-
Notifications
You must be signed in to change notification settings - Fork 1
/
memmap.c
291 lines (270 loc) · 8.7 KB
/
memmap.c
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
#include <stdlib.h>
#include <sys/mman.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <strings.h>
#include <assert.h>
#include <error.h>
#include <string.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <errno.h>
#include <signal.h>
#include <math.h>
#include "ipa.h"
#include "memmap.h"
#include "ipa_utils.h"
#include "file_io.h"
static void map_now(volatile ipa_page_t *);
extern shared_data_t * shared;
extern bool speculating(void);
extern void ipa_init(void);
static volatile ipa_page_t * map_info;
static bool is_mapped_bool;
header_page_t * seq_headers, * seq_headers_last;
void map_handler(int x) {
is_mapped_bool = false;
map_now(map_info);
}
void segv_bool_test(int x) {
is_mapped_bool = false;
}
bool is_addr_mapped(volatile void * address) {
is_mapped_bool = true;
typeof(&segv_bool_test) old = signal(SIGSEGV, segv_bool_test);
if (old == SIG_ERR) {
ipa_perror("Unable to install signal handler");
abort();
}
__sync_synchronize();
volatile int __attribute__((__unused__)) data = *(int *) address;
__sync_synchronize();
signal(SIGSEGV, old);
return is_mapped_bool;
}
bool is_mapped_segv_check(volatile ipa_page_t * prev) {
map_info = prev;
is_mapped_bool = true;
typeof(&map_handler) old = signal(SIGSEGV, map_handler);
if (old == SIG_ERR) {
ipa_perror("Unable to install signal handler");
abort();
}
__sync_synchronize();
volatile ipa_page_t __attribute__((__unused__)) data = *(ipa_page_t *) prev->next_page;
__sync_synchronize();
signal(SIGSEGV, old);
return is_mapped_bool;
}
bool is_mapped(void * ptr) {
ipa_page_t * aligned = (ipa_page_t *) (((intptr_t) ptr) & ~(PAGE_SIZE - 1));
bool b = !(msync((void *) aligned, 1, MS_ASYNC) == -1 && errno == ENOMEM);
// errno = 0;
return b;
}
static void map_now (volatile ipa_page_t * last_page) {
if (last_page->next_pg_name == (unsigned) -1) {
abort();
}
int fd = mmap_existing_fd(last_page->next_pg_name);
if (fd == -1) {
// TODO handle error
abort();
} else {
size_t size = get_size_fd(fd);
if (mmap((void *) last_page->next_page, size, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd, 0) == MAP_FAILED) {
// TODO handle error
abort();
}
close(fd); // think can do here
assert(is_mapped((void *) last_page->next_page));
if ((ipa_page_t *) last_page->next_page == last_page) {
abort();
}
}
}
static inline bool full_map_check(bool faulting, volatile ipa_page_t * prev) {
if (prev->next_pg_name == -1) {
return true;
} else if (!faulting) {
return is_mapped_segv_check(prev);
} else {
return false;
}
}
// Map all missing pages
// Returns the last non-null page (eg. the one with the next page filed set to null)
static inline volatile ipa_page_t * map_missing_pages_safe(bool faulting) {
static volatile ipa_page_t * cache = NULL;
volatile ipa_page_t * start = cache == NULL ? &shared->next_page : cache;
volatile ipa_page_t * last_page = start;
// These are for debugging, volatile so they can't be removed by compiler
volatile int rounds = 0, maps = 0;
if (start->next_page != NULL) {
if (!full_map_check(faulting, start)) {
maps++;
map_now(start);
assert(is_mapped((void *) start->next_page));
}
for (last_page = start; last_page->next_page != NULL; last_page = (volatile ipa_page_t *) last_page->next_page) {
rounds++;
if (!full_map_check(faulting, last_page)) {
maps++;
map_now(last_page);
}
if ((ipa_page_t *) last_page->next_page == last_page) {
// last_page->next_page = NULL;
// last_page->next_pg_name = 0;
// break;
abort();
}
}
}
return last_page;
}
volatile ipa_page_t * map_missing_pages() {
return map_missing_pages_safe(false);
}
void map_missing_pages_handler() {
map_missing_pages_safe(true);
}
/**
* MMap a page according to @type
*
* @method allocate_ipa_page
*
* @param minsize the minimal size of the allocation, including any headers
* @param flags baseline flags to pass to mmap, MAP_FIXED always added,
* MAP_ANONYMOUS while not speculating
