~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/tools/vm/page-types.c

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.77 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * page-types: Tool for querying page flags
  3  *
  4  * This program is free software; you can redistribute it and/or modify it
  5  * under the terms of the GNU General Public License as published by the Free
  6  * Software Foundation; version 2.
  7  *
  8  * This program is distributed in the hope that it will be useful, but WITHOUT
  9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 11  * more details.
 12  *
 13  * You should find a copy of v2 of the GNU General Public License somewhere on
 14  * your Linux system; if not, write to the Free Software Foundation, Inc., 59
 15  * Temple Place, Suite 330, Boston, MA 02111-1307 USA.
 16  *
 17  * Copyright (C) 2009 Intel corporation
 18  *
 19  * Authors: Wu Fengguang <fengguang.wu@intel.com>
 20  */
 21 
 22 #define _FILE_OFFSET_BITS 64
 23 #define _GNU_SOURCE
 24 #include <stdio.h>
 25 #include <stdlib.h>
 26 #include <unistd.h>
 27 #include <stdint.h>
 28 #include <stdarg.h>
 29 #include <string.h>
 30 #include <getopt.h>
 31 #include <limits.h>
 32 #include <assert.h>
 33 #include <ftw.h>
 34 #include <time.h>
 35 #include <setjmp.h>
 36 #include <signal.h>
 37 #include <sys/types.h>
 38 #include <sys/errno.h>
 39 #include <sys/fcntl.h>
 40 #include <sys/mount.h>
 41 #include <sys/statfs.h>
 42 #include <sys/mman.h>
 43 #include "../../include/uapi/linux/magic.h"
 44 #include "../../include/uapi/linux/kernel-page-flags.h"
 45 #include <api/fs/fs.h>
 46 
 47 #ifndef MAX_PATH
 48 # define MAX_PATH 256
 49 #endif
 50 
 51 #ifndef STR
 52 # define _STR(x) #x
 53 # define STR(x) _STR(x)
 54 #endif
 55 
 56 /*
 57  * pagemap kernel ABI bits
 58  */
 59 
 60 #define PM_ENTRY_BYTES          8
 61 #define PM_PFRAME_BITS          55
 62 #define PM_PFRAME_MASK          ((1LL << PM_PFRAME_BITS) - 1)
 63 #define PM_PFRAME(x)            ((x) & PM_PFRAME_MASK)
 64 #define MAX_SWAPFILES_SHIFT     5
 65 #define PM_SWAP_OFFSET(x)       (((x) & PM_PFRAME_MASK) >> MAX_SWAPFILES_SHIFT)
 66 #define PM_SOFT_DIRTY           (1ULL << 55)
 67 #define PM_MMAP_EXCLUSIVE       (1ULL << 56)
 68 #define PM_FILE                 (1ULL << 61)
 69 #define PM_SWAP                 (1ULL << 62)
 70 #define PM_PRESENT              (1ULL << 63)
 71 
 72 /*
 73  * kernel page flags
 74  */
 75 
 76 #define KPF_BYTES               8
 77 #define PROC_KPAGEFLAGS         "/proc/kpageflags"
 78 #define PROC_KPAGECGROUP        "/proc/kpagecgroup"
 79 
 80 /* [32-] kernel hacking assistances */
 81 #define KPF_RESERVED            32
 82 #define KPF_MLOCKED             33
 83 #define KPF_MAPPEDTODISK        34
 84 #define KPF_PRIVATE             35
 85 #define KPF_PRIVATE_2           36
 86 #define KPF_OWNER_PRIVATE       37
 87 #define KPF_ARCH                38
 88 #define KPF_UNCACHED            39
 89 #define KPF_SOFTDIRTY           40
 90 
 91 /* [48-] take some arbitrary free slots for expanding overloaded flags
 92  * not part of kernel API
 93  */
 94 #define KPF_READAHEAD           48
 95 #define KPF_SLOB_FREE           49
 96 #define KPF_SLUB_FROZEN         50
 97 #define KPF_SLUB_DEBUG          51
 98 #define KPF_FILE                61
 99 #define KPF_SWAP                62
100 #define KPF_MMAP_EXCLUSIVE      63
101 
102 #define KPF_ALL_BITS            ((uint64_t)~0ULL)
103 #define KPF_HACKERS_BITS        (0xffffULL << 32)
104 #define KPF_OVERLOADED_BITS     (0xffffULL << 48)
105 #define BIT(name)               (1ULL << KPF_##name)
106 #define BITS_COMPOUND           (BIT(COMPOUND_HEAD) | BIT(COMPOUND_TAIL))
107 
108 static const char * const page_flag_names[] = {
109         [KPF_LOCKED]            = "L:locked",
110         [KPF_ERROR]             = "E:error",
111         [KPF_REFERENCED]        = "R:referenced",
112         [KPF_UPTODATE]          = "U:uptodate",
113         [KPF_DIRTY]             = "D:dirty",
114         [KPF_LRU]               = "l:lru",
115         [KPF_ACTIVE]            = "A:active",
116         [KPF_SLAB]              = "S:slab",
117         [KPF_WRITEBACK]         = "W:writeback",
118         [KPF_RECLAIM]           = "I:reclaim",
119         [KPF_BUDDY]             = "B:buddy",
120 
