kernel_5.10 no rt

hc

2023-12-11 6778948f9de86c3cfaf36725a7c87dcff9ba247f

 kernel/include/linux/sched/mm.h
..
..
@@ -23,7 +23,7 @@
23
23
  * will still exist later on and mmget_not_zero() has to be used before
24
24
  * accessing it.
25
25
  *
26
- * This is a preferred way to to pin @mm for a longer/unbounded amount
26
+ * This is a preferred way to pin @mm for a longer/unbounded amount
27
27
  * of time.
28
28
  *
29
29
  * Use mmdrop() to release the reference acquired by mmgrab().
..
..
@@ -47,33 +47,6 @@
47
47
 	 */
48
48
 	if (unlikely(atomic_dec_and_test(&mm->mm_count)))
49
49
 		__mmdrop(mm);
50
-}
51
-
52
-void mmdrop(struct mm_struct *mm);
53
-
54
-/*
55
- * This has to be called after a get_task_mm()/mmget_not_zero()
56
- * followed by taking the mmap_sem for writing before modifying the
57
- * vmas or anything the coredump pretends not to change from under it.
58
- *
59
- * It also has to be called when mmgrab() is used in the context of
60
- * the process, but then the mm_count refcount is transferred outside
61
- * the context of the process to run down_write() on that pinned mm.
62
- *
63
- * NOTE: find_extend_vma() called from GUP context is the only place
64
- * that can modify the "mm" (notably the vm_start/end) under mmap_sem
65
- * for reading and outside the context of the process, so it is also
66
- * the only case that holds the mmap_sem for reading that must call
67
- * this function. Generally if the mmap_sem is hold for reading
68
- * there's no need of this check after get_task_mm()/mmget_not_zero().
69
- *
70
- * This function can be obsoleted and the check can be removed, after
71
- * the coredump code will hold the mmap_sem for writing before
72
- * invoking the ->core_dump methods.
73
- */
74
-static inline bool mmget_still_valid(struct mm_struct *mm)
75
-{
76
-	return likely(!mm->core_state);
77
50
 }
78
51
 
79
52
 /**
..
..
@@ -133,6 +106,14 @@
133
106
 #endif /* CONFIG_MEMCG */
134
107
 
135
108
 #ifdef CONFIG_MMU
109
+#ifndef arch_get_mmap_end
110
+#define arch_get_mmap_end(addr)	(TASK_SIZE)
111
+#endif
112
+
113
+#ifndef arch_get_mmap_base
114
+#define arch_get_mmap_base(addr, base) (base)
115
+#endif
116
+
136
117
 extern void arch_pick_mmap_layout(struct mm_struct *mm,
137
118
 				  struct rlimit *rlim_stack);
138
119
 extern unsigned long
..
..
@@ -181,14 +162,18 @@
181
162
  */
182
163
 static inline gfp_t current_gfp_context(gfp_t flags)
183
164
 {
184
-	/*
185
-	 * NOIO implies both NOIO and NOFS and it is a weaker context
186
-	 * so always make sure it makes precendence
187
-	 */
188
-	if (unlikely(current->flags & PF_MEMALLOC_NOIO))
189
-		flags &= ~(__GFP_IO | __GFP_FS);
190
-	else if (unlikely(current->flags & PF_MEMALLOC_NOFS))
191
-		flags &= ~__GFP_FS;
165
+	unsigned int pflags = READ_ONCE(current->flags);
166
+
167
+	if (unlikely(pflags & (PF_MEMALLOC_NOIO | PF_MEMALLOC_NOFS))) {
168
+		/*
169
+		 * NOIO implies both NOIO and NOFS and it is a weaker context
170
+		 * so always make sure it makes precedence
171
+		 */
172
+		if (pflags & PF_MEMALLOC_NOIO)
173
+			flags &= ~(__GFP_IO | __GFP_FS);
174
+		else if (pflags & PF_MEMALLOC_NOFS)
175
+			flags &= ~__GFP_FS;
176
+	}
192
177
 	return flags;
193
178
 }
194
179
 
..
..
@@ -227,7 +212,7 @@
227
212
  * @flags: Flags to restore.
228
213
  *
229
214
  * Ends the implicit GFP_NOIO scope started by memalloc_noio_save function.
230
- * Always make sure that that the given flags is the return value from the
215
+ * Always make sure that the given flags is the return value from the
231
216
  * pairing memalloc_noio_save call.
232
217
  */
233
218
 static inline void memalloc_noio_restore(unsigned int flags)
..
..
@@ -258,7 +243,7 @@
258
243
  * @flags: Flags to restore.
259
244
  *
260
245
  * Ends the implicit GFP_NOFS scope started by memalloc_nofs_save function.
261
- * Always make sure that that the given flags is the return value from the
246
+ * Always make sure that the given flags is the return value from the
262
247
  * pairing memalloc_nofs_save call.
263
248
  */
264
249
 static inline void memalloc_nofs_restore(unsigned int flags)
..
..
@@ -278,40 +263,63 @@
278
263
 	current->flags = (current->flags & ~PF_MEMALLOC) | flags;
279
264
 }
280
265
 
