linux/Documentation/kprobes.txt
<<
>>
Prefs
   1Title   : Kernel Probes (Kprobes)
   2Authors : Jim Keniston <jkenisto@us.ibm.com>
   3        : Prasanna S Panchamukhi <prasanna.panchamukhi@gmail.com>
   4        : Masami Hiramatsu <mhiramat@redhat.com>
   5
   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
  18Appendix A: The kprobes debugfs interface
  19Appendix B: The kprobes sysctl interface
  20
  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(*: some parts of the kernel code can not be trapped, see 1.5 Blacklist)
  28
  29There are currently three types of probes: kprobes, jprobes, and
  30kretprobes (also called return probes).  A kprobe can be inserted
  31on virtually any instruction in the kernel.  A jprobe is inserted at
  32the entry to a kernel function, and provides convenient access to the
  33function's arguments.  A return probe fires when a specified function
  34returns.
  35
  36In the typical case, Kprobes-based instrumentation is packaged as
  37a kernel module.  The module's init function installs ("registers")
  38one or more probes, and the exit function unregisters them.  A
  39registration function such as register_kprobe() specifies where
  40the probe is to be inserted and what handler is to be called when
  41the probe is hit.
  42
  43There are also register_/unregister_*probes() functions for batch
  44registration/unregistration of a group of *probes. These functions
  45can speed up unregistration process when you have to unregister
  46a lot of probes at once.
  47
  48The next four subsections explain how the different types of
  49probes work and how jump optimization works.  They explain certain
  50things that you'll need to know in order to make the best use of
  51Kprobes -- e.g., the difference between a pre_handler and
  52a post_handler, and how to use the maxactive and nmissed fields of
  53a kretprobe.  But if you're in a hurry to start using Kprobes, you
  54can skip ahead to section 2.
  55
  561.1 How Does a Kprobe Work?
  57
  58When a kprobe is registered, Kprobes makes a copy of the probed
  59instruction and replaces the first byte(s) of the probed instruction
  60with a breakpoint instruction (e.g., int3 on i386 and x86_64).
  61
  62When a CPU hits the breakpoint instruction, a trap occurs, the CPU's
  63registers are saved, and control passes to Kprobes via the
  64notifier_call_chain mechanism.  Kprobes executes the "pre_handler"
  65associated with the kprobe, passing the handler the addresses of the
  66kprobe struct and the saved registers.
  67
  68Next, Kprobes single-steps its copy of the probed instruction.
  69(It would be simpler to single-step the actual instruction in place,
  70but then Kprobes would have to temporarily remove the breakpoint
  71instruction.  This would open a small time window when another CPU
  72could sail right past the probepoint.)
  73
  74After the instruction is single-stepped, Kprobes executes the
  75"post_handler," if any, that is associated with the kprobe.
  76Execution then continues with the instruction following the probepoint.
  77
  781.2 How Does a Jprobe Work?
  79
  80A jprobe is implemented using a kprobe that is placed on a function's
  81entry point.  It employs a simple mirroring principle to allow
  82seamless access to the probed function's arguments.  The jprobe
  83handler routine should have the same signature (arg list and return
  84type) as the function being probed, and must always end by calling
  85the Kprobes function jprobe_return().
  86
  87Here's how it works.  When the probe is hit, Kprobes makes a copy of
  88the saved registers and a generous portion of the stack (see below).
  89Kprobes then points the saved instruction pointer at the jprobe's
  90handler routine, and returns from the trap.  As a result, control
  91passes to the handler, which is presented with the same register and
  92stack contents as the probed function.  When it is done, the handler
  93calls jprobe_return(), which traps again to restore the original stack
  94contents and processor state and switch to the probed function.
  95
  96By convention, the callee owns its arguments, so gcc may produce code
  97that unexpectedly modifies that portion of the stack.  This is why
  98Kprobes saves a copy of the stack and restores it after the jprobe
  99handler has run.  Up to MAX_STACK_SIZE bytes are copied -- e.g.,
 10064 bytes on i386.
