction. 604 605 IA 606 | 607 [-2][-1] 608 [ins1][ins2][ ins3 ] 609 [<- DCR ->] 610 [<- JTPR ->] 611 612 ins1: 1st Instruction 613 ins2: 2nd Instruction 614 ins3: 3rd Instruction 615 IA: Insertion Address 616 JTPR: Jump Target Prohibition Region 617 DCR: Detoured Code Region 618 619 The instructions in DCR are copied to the out-of-line buffer 620 of the kprobe, because the bytes in DCR are replaced by 621 a 5-byte jump instruction. So there are several limitations. 622 623 a) The instructions in DCR must be relocatable. 624 b) The instructions in DCR must not include a call instruction. 625 c) JTPR must not be targeted by any jump or call instruction. 626 d) DCR must not straddle the border between functions. 627 628 Anyway, these limitations are checked by the in-kernel instruction 629 decoder, so you don't need to worry about that. 630 631 6. Probe Overhead 632 633 On a typical CPU in use in 2005, a kprobe hit takes 0.5 to 1.0 634 microseconds to process. Specifically, a benchmark that hits the same 635 probepoint repeatedly, firing a simple handler each time, reports 1-2 636 million hits per second, depending on the architecture. A jprobe or 637 return-probe hit typically takes 50-75% longer than a kprobe hit. 638 When you have a return probe set on a function, adding a kprobe at 639 the entry to that function adds essentially no overhead. 640 641 Here are sample overhead figures (in usec) for different architectures. 642 k = kprobe; j = jprobe; r = return probe; kr = kprobe + return probe 643 on same function; jr = jprobe + return probe on same function 644 645 i386: Intel Pentium M, 1495 MHz, 2957.31 bogomips 646 k = 0.57 usec; j = 1.00; r = 0.92; kr = 0.99; jr = 1.40 647 648 x86_64: AMD Opteron 246, 1994 MHz, 3971.48 bogomips 649 k = 0.49 usec; j = 0.76; r = 0.80; kr = 0.82; jr = 1.07 650 651 ppc64: POWER5 (gr), 1656 MHz (SMT disabled, 1 virtual CPU per physical CPU) 652 k = 0.77 usec; j = 1.31; r = 1.26; kr = 1.45; jr = 1.99 653 654 6.1 Optimized Probe Overhead 655 656 Typically, an optimized kprobe hit takes 0.07 to 0.1 microseconds to 657 process. Here are sample overhead figures (in usec) for x86 architectures. 658 k = unoptimized kprobe, b = boosted (single-step skipped), o = optimized kprobe, 659 r = unoptimized kretprobe, rb = boosted kretprobe, ro = optimized kretprobe. 660 661 i386: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips 662 k = 0.80 usec; b = 0.33; o = 0.05; r = 1.10; rb = 0.61; ro = 0.33 663 664 x86-64: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips 665 k = 0.99 usec; b = 0.43; o = 0.06; r = 1.24; rb = 0.68; ro = 0.30 666 667 7. TODO 668 669 a. SystemTap (http://sourceware.org/systemtap): Provides a simplified 670 programming interface for probe-based instrumentation. Try it out. 671 b. Kernel return probes for sparc64. 672 c. Support for other architectures. 673 d. User-space probes. 674 e. Watchpoint probes (which fire on data references). 675 676 8. Kprobes Example 677 678 See samples/kprobes/kprobe_example.c 679 680 9. Jprobes Example 681 682 See samples/kprobes/jprobe_example.c 683 684 10. Kretprobes Example 685 686 See samples/kprobes/kretprobe_example.c 687 688 For additional information on Kprobes, refer to the following URLs: 689 http://www-106.ibm.com/developerworks/library/l-kprobes.html?ca=dgr-lnxw42Kprobe 690 http://www.redhat.com/magazine/005mar05/features/kprobes/ 691 http://www-users.cs.umn.edu/~boutcher/kprobes/ 692 http://www.linuxsymposium.org/2006/linuxsymposium_procv2.pdf (pages 101-115) 693 694 695 Appendix A: The kprobes debugfs interface 696 697 With recent kernels (> 2.6.20) the list of registered kprobes is visible 698 under the /sys/kernel/debug/kprobes/ directory (assuming debugfs is mounted at //sys/kernel/debug). 699 700 /sys/kernel/debug/kprobes/list: Lists all registered probes on the system 701 702 c015d71a k vfs_read+0x0 703 c011a316 j do_fork+0x0 704 c03dedc5 r tcp_v4_rcv+0x0 705 706 The first column provides the kernel address where the probe is inserted. 707 The second column identifies the type of probe (k - kprobe, r - kretprobe 708 and j - jprobe), while the third column specifies the symbol+offset of 709 the probe. If the probed function belongs to a module, the module name 710 is also specified. Following columns show probe status. If the probe is on 711 a virtual address that is no longer valid (module init sections, module 712 virtual addresses that correspond to modules that've been unloaded), 713 such probes are marked with [GONE]. If the probe is temporarily disabled, 714 such probes are marked with [DISABLED]. If the probe is optimized, it is 715 marked with [OPTIMIZED]. If the probe is ftrace-based, it is marked with 716 [FTRACE]. 717 718 /sys/kernel/debug/kprobes/enabled: Turn kprobes ON/OFF forcibly. 719 720 Provides a knob to globally and forcibly turn registered kprobes ON or OFF. 721 By default, all kprobes are enabled. By echoing "0" to this file, all 722 registered probes will be disarmed, till such time a "1" is echoed to this 723 file. Note that this knob just disarms and arms all kprobes and doesn't 724 change each probe's disabling state. This means that disabled kprobes (marked 725 [DISABLED]) will be not enabled if you turn ON all kprobes by this knob. 726 727 728 Appendix B: The kprobes sysctl interface 729 730 /proc/sys/debug/kprobes-optimization: Turn kprobes optimization ON/OFF. 731 732 When CONFIG_OPTPROBES=y, this sysctl interface appears and it provides 733 a knob to globally and forcibly turn jump optimization (see section 734 1.4) ON or OFF. By default, jump optimization is allowed (ON). 735 If you echo "0" to this file or set "debug.kprobes_optimization" to 736 0 via sysctl, all optimized probes will be unoptimized, and any new 737 probes registered after that will not be optimized. Note that this 738 knob *changes* the optimized state. This means that optimized probes 739 (marked [OPTIMIZED]) will be unoptimized ([OPTIMIZED] tag will be 740 removed). If the knob is turned on, they will be optimized again.