*/
static inline ipa_page_t * allocate_ipa_page(int file_no, size_t minsize, int flags) {
ipa_page_t * allocation = NULL;
ipa_page_t alloc = {0};
// Reserve the resources in shared for this allocation
int file_descriptor;
size_t allocation_size = MAX(minsize, PAGE_SIZE);
assert(allocation_size % PAGE_SIZE == 0);
assert(shared != NULL);
if (file_no == -1) {
flags |= MAP_ANONYMOUS;
} else {
flags &= ~MAP_PRIVATE;
flags |= MAP_SHARED;
}
volatile ipa_page_t * last_page;
if (file_no == -1) {
allocation = mmap(NULL, allocation_size, PROT_READ | PROT_WRITE, flags, -1, 0);
} else {
assert(file_no != -1);
file_descriptor = mmap_fd(file_no, allocation_size);
while (true) {
last_page = map_missing_pages();
/**
* Let the kernel decide where to put the new page(s)
* Tasks communicate by requiring the CAS to succeed. If it fails
* then some other task allocated its page
*/
allocation = mmap(NULL, allocation_size, PROT_READ | PROT_WRITE, flags, file_descriptor, 0);
if (allocation == MAP_FAILED) {
ipa_perror("Unable to set up mmap page");
abort();
}
#ifdef MANUAL_ZERO
memset(allocation, 0, allocation_size);
#endif
allocation->next_page = NULL;
allocation->next_pg_name = 0;
alloc.next_page = (struct ipa_page_t *) allocation;
alloc.next_pg_name = file_no;
if (last_page == (volatile ipa_page_t *) last_page->next_page) {
abort();
}
if (__sync_bool_compare_and_swap(&last_page->combined, 0, alloc.combined)) {
assert( (ipa_page_t *) last_page->next_page == allocation);
assert( (ipa_page_t *) allocation->next_page != last_page);
assert( (ipa_page_t *) allocation->next_page != allocation);
break;
} else {
munmap(allocation, allocation_size);
stats_collect(&shared->num_unmaps, 1);
}
}
if (file_descriptor != -1) {
if (close(file_descriptor)) {
ipa_perror("Unable to close file descriptor");
}
}
}
stats_collect(&shared->total_alloc, allocation_size);
page_collect(&shared->total_frames, allocation_size);
stats_collect(&shared->number_mmap, 1);
return allocation;
}
header_page_t * allocate_header_page() {
// headers are always shared -- always increment name
const int file_no = __sync_add_and_fetch(&shared->next_name, 1);
header_page_t * headers = (header_page_t *) allocate_ipa_page(file_no, MAX(PAGE_SIZE, sizeof(header_page_t)), MAP_SHARED);
if (headers == (header_page_t *) -1) {
exit(-1);
}
headers->next_free = 1;
if (speculating()) {
do {
if (shared->header_pg == NULL) {
headers->next_header = NULL;
} else {
headers->next_header = (volatile struct header_page_t *) shared->header_pg;
}
} while (!__sync_bool_compare_and_swap(&shared->header_pg,
(header_page_t *) headers->next_header,
headers));
} else {
#ifdef SUPPORT_THREADS
do {
if (seq_headers == NULL) {
headers->next_header = NULL;
} else {
headers->next_header = (volatile struct header_page_t *) seq_headers;
}
} while (!__sync_bool_compare_and_swap(&seq_headers,
(header_page_t *) headers->next_header,
headers));
#else
if (seq_headers == NULL) {
seq_headers_last = seq_headers = headers;
} else {
headers->next_header = (volatile struct header_page_t *) seq_headers;
seq_headers = headers;
}
#endif
}
stats_collect(&shared->header_pages, 1);
return headers;
}
huge_block_t * allocate_large(size_t size) {
const int file_no = !speculating() ? -1 : __sync_add_and_fetch(&shared->next_name, 1);
// Align to a page size
size_t alloc_size = PAGE_ALIGN((size + sizeof(huge_block_t)));
assert(alloc_size > size);
assert(alloc_size % PAGE_SIZE == 0);
huge_block_t * block = (huge_block_t *) allocate_ipa_page(file_no, alloc_size, MAP_PRIVATE);
if (block == (huge_block_t *) -1) {
return NULL;
}
block->huge_block_sz = alloc_size;
block->file_name = file_no;
block->is_shared = file_no == -1;
if (speculating()) {
do {
block->next_block = (volatile struct huge_block_t *) shared->large_block;
} while(!__sync_bool_compare_and_swap(&shared->large_block, (huge_block_t*) block->next_block, block));
}
if (block == (huge_block_t *) block->next_page.next_page) {
abort();
}
stats_collect(&shared->huge_allocations, 1);
return block;
}