 101
 102Note that the probed function's args may be passed on the stack
 103or in registers.  The jprobe will work in either case, so long as the
 104handler's prototype matches that of the probed function.
 105
 1061.3 Return Probes
 107
 1081.3.1 How Does a Return Probe Work?
 109
 110When you call register_kretprobe(), Kprobes establishes a kprobe at
 111the entry to the function.  When the probed function is called and this
 112probe is hit, Kprobes saves a copy of the return address, and replaces
 113the return address with the address of a "trampoline."  The trampoline
 114is an arbitrary piece of code -- typically just a nop instruction.
 115At boot time, Kprobes registers a kprobe at the trampoline.
 116
 117When the probed function executes its return instruction, control
 118passes to the trampoline and that probe is hit.  Kprobes' trampoline
 119handler calls the user-specified return handler associated with the
 120kretprobe, then sets the saved instruction pointer to the saved return
 121address, and that's where execution resumes upon return from the trap.
 122
 123While the probed function is executing, its return address is
 124stored in an object of type kretprobe_instance.  Before calling
 125register_kretprobe(), the user sets the maxactive field of the
 126kretprobe struct to specify how many instances of the specified
 127function can be probed simultaneously.  register_kretprobe()
 128pre-allocates the indicated number of kretprobe_instance objects.
 129
 130For example, if the function is non-recursive and is called with a
 131spinlock held, maxactive = 1 should be enough.  If the function is
 132non-recursive and can never relinquish the CPU (e.g., via a semaphore
 133or preemption), NR_CPUS should be enough.  If maxactive <= 0, it is
 134set to a default value.  If CONFIG_PREEMPT is enabled, the default
 135is max(10, 2*NR_CPUS).  Otherwise, the default is NR_CPUS.
 136
 137It's not a disaster if you set maxactive too low; you'll just miss
 138some probes.  In the kretprobe struct, the nmissed field is set to
 139zero when the return probe is registered, and is incremented every
 140time the probed function is entered but there is no kretprobe_instance
 141object available for establishing the return probe.
 142
 1431.3.2 Kretprobe entry-handler
 144
 145Kretprobes also provides an optional user-specified handler which runs
 146on function entry. This handler is specified by setting the entry_handler
 147field of the kretprobe struct. Whenever the kprobe placed by kretprobe at the
 148function entry is hit, the user-defined entry_handler, if any, is invoked.
 149If the entry_handler returns 0 (success) then a corresponding return handler
 150is guaranteed to be called upon function return. If the entry_handler
 151returns a non-zero error then Kprobes leaves the return address as is, and
 152the kretprobe has no further effect for that particular function instance.
 153
 154Multiple entry and return handler invocations are matched using the unique
 155kretprobe_instance object associated with them. Additionally, a user
 156may also specify per return-instance private data to be part of each
 157kretprobe_instance object. This is especially useful when sharing private
 158data between corresponding user entry and return handlers. The size of each
 159private data object can be specified at kretprobe registration time by
 160setting the data_size field of the kretprobe struct. This data can be
 161accessed through the data field of each kretprobe_instance object.
 162
 163In case probed function is entered but there is no kretprobe_instance
 164object available, then in addition to incrementing the nmissed count,
 165the user entry_handler invocation is also skipped.
 166
 1671.4 How Does Jump Optimization Work?
 168
 169If your kernel is built with CONFIG_OPTPROBES=y (currently this flag
 170is automatically set 'y' on x86/x86-64, non-preemptive kernel) and
 171the "debug.kprobes_optimization" kernel parameter is set to 1 (see
 172sysctl(8)), Kprobes tries to reduce probe-hit overhead by using a jump
 173instruction instead of a breakpoint instruction at each probepoint.
 174
 1751.4.1 Init a Kprobe
 176
 177When a probe is registered, before attempting this optimization,
 178Kprobes inserts an ordinary, breakpoint-based kprobe at the specified
 179address. So, even if it's not possible to optimize this particular
 180probepoint, there'll be a probe there.
 181
 1821.4.2 Safety Check
 183
 184Before optimizing a probe, Kprobes performs the following safety checks:
 185
 186- Kprobes verifies that the region that will be replaced by the jump
 187instruction (the "optimized region") lies entirely within one function.
