diff options
author | Hien Nguyen <hien@us.ibm.com> | 2005-06-23 03:09:19 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-06-23 12:45:21 -0400 |
commit | b94cce926b2b902b79380ccba370d6f9f2980de0 (patch) | |
tree | da2680b1ec36eae6423ba446d09284d2642ae82b /kernel/kprobes.c | |
parent | 2fa389c5eb8c97d621653184d2adf5fdbd4a3167 (diff) |
[PATCH] kprobes: function-return probes
This patch adds function-return probes to kprobes for the i386
architecture. This enables you to establish a handler to be run when a
function returns.
1. API
Two new functions are added to kprobes:
int register_kretprobe(struct kretprobe *rp);
void unregister_kretprobe(struct kretprobe *rp);
2. Registration and unregistration
2.1 Register
To register a function-return probe, the user populates the following
fields in a kretprobe object and calls register_kretprobe() with the
kretprobe address as an argument:
kp.addr - the function's address
handler - this function is run after the ret instruction executes, but
before control returns to the return address in the caller.
maxactive - The maximum number of instances of the probed function that
can be active concurrently. For example, if the function is non-
recursive and is called with a spinlock or mutex held, maxactive = 1
should be enough. If the function is non-recursive and can never
relinquish the CPU (e.g., via a semaphore or preemption), NR_CPUS should
be enough. maxactive is used to determine how many kretprobe_instance
objects to allocate for this particular probed function. If maxactive <=
0, it is set to a default value (if CONFIG_PREEMPT maxactive=max(10, 2 *
NR_CPUS) else maxactive=NR_CPUS)
For example:
struct kretprobe rp;
rp.kp.addr = /* entrypoint address */
rp.handler = /*return probe handler */
rp.maxactive = /* e.g., 1 or NR_CPUS or 0, see the above explanation */
register_kretprobe(&rp);
The following field may also be of interest:
nmissed - Initialized to zero when the function-return probe is
registered, and incremented every time the probed function is entered but
there is no kretprobe_instance object available for establishing the
function-return probe (i.e., because maxactive was set too low).
2.2 Unregister
To unregiter a function-return probe, the user calls
unregister_kretprobe() with the same kretprobe object as registered
previously. If a probed function is running when the return probe is
unregistered, the function will return as expected, but the handler won't
be run.
3. Limitations
3.1 This patch supports only the i386 architecture, but patches for
x86_64 and ppc64 are anticipated soon.
3.2 Return probes operates by replacing the return address in the stack
(or in a known register, such as the lr register for ppc). This may
cause __builtin_return_address(0), when invoked from the return-probed
function, to return the address of the return-probes trampoline.
3.3 This implementation uses the "Multiprobes at an address" feature in
2.6.12-rc3-mm3.
3.4 Due to a limitation in multi-probes, you cannot currently establish
a return probe and a jprobe on the same function. A patch to remove
this limitation is being tested.
This feature is required by SystemTap (http://sourceware.org/systemtap),
and reflects ideas contributed by several SystemTap developers, including
Will Cohen and Ananth Mavinakayanahalli.
Signed-off-by: Hien Nguyen <hien@us.ibm.com>
Signed-off-by: Prasanna S Panchamukhi <prasanna@in.ibm.com>
Signed-off-by: Frederik Deweerdt <frederik.deweerdt@laposte.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'kernel/kprobes.c')
-rw-r--r-- | kernel/kprobes.c | 213 |
1 files changed, 208 insertions, 5 deletions
diff --git a/kernel/kprobes.c b/kernel/kprobes.c index 037142b72a4..692fbf75ab4 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c | |||
@@ -27,6 +27,9 @@ | |||
