1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
|
/*
* plugin-gen.c - TCG-related bits of plugin infrastructure
*
* Copyright (C) 2018, Emilio G. Cota <cota@braap.org>
* License: GNU GPL, version 2 or later.
* See the COPYING file in the top-level directory.
*
* We support instrumentation at an instruction granularity. That is,
* if a plugin wants to instrument the memory accesses performed by a
* particular instruction, it can just do that instead of instrumenting
* all memory accesses. Thus, in order to do this we first have to
* translate a TB, so that plugins can decide what/where to instrument.
*
* Injecting the desired instrumentation could be done with a second
* translation pass that combined the instrumentation requests, but that
* would be ugly and inefficient since we would decode the guest code twice.
* Instead, during TB translation we add "plugin_cb" marker opcodes
* for all possible instrumentation events, and then once we collect the
* instrumentation requests from plugins, we generate code for those markers
* or remove them if they have no requests.
*/
#include "qemu/osdep.h"
#include "qemu/plugin.h"
#include "qemu/log.h"
#include "cpu.h"
#include "tcg/tcg.h"
#include "tcg/tcg-temp-internal.h"
#include "tcg/tcg-op.h"
#include "exec/exec-all.h"
#include "exec/plugin-gen.h"
#include "exec/translator.h"
enum plugin_gen_from {
PLUGIN_GEN_FROM_TB,
PLUGIN_GEN_FROM_INSN,
PLUGIN_GEN_AFTER_INSN,
PLUGIN_GEN_AFTER_TB,
};
/* called before finishing a TB with exit_tb, goto_tb or goto_ptr */
void plugin_gen_disable_mem_helpers(void)
{
if (tcg_ctx->plugin_insn) {
tcg_gen_plugin_cb(PLUGIN_GEN_AFTER_TB);
}
}
static void gen_enable_mem_helper(struct qemu_plugin_tb *ptb,
struct qemu_plugin_insn *insn)
{
GArray *arr;
size_t len;
/*
* Tracking memory accesses performed from helpers requires extra work.
* If an instruction is emulated with helpers, we do two things:
* (1) copy the CB descriptors, and keep track of it so that they can be
* freed later on, and (2) point CPUState.neg.plugin_mem_cbs to the
* descriptors, so that we can read them at run-time
* (i.e. when the helper executes).
* This run-time access is performed from qemu_plugin_vcpu_mem_cb.
*
* Note that plugin_gen_disable_mem_helpers undoes (2). Since it
* is possible that the code we generate after the instruction is
* dead, we also add checks before generating tb_exit etc.
*/
if (!insn->calls_helpers) {
return;
}
if (!insn->mem_cbs || !insn->mem_cbs->len) {
insn->mem_helper = false;
return;
}
insn->mem_helper = true;
ptb->mem_helper = true;
/*
* TODO: It seems like we should be able to use ref/unref
* to avoid needing to actually copy this array.
* Alternately, perhaps we could allocate new memory adjacent
* to the TranslationBlock itself, so that we do not have to
* actively manage the lifetime after this.
*/
len = insn->mem_cbs->len;
arr = g_array_sized_new(false, false,
sizeof(struct qemu_plugin_dyn_cb), len);
memcpy(arr->data, insn->mem_cbs->data,
