forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-05-10 10ebd8556b7990499c896a550e3d416b444211e6 
 kernel/fs/buffer.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-only
12 /*
23  *  linux/fs/buffer.c
34  *
....@@ -47,6 +48,8 @@
4748 #include <linux/sched/mm.h>
4849 #include <trace/events/block.h>
4950 #include <linux/fscrypt.h>
51
+
52
+#include "internal.h"
5053 
5154 static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);
5255 static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
....@@ -120,14 +123,6 @@
120123 }
121124 EXPORT_SYMBOL(__wait_on_buffer);
122125 
123
-static void
124
-__clear_page_buffers(struct page *page)
125
-{
126
-	ClearPagePrivate(page);
127
-	set_page_private(page, 0);
128
-	put_page(page);
129
-}
130
-
131126 static void buffer_io_error(struct buffer_head *bh, char *msg)
132127 {
133128 	if (!test_bit(BH_Quiet, &bh->b_state))
....@@ -178,7 +173,7 @@
178173 	unlock_buffer(bh);
179174 	put_bh(bh);
180175 }
181
-EXPORT_SYMBOL(end_buffer_write_sync);
176
+EXPORT_SYMBOL_NS(end_buffer_write_sync, ANDROID_GKI_VFS_EXPORT_ONLY);
182177 
183178 /*
184179  * Various filesystems appear to want __find_get_block to be non-blocking.
....@@ -246,10 +241,6 @@
246241 	return ret;
247242 }
248243 
249
-/*
250
- * I/O completion handler for block_read_full_page() - pages
251
- * which come unlocked at the end of I/O.
252
- */
253244 static void end_buffer_async_read(struct buffer_head *bh, int uptodate)
254245 {
255246 	unsigned long flags;
....@@ -275,7 +266,7 @@
275266 	 * decide that the page is now completely done.
276267 	 */
277268 	first = page_buffers(page);
278
-	flags = bh_uptodate_lock_irqsave(first);
269
+	spin_lock_irqsave(&first->b_uptodate_lock, flags);
279270 	clear_buffer_async_read(bh);
280271 	unlock_buffer(bh);
281272 	tmp = bh;
....@@ -288,7 +279,7 @@
288279 		}
289280 		tmp = tmp->b_this_page;
290281 	} while (tmp != bh);
291
-	bh_uptodate_unlock_irqrestore(first, flags);
282
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
292283 
293284 	/*
294285 	 * If none of the buffers had errors and they are all
....@@ -300,7 +291,48 @@
300291 	return;
301292 
302293 still_busy:
303
-	bh_uptodate_unlock_irqrestore(first, flags);
294
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
295
+	return;
296
+}
297
+
298
+struct decrypt_bh_ctx {
299
+	struct work_struct work;
300
+	struct buffer_head *bh;
301
+};
302
+
303
+static void decrypt_bh(struct work_struct *work)
304
+{
305
+	struct decrypt_bh_ctx *ctx =
306
+		container_of(work, struct decrypt_bh_ctx, work);
307
+	struct buffer_head *bh = ctx->bh;
308
+	int err;
309
+
310
+	err = fscrypt_decrypt_pagecache_blocks(bh->b_page, bh->b_size,
311
+					       bh_offset(bh));
312
+	end_buffer_async_read(bh, err == 0);
313
+	kfree(ctx);
314
+}
315
+
316
+/*
317
+ * I/O completion handler for block_read_full_page() - pages
318
+ * which come unlocked at the end of I/O.