 188(A jump instruction is multiple bytes, and so may overlay multiple
 189instructions.)
 190
 191- Kprobes analyzes the entire function and verifies that there is no
 192jump into the optimized region.  Specifically:
 193  - the function contains no indirect jump;
 194  - the function contains no instruction that causes an exception (since
 195  the fixup code triggered by the exception could jump back into the
 196  optimized region -- Kprobes checks the exception tables to verify this);
 197  and
 198  - there is no near jump to the optimized region (other than to the first
 199  byte).
 200
 201- For each instruction in the optimized region, Kprobes verifies that
 202the instruction can be executed out of line.
 203
 2041.4.3 Preparing Detour Buffer
 205
 206Next, Kprobes prepares a "detour" buffer, which contains the following
 207instruction sequence:
 208- code to push the CPU's registers (emulating a breakpoint trap)
 209- a call to the trampoline code which calls user's probe handlers.
 210- code to restore registers
 211- the instructions from the optimized region
 212- a jump back to the original execution path.
 213
 2141.4.4 Pre-optimization
 215
 216After preparing the detour buffer, Kprobes verifies that none of the
 217following situations exist:
 218- The probe has either a break_handler (i.e., it's a jprobe) or a
 219post_handler.
 220- Other instructions in the optimized region are probed.
 221- The probe is disabled.
 222In any of the above cases, Kprobes won't start optimizing the probe.
 223Since these are temporary situations, Kprobes tries to start
 224optimizing it again if the situation is changed.
 225
 226If the kprobe can be optimized, Kprobes enqueues the kprobe to an
 227optimizing list, and kicks the kprobe-optimizer workqueue to optimize
 228it.  If the to-be-optimized probepoint is hit before being optimized,
 229Kprobes returns control to the original instruction path by setting
 230the CPU's instruction pointer to the copied code in the detour buffer
 231-- thus at least avoiding the single-step.
 232
 2331.4.5 Optimization
 234
 235The Kprobe-optimizer doesn't insert the jump instruction immediately;
 236rather, it calls synchronize_sched() for safety first, because it's
 237possible for a CPU to be interrupted in the middle of executing the
 238optimized region(*).  As you know, synchronize_sched() can ensure
 239that all interruptions that were active when synchronize_sched()
 240was called are done, but only if CONFIG_PREEMPT=n.  So, this version
 241of kprobe optimization supports only kernels with CONFIG_PREEMPT=n.(**)
 242
 243After that, the Kprobe-optimizer calls stop_machine() to replace
 244the optimized region with a jump instruction to the detour buffer,
 245using text_poke_smp().
 246
 2471.4.6 Unoptimization
 248
 249When an optimized kprobe is unregistered, disabled, or blocked by
 250another kprobe, it will be unoptimized.  If this happens before
 251the optimization is complete, the kprobe is just dequeued from the
 252optimized list.  If the optimization has been done, the jump is
 253replaced with the original code (except for an int3 breakpoint in
 254the first byte) by using text_poke_smp().
 255
 256(*)Please imagine that the 2nd instruction is interrupted and then
 257the optimizer replaces the 2nd instruction with the jump *address*
 258while the interrupt handler is running. When the interrupt
 259returns to original address, there is no valid instruction,
 260and it causes an unexpected result.
 261
 262(**)This optimization-safety checking may be replaced with the
 263stop-machine method that ksplice uses for supporting a CONFIG_PREEMPT=y
 264kernel.
 265
 266NOTE for geeks:
 267The jump optimization changes the kprobe's pre_handler behavior.
 268Without optimization, the pre_handler can change the kernel's execution
 269path by changing regs->ip and returning 1.  However, when the probe
 270is optimized, that modification is ignored.  Thus, if you want to
 271tweak the kernel's execution path, you need to suppress optimization,
 272using one of the following techniques:
 273- Specify an empty function for the kprobe's post_handler or break_handler.