121         [KPF_MMAP]              = "M:mmap",
122         [KPF_ANON]              = "a:anonymous",
123         [KPF_SWAPCACHE]         = "s:swapcache",
124         [KPF_SWAPBACKED]        = "b:swapbacked",
125         [KPF_COMPOUND_HEAD]     = "H:compound_head",
126         [KPF_COMPOUND_TAIL]     = "T:compound_tail",
127         [KPF_HUGE]              = "G:huge",
128         [KPF_UNEVICTABLE]       = "u:unevictable",
129         [KPF_HWPOISON]          = "X:hwpoison",
130         [KPF_NOPAGE]            = "n:nopage",
131         [KPF_KSM]               = "x:ksm",
132         [KPF_THP]               = "t:thp",
133         [KPF_BALLOON]           = "o:balloon",
134         [KPF_ZERO_PAGE]         = "z:zero_page",
135         [KPF_IDLE]              = "i:idle_page",
136 
137         [KPF_RESERVED]          = "r:reserved",
138         [KPF_MLOCKED]           = "m:mlocked",
139         [KPF_MAPPEDTODISK]      = "d:mappedtodisk",
140         [KPF_PRIVATE]           = "P:private",
141         [KPF_PRIVATE_2]         = "p:private_2",
142         [KPF_OWNER_PRIVATE]     = "O:owner_private",
143         [KPF_ARCH]              = "h:arch",
144         [KPF_UNCACHED]          = "c:uncached",
145         [KPF_SOFTDIRTY]         = "f:softdirty",
146 
147         [KPF_READAHEAD]         = "I:readahead",
148         [KPF_SLOB_FREE]         = "P:slob_free",
149         [KPF_SLUB_FROZEN]       = "A:slub_frozen",
150         [KPF_SLUB_DEBUG]        = "E:slub_debug",
151 
152         [KPF_FILE]              = "F:file",
153         [KPF_SWAP]              = "w:swap",
154         [KPF_MMAP_EXCLUSIVE]    = "1:mmap_exclusive",
155 };
156 
157 
158 static const char * const debugfs_known_mountpoints[] = {
159         "/sys/kernel/debug",
160         "/debug",
161         0,
162 };
163 
164 /*
165  * data structures
166  */
167 
168 static int              opt_raw;        /* for kernel developers */
169 static int              opt_list;       /* list pages (in ranges) */
170 static int              opt_no_summary; /* don't show summary */
171 static pid_t            opt_pid;        /* process to walk */
172 const char *            opt_file;       /* file or directory path */
173 static uint64_t         opt_cgroup;     /* cgroup inode */
174 static int              opt_list_cgroup;/* list page cgroup */
175 
176 #define MAX_ADDR_RANGES 1024
177 static int              nr_addr_ranges;
178 static unsigned long    opt_offset[MAX_ADDR_RANGES];
179 static unsigned long    opt_size[MAX_ADDR_RANGES];
180 
181 #define MAX_VMAS        10240
182 static int              nr_vmas;
183 static unsigned long    pg_start[MAX_VMAS];
184 static unsigned long    pg_end[MAX_VMAS];
185 
186 #define MAX_BIT_FILTERS 64
187 static int              nr_bit_filters;
188 static uint64_t         opt_mask[MAX_BIT_FILTERS];
189 static uint64_t         opt_bits[MAX_BIT_FILTERS];
190 
191 static int              page_size;
192 
193 static int              pagemap_fd;
194 static int              kpageflags_fd;
195 static int              kpagecgroup_fd = -1;
196 
197 static int              opt_hwpoison;
198 static int              opt_unpoison;
199 
200 static const char       *hwpoison_debug_fs;
201 static int              hwpoison_inject_fd;
202 static int              hwpoison_forget_fd;
203 
204 #define HASH_SHIFT      13
205 #define HASH_SIZE       (1 << HASH_SHIFT)
206 #define HASH_MASK       (HASH_SIZE - 1)
207 #define HASH_KEY(flags) (flags & HASH_MASK)
208 
209 static unsigned long    total_pages;
210 static unsigned long    nr_pages[HASH_SIZE];
211 static uint64_t         page_flags[HASH_SIZE];
212 
213 
214 /*
215  * helper functions
216  */
217 
218 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
219 
220 #define min_t(type, x, y) ({                    \
221         type __min1 = (x);                      \
222         type __min2 = (y);                      \
223         __min1 < __min2 ? __min1 : __min2; })
224 
225 #define max_t(type, x, y) ({                    \
226         type __max1 = (x);                      \
227         type __max2 = (y);                      \
228         __max1 > __max2 ? __max1 : __max2; })
229 
230 static unsigned long pages2mb(unsigned long pages)
231 {
232         return (pages * page_size) >> 20;
233 }
234 
235 static void fatal(const char *x, ...)
