blob: f47d687e32521e8e1117a0d540744e7e3cffc998 [file] [log] [blame]
/* ******************************************************************************
* Copyright (c) 2013-2018 Google, Inc. All rights reserved.
* Copyright (c) 2011 Massachusetts Institute of Technology All rights reserved.
* Copyright (c) 2008 VMware, Inc. All rights reserved.
* ******************************************************************************/
/*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* * Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* * Neither the name of VMware, Inc. nor the names of its contributors may be
* used to endorse or promote products derived from this software without
* specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL VMWARE, INC. OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
* DAMAGE.
*/
/* Code Manipulation API Sample:
* modxfer.c
*
* - Reports the dynamic count of total number of instructions executed
* and the number of transfers between modules via indirect branches.
* - This is different from modxfer_app2lib.c as it counts all transfers between
* any modules while modxfer.c only counts transfers between app and other
* modules, i.e., not counting transfers between two libraries.
* - modxfer and modxfer_app2lib are to demonstrate how we can simplify the client
* and improve the performance if we collect less information (modxfer_app2lib).
* - We assume that most of cross module transfers happen via indirect branches,
* and most of them are paired, so we only instrument indrect branches
* but not returns for better performance.
* - It is possible a direct branch with DGC or self-mod jumps between modules,
* they are ignored for now. They can be handled differently from indirect
* branch since the src and target are known at instrumentation time.
*/
#include "utils.h"
#include "drmgr.h"
#include "drreg.h"
#include "drx.h"
#include <string.h>
#define MAX_NUM_MODULES 0x1000
#define UNKNOW_MODULE_IDX (MAX_NUM_MODULES - 1)
typedef struct _module_array_t {
app_pc base;
app_pc end;
bool loaded;
module_data_t *info;
} module_array_t;
/* the last slot is where all non-module address go */
static module_array_t mod_array[MAX_NUM_MODULES];
static uint64 ins_count; /* number of instructions executed in total */
static void *mod_lock;
static int num_mods;
static uint xfer_cnt[MAX_NUM_MODULES][MAX_NUM_MODULES];
static uint64 mod_cnt[MAX_NUM_MODULES];
static file_t logfile;
static bool
module_data_same(const module_data_t *d1, const module_data_t *d2)
{
if (d1->start == d2->start && d1->end == d2->end &&
d1->entry_point == d2->entry_point &&
#ifdef WINDOWS
d1->checksum == d2->checksum && d1->timestamp == d2->timestamp &&
#endif
/* treat two modules w/ no name (there are some) as different */
dr_module_preferred_name(d1) != NULL && dr_module_preferred_name(d2) != NULL &&
strcmp(dr_module_preferred_name(d1), dr_module_preferred_name(d2)) == 0)
return true;
return false;
}
/* Simple clean calls with two arguments will not be inlined, but the context
* switch can be optimized for better performance.
*/
static void
mbr_update(app_pc instr_addr, app_pc target_addr)
{
int i, j;
/* XXX: this can be optimized by walking a tree instead */
/* find the source module */
for (i = 0; i < num_mods; i++) {
if (mod_array[i].loaded && mod_array[i].base <= instr_addr &&
mod_array[i].end > instr_addr)
break;
}
/* if cannot find a module, put it to the last */
if (i == num_mods)
i = UNKNOW_MODULE_IDX;
/* find the target module */
/* quick check if it is the same module */
if (i < UNKNOW_MODULE_IDX && mod_array[i].base <= target_addr &&
mod_array[i].end > target_addr) {
j = i;
} else {
for (j = 0; j < num_mods; j++) {
if (mod_array[j].loaded && mod_array[j].base <= target_addr &&
mod_array[j].end > target_addr)
break;
}
/* if cannot find a module, put it to the last */
if (j == num_mods)
j = UNKNOW_MODULE_IDX;
}
/* this is a racy update, but should be ok to be a few number off */
xfer_cnt[i][j]++;
}
static void
event_exit(void);
static dr_emit_flags_t
event_analyze_bb(void *drcontext, void *tag, instrlist_t *bb, bool for_trace,
bool translating, void **user_data);
static dr_emit_flags_t
event_insert_instrumentation(void *drcontext, void *tag, instrlist_t *bb, instr_t *instr,
bool for_trace, bool translating, void *user_data);
static void
event_module_load(void *drcontext, const module_data_t *info, bool loaded);
static void
event_module_unload(void *drcontext, const module_data_t *info);
DR_EXPORT void
dr_client_main(client_id_t id, int argc, const char *argv[])
{
/* We need no drreg slots ourselves, but we initialize drreg as we call
* drreg_restore_app_values(), required since drx_insert_counter_update()
* uses drreg when drmgr is used.
