1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * DAMON Primitives for The Physical Address Space
4 *
5 * Author: SeongJae Park <sj@kernel.org>
6 */
7
8 #define pr_fmt(fmt) "damon-pa: " fmt
9
10 #include <linux/mmu_notifier.h>
11 #include <linux/page_idle.h>
12 #include <linux/pagemap.h>
13 #include <linux/rmap.h>
14 #include <linux/swap.h>
15
16 #include "../internal.h"
17 #include "ops-common.h"
18
__damon_pa_mkold(struct folio * folio,struct vm_area_struct * vma,unsigned long addr,void * arg)19 static bool __damon_pa_mkold(struct folio *folio, struct vm_area_struct *vma,
20 unsigned long addr, void *arg)
21 {
22 DEFINE_FOLIO_VMA_WALK(pvmw, folio, vma, addr, 0);
23
24 while (page_vma_mapped_walk(&pvmw)) {
25 addr = pvmw.address;
26 if (pvmw.pte)
27 damon_ptep_mkold(pvmw.pte, vma->vm_mm, addr);
28 else
29 damon_pmdp_mkold(pvmw.pmd, vma->vm_mm, addr);
30 }
31 return true;
32 }
33
damon_pa_mkold(unsigned long paddr)34 static void damon_pa_mkold(unsigned long paddr)
35 {
36 struct folio *folio = damon_get_folio(PHYS_PFN(paddr));
37 struct rmap_walk_control rwc = {
38 .rmap_one = __damon_pa_mkold,
39 .anon_lock = folio_lock_anon_vma_read,
40 };
41 bool need_lock;
42
43 if (!folio)
44 return;
45
46 if (!folio_mapped(folio) || !folio_raw_mapping(folio)) {
47 folio_set_idle(folio);
48 goto out;
49 }
50
51 need_lock = !folio_test_anon(folio) || folio_test_ksm(folio);
52 if (need_lock && !folio_trylock(folio))
53 goto out;
54
55 rmap_walk(folio, &rwc);
56
57 if (need_lock)
58 folio_unlock(folio);
59
60 out:
61 folio_put(folio);
62 }
63
__damon_pa_prepare_access_check(struct damon_region * r)64 static void __damon_pa_prepare_access_check(struct damon_region *r)
65 {
66 r->sampling_addr = damon_rand(r->ar.start, r->ar.end);
67
68 damon_pa_mkold(r->sampling_addr);
69 }
70
damon_pa_prepare_access_checks(struct damon_ctx * ctx)71 static void damon_pa_prepare_access_checks(struct damon_ctx *ctx)
72 {
73 struct damon_target *t;
74 struct damon_region *r;
75
76 damon_for_each_target(t, ctx) {
77 damon_for_each_region(r, t)
78 __damon_pa_prepare_access_check(r);
79 }
80 }
81
__damon_pa_young(struct folio * folio,struct vm_area_struct * vma,unsigned long addr,void * arg)82 static bool __damon_pa_young(struct folio *folio, struct vm_area_struct *vma,
83 unsigned long addr, void *arg)
84 {
85 bool *accessed = arg;
86 DEFINE_FOLIO_VMA_WALK(pvmw, folio, vma, addr, 0);
87
88 *accessed = false;
89 while (page_vma_mapped_walk(&pvmw)) {
90 addr = pvmw.address;
91 if (pvmw.pte) {
92 *accessed = pte_young(*pvmw.pte) ||
93 !folio_test_idle(folio) ||
94 mmu_notifier_test_young(vma->vm_mm, addr);
95 } else {
96 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
97 *accessed = pmd_young(*pvmw.pmd) ||
98 !folio_test_idle(folio) ||
99 mmu_notifier_test_young(vma->vm_mm, addr);
100 #else
101 WARN_ON_ONCE(1);
102 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
103 }
104 if (*accessed) {