236 {
237         va_list ap;
238 
239         va_start(ap, x);
240         vfprintf(stderr, x, ap);
241         va_end(ap);
242         exit(EXIT_FAILURE);
243 }
244 
245 static int checked_open(const char *pathname, int flags)
246 {
247         int fd = open(pathname, flags);
248 
249         if (fd < 0) {
250                 perror(pathname);
251                 exit(EXIT_FAILURE);
252         }
253 
254         return fd;
255 }
256 
257 /*
258  * pagemap/kpageflags routines
259  */
260 
261 static unsigned long do_u64_read(int fd, char *name,
262                                  uint64_t *buf,
263                                  unsigned long index,
264                                  unsigned long count)
265 {
266         long bytes;
267 
268         if (index > ULONG_MAX / 8)
269                 fatal("index overflow: %lu\n", index);
270 
271         bytes = pread(fd, buf, count * 8, (off_t)index * 8);
272         if (bytes < 0) {
273                 perror(name);
274                 exit(EXIT_FAILURE);
275         }
276         if (bytes % 8)
277                 fatal("partial read: %lu bytes\n", bytes);
278 
279         return bytes / 8;
280 }
281 
282 static unsigned long kpageflags_read(uint64_t *buf,
283                                      unsigned long index,
284                                      unsigned long pages)
285 {
286         return do_u64_read(kpageflags_fd, PROC_KPAGEFLAGS, buf, index, pages);
287 }
288 
289 static unsigned long kpagecgroup_read(uint64_t *buf,
290                                       unsigned long index,
291                                       unsigned long pages)
292 {
293         if (kpagecgroup_fd < 0)
294                 return pages;
295 
296         return do_u64_read(kpagecgroup_fd, PROC_KPAGEFLAGS, buf, index, pages);
297 }
298 
299 static unsigned long pagemap_read(uint64_t *buf,
300                                   unsigned long index,
301                                   unsigned long pages)
302 {
303         return do_u64_read(pagemap_fd, "/proc/pid/pagemap", buf, index, pages);
304 }
305 
306 static unsigned long pagemap_pfn(uint64_t val)
307 {
308         unsigned long pfn;
309 
310         if (val & PM_PRESENT)
311                 pfn = PM_PFRAME(val);
312         else
313                 pfn = 0;
314 
315         return pfn;
316 }
317 
318 static unsigned long pagemap_swap_offset(uint64_t val)
319 {
320         return val & PM_SWAP ? PM_SWAP_OFFSET(val) : 0;
321 }
322 
323 /*
324  * page flag names
325  */
326 
327 static char *page_flag_name(uint64_t flags)
328 {
329         static char buf[65];
330         int present;
331         size_t i, j;
332 
333         for (i = 0, j = 0; i < ARRAY_SIZE(page_flag_names); i++) {
334                 present = (flags >> i) & 1;
335                 if (!page_flag_names[i]) {
336                         if (present)
337                                 fatal("unknown flag bit %d\n", i);
338                         continue;
339                 }
340                 buf[j++] = present ? page_flag_names[i][0] : '_';
341         }
342 
343         return buf;
344 }
345 
346 static char *page_flag_longname(uint64_t flags)
347 {
348         static char buf[1024];
349         size_t i, n;
350 
351         for (i = 0, n = 0; i < ARRAY_SIZE(page_flag_names); i++) {
352                 if (!page_flag_names[i])
353                         continue;
354                 if ((flags >> i) & 1)
355                         n += snprintf(buf + n, sizeof(buf) - n, "%s,",
356                                         page_flag_names[i] + 2);
357         }
358         if (n)
359                 n--;
360         buf[n] = '\0';
361 
362         return buf;
363 }
364 
365 
366 /*
367  * page list and summary
368  */
369 
370 static void show_page_range(unsigned long voffset, unsigned long offset,
371                             unsigned long size, uint64_t flags, uint64_t cgroup)
372 {
373         static uint64_t      flags0;
374         static uint64_t      cgroup0;
375         static unsigned long voff;
376         static unsigned long index;
377         static unsigned long count;
378 
379         if (flags == flags0 && cgroup == cgroup0 && offset == index + count &&
380             size && voffset == voff + count) {
381                 count += size;
382                 return;
383         }
384 
385         if (count) {
386                 if (opt_pid)
387                         printf("%lx\t", voff);
388                 if (opt_file)
389                         printf("%lu\t", voff);
390                 if (opt_list_cgroup)
391                         printf("@%llu\t", (unsigned long long)cgroup0);
392                 printf("%lx\t%lx\t%s\n",
393                                 index, count, page_flag_name(flags0));
394         }
395 
396         flags0 = flags;
397         cgroup0= cgroup;
398         index  = offset;
399         voff   = voffset;
400         count  = size;
401 }
402 
403 static void flush_page_range(void)
404 {
405         show_page_range(0, 0, 0, 0, 0);
406 }
407 
408 static void show_page(unsigned long voffset, unsigned long offset,
409                       uint64_t flags, uint64_t cgroup)
410 {
411         if (opt_pid)
412                 printf("%lx\t", voffset);
413         if (opt_file)
414                 printf("%lu\t", voffset);
415         if (opt_list_cgroup)
416                 printf("@%llu\t", (unsigned long long)cgroup);
417         printf("%lx\t%s\n", offset, page_flag_name(flags));
418 }
419 
420 static void show_summary(void)
421 {
422         size_t i;
423 
424         printf("             flags\tpage-count       MB"
425                 "  symbolic-flags\t\t\tlong-symbolic-flags\n");
426 
427         for (i = 0; i < ARRAY_SIZE(nr_pages); i++) {
428                 if (nr_pages[i])
429                         printf("0x%016llx\t%10lu %8lu  %s\t%s\n",
430                                 (unsigned long long)page_flags[i],
431                                 nr_pages[i],
432                                 pages2mb(nr_pages[i]),
433                                 page_flag_name(page_flags[i]),
434                                 page_flag_longname(page_flags[i]));
435         }
436 
437         printf("             total\t%10lu %8lu\n",
438                         total_pages, pages2mb(total_pages));