27 | * interface to access function arguments. | 27 | * interface to access function arguments. |
28 | * 2004-Sep Prasanna S Panchamukhi <prasanna@in.ibm.com> Changed Kprobes | 28 | * 2004-Sep Prasanna S Panchamukhi <prasanna@in.ibm.com> Changed Kprobes |
29 | * exceptions notifier to be first on the priority list. | 29 | * exceptions notifier to be first on the priority list. |
30 | * 2005-May Hien Nguyen <hien@us.ibm.com>, Jim Keniston | ||
31 | * <jkenisto@us.ibm.com> and Prasanna S Panchamukhi | ||
32 | * <prasanna@in.ibm.com> added function-return probes. | ||
30 | */ | 33 | */ |
31 | #include <linux/kprobes.h> | 34 | #include <linux/kprobes.h> |
32 | #include <linux/spinlock.h> | 35 | #include <linux/spinlock.h> |
@@ -41,6 +44,7 @@ | |||
41 | #define KPROBE_TABLE_SIZE (1 << KPROBE_HASH_BITS) | 44 | #define KPROBE_TABLE_SIZE (1 << KPROBE_HASH_BITS) |
42 | 45 | ||
43 | static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE]; | 46 | static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE]; |
47 | static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE]; | ||
44 | 48 | ||
45 | unsigned int kprobe_cpu = NR_CPUS; | 49 | unsigned int kprobe_cpu = NR_CPUS; |
46 | static DEFINE_SPINLOCK(kprobe_lock); | 50 | static DEFINE_SPINLOCK(kprobe_lock); |
@@ -78,7 +82,7 @@ struct kprobe *get_kprobe(void *addr) | |||
78 | * Aggregate handlers for multiple kprobes support - these handlers | 82 | * Aggregate handlers for multiple kprobes support - these handlers |
79 | * take care of invoking the individual kprobe handlers on p->list | 83 | * take care of invoking the individual kprobe handlers on p->list |
80 | */ | 84 | */ |
81 | int aggr_pre_handler(struct kprobe *p, struct pt_regs *regs) | 85 | static int aggr_pre_handler(struct kprobe *p, struct pt_regs *regs) |
82 | { | 86 | { |
83 | struct kprobe *kp; | 87 | struct kprobe *kp; |
84 | 88 | ||
@@ -92,8 +96,8 @@ int aggr_pre_handler(struct kprobe *p, struct pt_regs *regs) | |||
92 | return 0; | 96 | return 0; |
93 | } | 97 | } |
94 | 98 | ||
95 | void aggr_post_handler(struct kprobe *p, struct pt_regs *regs, | 99 | static void aggr_post_handler(struct kprobe *p, struct pt_regs *regs, |
96 | unsigned long flags) | 100 | unsigned long flags) |
97 | { | 101 | { |
98 | struct kprobe *kp; | 102 | struct kprobe *kp; |
99 | 103 | ||
@@ -107,7 +111,8 @@ void aggr_post_handler(struct kprobe *p, struct pt_regs *regs, | |||
107 | return; | 111 | return; |
108 | } | 112 | } |
109 | 113 | ||
110 | int aggr_fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr) | 114 | static int aggr_fault_handler(struct kprobe *p, struct pt_regs *regs, |
115 | int trapnr) | ||
111 | { | 116 | { |
112 | /* | 117 | /* |
113 | * if we faulted "during" the execution of a user specified | 118 | * if we faulted "during" the execution of a user specified |
@@ -120,6 +125,135 @@ int aggr_fault_handler(struct kprobe *p, struct pt_regs *regs, int trapnr) | |||
120 | return 0; | 125 | return 0; |
121 | } | 126 | } |
122 | 127 | ||
128 | struct kprobe trampoline_p = { | ||
129 | .addr = (kprobe_opcode_t *) &kretprobe_trampoline, | ||
130 | .pre_handler = trampoline_probe_handler, | ||
131 | .post_handler = trampoline_post_handler | ||
132 | }; | ||
133 | |||
134 | struct kretprobe_instance *get_free_rp_inst(struct kretprobe *rp) | ||
135 | { | ||
136 | struct hlist_node *node; | ||
137 | struct kretprobe_instance *ri; | ||
138 | hlist_for_each_entry(ri, node, &rp->free_instances, uflist) | ||
139 | return ri; | ||
140 | return NULL; | ||
141 | } | ||
142 | |||
143 | static struct kretprobe_instance *get_used_rp_inst(struct kretprobe *rp) | ||
144 | { | ||
145 | struct hlist_node *node; | ||
146 | struct kretprobe_instance *ri; | ||
147 | hlist_for_each_entry(ri, node, &rp->used_instances, uflist) | ||
148 | return ri; | ||
149 | return NULL; | ||
150 | } | ||
151 | |||