 274 or
 275- Execute 'sysctl -w debug.kprobes_optimization=n'
 276
 2771.5 Blacklist
 278
 279Kprobes can probe most of the kernel except itself. This means
 280that there are some functions where kprobes cannot probe. Probing
 281(trapping) such functions can cause a recursive trap (e.g. double
 282fault) or the nested probe handler may never be called.
 283Kprobes manages such functions as a blacklist.
 284If you want to add a function into the blacklist, you just need
 285to (1) include linux/kprobes.h and (2) use NOKPROBE_SYMBOL() macro
 286to specify a blacklisted function.
 287Kprobes checks the given probe address against the blacklist and
 288rejects registering it, if the given address is in the blacklist.
 289
 2902. Architectures Supported
 291
 292Kprobes, jprobes, and return probes are implemented on the following
 293architectures:
 294
 295- i386 (Supports jump optimization)
 296- x86_64 (AMD-64, EM64T) (Supports jump optimization)
 297- ppc64
 298- ia64 (Does not support probes on instruction slot1.)
 299- sparc64 (Return probes not yet implemented.)
 300- arm
 301- ppc
 302- mips
 303- s390
 304
 3053. Configuring Kprobes
 306
 307When configuring the kernel using make menuconfig/xconfig/oldconfig,
 308ensure that CONFIG_KPROBES is set to "y".  Under "Instrumentation
 309Support", look for "Kprobes".
 310
 311So that you can load and unload Kprobes-based instrumentation modules,
 312make sure "Loadable module support" (CONFIG_MODULES) and "Module
 313unloading" (CONFIG_MODULE_UNLOAD) are set to "y".
 314
 315Also make sure that CONFIG_KALLSYMS and perhaps even CONFIG_KALLSYMS_ALL
 316are set to "y", since kallsyms_lookup_name() is used by the in-kernel
 317kprobe address resolution code.
 318
 319If you need to insert a probe in the middle of a function, you may find
 320it useful to "Compile the kernel with debug info" (CONFIG_DEBUG_INFO),
 321so you can use "objdump -d -l vmlinux" to see the source-to-object
 322code mapping.
 323
 3244. API Reference
 325
 326The Kprobes API includes a "register" function and an "unregister"
 327function for each type of probe. The API also includes "register_*probes"
 328and "unregister_*probes" functions for (un)registering arrays of probes.
 329Here are terse, mini-man-page specifications for these functions and
 330the associated probe handlers that you'll write. See the files in the
 331samples/kprobes/ sub-directory for examples.
 332
 3334.1 register_kprobe
 334
 335#include <linux/kprobes.h>
 336int register_kprobe(struct kprobe *kp);
 337
 338Sets a breakpoint at the address kp->addr.  When the breakpoint is
 339hit, Kprobes calls kp->pre_handler.  After the probed instruction
 340is single-stepped, Kprobe calls kp->post_handler.  If a fault
 341occurs during execution of kp->pre_handler or kp->post_handler,
 342or during single-stepping of the probed instruction, Kprobes calls
 343kp->fault_handler.  Any or all handlers can be NULL. If kp->flags
 344is set KPROBE_FLAG_DISABLED, that kp will be registered but disabled,
 345so, its handlers aren't hit until calling enable_kprobe(kp).
 346
 347NOTE:
 3481. With the introduction of the "symbol_name" field to struct kprobe,
 349the probepoint address resolution will now be taken care of by the kernel.
 350The following will now work:
 351
 352        kp.symbol_name = "symbol_name";
 353
 354(64-bit powerpc intricacies such as function descriptors are handled
 355transparently)
 356
 3572. Use the "offset" field of struct kprobe if the offset into the symbol
 358to install a probepoint is known. This field is used to calculate the
 359probepoint.
 360
 3613. Specify either the kprobe "symbol_name" OR the "addr". If both are
 362specified, kprobe registration will fail with -EINVAL.
 363
 3644. With CISC architectures (such as i386 and x86_64), the kprobes code
 365does not validate if the kprobe.addr is at an instruction boundary.
 366Use "offset" with caution.
 367
 368register_kprobe() returns 0 on success, or a negative errno otherwise.