439 }
440 
441 
442 /*
443  * page flag filters
444  */
445 
446 static int bit_mask_ok(uint64_t flags)
447 {
448         int i;
449 
450         for (i = 0; i < nr_bit_filters; i++) {
451                 if (opt_bits[i] == KPF_ALL_BITS) {
452                         if ((flags & opt_mask[i]) == 0)
453                                 return 0;
454                 } else {
455                         if ((flags & opt_mask[i]) != opt_bits[i])
456                                 return 0;
457                 }
458         }
459 
460         return 1;
461 }
462 
463 static uint64_t expand_overloaded_flags(uint64_t flags, uint64_t pme)
464 {
465         /* SLOB/SLUB overload several page flags */
466         if (flags & BIT(SLAB)) {
467                 if (flags & BIT(PRIVATE))
468                         flags ^= BIT(PRIVATE) | BIT(SLOB_FREE);
469                 if (flags & BIT(ACTIVE))
470                         flags ^= BIT(ACTIVE) | BIT(SLUB_FROZEN);
471                 if (flags & BIT(ERROR))
472                         flags ^= BIT(ERROR) | BIT(SLUB_DEBUG);
473         }
474 
475         /* PG_reclaim is overloaded as PG_readahead in the read path */
476         if ((flags & (BIT(RECLAIM) | BIT(WRITEBACK))) == BIT(RECLAIM))
477                 flags ^= BIT(RECLAIM) | BIT(READAHEAD);
478 
479         if (pme & PM_SOFT_DIRTY)
480                 flags |= BIT(SOFTDIRTY);
481         if (pme & PM_FILE)
482                 flags |= BIT(FILE);
483         if (pme & PM_SWAP)
484                 flags |= BIT(SWAP);
485         if (pme & PM_MMAP_EXCLUSIVE)
486                 flags |= BIT(MMAP_EXCLUSIVE);
487 
488         return flags;
489 }
490 
491 static uint64_t well_known_flags(uint64_t flags)
492 {
493         /* hide flags intended only for kernel hacker */
494         flags &= ~KPF_HACKERS_BITS;
495 
496         /* hide non-hugeTLB compound pages */
497         if ((flags & BITS_COMPOUND) && !(flags & BIT(HUGE)))
498                 flags &= ~BITS_COMPOUND;
499 
500         return flags;
501 }
502 
503 static uint64_t kpageflags_flags(uint64_t flags, uint64_t pme)
504 {
505         if (opt_raw)
506                 flags = expand_overloaded_flags(flags, pme);
507         else
508                 flags = well_known_flags(flags);
509 
510         return flags;
511 }
512 
513 /*
514  * page actions
515  */
516 
517 static void prepare_hwpoison_fd(void)
518 {
519         char buf[MAX_PATH + 1];
520 
521         hwpoison_debug_fs = debugfs__mount();
522         if (!hwpoison_debug_fs) {
523                 perror("mount debugfs");
524                 exit(EXIT_FAILURE);
525         }
526 
527         if (opt_hwpoison && !hwpoison_inject_fd) {
528                 snprintf(buf, MAX_PATH, "%s/hwpoison/corrupt-pfn",
529                         hwpoison_debug_fs);
530                 hwpoison_inject_fd = checked_open(buf, O_WRONLY);
531         }
532 
533         if (opt_unpoison && !hwpoison_forget_fd) {
534                 snprintf(buf, MAX_PATH, "%s/hwpoison/unpoison-pfn",
535                         hwpoison_debug_fs);
536                 hwpoison_forget_fd = checked_open(buf, O_WRONLY);
537         }
538 }
539 
540 static int hwpoison_page(unsigned long offset)
541 {
542         char buf[100];
543         int len;
544 
545         len = sprintf(buf, "0x%lx\n", offset);
546         len = write(hwpoison_inject_fd, buf, len);
547         if (len < 0) {
548                 perror("hwpoison inject");
549                 return len;
550         }
551         return 0;
552 }
553 
554 static int unpoison_page(unsigned long offset)
555 {
556         char buf[100];
557         int len;
558 
559         len = sprintf(buf, "0x%lx\n", offset);
560         len = write(hwpoison_forget_fd, buf, len);
561         if (len < 0) {
562                 perror("hwpoison forget");
563                 return len;
564         }
565         return 0;
566 }
567 
568 /*
569  * page frame walker
570  */
571 
572 static size_t hash_slot(uint64_t flags)
573 {
574         size_t k = HASH_KEY(flags);
575         size_t i;
576 
577         /* Explicitly reserve slot 0 for flags 0: the following logic
578          * cannot distinguish an unoccupied slot from slot (flags==0).
579          */
580         if (flags == 0)
581                 return 0;
582 
583         /* search through the remaining (HASH_SIZE-1) slots */
584         for (i = 1; i < ARRAY_SIZE(page_flags); i++, k++) {
585                 if (!k || k >= ARRAY_SIZE(page_flags))
586                         k = 1;
587                 if (page_flags[k] == 0) {
588                         page_flags[k] = flags;
589                         return k;
590                 }
591                 if (page_flags[k] == flags)
592                         return k;
593         }
594 
595         fatal("hash table full: bump up HASH_SHIFT?\n");
596         exit(EXIT_FAILURE);
597 }
598 
599 static void add_page(unsigned long voffset, unsigned long offset,
600                      uint64_t flags, uint64_t cgroup, uint64_t pme)
601 {
602         flags = kpageflags_flags(flags, pme);
603 
604         if (!bit_mask_ok(flags))
605                 return;
606 
607         if (opt_cgroup && cgroup != (uint64_t)opt_cgroup)
608                 return;
609 
610         if (opt_hwpoison)
611                 hwpoison_page(offset);
612         if (opt_unpoison)
613                 unpoison_page(offset);
614 
615         if (opt_list == 1)
616                 show_page_range(voffset, offset, 1, flags, cgroup);
617         else if (opt_list == 2)
618                 show_page(voffset, offset, flags, cgroup);
619 
620         nr_pages[hash_slot(flags)]++;
621         total_pages++;
622 }
623 
624 #define KPAGEFLAGS_BATCH        (64 << 10)      /* 64k pages */
625 static void walk_pfn(unsigned long voffset,
626                      unsigned long index,
627                      unsigned long count,
628                      uint64_t pme)
629 {
630         uint64_t buf[KPAGEFLAGS_BATCH];
631         uint64_t cgi[KPAGEFLAGS_BATCH];
632         unsigned long batch;
633         unsigned long pages;
634         unsigned long i;
635 
636         /*
637          * kpagecgroup_read() reads only if kpagecgroup were opened, but
638          * /proc/kpagecgroup might even not exist, so it's better to fill
639          * them with zeros here.
