1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 /* 2 /*
3 * mm/debug.c 3 * mm/debug.c
4 * 4 *
5 * mm/ specific debug routines. 5 * mm/ specific debug routines.
6 * 6 *
7 */ 7 */
8 8
9 #include <linux/kernel.h> 9 #include <linux/kernel.h>
10 #include <linux/mm.h> 10 #include <linux/mm.h>
11 #include <linux/trace_events.h> 11 #include <linux/trace_events.h>
12 #include <linux/memcontrol.h> 12 #include <linux/memcontrol.h>
13 #include <trace/events/mmflags.h> 13 #include <trace/events/mmflags.h>
14 #include <linux/migrate.h> 14 #include <linux/migrate.h>
15 #include <linux/page_owner.h> 15 #include <linux/page_owner.h>
16 #include <linux/ctype.h> 16 #include <linux/ctype.h>
17 17
18 #include "internal.h" 18 #include "internal.h"
19 #include <trace/events/migrate.h> <<
20 <<
21 /* <<
22 * Define EM() and EMe() so that MIGRATE_REASO <<
23 * be used to populate migrate_reason_names[]. <<
24 */ <<
25 #undef EM <<
26 #undef EMe <<
27 #define EM(a, b) b, <<
28 #define EMe(a, b) b <<
29 19
30 const char *migrate_reason_names[MR_TYPES] = { 20 const char *migrate_reason_names[MR_TYPES] = {
31 MIGRATE_REASON !! 21 "compaction",
>> 22 "memory_failure",
>> 23 "memory_hotplug",
>> 24 "syscall_or_cpuset",
>> 25 "mempolicy_mbind",
>> 26 "numa_misplaced",
>> 27 "cma",
32 }; 28 };
33 29
34 const struct trace_print_flags pageflag_names[ 30 const struct trace_print_flags pageflag_names[] = {
35 __def_pageflag_names, 31 __def_pageflag_names,
36 {0, NULL} 32 {0, NULL}
37 }; 33 };
38 34
39 const struct trace_print_flags gfpflag_names[] 35 const struct trace_print_flags gfpflag_names[] = {
40 __def_gfpflag_names, 36 __def_gfpflag_names,
41 {0, NULL} 37 {0, NULL}
42 }; 38 };
43 39
44 const struct trace_print_flags vmaflag_names[] 40 const struct trace_print_flags vmaflag_names[] = {
45 __def_vmaflag_names, 41 __def_vmaflag_names,
46 {0, NULL} 42 {0, NULL}
47 }; 43 };
48 44
49 #define DEF_PAGETYPE_NAME(_name) [PGTY_##_name !! 45 void __dump_page(struct page *page, const char *reason)
50 <<
51 static const char *page_type_names[] = { <<
52 DEF_PAGETYPE_NAME(slab), <<
53 DEF_PAGETYPE_NAME(hugetlb), <<
54 DEF_PAGETYPE_NAME(offline), <<
55 DEF_PAGETYPE_NAME(guard), <<
56 DEF_PAGETYPE_NAME(table), <<
57 DEF_PAGETYPE_NAME(buddy), <<
58 DEF_PAGETYPE_NAME(unaccepted), <<
59 }; <<
60 <<
61 static const char *page_type_name(unsigned int <<
62 { 46 {
63 unsigned i = (page_type >> 24) - 0xf0; !! 47 struct page *head = compound_head(page);
>> 48 struct address_space *mapping;
>> 49 bool page_poisoned = PagePoisoned(page);
>> 50 bool compound = PageCompound(page);
>> 51 /*
>> 52 * Accessing the pageblock without the zone lock. It could change to
>> 53 * "isolate" again in the meantime, but since we are just dumping the
>> 54 * state for debugging, it should be fine to accept a bit of
>> 55 * inaccuracy here due to racing.