 369
 370User's pre-handler (kp->pre_handler):
 371#include <linux/kprobes.h>
 372#include <linux/ptrace.h>
 373int pre_handler(struct kprobe *p, struct pt_regs *regs);
 374
 375Called with p pointing to the kprobe associated with the breakpoint,
 376and regs pointing to the struct containing the registers saved when
 377the breakpoint was hit.  Return 0 here unless you're a Kprobes geek.
 378
 379User's post-handler (kp->post_handler):
 380#include <linux/kprobes.h>
 381#include <linux/ptrace.h>
 382void post_handler(struct kprobe *p, struct pt_regs *regs,
 383        unsigned long flags);
 384
 385p and regs are as described for the pre_handler.  flags always seems
 386to be zero.
 387
 388User's fault-handler (kp->fault_handler):
 389#include <linux/kprobes.h>
 390#include <linux/ptrace.h>
 391int fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr);
 392
 393p and regs are as described for the pre_handler.  trapnr is the
 394architecture-specific trap number associated with the fault (e.g.,
 395on i386, 13 for a general protection fault or 14 for a page fault).
 396Returns 1 if it successfully handled the exception.
 397
 3984.2 register_jprobe
 399
 400#include <linux/kprobes.h>
 401int register_jprobe(struct jprobe *jp)
 402
 403Sets a breakpoint at the address jp->kp.addr, which must be the address
 404of the first instruction of a function.  When the breakpoint is hit,
 405Kprobes runs the handler whose address is jp->entry.
 406
 407The handler should have the same arg list and return type as the probed
 408function; and just before it returns, it must call jprobe_return().
 409(The handler never actually returns, since jprobe_return() returns
 410control to Kprobes.)  If the probed function is declared asmlinkage
 411or anything else that affects how args are passed, the handler's
 412declaration must match.
 413
 414register_jprobe() returns 0 on success, or a negative errno otherwise.
 415
 4164.3 register_kretprobe
 417
 418#include <linux/kprobes.h>
 419int register_kretprobe(struct kretprobe *rp);
 420
 421Establishes a return probe for the function whose address is
 422rp->kp.addr.  When that function returns, Kprobes calls rp->handler.
 423You must set rp->maxactive appropriately before you call
 424register_kretprobe(); see "How Does a Return Probe Work?" for details.
 425
 426register_kretprobe() returns 0 on success, or a negative errno
 427otherwise.
 428
 429User's return-probe handler (rp->handler):
 430#include <linux/kprobes.h>
 431#include <linux/ptrace.h>
 432int kretprobe_handler(struct kretprobe_instance *ri, struct pt_regs *regs);
 433
 434regs is as described for kprobe.pre_handler.  ri points to the
 435kretprobe_instance object, of which the following fields may be
 436of interest:
 437- ret_addr: the return address
 438- rp: points to the corresponding kretprobe object
 439- task: points to the corresponding task struct
 440- data: points to per return-instance private data; see "Kretprobe
 441        entry-handler" for details.
 442
 443The regs_return_value(regs) macro provides a simple abstraction to
 444extract the return value from the appropriate register as defined by
 445the architecture's ABI.
 446
 447The handler's return value is currently ignored.
 448
 4494.4 unregister_*probe
 450
 451#include <linux/kprobes.h>
 452void unregister_kprobe(struct kprobe *kp);
 453void unregister_jprobe(struct jprobe *jp);
 454void unregister_kretprobe(struct kretprobe *rp);
 455
 456Removes the specified probe.  The unregister function can be called
 457at any time after the probe has been registered.
 458
 459NOTE:
 460If the functions find an incorrect probe (ex. an unregistered probe),
 461they clear the addr field of the probe.
 462
 4634.5 register_*probes
 464
 465#include <linux/kprobes.h>
 466int register_kprobes(struct kprobe **kps, int num);
 467int register_kretprobes(struct kretprobe **rps, int num);
 468int register_jprobes(struct jprobe **jps, int num);
 469
 470Registers each of the num probes in the specified array.  If any
 471error occurs during registration, all probes in the array, up to
 472the bad probe, are safely unregistered before the register_*probes
 473function returns.