152 | struct kretprobe_instance *get_rp_inst(void *sara) | ||
153 | { | ||
154 | struct hlist_head *head; | ||
155 | struct hlist_node *node; | ||
156 | struct task_struct *tsk; | ||
157 | struct kretprobe_instance *ri; | ||
158 | |||
159 | tsk = arch_get_kprobe_task(sara); | ||
160 | head = &kretprobe_inst_table[hash_ptr(tsk, KPROBE_HASH_BITS)]; | ||
161 | hlist_for_each_entry(ri, node, head, hlist) { | ||
162 | if (ri->stack_addr == sara) | ||
163 | return ri; | ||
164 | } | ||
165 | return NULL; | ||
166 | } | ||
167 | |||
168 | void add_rp_inst(struct kretprobe_instance *ri) | ||
169 | { | ||
170 | struct task_struct *tsk; | ||
171 | /* | ||
172 | * Remove rp inst off the free list - | ||
173 | * Add it back when probed function returns | ||
174 | */ | ||
175 | hlist_del(&ri->uflist); | ||
176 | tsk = arch_get_kprobe_task(ri->stack_addr); | ||
177 | /* Add rp inst onto table */ | ||
178 | INIT_HLIST_NODE(&ri->hlist); | ||
179 | hlist_add_head(&ri->hlist, | ||
180 | &kretprobe_inst_table[hash_ptr(tsk, KPROBE_HASH_BITS)]); | ||
181 | |||
182 | /* Also add this rp inst to the used list. */ | ||
183 | INIT_HLIST_NODE(&ri->uflist); | ||
184 | hlist_add_head(&ri->uflist, &ri->rp->used_instances); | ||
185 | } | ||
186 | |||
187 | void recycle_rp_inst(struct kretprobe_instance *ri) | ||
188 | { | ||
189 | /* remove rp inst off the rprobe_inst_table */ | ||
190 | hlist_del(&ri->hlist); | ||
191 | if (ri->rp) { | ||
192 | /* remove rp inst off the used list */ | ||
193 | hlist_del(&ri->uflist); | ||
194 | /* put rp inst back onto the free list */ | ||
195 | INIT_HLIST_NODE(&ri->uflist); | ||
196 | hlist_add_head(&ri->uflist, &ri->rp->free_instances); | ||
197 | } else | ||
198 | /* Unregistering */ | ||
199 | kfree(ri); | ||
200 | } | ||
201 | |||
202 | struct hlist_head * kretprobe_inst_table_head(struct task_struct *tsk) | ||
203 | { | ||
204 | return &kretprobe_inst_table[hash_ptr(tsk, KPROBE_HASH_BITS)]; | ||
205 | } | ||
206 | |||
207 | struct kretprobe_instance *get_rp_inst_tsk(struct task_struct *tk) | ||
208 | { | ||
209 | struct task_struct *tsk; | ||
210 | struct hlist_head *head; | ||
211 | struct hlist_node *node; | ||
212 | struct kretprobe_instance *ri; | ||
213 | |||
214 | head = &kretprobe_inst_table[hash_ptr(tk, KPROBE_HASH_BITS)]; | ||
215 | |||
216 | hlist_for_each_entry(ri, node, head, hlist) { | ||
217 | tsk = arch_get_kprobe_task(ri->stack_addr); | ||
218 | if (tsk == tk) | ||
219 | return ri; | ||
220 | } | ||
221 | return NULL; | ||
222 | } | ||
223 | |||
224 | /* | ||
225 | * This function is called from do_exit or do_execv when task tk's stack is | ||
226 | * about to be recycled. Recycle any function-return probe instances | ||
227 | * associated with this task. These represent probed functions that have | ||
228 | * been called but may never return. | ||
229 | */ | ||
230 | void kprobe_flush_task(struct task_struct *tk) | ||
231 | { | ||
232 | arch_kprobe_flush_task(tk, &kprobe_lock); | ||
233 | } | ||
234 | |||
235 | /* | ||
236 | * This kprobe pre_handler is registered with every kretprobe. When probe | ||
237 | * hits it will set up the return probe. | ||
238 | */ | ||
239 | static int pre_handler_kretprobe(struct kprobe *p, struct pt_regs *regs) | ||
240 | { | ||
241 | struct kretprobe *rp = container_of(p, struct kretprobe, kp); | ||
242 | |||
243 | /*TODO: consider to only swap the RA after the last pre_handler fired */ | ||
244 | arch_prepare_kretprobe(rp, regs); | ||
245 | return 0; | ||
246 | } | ||
247 | |||
248 | static inline void free_rp_inst(struct kretprobe *rp) | ||
249 | { | ||
250 | struct kretprobe_instance *ri; | ||
251 | while ((ri = get_free_rp_inst(rp)) != NULL) { | ||
252 | hlist_del(&ri->uflist); | ||
253 | kfree(ri); | ||
254 | } | ||
255 | } | ||
256 | |||
123 | /* | 257 | /* |