>> 56 */
>> 57 bool page_cma = is_migrate_cma_page(page);
>> 58 int mapcount;
>> 59 char *type = "";
64 60
65 if (i >= ARRAY_SIZE(page_type_names)) !! 61 /*
66 return "unknown"; !! 62 * If struct page is poisoned don't access Page*() functions as that
67 return page_type_names[i]; !! 63 * leads to recursive loop. Page*() check for poisoned pages, and calls
68 } !! 64 * dump_page() when detected.
>> 65 */
>> 66 if (page_poisoned) {
>> 67 pr_warn("page:%px is uninitialized and poisoned", page);
>> 68 goto hex_only;
>> 69 }
69 70
70 static void __dump_folio(struct folio *folio, !! 71 if (page < head || (page >= head + MAX_ORDER_NR_PAGES)) {
71 unsigned long pfn, unsigned lo !! 72 /*
72 { !! 73 * Corrupt page, so we cannot call page_mapping. Instead, do a
73 struct address_space *mapping = folio_ !! 74 * safe subset of the steps that page_mapping() does. Caution:
74 int mapcount = atomic_read(&page->_map !! 75 * this will be misleading for tail pages, PageSwapCache pages,
75 char *type = ""; !! 76 * and potentially other situations. (See the page_mapping()
>> 77 * implementation for what's missing here.)
>> 78 */
>> 79 unsigned long tmp = (unsigned long)page->mapping;
>> 80
>> 81 if (tmp & PAGE_MAPPING_ANON)
>> 82 mapping = NULL;
>> 83 else
>> 84 mapping = (void *)(tmp & ~PAGE_MAPPING_FLAGS);
>> 85 head = page;
>> 86 compound = false;
>> 87 } else {
>> 88 mapping = page_mapping(page);
>> 89 }
76 90
77 mapcount = page_mapcount_is_type(mapco !! 91 /*
78 pr_warn("page: refcount:%d mapcount:%d !! 92 * Avoid VM_BUG_ON() in page_mapcount().
79 folio_ref_count(folio) !! 93 * page->_mapcount space in struct page is used by sl[aou]b pages to
80 folio->index + idx, pf !! 94 * encode own info.
81 if (folio_test_large(folio)) { !! 95 */
82 pr_warn("head: order:%u mapcou !! 96 mapcount = PageSlab(head) ? 0 : page_mapcount(page);
83 folio_order(fo !! 97
84 folio_mapcount !! 98 pr_warn("page:%p refcount:%d mapcount:%d mapping:%p index:%#lx pfn:%#lx\n",
85 folio_entire_m !! 99 page, page_ref_count(head), mapcount, mapping,
86 folio_nr_pages !! 100 page_to_pgoff(page), page_to_pfn(page));
87 atomic_read(&f !! 101 if (compound) {
>> 102 if (hpage_pincount_available(page)) {
>> 103 pr_warn("head:%p order:%u compound_mapcount:%d compound_pincount:%d\n",
>> 104 head, compound_order(head),
>> 105 head_compound_mapcount(head),
>> 106 head_compound_pincount(head));
>> 107 } else {
>> 108 pr_warn("head:%p order:%u compound_mapcount:%d\n",
>> 109 head, compound_order(head),
>> 110 head_compound_mapcount(head));
>> 111 }
88 } 112 }
89 113
90 #ifdef CONFIG_MEMCG 114 #ifdef CONFIG_MEMCG
91 if (folio->memcg_data) !! 115 if (head->memcg_data)
92 pr_warn("memcg:%lx\n", folio-> !! 116 pr_warn("memcg:%lx\n", head->memcg_data);
93 #endif 117 #endif
94 if (folio_test_ksm(folio)) !! 118 if (PageKsm(page))
95 type = "ksm "; 119 type = "ksm ";