 474- kps/rps/jps: an array of pointers to *probe data structures
 475- num: the number of the array entries.
 476
 477NOTE:
 478You have to allocate(or define) an array of pointers and set all
 479of the array entries before using these functions.
 480
 4814.6 unregister_*probes
 482
 483#include <linux/kprobes.h>
 484void unregister_kprobes(struct kprobe **kps, int num);
 485void unregister_kretprobes(struct kretprobe **rps, int num);
 486void unregister_jprobes(struct jprobe **jps, int num);
 487
 488Removes each of the num probes in the specified array at once.
 489
 490NOTE:
 491If the functions find some incorrect probes (ex. unregistered
 492probes) in the specified array, they clear the addr field of those
 493incorrect probes. However, other probes in the array are
 494unregistered correctly.
 495
 4964.7 disable_*probe
 497
 498#include <linux/kprobes.h>
 499int disable_kprobe(struct kprobe *kp);
 500int disable_kretprobe(struct kretprobe *rp);
 501int disable_jprobe(struct jprobe *jp);
 502
 503Temporarily disables the specified *probe. You can enable it again by using
 504enable_*probe(). You must specify the probe which has been registered.
 505
 5064.8 enable_*probe
 507
 508#include <linux/kprobes.h>
 509int enable_kprobe(struct kprobe *kp);
 510int enable_kretprobe(struct kretprobe *rp);
 511int enable_jprobe(struct jprobe *jp);
 512
 513Enables *probe which has been disabled by disable_*probe(). You must specify
 514the probe which has been registered.
 515
 5165. Kprobes Features and Limitations
 517
 518Kprobes allows multiple probes at the same address.  Currently,
 519however, there cannot be multiple jprobes on the same function at
 520the same time.  Also, a probepoint for which there is a jprobe or
 521a post_handler cannot be optimized.  So if you install a jprobe,
 522or a kprobe with a post_handler, at an optimized probepoint, the
 523probepoint will be unoptimized automatically.
 524
 525In general, you can install a probe anywhere in the kernel.
 526In particular, you can probe interrupt handlers.  Known exceptions
 527are discussed in this section.
 528
 529The register_*probe functions will return -EINVAL if you attempt
 530to install a probe in the code that implements Kprobes (mostly
 531kernel/kprobes.c and arch/*/kernel/kprobes.c, but also functions such
 532as do_page_fault and notifier_call_chain).
 533
 534If you install a probe in an inline-able function, Kprobes makes
 535no attempt to chase down all inline instances of the function and
 536install probes there.  gcc may inline a function without being asked,
 537so keep this in mind if you're not seeing the probe hits you expect.
 538
 539A probe handler can modify the environment of the probed function
 540-- e.g., by modifying kernel data structures, or by modifying the
 541contents of the pt_regs struct (which are restored to the registers
 542upon return from the breakpoint).  So Kprobes can be used, for example,
 543to install a bug fix or to inject faults for testing.  Kprobes, of
 544course, has no way to distinguish the deliberately injected faults
 545from the accidental ones.  Don't drink and probe.
 546
 547Kprobes makes no attempt to prevent probe handlers from stepping on
 548each other -- e.g., probing printk() and then calling printk() from a
 549probe handler.  If a probe handler hits a probe, that second probe's
 550handlers won't be run in that instance, and the kprobe.nmissed member
 551of the second probe will be incremented.
 552
 553As of Linux v2.6.15-rc1, multiple handlers (or multiple instances of
 554the same handler) may run concurrently on different CPUs.
 555
 556Kprobes does not use mutexes or allocate memory except during
 557registration and unregistration.
 558
 559Probe handlers are run with preemption disabled.  Depending on the
 560architecture and optimization state, handlers may also run with
 561interrupts disabled (e.g., kretprobe handlers and optimized kprobe
 562handlers run without interrupt disabled on x86/x86-64).  In any case,
 563your handler should not yield the CPU (e.g., by attempting to acquire
 564a semaphore).