105 page_vma_mapped_walk_done(&pvmw);
106 break;
107 }
108 }
109
110 /* If accessed, stop walking */
111 return *accessed == false;
112 }
113
damon_pa_young(unsigned long paddr,unsigned long * folio_sz)114 static bool damon_pa_young(unsigned long paddr, unsigned long *folio_sz)
115 {
116 struct folio *folio = damon_get_folio(PHYS_PFN(paddr));
117 bool accessed = false;
118 struct rmap_walk_control rwc = {
119 .arg = &accessed,
120 .rmap_one = __damon_pa_young,
121 .anon_lock = folio_lock_anon_vma_read,
122 };
123 bool need_lock;
124
125 if (!folio)
126 return false;
127
128 if (!folio_mapped(folio) || !folio_raw_mapping(folio)) {
129 if (folio_test_idle(folio))
130 accessed = false;
131 else
132 accessed = true;
133 folio_put(folio);
134 goto out;
135 }
136
137 need_lock = !folio_test_anon(folio) || folio_test_ksm(folio);
138 if (need_lock && !folio_trylock(folio)) {
139 folio_put(folio);
140 return false;
141 }
142
143 rmap_walk(folio, &rwc);
144
145 if (need_lock)
146 folio_unlock(folio);
147 folio_put(folio);
148
149 out:
150 *folio_sz = folio_size(folio);
151 return accessed;
152 }
153
__damon_pa_check_access(struct damon_region * r)154 static void __damon_pa_check_access(struct damon_region *r)
155 {
156 static unsigned long last_addr;
157 static unsigned long last_folio_sz = PAGE_SIZE;
158 static bool last_accessed;
159
160 /* If the region is in the last checked page, reuse the result */
161 if (ALIGN_DOWN(last_addr, last_folio_sz) ==
162 ALIGN_DOWN(r->sampling_addr, last_folio_sz)) {
163 if (last_accessed)
164 r->nr_accesses++;
165 return;
166 }
167
168 last_accessed = damon_pa_young(r->sampling_addr, &last_folio_sz);
169 if (last_accessed)
170 r->nr_accesses++;
171
172 last_addr = r->sampling_addr;
173 }
174
damon_pa_check_accesses(struct damon_ctx * ctx)175 static unsigned int damon_pa_check_accesses(struct damon_ctx *ctx)
176 {
177 struct damon_target *t;
178 struct damon_region *r;
179 unsigned int max_nr_accesses = 0;
180
181 damon_for_each_target(t, ctx) {
182 damon_for_each_region(r, t) {
183 __damon_pa_check_access(r);
184 max_nr_accesses = max(r->nr_accesses, max_nr_accesses);
185 }
186 }
187
188 return max_nr_accesses;
189 }
190
__damos_pa_filter_out(struct damos_filter * filter,struct folio * folio)191 static bool __damos_pa_filter_out(struct damos_filter *filter,
192 struct folio *folio)
193 {
194 bool matched = false;
195 struct mem_cgroup *memcg;
196
197 switch (filter->type) {
198 case DAMOS_FILTER_TYPE_ANON:
199 matched = folio_test_anon(folio);
200 break;
201 case DAMOS_FILTER_TYPE_MEMCG:
202 rcu_read_lock();
203 memcg = folio_memcg_check(folio);
204 if (!memcg)
205 matched = false;
206 else
207 matched = filter->memcg_id == mem_cgroup_id(memcg);
208 rcu_read_unlock();
209 break;
210 default:
211 break;
212 }
213
214 return matched == filter->matching;
215 }
216
217 /*
218 * damos_pa_filter_out - Return true if the page should be filtered out.