96 else if (folio_test_anon(folio)) !! 120 else if (PageAnon(page))
97 type = "anon "; 121 type = "anon ";
98 else if (mapping) !! 122 else if (mapping) {
99 dump_mapping(mapping); !! 123 struct inode *host;
>> 124 const struct address_space_operations *a_ops;
>> 125 struct hlist_node *dentry_first;
>> 126 struct dentry *dentry_ptr;
>> 127 struct dentry dentry;
>> 128 unsigned long ino;
>> 129
>> 130 /*
>> 131 * mapping can be invalid pointer and we don't want to crash
>> 132 * accessing it, so probe everything depending on it carefully
>> 133 */
>> 134 if (get_kernel_nofault(host, &mapping->host) ||
>> 135 get_kernel_nofault(a_ops, &mapping->a_ops)) {
>> 136 pr_warn("failed to read mapping contents, not a valid kernel address?\n");
>> 137 goto out_mapping;
>> 138 }
>> 139
>> 140 if (!host) {
>> 141 pr_warn("aops:%ps\n", a_ops);
>> 142 goto out_mapping;
>> 143 }
>> 144
>> 145 if (get_kernel_nofault(dentry_first, &host->i_dentry.first) ||
>> 146 get_kernel_nofault(ino, &host->i_ino)) {
>> 147 pr_warn("aops:%ps with invalid host inode %px\n",
>> 148 a_ops, host);
>> 149 goto out_mapping;
>> 150 }
>> 151
>> 152 if (!dentry_first) {
>> 153 pr_warn("aops:%ps ino:%lx\n", a_ops, ino);
>> 154 goto out_mapping;
>> 155 }
>> 156
>> 157 dentry_ptr = container_of(dentry_first, struct dentry, d_u.d_alias);
>> 158 if (get_kernel_nofault(dentry, dentry_ptr)) {
>> 159 pr_warn("aops:%ps ino:%lx with invalid dentry %px\n",
>> 160 a_ops, ino, dentry_ptr);
>> 161 } else {
>> 162 /*
>> 163 * if dentry is corrupted, the %pd handler may still
>> 164 * crash, but it's unlikely that we reach here with a
>> 165 * corrupted struct page
>> 166 */
>> 167 pr_warn("aops:%ps ino:%lx dentry name:\"%pd\"\n",
>> 168 a_ops, ino, &dentry);
>> 169 }
>> 170 }
>> 171 out_mapping:
100 BUILD_BUG_ON(ARRAY_SIZE(pageflag_names 172 BUILD_BUG_ON(ARRAY_SIZE(pageflag_names) != __NR_PAGEFLAGS + 1);
101 173
102 /* !! 174 pr_warn("%sflags: %#lx(%pGp)%s\n", type, head->flags, &head->flags,
103 * Accessing the pageblock without the !! 175 page_cma ? " CMA" : "");
104 * "isolate" again in the meantime, bu <<
105 * state for debugging, it should be f <<
106 * inaccuracy here due to racing. <<
107 */ <<
108 pr_warn("%sflags: %pGp%s\n", type, &fo <<
109 is_migrate_cma_folio(folio, pf <<
110 if (page_has_type(&folio->page)) <<
111 pr_warn("page_type: %x(%s)\n", <<
112 page_type_name <<
113 176
>> 177 hex_only:
114 print_hex_dump(KERN_WARNING, "raw: ", 178 print_hex_dump(KERN_WARNING, "raw: ", DUMP_PREFIX_NONE, 32,
115 sizeof(unsigned long), 179 sizeof(unsigned long), page,
116 sizeof(struct page), f 180 sizeof(struct page), false);
117 if (folio_test_large(folio)) !! 181 if (head != page)
118 print_hex_dump(KERN_WARNING, " 182 print_hex_dump(KERN_WARNING, "head: ", DUMP_PREFIX_NONE, 32,
119 sizeof(unsigned long), !! 183 sizeof(unsigned long), head,
120 2 * sizeof(struct page !! 184 sizeof(struct page), false);