640          */
641         if (count == 1)
642                 cgi[0] = 0;
643         else
644                 memset(cgi, 0, sizeof cgi);
645 
646         while (count) {
647                 batch = min_t(unsigned long, count, KPAGEFLAGS_BATCH);
648                 pages = kpageflags_read(buf, index, batch);
649                 if (pages == 0)
650                         break;
651 
652                 if (kpagecgroup_read(cgi, index, pages) != pages)
653                         fatal("kpagecgroup returned fewer pages than expected");
654 
655                 for (i = 0; i < pages; i++)
656                         add_page(voffset + i, index + i, buf[i], cgi[i], pme);
657 
658                 index += pages;
659                 count -= pages;
660         }
661 }
662 
663 static void walk_swap(unsigned long voffset, uint64_t pme)
664 {
665         uint64_t flags = kpageflags_flags(0, pme);
666 
667         if (!bit_mask_ok(flags))
668                 return;
669 
670         if (opt_cgroup)
671                 return;
672 
673         if (opt_list == 1)
674                 show_page_range(voffset, pagemap_swap_offset(pme), 1, flags, 0);
675         else if (opt_list == 2)
676                 show_page(voffset, pagemap_swap_offset(pme), flags, 0);
677 
678         nr_pages[hash_slot(flags)]++;
679         total_pages++;
680 }
681 
682 #define PAGEMAP_BATCH   (64 << 10)
683 static void walk_vma(unsigned long index, unsigned long count)
684 {
685         uint64_t buf[PAGEMAP_BATCH];
686         unsigned long batch;
687         unsigned long pages;
688         unsigned long pfn;
689         unsigned long i;
690 
691         while (count) {
692                 batch = min_t(unsigned long, count, PAGEMAP_BATCH);
693                 pages = pagemap_read(buf, index, batch);
694                 if (pages == 0)
695                         break;
696 
697                 for (i = 0; i < pages; i++) {
698                         pfn = pagemap_pfn(buf[i]);
699                         if (pfn)
700                                 walk_pfn(index + i, pfn, 1, buf[i]);
701                         if (buf[i] & PM_SWAP)
702                                 walk_swap(index + i, buf[i]);
703                 }
704 
705                 index += pages;
706                 count -= pages;
707         }
708 }
709 
710 static void walk_task(unsigned long index, unsigned long count)
711 {
712         const unsigned long end = index + count;
713         unsigned long start;
714         int i = 0;
715 
716         while (index < end) {
717 
718                 while (pg_end[i] <= index)
719                         if (++i >= nr_vmas)
720                                 return;
721                 if (pg_start[i] >= end)
722                         return;
723 
724                 start = max_t(unsigned long, pg_start[i], index);
725                 index = min_t(unsigned long, pg_end[i], end);
726 
727                 assert(start < index);
728                 walk_vma(start, index - start);
729         }
730 }
731 
732 static void add_addr_range(unsigned long offset, unsigned long size)
733 {
734         if (nr_addr_ranges >= MAX_ADDR_RANGES)
735                 fatal("too many addr ranges\n");
736 
737         opt_offset[nr_addr_ranges] = offset;
738         opt_size[nr_addr_ranges] = min_t(unsigned long, size, ULONG_MAX-offset);
739         nr_addr_ranges++;
740 }
741 
742 static void walk_addr_ranges(void)
743 {
744         int i;
745 
746         kpageflags_fd = checked_open(PROC_KPAGEFLAGS, O_RDONLY);
747 
748         if (!nr_addr_ranges)
749                 add_addr_range(0, ULONG_MAX);
750 
751         for (i = 0; i < nr_addr_ranges; i++)
752                 if (!opt_pid)
753                         walk_pfn(opt_offset[i], opt_offset[i], opt_size[i], 0);
754                 else
755                         walk_task(opt_offset[i], opt_size[i]);
756 
757         close(kpageflags_fd);
758 }
759 
760 
761 /*
762  * user interface
763  */
764 
765 static const char *page_flag_type(uint64_t flag)
766 {
767         if (flag & KPF_HACKERS_BITS)
768                 return "(r)";
769         if (flag & KPF_OVERLOADED_BITS)
770                 return "(o)";
771         return "   ";
772 }
773 
774 static void usage(void)
775 {
776         size_t i, j;
777 
778         printf(
779 "page-types [options]\n"
780 "            -r|--raw                   Raw mode, for kernel developers\n"
781 "            -d|--describe flags        Describe flags\n"
782 "            -a|--addr    addr-spec     Walk a range of pages\n"
783 "            -b|--bits    bits-spec     Walk pages with specified bits\n"
784 "            -c|--cgroup  path|@inode   Walk pages within memory cgroup\n"
785 "            -p|--pid     pid           Walk process address space\n"
786 "            -f|--file    filename      Walk file address space\n"
787 "            -l|--list                  Show page details in ranges\n"
788 "            -L|--list-each             Show page details one by one\n"
789 "            -C|--list-cgroup           Show cgroup inode for pages\n"
790 "            -N|--no-summary            Don't show summary info\n"
791 "            -X|--hwpoison              hwpoison pages\n"
792 "            -x|--unpoison              unpoison pages\n"
793 "            -h|--help                  Show this usage message\n"
794 "flags:\n"
795 "            0x10                       bitfield format, e.g.\n"
796 "            anon                       bit-name, e.g.\n"
797 "            0x10,anon                  comma-separated list, e.g.\n"
798 "addr-spec:\n"
799 "            N                          one page at offset N (unit: pages)\n"
800 "            N+M                        pages range from N to N+M-1\n"