219 */
damos_pa_filter_out(struct damos * scheme,struct folio * folio)220 static bool damos_pa_filter_out(struct damos *scheme, struct folio *folio)
221 {
222 struct damos_filter *filter;
223
224 damos_for_each_filter(filter, scheme) {
225 if (__damos_pa_filter_out(filter, folio))
226 return true;
227 }
228 return false;
229 }
230
damon_pa_pageout(struct damon_region * r,struct damos * s)231 static unsigned long damon_pa_pageout(struct damon_region *r, struct damos *s)
232 {
233 unsigned long addr, applied;
234 LIST_HEAD(folio_list);
235
236 for (addr = r->ar.start; addr < r->ar.end; addr += PAGE_SIZE) {
237 struct folio *folio = damon_get_folio(PHYS_PFN(addr));
238
239 if (!folio)
240 continue;
241
242 if (damos_pa_filter_out(s, folio)) {
243 folio_put(folio);
244 continue;
245 }
246
247 folio_clear_referenced(folio);
248 folio_test_clear_young(folio);
249 if (!folio_isolate_lru(folio)) {
250 folio_put(folio);
251 continue;
252 }
253 if (folio_test_unevictable(folio))
254 folio_putback_lru(folio);
255 else
256 list_add(&folio->lru, &folio_list);
257 folio_put(folio);
258 }
259 applied = reclaim_pages(&folio_list);
260 cond_resched();
261 return applied * PAGE_SIZE;
262 }
263
damon_pa_mark_accessed_or_deactivate(struct damon_region * r,struct damos * s,bool mark_accessed)264 static inline unsigned long damon_pa_mark_accessed_or_deactivate(
265 struct damon_region *r, struct damos *s, bool mark_accessed)
266 {
267 unsigned long addr, applied = 0;
268
269 for (addr = r->ar.start; addr < r->ar.end; addr += PAGE_SIZE) {
270 struct folio *folio = damon_get_folio(PHYS_PFN(addr));
271
272 if (!folio)
273 continue;
274
275 if (damos_pa_filter_out(s, folio)) {
276 folio_put(folio);
277 continue;
278 }
279
280 if (mark_accessed)
281 folio_mark_accessed(folio);
282 else
283 folio_deactivate(folio);
284 folio_put(folio);
285 applied += folio_nr_pages(folio);
286 }
287 return applied * PAGE_SIZE;
288 }
289
damon_pa_mark_accessed(struct damon_region * r,struct damos * s)290 static unsigned long damon_pa_mark_accessed(struct damon_region *r,
291 struct damos *s)
292 {
293 return damon_pa_mark_accessed_or_deactivate(r, s, true);
294 }
295
damon_pa_deactivate_pages(struct damon_region * r,struct damos * s)296 static unsigned long damon_pa_deactivate_pages(struct damon_region *r,
297 struct damos *s)
298 {
299 return damon_pa_mark_accessed_or_deactivate(r, s, false);
300 }
301
damon_pa_apply_scheme(struct damon_ctx * ctx,struct damon_target * t,struct damon_region * r,struct damos * scheme)302 static unsigned long damon_pa_apply_scheme(struct damon_ctx *ctx,
303 struct damon_target *t, struct damon_region *r,
304 struct damos *scheme)
305 {
306 switch (scheme->action) {
307 case DAMOS_PAGEOUT:
308 return damon_pa_pageout(r, scheme);
309 case DAMOS_LRU_PRIO:
310 return damon_pa_mark_accessed(r, scheme);
311 case DAMOS_LRU_DEPRIO:
312 return damon_pa_deactivate_pages(r, scheme);
313 case DAMOS_STAT:
314 break;
315 default:
316 /* DAMOS actions that not yet supported by 'paddr'. */
317 break;
318 }
319 return 0;
320 }
321
damon_pa_scheme_score(struct damon_ctx * context,struct damon_target * t,struct damon_region * r,struct damos * scheme)322 static int damon_pa_scheme_score(struct damon_ctx *context,
323 struct damon_target *t, struct damon_region *r,
324 struct damos *scheme)
325 {
326 switch (scheme->action) {
327 case DAMOS_PAGEOUT:
328 return damon_cold_score(context, r, scheme);
329 case DAMOS_LRU_PRIO:
330 return damon_hot_score(context, r, scheme);
331 case DAMOS_LRU_DEPRIO:
332 return damon_cold_score(context, r, scheme);
333 default:
334 break;
335 }
336
337 return DAMOS_MAX_SCORE;
338 }
339
damon_pa_initcall(void)340 static int __init damon_pa_initcall(void)
341 {
342 struct damon_operations ops = {
343 .id = DAMON_OPS_PADDR,
344 .init = NULL,
345 .update = NULL,
346 .prepare_access_checks = damon_pa_prepare_access_checks,
347 .check_accesses = damon_pa_check_accesses,
348 .reset_aggregated = NULL,
349 .target_valid = NULL,
350 .cleanup = NULL,
351 .apply_scheme = damon_pa_apply_scheme,
352 .get_scheme_score = damon_pa_scheme_score,
353 };
354
355 return damon_register_ops(&ops);
356 };
357
358 subsys_initcall(damon_pa_initcall);
359