121 } <<
122 <<
123 static void __dump_page(const struct page *pag <<
124 { <<
125 struct folio *foliop, folio; <<
126 struct page precise; <<
127 unsigned long pfn = page_to_pfn(page); <<
128 unsigned long idx, nr_pages = 1; <<
129 int loops = 5; <<
130 <<
131 again: <<
132 memcpy(&precise, page, sizeof(*page)); <<
133 foliop = page_folio(&precise); <<
134 if (foliop == (struct folio *)&precise <<
135 idx = 0; <<
136 if (!folio_test_large(foliop)) <<
137 goto dump; <<
138 foliop = (struct folio *)page; <<
139 } else { <<
140 idx = folio_page_idx(foliop, p <<
141 } <<
142 <<
143 if (idx < MAX_FOLIO_NR_PAGES) { <<
144 memcpy(&folio, foliop, 2 * siz <<
145 nr_pages = folio_nr_pages(&fol <<
146 foliop = &folio; <<
147 } <<
148 <<
149 if (idx > nr_pages) { <<
150 if (loops-- > 0) <<
151 goto again; <<
152 pr_warn("page does not match f <<
153 precise.compound_head &= ~1UL; <<
154 foliop = (struct folio *)&prec <<
155 idx = 0; <<
156 } <<
157 185
158 dump: !! 186 if (reason)
159 __dump_folio(foliop, &precise, pfn, id !! 187 pr_warn("page dumped because: %s\n", reason);
160 } 188 }
161 189
162 void dump_page(const struct page *page, const !! 190 void dump_page(struct page *page, const char *reason)
163 { 191 {
164 if (PagePoisoned(page)) !! 192 __dump_page(page, reason);
165 pr_warn("page:%p is uninitiali <<
166 else <<
167 __dump_page(page); <<
168 if (reason) <<
169 pr_warn("page dumped because: <<
170 dump_page_owner(page); 193 dump_page_owner(page);
171 } 194 }
172 EXPORT_SYMBOL(dump_page); 195 EXPORT_SYMBOL(dump_page);
173 196
174 #ifdef CONFIG_DEBUG_VM 197 #ifdef CONFIG_DEBUG_VM
175 198
176 void dump_vma(const struct vm_area_struct *vma 199 void dump_vma(const struct vm_area_struct *vma)
177 { 200 {
178 pr_emerg("vma %px start %px end %px mm !! 201 pr_emerg("vma %px start %px end %px\n"
>> 202 "next %px prev %px mm %px\n"
179 "prot %lx anon_vma %px vm_ops 203 "prot %lx anon_vma %px vm_ops %px\n"
180 "pgoff %lx file %px private_da 204 "pgoff %lx file %px private_data %px\n"
181 "flags: %#lx(%pGv)\n", 205 "flags: %#lx(%pGv)\n",
182 vma, (void *)vma->vm_start, (v !! 206 vma, (void *)vma->vm_start, (void *)vma->vm_end, vma->vm_next,
>> 207 vma->vm_prev, vma->vm_mm,
183 (unsigned long)pgprot_val(vma- 208 (unsigned long)pgprot_val(vma->vm_page_prot),
184 vma->anon_vma, vma->vm_ops, vm 209 vma->anon_vma, vma->vm_ops, vma->vm_pgoff,
185 vma->vm_file, vma->vm_private_ 210 vma->vm_file, vma->vm_private_data,
186 vma->vm_flags, &vma->vm_flags) 211 vma->vm_flags, &vma->vm_flags);
187 } 212 }
188 EXPORT_SYMBOL(dump_vma); 213 EXPORT_SYMBOL(dump_vma);
189 214
190 void dump_mm(const struct mm_struct *mm) 215 void dump_mm(const struct mm_struct *mm)
191 { 216 {
192 pr_emerg("mm %px task_size %lu\n" !! 217 pr_emerg("mm %px mmap %px seqnum %llu task_size %lu\n"
193 "mmap_base %lu mmap_legacy_bas !! 218 #ifdef CONFIG_MMU
>> 219 "get_unmapped_area %px\n"
>> 220 #endif
>> 221 "mmap_base %lu mmap_legacy_base %lu highest_vm_end %lu\n"
194 "pgd %px mm_users %d mm_count 222 "pgd %px mm_users %d mm_count %d pgtables_bytes %lu map_count %d\n"
195 "hiwater_rss %lx hiwater_vm %l 223 "hiwater_rss %lx hiwater_vm %lx total_vm %lx locked_vm %lx\n"
196 "pinned_vm %llx data_vm %lx ex 224 "pinned_vm %llx data_vm %lx exec_vm %lx stack_vm %lx\n"
197 "start_code %lx end_code %lx s 225 "start_code %lx end_code %lx start_data %lx end_data %lx\n"
198 "start_brk %lx brk %lx start_s 226 "start_brk %lx brk %lx start_stack %lx\n"
199 "arg_start %lx arg_end %lx env 227 "arg_start %lx arg_end %lx env_start %lx env_end %lx\n"
200 "binfmt %px flags %lx\n" !! 228 "binfmt %px flags %lx core_state %px\n"
201 #ifdef CONFIG_AIO 229 #ifdef CONFIG_AIO
202 "ioctx_table %px\n" 230 "ioctx_table %px\n"
203 #endif 231 #endif
204 #ifdef CONFIG_MEMCG 232 #ifdef CONFIG_MEMCG
205 "owner %px " 233 "owner %px "
206 #endif 234 #endif
207 "exe_file %px\n" 235 "exe_file %px\n"
208 #ifdef CONFIG_MMU_NOTIFIER 236 #ifdef CONFIG_MMU_NOTIFIER
209 "notifier_subscriptions %px\n" 237 "notifier_subscriptions %px\n"
210 #endif 238 #endif
211 #ifdef CONFIG_NUMA_BALANCING 239 #ifdef CONFIG_NUMA_BALANCING
212 "numa_next_scan %lu numa_scan_ 240 "numa_next_scan %lu numa_scan_offset %lu numa_scan_seq %d\n"
213 #endif 241 #endif
214 "tlb_flush_pending %d\n" 242 "tlb_flush_pending %d\n"
215 "def_flags: %#lx(%pGv)\n", 243 "def_flags: %#lx(%pGv)\n",
216 244
217 mm, mm->task_size, !! 245 mm, mm->mmap, (long long) mm->vmacache_seqnum, mm->task_size,
218 mm->mmap_base, mm->mmap_legacy !! 246 #ifdef CONFIG_MMU
>> 247 mm->get_unmapped_area,
>> 248 #endif
>> 249 mm->mmap_base, mm->mmap_legacy_base, mm->highest_vm_end,
219 mm->pgd, atomic_read(&mm->mm_u 250 mm->pgd, atomic_read(&mm->mm_users),
220 atomic_read(&mm->mm_count), 251 atomic_read(&mm->mm_count),
221 mm_pgtables_bytes(mm), 252 mm_pgtables_bytes(mm),
222 mm->map_count, 253 mm->map_count,
223 mm->hiwater_rss, mm->hiwater_v 254 mm->hiwater_rss, mm->hiwater_vm, mm->total_vm, mm->locked_vm,
224 (u64)atomic64_read(&mm->pinned 255 (u64)atomic64_read(&mm->pinned_vm),
225 mm->data_vm, mm->exec_vm, mm-> 256 mm->data_vm, mm->exec_vm, mm->stack_vm,
226 mm->start_code, mm->end_code, 257 mm->start_code, mm->end_code, mm->start_data, mm->end_data,
227 mm->start_brk, mm->brk, mm->st 258 mm->start_brk, mm->brk, mm->start_stack,
228 mm->arg_start, mm->arg_end, mm 259 mm->arg_start, mm->arg_end, mm->env_start, mm->env_end,
229 mm->binfmt, mm->flags, !! 260 mm->binfmt, mm->flags, mm->core_state,
230 #ifdef CONFIG_AIO 261 #ifdef CONFIG_AIO
231 mm->ioctx_table, 262 mm->ioctx_table,
232 #endif 263 #endif
233 #ifdef CONFIG_MEMCG 264 #ifdef CONFIG_MEMCG
234 mm->owner, 265 mm->owner,
235 #endif 266 #endif
236 mm->exe_file, 267 mm->exe_file,
237 #ifdef CONFIG_MMU_NOTIFIER 268 #ifdef CONFIG_MMU_NOTIFIER
238 mm->notifier_subscriptions, 269 mm->notifier_subscriptions,
239 #endif 270 #endif
240 #ifdef CONFIG_NUMA_BALANCING 271 #ifdef CONFIG_NUMA_BALANCING
241 mm->numa_next_scan, mm->numa_s 272 mm->numa_next_scan, mm->numa_scan_offset, mm->numa_scan_seq,
242 #endif 273 #endif
243 atomic_read(&mm->tlb_flush_pen 274 atomic_read(&mm->tlb_flush_pending),
244 mm->def_flags, &mm->def_flags 275 mm->def_flags, &mm->def_flags
245 ); 276 );
246 } 277 }
247 EXPORT_SYMBOL(dump_mm); <<
248 278
249 static bool page_init_poisoning __read_mostly 279 static bool page_init_poisoning __read_mostly = true;
250 280
251 static int __init setup_vm_debug(char *str) 281 static int __init setup_vm_debug(char *str)
252 { 282 {
253 bool __page_init_poisoning = true; 283 bool __page_init_poisoning = true;
254 284
255 /* 285 /*
256 * Calling vm_debug with no arguments 286 * Calling vm_debug with no arguments is equivalent to requesting
257 * to enable all debugging options we 287 * to enable all debugging options we can control.
258 */ 288 */
259 if (*str++ != '=' || !*str) 289 if (*str++ != '=' || !*str)
260 goto out; 290 goto out;
261 291
262 __page_init_poisoning = false; 292 __page_init_poisoning = false;
263 if (*str == '-') 293 if (*str == '-')
264 goto out; 294 goto out;
265 295
266 while (*str) { 296 while (*str) {
267 switch (tolower(*str)) { 297 switch (tolower(*str)) {
268 case'p': 298 case'p':
269 __page_init_poisoning 299 __page_init_poisoning = true;
270 break; 300 break;
271 default: 301 default:
272 pr_err("vm_debug optio 302 pr_err("vm_debug option '%c' unknown. skipped\n",
273 *str); 303 *str);
274 } 304 }
275 305
276 str++; 306 str++;
277 } 307 }
278 out: 308 out:
279 if (page_init_poisoning && !__page_ini 309 if (page_init_poisoning && !__page_init_poisoning)
280 pr_warn("Page struct poisoning 310 pr_warn("Page struct poisoning disabled by kernel command line option 'vm_debug'\n");
281 311
282 page_init_poisoning = __page_init_pois 312 page_init_poisoning = __page_init_poisoning;
283 313
284 return 1; 314 return 1;
285 } 315 }
286 __setup("vm_debug", setup_vm_debug); 316 __setup("vm_debug", setup_vm_debug);
287 317
288 void page_init_poison(struct page *page, size_ 318 void page_init_poison(struct page *page, size_t size)
289 { 319 {
290 if (page_init_poisoning) 320 if (page_init_poisoning)
291 memset(page, PAGE_POISON_PATTE 321 memset(page, PAGE_POISON_PATTERN, size);
292 } 322 }
293 !! 323 EXPORT_SYMBOL_GPL(page_init_poison);
294 void vma_iter_dump_tree(const struct vma_itera <<
295 { <<
296 #if defined(CONFIG_DEBUG_VM_MAPLE_TREE) <<
297 mas_dump(&vmi->mas); <<
298 mt_dump(vmi->mas.tree, mt_dump_hex); <<
299 #endif /* CONFIG_DEBUG_VM_MAPLE_TREE */ <<
300 } <<
301 <<
302 #endif /* CONFIG_DEBUG_VM */ 324 #endif /* CONFIG_DEBUG_VM */
303 325
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.