801 "            N,M                        pages range from N to M-1\n"
802 "            N,                         pages range from N to end\n"
803 "            ,M                         pages range from 0 to M-1\n"
804 "bits-spec:\n"
805 "            bit1,bit2                  (flags & (bit1|bit2)) != 0\n"
806 "            bit1,bit2=bit1             (flags & (bit1|bit2)) == bit1\n"
807 "            bit1,~bit2                 (flags & (bit1|bit2)) == bit1\n"
808 "            =bit1,bit2                 flags == (bit1|bit2)\n"
809 "bit-names:\n"
810         );
811 
812         for (i = 0, j = 0; i < ARRAY_SIZE(page_flag_names); i++) {
813                 if (!page_flag_names[i])
814                         continue;
815                 printf("%16s%s", page_flag_names[i] + 2,
816                                  page_flag_type(1ULL << i));
817                 if (++j > 3) {
818                         j = 0;
819                         putchar('\n');
820                 }
821         }
822         printf("\n                                   "
823                 "(r) raw mode bits  (o) overloaded bits\n");
824 }
825 
826 static unsigned long long parse_number(const char *str)
827 {
828         unsigned long long n;
829 
830         n = strtoll(str, NULL, 0);
831 
832         if (n == 0 && str[0] != '')
833                 fatal("invalid name or number: %s\n", str);
834 
835         return n;
836 }
837 
838 static void parse_pid(const char *str)
839 {
840         FILE *file;
841         char buf[5000];
842 
843         opt_pid = parse_number(str);
844 
845         sprintf(buf, "/proc/%d/pagemap", opt_pid);
846         pagemap_fd = checked_open(buf, O_RDONLY);
847 
848         sprintf(buf, "/proc/%d/maps", opt_pid);
849         file = fopen(buf, "r");
850         if (!file) {
851                 perror(buf);
852                 exit(EXIT_FAILURE);
853         }
854 
855         while (fgets(buf, sizeof(buf), file) != NULL) {
856                 unsigned long vm_start;
857                 unsigned long vm_end;
858                 unsigned long long pgoff;
859                 int major, minor;
860                 char r, w, x, s;
861                 unsigned long ino;
862                 int n;
863 
864                 n = sscanf(buf, "%lx-%lx %c%c%c%c %llx %x:%x %lu",
865                            &vm_start,
866                            &vm_end,
867                            &r, &w, &x, &s,
868                            &pgoff,
869                            &major, &minor,
870                            &ino);
871                 if (n < 10) {
872                         fprintf(stderr, "unexpected line: %s\n", buf);
873                         continue;
874                 }
875                 pg_start[nr_vmas] = vm_start / page_size;
876                 pg_end[nr_vmas] = vm_end / page_size;
877                 if (++nr_vmas >= MAX_VMAS) {
878                         fprintf(stderr, "too many VMAs\n");
879                         break;
880                 }
881         }
882         fclose(file);
883 }
884 
885 static void show_file(const char *name, const struct stat *st)
886 {
887         unsigned long long size = st->st_size;
888         char atime[64], mtime[64];
889         long now = time(NULL);
890 
891         printf("%s\tInode: %u\tSize: %llu (%llu pages)\n",
892                         name, (unsigned)st->st_ino,
893                         size, (size + page_size - 1) / page_size);
894 
895         strftime(atime, sizeof(atime), "%c", localtime(&st->st_atime));
896         strftime(mtime, sizeof(mtime), "%c", localtime(&st->st_mtime));
897 
898         printf("Modify: %s (%ld seconds ago)\nAccess: %s (%ld seconds ago)\n",
899                         mtime, now - st->st_mtime,
900                         atime, now - st->st_atime);
901 }
902 
903 static sigjmp_buf sigbus_jmp;
904 
905 static void * volatile sigbus_addr;
906 
907 static void sigbus_handler(int sig, siginfo_t *info, void *ucontex)
908 {
909         (void)sig;
910         (void)ucontex;
911         sigbus_addr = info ? info->si_addr : NULL;
912         siglongjmp(sigbus_jmp, 1);
913 }
914 
915 static struct sigaction sigbus_action = {
916         .sa_sigaction = sigbus_handler,
917         .sa_flags = SA_SIGINFO,
918 };
919 
920 static void walk_file(const char *name, const struct stat *st)
921 {
922         uint8_t vec[PAGEMAP_BATCH];
923         uint64_t buf[PAGEMAP_BATCH], flags;
924         uint64_t cgroup = 0;
925         unsigned long nr_pages, pfn, i;
926         off_t off, end = st->st_size;
927         int fd;
928         ssize_t len;
929         void *ptr;
930         int first = 1;
931 
932         fd = checked_open(name, O_RDONLY|O_NOATIME|O_NOFOLLOW);
933 
934         for (off = 0; off < end; off += len) {
935                 nr_pages = (end - off + page_size - 1) / page_size;
936                 if (nr_pages > PAGEMAP_BATCH)
937                         nr_pages = PAGEMAP_BATCH;
938                 len = nr_pages * page_size;
939 
940                 ptr = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, off);
941                 if (ptr == MAP_FAILED)
942                         fatal("mmap failed: %s", name);
943 
944                 /* determine cached pages */
945                 if (mincore(ptr, len, vec))
946                         fatal("mincore failed: %s", name);
947 
948                 /* turn off readahead */
949                 if (madvise(ptr, len, MADV_RANDOM))
950                         fatal("madvice failed: %s", name);
951 
952                 if (sigsetjmp(sigbus_jmp, 1)) {
953                         end = off + sigbus_addr ? sigbus_addr - ptr : 0;
954                         fprintf(stderr, "got sigbus at offset %lld: %s\n",