 565
 566Since a return probe is implemented by replacing the return
 567address with the trampoline's address, stack backtraces and calls
 568to __builtin_return_address() will typically yield the trampoline's
 569address instead of the real return address for kretprobed functions.
 570(As far as we can tell, __builtin_return_address() is used only
 571for instrumentation and error reporting.)
 572
 573If the number of times a function is called does not match the number
 574of times it returns, registering a return probe on that function may
 575produce undesirable results. In such a case, a line:
 576kretprobe BUG!: Processing kretprobe d000000000041aa8 @ c00000000004f48c
 577gets printed. With this information, one will be able to correlate the
 578exact instance of the kretprobe that caused the problem. We have the
 579do_exit() case covered. do_execve() and do_fork() are not an issue.
 580We're unaware of other specific cases where this could be a problem.
 581
 582If, upon entry to or exit from a function, the CPU is running on
 583a stack other than that of the current task, registering a return
 584probe on that function may produce undesirable results.  For this
 585reason, Kprobes doesn't support return probes (or kprobes or jprobes)
 586on the x86_64 version of __switch_to(); the registration functions
 587return -EINVAL.
 588
 589On x86/x86-64, since the Jump Optimization of Kprobes modifies
 590instructions widely, there are some limitations to optimization. To
 591explain it, we introduce some terminology. Imagine a 3-instruction
 592sequence consisting of a two 2-byte instructions and one 3-byte
 593instruction.
 594
 595        IA
 596         |
 597[-2][-1][0][1][2][3][4][5][6][7]
 598        [ins1][ins2][  ins3 ]
 599        [<-     DCR       ->]
 600           [<- JTPR ->]
 601
 602ins1: 1st Instruction
 603ins2: 2nd Instruction
 604ins3: 3rd Instruction
 605IA:  Insertion Address
 606JTPR: Jump Target Prohibition Region
 607DCR: Detoured Code Region
 608
 609The instructions in DCR are copied to the out-of-line buffer
 610of the kprobe, because the bytes in DCR are replaced by
 611a 5-byte jump instruction. So there are several limitations.
 612
 613a) The instructions in DCR must be relocatable.
 614b) The instructions in DCR must not include a call instruction.
 615c) JTPR must not be targeted by any jump or call instruction.
 616d) DCR must not straddle the border between functions.
 617
 618Anyway, these limitations are checked by the in-kernel instruction
 619decoder, so you don't need to worry about that.
 620
 6216. Probe Overhead
 622
 623On a typical CPU in use in 2005, a kprobe hit takes 0.5 to 1.0
 624microseconds to process.  Specifically, a benchmark that hits the same
 625probepoint repeatedly, firing a simple handler each time, reports 1-2
 626million hits per second, depending on the architecture.  A jprobe or
 627return-probe hit typically takes 50-75% longer than a kprobe hit.
 628When you have a return probe set on a function, adding a kprobe at
 629the entry to that function adds essentially no overhead.
 630
 631Here are sample overhead figures (in usec) for different architectures.
 632k = kprobe; j = jprobe; r = return probe; kr = kprobe + return probe
 633on same function; jr = jprobe + return probe on same function
 634
 635i386: Intel Pentium M, 1495 MHz, 2957.31 bogomips
 636k = 0.57 usec; j = 1.00; r = 0.92; kr = 0.99; jr = 1.40
 637
 638x86_64: AMD Opteron 246, 1994 MHz, 3971.48 bogomips
 639k = 0.49 usec; j = 0.76; r = 0.80; kr = 0.82; jr = 1.07
 640
 641ppc64: POWER5 (gr), 1656 MHz (SMT disabled, 1 virtual CPU per physical CPU)
 642k = 0.77 usec; j = 1.31; r = 1.26; kr = 1.45; jr = 1.99
 643
 6446.1 Optimized Probe Overhead
 645
 646Typically, an optimized kprobe hit takes 0.07 to 0.1 microseconds to
 647process. Here are sample overhead figures (in usec) for x86 architectures.
 648k = unoptimized kprobe, b = boosted (single-step skipped), o = optimized kprobe,
 649r = unoptimized kretprobe, rb = boosted kretprobe, ro = optimized kretprobe.