124 | * Fill in the required fields of the "manager kprobe". Replace the | 258 | * Fill in the required fields of the "manager kprobe". Replace the |
125 | * earlier kprobe in the hlist with the manager kprobe | 259 | * earlier kprobe in the hlist with the manager kprobe |
@@ -257,16 +391,82 @@ void unregister_jprobe(struct jprobe *jp) | |||
257 | unregister_kprobe(&jp->kp); | 391 | unregister_kprobe(&jp->kp); |
258 | } | 392 | } |
259 | 393 | ||
394 | #ifdef ARCH_SUPPORTS_KRETPROBES | ||
395 | |||
396 | int register_kretprobe(struct kretprobe *rp) | ||
397 | { | ||
398 | int ret = 0; | ||
399 | struct kretprobe_instance *inst; | ||
400 | int i; | ||
401 | |||
402 | rp->kp.pre_handler = pre_handler_kretprobe; | ||
403 | |||
404 | /* Pre-allocate memory for max kretprobe instances */ | ||
405 | if (rp->maxactive <= 0) { | ||
406 | #ifdef CONFIG_PREEMPT | ||
407 | rp->maxactive = max(10, 2 * NR_CPUS); | ||
408 | #else | ||
409 | rp->maxactive = NR_CPUS; | ||
410 | #endif | ||
411 | } | ||
412 | INIT_HLIST_HEAD(&rp->used_instances); | ||
413 | INIT_HLIST_HEAD(&rp->free_instances); | ||
414 | for (i = 0; i < rp->maxactive; i++) { | ||
415 | inst = kmalloc(sizeof(struct kretprobe_instance), GFP_KERNEL); | ||
416 | if (inst == NULL) { | ||
417 | free_rp_inst(rp); | ||
418 | return -ENOMEM; | ||
419 | } | ||
420 | INIT_HLIST_NODE(&inst->uflist); | ||
421 | hlist_add_head(&inst->uflist, &rp->free_instances); | ||
422 | } | ||
423 | |||
424 | rp->nmissed = 0; | ||
425 | /* Establish function entry probe point */ | ||
426 | if ((ret = register_kprobe(&rp->kp)) != 0) | ||
427 | free_rp_inst(rp); | ||
428 | return ret; | ||
429 | } | ||
430 | |||
431 | #else /* ARCH_SUPPORTS_KRETPROBES */ | ||
432 | |||
433 | int register_kretprobe(struct kretprobe *rp) | ||
434 | { | ||
435 | return -ENOSYS; | ||
436 | } | ||
437 | |||
438 | #endif /* ARCH_SUPPORTS_KRETPROBES */ | ||
439 | |||
440 | void unregister_kretprobe(struct kretprobe *rp) | ||
441 | { | ||
442 | unsigned long flags; | ||
443 | struct kretprobe_instance *ri; | ||
444 | |||
445 | unregister_kprobe(&rp->kp); | ||
446 | /* No race here */ | ||
447 | spin_lock_irqsave(&kprobe_lock, flags); | ||
448 | free_rp_inst(rp); | ||
449 | while ((ri = get_used_rp_inst(rp)) != NULL) { | ||
450 | ri->rp = NULL; | ||
451 | hlist_del(&ri->uflist); | ||
452 | } | ||
453 | spin_unlock_irqrestore(&kprobe_lock, flags); | ||
454 | } | ||
455 | |||
260 | static int __init init_kprobes(void) | 456 | static int __init init_kprobes(void) |
261 | { | 457 | { |
262 | int i, err = 0; | 458 | int i, err = 0; |
263 | 459 | ||
264 | /* FIXME allocate the probe table, currently defined statically */ | 460 | /* FIXME allocate the probe table, currently defined statically */ |
265 | /* initialize all list heads */ | 461 | /* initialize all list heads */ |
266 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) | 462 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) { |
267 | INIT_HLIST_HEAD(&kprobe_table[i]); | 463 | INIT_HLIST_HEAD(&kprobe_table[i]); |
464 | INIT_HLIST_HEAD(&kretprobe_inst_table[i]); | ||
465 | } | ||
268 | 466 | ||
269 | err = register_die_notifier(&kprobe_exceptions_nb); | 467 | err = register_die_notifier(&kprobe_exceptions_nb); |
468 | /* Register the trampoline probe for return probe */ | ||
469 | register_kprobe(&trampoline_p); | ||
270 | return err; | 470 | return err; |
271 | } | 471 | } |
272 | 472 | ||
@@ -277,3 +477,6 @@ EXPORT_SYMBOL_GPL(unregister_kprobe); | |||
277 | EXPORT_SYMBOL_GPL(register_jprobe); | 477 | EXPORT_SYMBOL_GPL(register_jprobe); |
278 | EXPORT_SYMBOL_GPL(unregister_jprobe); | 478 | EXPORT_SYMBOL_GPL(unregister_jprobe); |
279 | EXPORT_SYMBOL_GPL(jprobe_return); | 479 | EXPORT_SYMBOL_GPL(jprobe_return); |
480 | EXPORT_SYMBOL_GPL(register_kretprobe); | ||
481 | EXPORT_SYMBOL_GPL(unregister_kretprobe); | ||
482 | |||