955                                         (long long)end, name);
956                         goto got_sigbus;
957                 }
958 
959                 /* populate ptes */
960                 for (i = 0; i < nr_pages ; i++) {
961                         if (vec[i] & 1)
962                                 (void)*(volatile int *)(ptr + i * page_size);
963                 }
964 got_sigbus:
965 
966                 /* turn off harvesting reference bits */
967                 if (madvise(ptr, len, MADV_SEQUENTIAL))
968                         fatal("madvice failed: %s", name);
969 
970                 if (pagemap_read(buf, (unsigned long)ptr / page_size,
971                                         nr_pages) != nr_pages)
972                         fatal("cannot read pagemap");
973 
974                 munmap(ptr, len);
975 
976                 for (i = 0; i < nr_pages; i++) {
977                         pfn = pagemap_pfn(buf[i]);
978                         if (!pfn)
979                                 continue;
980                         if (!kpageflags_read(&flags, pfn, 1))
981                                 continue;
982                         if (!kpagecgroup_read(&cgroup, pfn, 1))
983                                 fatal("kpagecgroup_read failed");
984                         if (first && opt_list) {
985                                 first = 0;
986                                 flush_page_range();
987                                 show_file(name, st);
988                         }
989                         add_page(off / page_size + i, pfn,
990                                  flags, cgroup, buf[i]);
991                 }
992         }
993 
994         close(fd);
995 }
996 
997 int walk_tree(const char *name, const struct stat *st, int type, struct FTW *f)
998 {
999         (void)f;
1000         switch (type) {
1001         case FTW_F:
1002                 if (S_ISREG(st->st_mode))
1003                         walk_file(name, st);
1004                 break;
1005         case FTW_DNR:
1006                 fprintf(stderr, "cannot read dir: %s\n", name);
1007                 break;
1008         }
1009         return 0;
1010 }
1011 
1012 static void walk_page_cache(void)
1013 {
1014         struct stat st;
1015 
1016         kpageflags_fd = checked_open(PROC_KPAGEFLAGS, O_RDONLY);
1017         pagemap_fd = checked_open("/proc/self/pagemap", O_RDONLY);
1018         sigaction(SIGBUS, &sigbus_action, NULL);
1019 
1020         if (stat(opt_file, &st))
1021                 fatal("stat failed: %s\n", opt_file);
1022 
1023         if (S_ISREG(st.st_mode)) {
1024                 walk_file(opt_file, &st);
1025         } else if (S_ISDIR(st.st_mode)) {
1026                 /* do not follow symlinks and mountpoints */
1027                 if (nftw(opt_file, walk_tree, 64, FTW_MOUNT | FTW_PHYS) < 0)
1028                         fatal("nftw failed: %s\n", opt_file);
1029         } else
1030                 fatal("unhandled file type: %s\n", opt_file);
1031 
1032         close(kpageflags_fd);
1033         close(pagemap_fd);
1034         signal(SIGBUS, SIG_DFL);
1035 }
1036 
1037 static void parse_file(const char *name)
1038 {
1039         opt_file = name;
1040 }
1041 
1042 static void parse_cgroup(const char *path)
1043 {
1044         if (path[0] == '@') {
1045                 opt_cgroup = parse_number(path + 1);
1046                 return;
1047         }
1048 
1049         struct stat st;
1050 
1051         if (stat(path, &st))
1052                 fatal("stat failed: %s: %m\n", path);
1053 
1054         if (!S_ISDIR(st.st_mode))
1055                 fatal("cgroup supposed to be a directory: %s\n", path);
1056 
1057         opt_cgroup = st.st_ino;
1058 }
1059 
1060 static void parse_addr_range(const char *optarg)
1061 {
1062         unsigned long offset;
1063         unsigned long size;
1064         char *p;
1065 
1066         p = strchr(optarg, ',');
1067         if (!p)
1068                 p = strchr(optarg, '+');
1069 
1070         if (p == optarg) {
1071                 offset = 0;
1072                 size   = parse_number(p + 1);
1073         } else if (p) {
1074                 offset = parse_number(optarg);
1075                 if (p[1] == '\0')
1076                         size = ULONG_MAX;
1077                 else {
1078                         size = parse_number(p + 1);
1079                         if (*p == ',') {
1080                                 if (size < offset)
1081                                         fatal("invalid range: %lu,%lu\n",
1082                                                         offset, size);
1083                                 size -= offset;
1084                         }
1085                 }
1086         } else {
1087                 offset = parse_number(optarg);
1088                 size   = 1;
1089         }
1090 
1091         add_addr_range(offset, size);
1092 }
1093 
1094 static void add_bits_filter(uint64_t mask, uint64_t bits)
1095 {
1096         if (nr_bit_filters >= MAX_BIT_FILTERS)
1097                 fatal("too much bit filters\n");
1098 
1099         opt_mask[nr_bit_filters] = mask;
1100         opt_bits[nr_bit_filters] = bits;
1101         nr_bit_filters++;
1102 }
1103 
1104 static uint64_t parse_flag_name(const char *str, int len)
1105 {
1106         size_t i;
1107 
1108         if (!*str || !len)
1109                 return 0;
1110 
1111         if (len <= 8 && !strncmp(str, "compound", len))
1112                 return BITS_COMPOUND;
1113 
1114         for (i = 0; i < ARRAY_SIZE(page_flag_names); i++) {
1115                 if (!page_flag_names[i])
1116                         continue;
1117                 if (!strncmp(str, page_flag_names[i] + 2, len))
1118                         return 1ULL << i;