 650
 651i386: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips
 652k = 0.80 usec; b = 0.33; o = 0.05; r = 1.10; rb = 0.61; ro = 0.33
 653
 654x86-64: Intel(R) Xeon(R) E5410, 2.33GHz, 4656.90 bogomips
 655k = 0.99 usec; b = 0.43; o = 0.06; r = 1.24; rb = 0.68; ro = 0.30
 656
 6577. TODO
 658
 659a. SystemTap (http://sourceware.org/systemtap): Provides a simplified
 660programming interface for probe-based instrumentation.  Try it out.
 661b. Kernel return probes for sparc64.
 662c. Support for other architectures.
 663d. User-space probes.
 664e. Watchpoint probes (which fire on data references).
 665
 6668. Kprobes Example
 667
 668See samples/kprobes/kprobe_example.c
 669
 6709. Jprobes Example
 671
 672See samples/kprobes/jprobe_example.c
 673
 67410. Kretprobes Example
 675
 676See samples/kprobes/kretprobe_example.c
 677
 678For additional information on Kprobes, refer to the following URLs:
 679http://www-106.ibm.com/developerworks/library/l-kprobes.html?ca=dgr-lnxw42Kprobe
 680http://www.redhat.com/magazine/005mar05/features/kprobes/
 681http://www-users.cs.umn.edu/~boutcher/kprobes/
 682http://www.linuxsymposium.org/2006/linuxsymposium_procv2.pdf (pages 101-115)
 683
 684
 685Appendix A: The kprobes debugfs interface
 686
 687With recent kernels (> 2.6.20) the list of registered kprobes is visible
 688under the /sys/kernel/debug/kprobes/ directory (assuming debugfs is mounted at //sys/kernel/debug).
 689
 690/sys/kernel/debug/kprobes/list: Lists all registered probes on the system
 691
 692c015d71a  k  vfs_read+0x0
 693c011a316  j  do_fork+0x0
 694c03dedc5  r  tcp_v4_rcv+0x0
 695
 696The first column provides the kernel address where the probe is inserted.
 697The second column identifies the type of probe (k - kprobe, r - kretprobe
 698and j - jprobe), while the third column specifies the symbol+offset of
 699the probe. If the probed function belongs to a module, the module name
 700is also specified. Following columns show probe status. If the probe is on
 701a virtual address that is no longer valid (module init sections, module
 702virtual addresses that correspond to modules that've been unloaded),
 703such probes are marked with [GONE]. If the probe is temporarily disabled,
 704such probes are marked with [DISABLED]. If the probe is optimized, it is
 705marked with [OPTIMIZED].
 706
 707/sys/kernel/debug/kprobes/enabled: Turn kprobes ON/OFF forcibly.
 708
 709Provides a knob to globally and forcibly turn registered kprobes ON or OFF.
 710By default, all kprobes are enabled. By echoing "0" to this file, all
 711registered probes will be disarmed, till such time a "1" is echoed to this
 712file. Note that this knob just disarms and arms all kprobes and doesn't
 713change each probe's disabling state. This means that disabled kprobes (marked
 714[DISABLED]) will be not enabled if you turn ON all kprobes by this knob.
 715
 716
 717Appendix B: The kprobes sysctl interface
 718
 719/proc/sys/debug/kprobes-optimization: Turn kprobes optimization ON/OFF.
 720
 721When CONFIG_OPTPROBES=y, this sysctl interface appears and it provides
 722a knob to globally and forcibly turn jump optimization (see section
 7231.4) ON or OFF. By default, jump optimization is allowed (ON).
 724If you echo "0" to this file or set "debug.kprobes_optimization" to
 7250 via sysctl, all optimized probes will be unoptimized, and any new
 726probes registered after that will not be optimized.  Note that this
 727knob *changes* the optimized state. This means that optimized probes
 728(marked [OPTIMIZED]) will be unoptimized ([OPTIMIZED] tag will be
 729removed). If the knob is turned on, they will be optimized again.
 730
 731
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.