319
+ */
320
+static void end_buffer_async_read_io(struct buffer_head *bh, int uptodate)
321
+{
322
+	/* Decrypt if needed */
323
+	if (uptodate &&
324
+	    fscrypt_inode_uses_fs_layer_crypto(bh->b_page->mapping->host)) {
325
+		struct decrypt_bh_ctx *ctx = kmalloc(sizeof(*ctx), GFP_ATOMIC);
326
+
327
+		if (ctx) {
328
+			INIT_WORK(&ctx->work, decrypt_bh);
329
+			ctx->bh = bh;
330
+			fscrypt_enqueue_decrypt_work(&ctx->work);
331
+			return;
332
+		}
333
+		uptodate = 0;
334
+	}
335
+	end_buffer_async_read(bh, uptodate);
304336 }
305337 
306338 /*
....@@ -327,7 +359,7 @@
327359 	}
328360 
329361 	first = page_buffers(page);
330
-	flags = bh_uptodate_lock_irqsave(first);
362
+	spin_lock_irqsave(&first->b_uptodate_lock, flags);
331363 
332364 	clear_buffer_async_write(bh);
333365 	unlock_buffer(bh);
....@@ -339,12 +371,13 @@
339371 		}
340372 		tmp = tmp->b_this_page;
341373 	}
342
-	bh_uptodate_unlock_irqrestore(first, flags);
374
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
343375 	end_page_writeback(page);
344376 	return;
345377 
346378 still_busy:
347
-	bh_uptodate_unlock_irqrestore(first, flags);
379
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
380
+	return;
348381 }
349382 EXPORT_SYMBOL(end_buffer_async_write);
350383 
....@@ -371,7 +404,7 @@
371404  */
372405 static void mark_buffer_async_read(struct buffer_head *bh)
373406 {
374
-	bh->b_end_io = end_buffer_async_read;
407
+	bh->b_end_io = end_buffer_async_read_io;
375408 	set_buffer_async_read(bh);
376409 }
377410 
....@@ -386,7 +419,7 @@
386419 {
387420 	mark_buffer_async_write_endio(bh, end_buffer_async_write);
388421 }
389
-EXPORT_SYMBOL(mark_buffer_async_write);
422
+EXPORT_SYMBOL_NS(mark_buffer_async_write, ANDROID_GKI_VFS_EXPORT_ONLY);
390423 
391424 
392425 /*
....@@ -490,7 +523,7 @@
490523 
491524 void emergency_thaw_bdev(struct super_block *sb)
492525 {
493
-	while (sb->s_bdev && !thaw_bdev(sb->s_bdev, sb))
526
+	while (sb->s_bdev && !thaw_bdev(sb->s_bdev))
494527 		printk(KERN_WARNING "Emergency Thaw on %pg\n", sb->s_bdev);
495528 }
496529 
....@@ -556,7 +589,7 @@
556589 EXPORT_SYMBOL(mark_buffer_dirty_inode);
557590 
558591 /*
559
- * Mark the page dirty, and set it dirty in the radix tree, and mark the inode
592
+ * Mark the page dirty, and set it dirty in the page cache, and mark the inode
560593  * dirty.
561594  *
562595  * If warn is true, then emit a warning if the page is not uptodate and has
....@@ -573,8 +606,8 @@
573606 	if (page->mapping) {	/* Race with truncate? */
574607 		WARN_ON_ONCE(warn && !PageUptodate(page));
575608 		account_page_dirtied(page, mapping);
576
-		radix_tree_tag_set(&mapping->i_pages,
577
-				page_index(page), PAGECACHE_TAG_DIRTY);
609
+		__xa_set_mark(&mapping->i_pages, page_index(page),
610
+				PAGECACHE_TAG_DIRTY);
578611 	}
579612 	xa_unlock_irqrestore(&mapping->i_pages, flags);
580613 }
....@@ -641,7 +674,7 @@
641674 
642675 	return newly_dirty;
643676 }
644
-EXPORT_SYMBOL(__set_page_dirty_buffers);
677
+EXPORT_SYMBOL_NS(__set_page_dirty_buffers, ANDROID_GKI_VFS_EXPORT_ONLY);
645678 
646679 /*
647680  * Write out and wait upon a list of buffers.
....@@ -809,13 +842,13 @@
809842 	struct buffer_head *bh, *head;
810843 	gfp_t gfp = GFP_NOFS | __GFP_ACCOUNT;
811844 	long offset;
812
-	struct mem_cgroup *memcg;
845
+	struct mem_cgroup *memcg, *old_memcg;
813846 
814847 	if (retry)
815848 		gfp |= __GFP_NOFAIL;
816849 
817850 	memcg = get_mem_cgroup_from_page(page);
818
-	memalloc_use_memcg(memcg);
851
+	old_memcg = set_active_memcg(memcg);
819852 
820853 	head = NULL;
821854 	offset = PAGE_SIZE;
....@@ -834,7 +867,7 @@
834867 		set_bh_page(bh, page, offset);
835868 	}
836869 out:
837
-	memalloc_unuse_memcg();
870
+	set_active_memcg(old_memcg);
838871 	mem_cgroup_put(memcg);
839872 	return head;
840873 /*
....@@ -864,7 +897,7 @@
864897 		bh = bh->b_this_page;
865898 	} while (bh);
866899 	tail->b_this_page = head;
867
-	attach_page_buffers(page, head);
900
+	attach_page_private(page, head);
868901 }
869902 
870903 static sector_t blkdev_max_block(struct block_device *bdev, unsigned int size)
....@@ -925,7 +958,7 @@
925958 	struct page *page;
926959 	struct buffer_head *bh;
927960 	sector_t end_block;
928
-	int ret = 0;		/* Will call free_more_memory() */
961
+	int ret = 0;
929962 	gfp_t gfp_mask;
930963 
931964 	gfp_mask = mapping_gfp_constraint(inode->i_mapping, ~__GFP_FS) | gfp;
....@@ -1044,7 +1077,7 @@
10441077  * The relationship between dirty buffers and dirty pages:
10451078  *
10461079  * Whenever a page has any dirty buffers, the page's dirty bit is set, and
1047
- * the page is tagged dirty in its radix tree.
1080
+ * the page is tagged dirty in the page cache.
10481081  *
10491082  * At all times, the dirtiness of the buffers represents the dirtiness of
10501083  * subsections of the page.  If the page has buffers, the page dirty bit is
....@@ -1067,9 +1100,9 @@
10671100  * mark_buffer_dirty - mark a buffer_head as needing writeout
10681101  * @bh: the buffer_head to mark dirty
10691102  *
1070
- * mark_buffer_dirty() will set the dirty bit against the buffer, then set its
1071
- * backing page dirty, then tag the page as dirty in its address_space's radix
1072
- * tree and then attach the address_space's inode to its superblock's dirty
1103
+ * mark_buffer_dirty() will set the dirty bit against the buffer, then set
1104
+ * its backing page dirty, then tag the page as dirty in the page cache
1105
+ * and then attach the address_space's inode to its superblock's dirty
10731106  * inode list.
10741107  *
10751108  * mark_buffer_dirty() is atomic.  It takes bh->b_page->mapping->private_lock,
....@@ -1108,18 +1141,25 @@
11081141 			__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
11091142 	}
11101143 }
1111
-EXPORT_SYMBOL(mark_buffer_dirty);
1144
+EXPORT_SYMBOL_NS(mark_buffer_dirty, ANDROID_GKI_VFS_EXPORT_ONLY);
11121145 
11131146 void mark_buffer_write_io_error(struct buffer_head *bh)
11141147 {
1148
+	struct super_block *sb;
1149
+
11151150 	set_buffer_write_io_error(bh);
11161151 	/* FIXME: do we need to set this in both places? */
11171152 	if (bh->b_page && bh->b_page->mapping)
11181153 		mapping_set_error(bh->b_page->mapping, -EIO);
11191154 	if (bh->b_assoc_map)
11201155 		mapping_set_error(bh->b_assoc_map, -EIO);
1156
+	rcu_read_lock();
1157
+	sb = READ_ONCE(bh->b_bdev->bd_super);
1158
+	if (sb)
1159
+		errseq_set(&sb->s_wb_err, -EIO);
1160
+	rcu_read_unlock();
11211161 }
1122
-EXPORT_SYMBOL(mark_buffer_write_io_error);
1162
+EXPORT_SYMBOL_NS(mark_buffer_write_io_error, ANDROID_GKI_VFS_EXPORT_ONLY);
11231163 
11241164 /*
11251165  * Decrement a buffer_head's reference count.  If all buffers against a page
....@@ -1136,7 +1176,7 @@
11361176 	}
11371177 	WARN(1, KERN_ERR "VFS: brelse: Trying to free free buffer\n");
11381178 }
1139
-EXPORT_SYMBOL(__brelse);
1179
+EXPORT_SYMBOL_NS(__brelse, ANDROID_GKI_VFS_EXPORT_ONLY);
11401180 
11411181 /*
11421182  * bforget() is like brelse(), except it discards any
....@@ -1155,7 +1195,7 @@
11551195 	}
11561196 	__brelse(bh);
11571197 }
1158
-EXPORT_SYMBOL(__bforget);
1198
+EXPORT_SYMBOL_NS(__bforget, ANDROID_GKI_VFS_EXPORT_ONLY);
11591199 
11601200 static struct buffer_head *__bread_slow(struct buffer_head *bh)
11611201 {
....@@ -1224,6 +1264,15 @@
12241264 	int i;
12251265 
12261266 	check_irqs_on();
1267
+	/*
1268
+	 * the refcount of buffer_head in bh_lru prevents dropping the
1269
+	 * attached page(i.e., try_to_free_buffers) so it could cause
1270
+	 * failing page migration.
1271
+	 * Skip putting upcoming bh into bh_lru until migration is done.
1272
+	 */
1273
+	if (lru_cache_disabled())
1274
+		return;
1275
+
12271276 	bh_lru_lock();
12281277 
12291278 	b = this_cpu_ptr(&bh_lrus);
....@@ -1327,7 +1376,7 @@
13271376 		brelse(bh);
13281377 	}
13291378 }
1330
-EXPORT_SYMBOL(__breadahead);
1379
+EXPORT_SYMBOL_NS(__breadahead, ANDROID_GKI_VFS_EXPORT_ONLY);
13311380 
13321381 void __breadahead_gfp(struct block_device *bdev, sector_t block, unsigned size,
13331382 		      gfp_t gfp)
....@@ -1362,8 +1411,17 @@
13621411 		bh = __bread_slow(bh);
13631412 	return bh;
13641413 }
1365
-EXPORT_SYMBOL(__bread_gfp);
1414
+EXPORT_SYMBOL_NS(__bread_gfp, ANDROID_GKI_VFS_EXPORT_ONLY);
13661415 
1416
+static void __invalidate_bh_lrus(struct bh_lru *b)
1417
+{
1418
+	int i;
1419
+
1420
+	for (i = 0; i < BH_LRU_SIZE; i++) {
1421
+		brelse(b->bhs[i]);
1422
+		b->bhs[i] = NULL;
1423
+	}
1424
+}
13671425 /*
13681426  * invalidate_bh_lrus() is called rarely - but not only at unmount.
13691427  * This doesn't race because it runs in each cpu either in irq
....@@ -1372,33 +1430,43 @@
13721430 static void invalidate_bh_lru(void *arg)
13731431 {
13741432 	struct bh_lru *b = &get_cpu_var(bh_lrus);
1375
-	int i;
13761433 
1377
-	for (i = 0; i < BH_LRU_SIZE; i++) {
1378
-		brelse(b->bhs[i]);
1379
-		b->bhs[i] = NULL;
1380
-	}
1434
+	__invalidate_bh_lrus(b);
13811435 	put_cpu_var(bh_lrus);
13821436 }
13831437 
1384
-static bool has_bh_in_lru(int cpu, void *dummy)
1438
+bool has_bh_in_lru(int cpu, void *dummy)
13851439 {
13861440 	struct bh_lru *b = per_cpu_ptr(&bh_lrus, cpu);
13871441 	int i;
13881442 	
13891443 	for (i = 0; i < BH_LRU_SIZE; i++) {
13901444 		if (b->bhs[i])
1391
-			return 1;
1445
+			return true;
13921446 	}
13931447 
1394
-	return 0;
1448
+	return false;
13951449 }
13961450 
13971451 void invalidate_bh_lrus(void)
13981452 {
1399
-	on_each_cpu_cond(has_bh_in_lru, invalidate_bh_lru, NULL, 1, GFP_KERNEL);
1453
+	on_each_cpu_cond(has_bh_in_lru, invalidate_bh_lru, NULL, 1);
14001454 }
14011455 EXPORT_SYMBOL_GPL(invalidate_bh_lrus);
1456
+
1457
+/*
1458
+ * It's called from workqueue context so we need a bh_lru_lock to close
1459
+ * the race with preemption/irq.
1460
+ */
1461
+void invalidate_bh_lrus_cpu(void)
1462
+{
1463
+	struct bh_lru *b;
1464
+
1465
+	bh_lru_lock();
1466
+	b = this_cpu_ptr(&bh_lrus);
1467
+	__invalidate_bh_lrus(b);
1468
+	bh_lru_unlock();
1469
+}
14021470 
14031471 void set_bh_page(struct buffer_head *bh,
14041472 		struct page *page, unsigned long offset)
....@@ -1505,7 +1573,7 @@
15051573 out:
15061574 	return;
15071575 }
1508
-EXPORT_SYMBOL(block_invalidatepage);
1576
+EXPORT_SYMBOL_NS(block_invalidatepage, ANDROID_GKI_VFS_EXPORT_ONLY);
15091577 
15101578 
15111579 /*
....@@ -1538,10 +1606,10 @@
15381606 			bh = bh->b_this_page;
15391607 		} while (bh != head);
15401608 	}
1541
-	attach_page_buffers(page, head);
1609
+	attach_page_private(page, head);
15421610 	spin_unlock(&page->mapping->private_lock);
15431611 }
1544
-EXPORT_SYMBOL(create_empty_buffers);
1612
+EXPORT_SYMBOL_NS(create_empty_buffers, ANDROID_GKI_VFS_EXPORT_ONLY);
15451613 
15461614 /**
15471615  * clean_bdev_aliases: clean a range of buffers in block device
....@@ -1615,7 +1683,7 @@
16151683 			break;
16161684 	}
16171685 }
1618
-EXPORT_SYMBOL(clean_bdev_aliases);
1686
+EXPORT_SYMBOL_NS(clean_bdev_aliases, ANDROID_GKI_VFS_EXPORT_ONLY);
16191687 
16201688 /*
16211689  * Size is a power-of-two in the range 512..PAGE_SIZE,
....@@ -1873,7 +1941,7 @@
18731941 		bh = bh->b_this_page;
18741942 	} while (bh != head);
18751943 }
1876
-EXPORT_SYMBOL(page_zero_new_buffers);
1944
+EXPORT_SYMBOL_NS(page_zero_new_buffers, ANDROID_GKI_VFS_EXPORT_ONLY);
18771945 
18781946 static void
18791947 iomap_to_bh(struct inode *inode, sector_t block, struct buffer_head *bh,
....@@ -1918,7 +1986,7 @@
19181986 		 */
19191987 		set_buffer_new(bh);
19201988 		set_buffer_unwritten(bh);
1921
-		/* FALLTHRU */
1989
+		fallthrough;
19221990 	case IOMAP_MAPPED:
19231991 		if ((iomap->flags & IOMAP_F_NEW) ||
19241992 		    offset >= i_size_read(inode))
....@@ -2089,40 +2157,6 @@
20892157 }
20902158 EXPORT_SYMBOL(block_write_begin);
20912159 
2092
-int __generic_write_end(struct inode *inode, loff_t pos, unsigned copied,
2093
-		struct page *page)
2094
-{
2095
-	loff_t old_size = inode->i_size;
2096
-	bool i_size_changed = false;
2097
-
2098
-	/*
2099
-	 * No need to use i_size_read() here, the i_size cannot change under us
2100
-	 * because we hold i_rwsem.
2101
-	 *
2102
-	 * But it's important to update i_size while still holding page lock:
2103
-	 * page writeout could otherwise come in and zero beyond i_size.
2104
-	 */
2105
-	if (pos + copied > inode->i_size) {
2106
-		i_size_write(inode, pos + copied);
2107
-		i_size_changed = true;
2108
-	}
2109
-
2110
-	unlock_page(page);
2111
-	put_page(page);
2112
-
2113
-	if (old_size < pos)
2114
-		pagecache_isize_extended(inode, old_size, pos);
2115
-	/*
2116
-	 * Don't mark the inode dirty under page lock. First, it unnecessarily
2117
-	 * makes the holding time of page lock longer. Second, it forces lock
2118
-	 * ordering of page lock and transaction start for journaling
2119
-	 * filesystems.
2120
-	 */
2121
-	if (i_size_changed)
2122
-		mark_inode_dirty(inode);
2123
-	return copied;
2124
-}
2125
-
21262160 int block_write_end(struct file *file, struct address_space *mapping,
21272161 			loff_t pos, unsigned len, unsigned copied,
21282162 			struct page *page, void *fsdata)
....@@ -2163,8 +2197,38 @@
21632197 			loff_t pos, unsigned len, unsigned copied,
21642198 			struct page *page, void *fsdata)
21652199 {
2200
+	struct inode *inode = mapping->host;
2201
+	loff_t old_size = inode->i_size;
2202
+	bool i_size_changed = false;
2203
+
21662204 	copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);
2167
-	return __generic_write_end(mapping->host, pos, copied, page);
2205
+
2206
+	/*
2207
+	 * No need to use i_size_read() here, the i_size cannot change under us
2208
+	 * because we hold i_rwsem.