len * sizeof(struct qemu_plugin_dyn_cb));
qemu_plugin_add_dyn_cb_arr(arr);
tcg_gen_st_ptr(tcg_constant_ptr((intptr_t)arr), tcg_env,
offsetof(CPUState, neg.plugin_mem_cbs) -
offsetof(ArchCPU, env));
}
static void gen_disable_mem_helper(void)
{
tcg_gen_st_ptr(tcg_constant_ptr(0), tcg_env,
offsetof(CPUState, neg.plugin_mem_cbs) -
offsetof(ArchCPU, env));
}
static void gen_udata_cb(struct qemu_plugin_dyn_cb *cb)
{
TCGv_i32 cpu_index = tcg_temp_ebb_new_i32();
tcg_gen_ld_i32(cpu_index, tcg_env,
-offsetof(ArchCPU, env) + offsetof(CPUState, cpu_index));
tcg_gen_call2(cb->regular.f.vcpu_udata, cb->regular.info, NULL,
tcgv_i32_temp(cpu_index),
tcgv_ptr_temp(tcg_constant_ptr(cb->userp)));
tcg_temp_free_i32(cpu_index);
}
static void gen_inline_cb(struct qemu_plugin_dyn_cb *cb)
{
GArray *arr = cb->inline_insn.entry.score->data;
size_t offset = cb->inline_insn.entry.offset;
TCGv_i32 cpu_index = tcg_temp_ebb_new_i32();
TCGv_i64 val = tcg_temp_ebb_new_i64();
TCGv_ptr ptr = tcg_temp_ebb_new_ptr();
tcg_gen_ld_i32(cpu_index, tcg_env,
-offsetof(ArchCPU, env) + offsetof(CPUState, cpu_index));
tcg_gen_muli_i32(cpu_index, cpu_index, g_array_get_element_size(arr));
tcg_gen_ext_i32_ptr(ptr, cpu_index);
tcg_temp_free_i32(cpu_index);
tcg_gen_addi_ptr(ptr, ptr, (intptr_t)arr->data);
tcg_gen_ld_i64(val, ptr, offset);
tcg_gen_addi_i64(val, val, cb->inline_insn.imm);
tcg_gen_st_i64(val, ptr, offset);
tcg_temp_free_i64(val);
tcg_temp_free_ptr(ptr);
}
static void gen_mem_cb(struct qemu_plugin_dyn_cb *cb,
qemu_plugin_meminfo_t meminfo, TCGv_i64 addr)
{
TCGv_i32 cpu_index = tcg_temp_ebb_new_i32();
tcg_gen_ld_i32(cpu_index, tcg_env,
-offsetof(ArchCPU, env) + offsetof(CPUState, cpu_index));
tcg_gen_call4(cb->regular.f.vcpu_mem, cb->regular.info, NULL,
tcgv_i32_temp(cpu_index),
tcgv_i32_temp(tcg_constant_i32(meminfo)),
tcgv_i64_temp(addr),
tcgv_ptr_temp(tcg_constant_ptr(cb->userp)));
tcg_temp_free_i32(cpu_index);
}
static void inject_cb(struct qemu_plugin_dyn_cb *cb)
{
switch (cb->type) {
case PLUGIN_CB_REGULAR:
gen_udata_cb(cb);
break;
case PLUGIN_CB_INLINE:
gen_inline_cb(cb);
break;
default:
g_assert_not_reached();
}
}
static void inject_mem_cb(struct qemu_plugin_dyn_cb *cb,
enum qemu_plugin_mem_rw rw,
qemu_plugin_meminfo_t meminfo, TCGv_i64 addr)
{
if (cb->rw & rw) {
switch (cb->type) {
case PLUGIN_CB_MEM_REGULAR:
gen_mem_cb(cb, meminfo, addr);
break;
default:
inject_cb(cb);
break;
}
}
}
static void plugin_gen_inject(struct qemu_plugin_tb *plugin_tb)
{
TCGOp *op, *next;
int insn_idx = -1;
if (unlikely(qemu_loglevel_mask(LOG_TB_OP_PLUGIN)
&& qemu_log_in_addr_range(tcg_ctx->plugin_db->pc_first))) {
FILE *logfile = qemu_log_trylock();
if (logfile) {
fprintf(logfile, "OP before plugin injection:\n");
tcg_dump_ops(tcg_ctx, logfile, false);
fprintf(logfile, "\n");
qemu_log_unlock(logfile);
}
}
/*
* While injecting code, we cannot afford to reuse any ebb temps
* that might be live within the existing opcode stream.
* The simplest solution is to release them all and create new.