266
+#ifdef CONFIG_CMA
267
+static inline unsigned int memalloc_nocma_save(void)
268
+{
269
+	unsigned int flags = current->flags & PF_MEMALLOC_NOCMA;
270
+
271
+	current->flags |= PF_MEMALLOC_NOCMA;
272
+	return flags;
273
+}
274
+
275
+static inline void memalloc_nocma_restore(unsigned int flags)
276
+{
277
+	current->flags = (current->flags & ~PF_MEMALLOC_NOCMA) | flags;
278
+}
279
+#else
280
+static inline unsigned int memalloc_nocma_save(void)
281
+{
282
+	return 0;
283
+}
284
+
285
+static inline void memalloc_nocma_restore(unsigned int flags)
286
+{
287
+}
288
+#endif
289
+
281
290
 #ifdef CONFIG_MEMCG
291
+DECLARE_PER_CPU(struct mem_cgroup *, int_active_memcg);
282
292
 /**
283
- * memalloc_use_memcg - Starts the remote memcg charging scope.
293
+ * set_active_memcg - Starts the remote memcg charging scope.
284
294
  * @memcg: memcg to charge.
285
295
  *
286
296
  * This function marks the beginning of the remote memcg charging scope. All the
287
297
  * __GFP_ACCOUNT allocations till the end of the scope will be charged to the
288
298
  * given memcg.
289
299
  *
290
- * NOTE: This function is not nesting safe.
300
+ * NOTE: This function can nest. Users must save the return value and
301
+ * reset the previous value after their own charging scope is over.
291
302
  */
292
-static inline void memalloc_use_memcg(struct mem_cgroup *memcg)
303
+static inline struct mem_cgroup *
304
+set_active_memcg(struct mem_cgroup *memcg)
293
305
 {
294
-	WARN_ON_ONCE(current->active_memcg);
295
-	current->active_memcg = memcg;
296
-}
306
+	struct mem_cgroup *old;
297
307
 
298
-/**
299
- * memalloc_unuse_memcg - Ends the remote memcg charging scope.
300
- *
301
- * This function marks the end of the remote memcg charging scope started by
302
- * memalloc_use_memcg().
303
- */
304
-static inline void memalloc_unuse_memcg(void)
305
-{
306
-	current->active_memcg = NULL;
308
+	if (in_interrupt()) {
309
+		old = this_cpu_read(int_active_memcg);
310
+		this_cpu_write(int_active_memcg, memcg);
311
+	} else {
312
+		old = current->active_memcg;
313
+		current->active_memcg = memcg;
314
+	}
315
+
316
+	return old;
307
317
 }
308
318
 #else
309
-static inline void memalloc_use_memcg(struct mem_cgroup *memcg)
319
+static inline struct mem_cgroup *
320
+set_active_memcg(struct mem_cgroup *memcg)
310
321
 {
311
-}
312
-
313
-static inline void memalloc_unuse_memcg(void)
314
-{
322
+	return NULL;
315
323
 }
316
324
 #endif
317
325
 
..
..
@@ -323,10 +331,13 @@
323
331
 	MEMBARRIER_STATE_GLOBAL_EXPEDITED			= (1U << 3),
324
332
 	MEMBARRIER_STATE_PRIVATE_EXPEDITED_SYNC_CORE_READY	= (1U << 4),
325
333
 	MEMBARRIER_STATE_PRIVATE_EXPEDITED_SYNC_CORE		= (1U << 5),
334
+	MEMBARRIER_STATE_PRIVATE_EXPEDITED_RSEQ_READY		= (1U << 6),
335
+	MEMBARRIER_STATE_PRIVATE_EXPEDITED_RSEQ			= (1U << 7),
326
336
 };
327
337
 
328
338
 enum {
329
339
 	MEMBARRIER_FLAG_SYNC_CORE	= (1U << 0),
340
+	MEMBARRIER_FLAG_RSEQ		= (1U << 1),
330
341
 };
331
342
 
332
343
 #ifdef CONFIG_ARCH_HAS_MEMBARRIER_CALLBACKS
..
..
@@ -343,10 +354,8 @@
343
354
 	sync_core_before_usermode();
344
355
 }
345
356
 
346
-static inline void membarrier_execve(struct task_struct *t)
347
-{
348
-	atomic_set(&t->mm->membarrier_state, 0);
349
-}
357
+extern void membarrier_exec_mmap(struct mm_struct *mm);
358
+
350
359
 #else
351
360
 #ifdef CONFIG_ARCH_HAS_MEMBARRIER_CALLBACKS
352
361
 static inline void membarrier_arch_switch_mm(struct mm_struct *prev,
..
..
@@ -355,7 +364,7 @@
355
364
 {
356
365
 }
357
366
 #endif
358
-static inline void membarrier_execve(struct task_struct *t)
367
+static inline void membarrier_exec_mmap(struct mm_struct *mm)
359
368
 {
360
369
 }
361
370
 static inline void membarrier_mm_sync_core_before_usermode(struct mm_struct *mm)