*/
drreg_options_t ops = { sizeof(ops) };
dr_set_client_name("DynamoRIO Sample Client 'modxfer'",
"http://dynamorio.org/issues");
if (!drmgr_init() || drreg_init(&ops) != DRREG_SUCCESS)
DR_ASSERT(false);
drx_init();
/* register events */
dr_register_exit_event(event_exit);
if (!drmgr_register_bb_instrumentation_event(event_analyze_bb,
event_insert_instrumentation, NULL))
DR_ASSERT(false);
drmgr_register_module_load_event(event_module_load);
drmgr_register_module_unload_event(event_module_unload);
mod_lock = dr_mutex_create();
logfile = log_file_open(id, NULL /* drcontext */, NULL /* path */, "modxfer",
#ifndef WINDOWS
DR_FILE_CLOSE_ON_FORK |
#endif
DR_FILE_ALLOW_LARGE);
DR_ASSERT(logfile != INVALID_FILE);
/* make it easy to tell, by looking at log file, which client executed */
dr_log(NULL, DR_LOG_ALL, 1, "Client 'modxfer' initializing\n");
#ifdef SHOW_RESULTS
/* also give notification to stderr */
if (dr_is_notify_on()) {
# ifdef WINDOWS
/* ask for best-effort printing to cmd window. must be called at init. */
dr_enable_console_printing();
# endif
dr_fprintf(STDERR, "Client modxfer is running\n");
}
#endif
}
static void
event_exit(void)
{
int i;
char msg[512];
int len;
int j;
uint64 xmod_xfer = 0;
uint64 self_xfer = 0;
for (i = 0; i < num_mods; i++) {
dr_fprintf(logfile, "module %3d: %s\n", i,
dr_module_preferred_name(mod_array[i].info) == NULL
? "<unknown>"
: dr_module_preferred_name(mod_array[i].info));
dr_fprintf(logfile, "%20llu instruction executed\n", mod_cnt[i]);
}
if (mod_cnt[UNKNOW_MODULE_IDX] != 0) {
dr_fprintf(logfile, "unknown modules:\n%20llu instruction executed\n",
mod_cnt[UNKNOW_MODULE_IDX]);
}
for (i = 0; i < MAX_NUM_MODULES; i++) {
for (j = 0; j < num_mods; j++) {
if (xfer_cnt[i][j] != 0) {
dr_fprintf(logfile, "mod %3d => mod %3d: %8u\n", i, j, xfer_cnt[i][j]);
if (i == j)
self_xfer += xfer_cnt[i][j];
else
xmod_xfer += xfer_cnt[i][j];
}
}
}
len = dr_snprintf(msg, sizeof(msg) / sizeof(msg[0]),
"Instrumentation results:\n"
"\t%10llu instructions executed\n"
"\t%10llu (%2.3f%%) cross module indirect branches\n"
"\t%10llu (%2.3f%%) intra-module indirect branches\n",
ins_count, xmod_xfer, 100 * (float)xmod_xfer / ins_count, self_xfer,
100 * (float)self_xfer / ins_count);
DR_ASSERT(len > 0);
NULL_TERMINATE_BUFFER(msg);
#ifdef SHOW_RESULTS
DISPLAY_STRING(msg);
#endif /* SHOW_RESULTS */
dr_fprintf(logfile, "%s\n", msg);
dr_mutex_lock(mod_lock);
for (i = 0; i < num_mods; i++) {
DR_ASSERT(mod_array[i].info != NULL);
dr_free_module_data(mod_array[i].info);
}
dr_mutex_unlock(mod_lock);
dr_mutex_destroy(mod_lock);
log_file_close(logfile);
drx_exit();
if (!drmgr_unregister_bb_instrumentation_event(event_analyze_bb) ||
!drmgr_unregister_module_load_event(event_module_load) ||
!drmgr_unregister_module_unload_event(event_module_unload) ||
drreg_exit() != DRREG_SUCCESS)
DR_ASSERT(false);
drmgr_exit();
}
/* This event is passed the instruction list for the whole bb. */
static dr_emit_flags_t
event_analyze_bb(void *drcontext, void *tag, instrlist_t *bb, bool for_trace,
bool translating, void **user_data)
{
/* Count the instructions and pass the result to event_insert_instrumentation. */
instr_t *instr;
uint num_instrs;
for (instr = instrlist_first_app(bb), num_instrs = 0; instr != NULL;
instr = instr_get_next_app(instr)) {
num_instrs++;
}
*(uint *)user_data = num_instrs;
return DR_EMIT_DEFAULT;
}
/* This event is called separately for each individual instruction in the bb. */
static dr_emit_flags_t
event_insert_instrumentation(void *drcontext, void *tag, instrlist_t *bb, instr_t *instr,
bool for_trace, bool translating, void *user_data)
{
if (drmgr_is_first_instr(drcontext, instr)) {
uint num_instrs = (uint)(ptr_uint_t)user_data;
int i;
app_pc bb_addr = dr_fragment_app_pc(tag);
for (i = 0; i < num_mods; i++) {
if (mod_array[i].loaded && mod_array[i].base <= bb_addr &&
mod_array[i].end > bb_addr)
break;
}
if (i == num_mods)
i = UNKNOW_MODULE_IDX;
/* We pass SPILL_SLOT_MAX+1 as drx will use drreg for spilling. */
drx_insert_counter_update(drcontext, bb, instr, SPILL_SLOT_MAX + 1,
(void *)&mod_cnt[i], num_instrs, DRX_COUNTER_64BIT);
drx_insert_counter_update(drcontext, bb, instr, SPILL_SLOT_MAX + 1,
(void *)&ins_count, num_instrs, DRX_COUNTER_64BIT);
}
if (instr_is_mbr(instr) && !instr_is_return(instr)) {
/* Assuming most of the transfers between modules are paired, we
* instrument indirect branches but not returns for better performance.
* We assume that most cross module transfers happens via indirect
* branches.
* Direct branch with DGC or self-modify may also cross modules, but
* it should be ok to ignore, and we can handle them more efficiently.
*/
/* dr_insert_mbr_instrumentation is going to read app values, so we need a
* drreg lazy restore "barrier" here.
*/
drreg_status_t res =
drreg_restore_app_values(drcontext, bb, instr, instr_get_target(instr), NULL);
DR_ASSERT(res == DRREG_SUCCESS || res == DRREG_ERROR_NO_APP_VALUE);
dr_insert_mbr_instrumentation(drcontext, bb, instr, (void *)mbr_update,
SPILL_SLOT_1);
}
return DR_EMIT_DEFAULT;
}
static void
event_module_load(void *drcontext, const module_data_t *info, bool loaded)
{
int i;
dr_mutex_lock(mod_lock);
for (i = 0; i < num_mods; i++) {
/* check if it is the same as any unloaded module */
if (!mod_array[i].loaded && module_data_same(mod_array[i].info, info)) {
mod_array[i].loaded = true;
break;
}
}
if (i == num_mods) {
/* new module */
mod_array[i].base = info->start;
mod_array[i].end = info->end;
mod_array[i].loaded = true;
mod_array[i].info = dr_copy_module_data(info);
num_mods++;
}
DR_ASSERT(num_mods < UNKNOW_MODULE_IDX);
dr_mutex_unlock(mod_lock);
}
static void
event_module_unload(void *drcontext, const module_data_t *info)
{
int i;
dr_mutex_lock(mod_lock);
for (i = 0; i < num_mods; i++) {
if (mod_array[i].loaded && module_data_same(mod_array[i].info, info)) {
/* Some module might be repeatedly loaded and unloaded, so instead
* of clearing out the array entry, we keep the data for possible
* reuse.
*/
mod_array[i].loaded = false;
break;
}
}
DR_ASSERT(i < num_mods);
dr_mutex_unlock(mod_lock);
}