1119         }
1120 
1121         return parse_number(str);
1122 }
1123 
1124 static uint64_t parse_flag_names(const char *str, int all)
1125 {
1126         const char *p    = str;
1127         uint64_t   flags = 0;
1128 
1129         while (1) {
1130                 if (*p == ',' || *p == '=' || *p == '\0') {
1131                         if ((*str != '~') || (*str == '~' && all && *++str))
1132                                 flags |= parse_flag_name(str, p - str);
1133                         if (*p != ',')
1134                                 break;
1135                         str = p + 1;
1136                 }
1137                 p++;
1138         }
1139 
1140         return flags;
1141 }
1142 
1143 static void parse_bits_mask(const char *optarg)
1144 {
1145         uint64_t mask;
1146         uint64_t bits;
1147         const char *p;
1148 
1149         p = strchr(optarg, '=');
1150         if (p == optarg) {
1151                 mask = KPF_ALL_BITS;
1152                 bits = parse_flag_names(p + 1, 0);
1153         } else if (p) {
1154                 mask = parse_flag_names(optarg, 0);
1155                 bits = parse_flag_names(p + 1, 0);
1156         } else if (strchr(optarg, '~')) {
1157                 mask = parse_flag_names(optarg, 1);
1158                 bits = parse_flag_names(optarg, 0);
1159         } else {
1160                 mask = parse_flag_names(optarg, 0);
1161                 bits = KPF_ALL_BITS;
1162         }
1163 
1164         add_bits_filter(mask, bits);
1165 }
1166 
1167 static void describe_flags(const char *optarg)
1168 {
1169         uint64_t flags = parse_flag_names(optarg, 0);
1170 
1171         printf("0x%016llx\t%s\t%s\n",
1172                 (unsigned long long)flags,
1173                 page_flag_name(flags),
1174                 page_flag_longname(flags));
1175 }
1176 
1177 static const struct option opts[] = {
1178         { "raw"       , 0, NULL, 'r' },
1179         { "pid"       , 1, NULL, 'p' },
1180         { "file"      , 1, NULL, 'f' },
1181         { "addr"      , 1, NULL, 'a' },
1182         { "bits"      , 1, NULL, 'b' },
1183         { "cgroup"    , 1, NULL, 'c' },
1184         { "describe"  , 1, NULL, 'd' },
1185         { "list"      , 0, NULL, 'l' },
1186         { "list-each" , 0, NULL, 'L' },
1187         { "list-cgroup", 0, NULL, 'C' },
1188         { "no-summary", 0, NULL, 'N' },
1189         { "hwpoison"  , 0, NULL, 'X' },
1190         { "unpoison"  , 0, NULL, 'x' },
1191         { "help"      , 0, NULL, 'h' },
1192         { NULL        , 0, NULL, 0 }
1193 };
1194 
1195 int main(int argc, char *argv[])
1196 {
1197         int c;
1198 
1199         page_size = getpagesize();
1200 
1201         while ((c = getopt_long(argc, argv,
1202                                 "rp:f:a:b:d:c:ClLNXxh", opts, NULL)) != -1) {
1203                 switch (c) {
1204                 case 'r':
1205                         opt_raw = 1;
1206                         break;
1207                 case 'p':
1208                         parse_pid(optarg);
1209                         break;
1210                 case 'f':
1211                         parse_file(optarg);
1212                         break;
1213                 case 'a':
1214                         parse_addr_range(optarg);
1215                         break;
1216                 case 'b':
1217                         parse_bits_mask(optarg);
1218                         break;
1219                 case 'c':
1220                         parse_cgroup(optarg);
1221                         break;
1222                 case 'C':
1223                         opt_list_cgroup = 1;
1224                         break;
1225                 case 'd':
1226                         describe_flags(optarg);
1227                         exit(0);
1228                 case 'l':
1229                         opt_list = 1;
1230                         break;
1231                 case 'L':
1232                         opt_list = 2;
1233                         break;
1234                 case 'N':
1235                         opt_no_summary = 1;
1236                         break;
1237                 case 'X':
1238                         opt_hwpoison = 1;
1239                         prepare_hwpoison_fd();
1240                         break;
1241                 case 'x':
1242                         opt_unpoison = 1;
1243                         prepare_hwpoison_fd();
1244                         break;
1245                 case 'h':
1246                         usage();
1247                         exit(0);
1248                 default:
1249                         usage();
1250                         exit(1);
1251                 }
1252         }
1253 
1254         if (opt_cgroup || opt_list_cgroup)
1255                 kpagecgroup_fd = checked_open(PROC_KPAGECGROUP, O_RDONLY);
1256 
1257         if (opt_list && opt_pid)
1258                 printf("voffset\t");
1259         if (opt_list && opt_file)
1260                 printf("foffset\t");
1261         if (opt_list && opt_list_cgroup)
1262                 printf("cgroup\t");
1263         if (opt_list == 1)
1264                 printf("offset\tlen\tflags\n");
1265         if (opt_list == 2)
1266                 printf("offset\tflags\n");
1267 
1268         if (opt_file)
1269                 walk_page_cache();
1270         else
1271                 walk_addr_ranges();
1272 
1273         if (opt_list == 1)
1274                 flush_page_range();
1275 
1276         if (opt_no_summary)
1277                 return 0;
1278 
1279         if (opt_list)
1280                 printf("\n\n");
1281 
1282         show_summary();
1283 
1284         return 0;
1285 }
1286 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp