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