*/
memset(tcg_ctx->free_temps, 0, sizeof(tcg_ctx->free_temps));
QTAILQ_FOREACH_SAFE(op, &tcg_ctx->ops, link, next) {
switch (op->opc) {
case INDEX_op_insn_start:
insn_idx++;
break;
case INDEX_op_plugin_cb:
{
enum plugin_gen_from from = op->args[0];
struct qemu_plugin_insn *insn = NULL;
const GArray *cbs;
int i, n;
if (insn_idx >= 0) {
insn = g_ptr_array_index(plugin_tb->insns, insn_idx);
}
tcg_ctx->emit_before_op = op;
switch (from) {
case PLUGIN_GEN_AFTER_TB:
if (plugin_tb->mem_helper) {
gen_disable_mem_helper();
}
break;
case PLUGIN_GEN_AFTER_INSN:
assert(insn != NULL);
if (insn->mem_helper) {
gen_disable_mem_helper();
}
break;
case PLUGIN_GEN_FROM_TB:
assert(insn == NULL);
cbs = plugin_tb->cbs;
for (i = 0, n = (cbs ? cbs->len : 0); i < n; i++) {
inject_cb(
&g_array_index(cbs, struct qemu_plugin_dyn_cb, i));
}
break;
case PLUGIN_GEN_FROM_INSN:
assert(insn != NULL);
gen_enable_mem_helper(plugin_tb, insn);
cbs = insn->insn_cbs;
for (i = 0, n = (cbs ? cbs->len : 0); i < n; i++) {
inject_cb(
&g_array_index(cbs, struct qemu_plugin_dyn_cb, i));
}
break;
default:
g_assert_not_reached();
}
tcg_ctx->emit_before_op = NULL;
tcg_op_remove(tcg_ctx, op);
break;
}
case INDEX_op_plugin_mem_cb:
{
TCGv_i64 addr = temp_tcgv_i64(arg_temp(op->args[0]));
qemu_plugin_meminfo_t meminfo = op->args[1];
enum qemu_plugin_mem_rw rw =
(qemu_plugin_mem_is_store(meminfo)
? QEMU_PLUGIN_MEM_W : QEMU_PLUGIN_MEM_R);
struct qemu_plugin_insn *insn;
const GArray *cbs;
int i, n;
assert(insn_idx >= 0);
insn = g_ptr_array_index(plugin_tb->insns, insn_idx);
tcg_ctx->emit_before_op = op;
cbs = insn->mem_cbs;
for (i = 0, n = (cbs ? cbs->len : 0); i < n; i++) {
inject_mem_cb(&g_array_index(cbs, struct qemu_plugin_dyn_cb, i),
rw, meminfo, addr);
}
tcg_ctx->emit_before_op = NULL;
tcg_op_remove(tcg_ctx, op);
break;
}
default:
/* plugins don't care about any other ops */
break;
}
}
}
bool plugin_gen_tb_start(CPUState *cpu, const DisasContextBase *db)
{
struct qemu_plugin_tb *ptb;
if (!test_bit(QEMU_PLUGIN_EV_VCPU_TB_TRANS,
cpu->plugin_state->event_mask)) {
return false;
}
tcg_ctx->plugin_db = db;
tcg_ctx->plugin_insn = NULL;
ptb = tcg_ctx->plugin_tb;
if (ptb) {
/* Reset callbacks */
if (ptb->cbs) {
g_array_set_size(ptb->cbs, 0);
}
ptb->n = 0;
ptb->mem_helper = false;
} else {
ptb = g_new0(struct qemu_plugin_tb, 1);
tcg_ctx->plugin_tb = ptb;
ptb->insns = g_ptr_array_new();
}
tcg_gen_plugin_cb(PLUGIN_GEN_FROM_TB);
return true;
}
void plugin_gen_insn_start(CPUState *cpu, const DisasContextBase *db)
{
struct qemu_plugin_tb *ptb = tcg_ctx->plugin_tb;
struct qemu_plugin_insn *insn;
size_t n = db->num_insns;
vaddr pc;
assert(n >= 1);
ptb->n = n;
if (n <= ptb->insns->len) {
insn = g_ptr_array_index(ptb->insns, n - 1);
} else {
assert(n - 1 == ptb->insns->len);
insn = g_new0(struct qemu_plugin_insn, 1);
g_ptr_array_add(ptb->insns, insn);
}
tcg_ctx->plugin_insn = insn;
insn->calls_helpers = false;
insn->mem_helper = false;
if (insn->insn_cbs) {
g_array_set_size(insn->insn_cbs, 0);
}
if (insn->mem_cbs) {
g_array_set_size(insn->mem_cbs, 0);
}
pc = db->pc_next;
insn->vaddr = pc;
tcg_gen_plugin_cb(PLUGIN_GEN_FROM_INSN);
}
void plugin_gen_insn_end(void)
{
const DisasContextBase *db = tcg_ctx->plugin_db;
struct qemu_plugin_insn *pinsn = tcg_ctx->plugin_insn;
pinsn->len = db->fake_insn ? db->record_len : db->pc_next - pinsn->vaddr;
tcg_gen_plugin_cb(PLUGIN_GEN_AFTER_INSN);
}
/*
* There are cases where we never get to finalise a translation - for
* example a page fault during translation. As a result we shouldn't
* do any clean-up here and make sure things are reset in
* plugin_gen_tb_start.
*/
void plugin_gen_tb_end(CPUState *cpu, size_t num_insns)
{
struct qemu_plugin_tb *ptb = tcg_ctx->plugin_tb;
/* translator may have removed instructions, update final count */
g_assert(num_insns <= ptb->n);
ptb->n = num_insns;
/* collect instrumentation requests */
qemu_plugin_tb_trans_cb(cpu, ptb);
/* inject the instrumentation at the appropriate places */
plugin_gen_inject(ptb);
}
|