2209
+	 *
2210
+	 * But it's important to update i_size while still holding page lock:
2211
+	 * page writeout could otherwise come in and zero beyond i_size.
2212
+	 */
2213
+	if (pos + copied > inode->i_size) {
2214
+		i_size_write(inode, pos + copied);
2215
+		i_size_changed = true;
2216
+	}
2217
+
2218
+	unlock_page(page);
2219
+	put_page(page);
2220
+
2221
+	if (old_size < pos)
2222
+		pagecache_isize_extended(inode, old_size, pos);
2223
+	/*
2224
+	 * Don't mark the inode dirty under page lock. First, it unnecessarily
2225
+	 * makes the holding time of page lock longer. Second, it forces lock
2226
+	 * ordering of page lock and transaction start for journaling
2227
+	 * filesystems.
2228
+	 */
2229
+	if (i_size_changed)
2230
+		mark_inode_dirty(inode);
2231
+	return copied;
21682232 }
21692233 EXPORT_SYMBOL(generic_write_end);
21702234 
....@@ -2211,7 +2275,7 @@
22112275 
22122276 	return ret;
22132277 }
2214
-EXPORT_SYMBOL(block_is_partially_uptodate);
2278
+EXPORT_SYMBOL_NS(block_is_partially_uptodate, ANDROID_GKI_VFS_EXPORT_ONLY);
22152279 
22162280 /*
22172281  * Generic "read page" function for block devices that have the normal
....@@ -2314,7 +2378,7 @@
23142378 {
23152379 	struct address_space *mapping = inode->i_mapping;
23162380 	struct page *page;
2317
-	void *fsdata;
2381
+	void *fsdata = NULL;
23182382 	int err;
23192383 
23202384 	err = inode_newsize_ok(inode, size);
....@@ -2340,7 +2404,7 @@
23402404 	struct inode *inode = mapping->host;
23412405 	unsigned int blocksize = i_blocksize(inode);
23422406 	struct page *page;
2343
-	void *fsdata;
2407
+	void *fsdata = NULL;
23442408 	pgoff_t index, curidx;
23452409 	loff_t curpos;
23462410 	unsigned zerofrom, offset, len;
....@@ -2371,7 +2435,7 @@
23712435 
23722436 		balance_dirty_pages_ratelimited(mapping);
23732437 
2374
-		if (unlikely(fatal_signal_pending(current))) {
2438
+		if (fatal_signal_pending(current)) {
23752439 			err = -EINTR;
23762440 			goto out;
23772441 		}
....@@ -2529,7 +2593,7 @@
25292593 			bh->b_this_page = head;
25302594 		bh = bh->b_this_page;
25312595 	} while (bh != head);
2532
-	attach_page_buffers(page, head);
2596
+	attach_page_private(page, head);
25332597 	spin_unlock(&page->mapping->private_lock);
25342598 }
25352599 
....@@ -2970,69 +3034,6 @@
29703034 	bio_put(bio);
29713035 }
29723036 
2973
-/*
2974
- * This allows us to do IO even on the odd last sectors
2975
- * of a device, even if the block size is some multiple
2976
- * of the physical sector size.
2977
- *
2978
- * We'll just truncate the bio to the size of the device,
2979
- * and clear the end of the buffer head manually.
2980
- *
2981
- * Truly out-of-range accesses will turn into actual IO
2982
- * errors, this only handles the "we need to be able to
2983
- * do IO at the final sector" case.
2984
- */
2985
-void guard_bio_eod(int op, struct bio *bio)
2986
-{
2987
-	sector_t maxsector;
2988
-	struct bio_vec *bvec = bio_last_bvec_all(bio);
2989
-	unsigned truncated_bytes;
2990
-	struct hd_struct *part;
2991
-
2992
-	rcu_read_lock();
2993
-	part = __disk_get_part(bio->bi_disk, bio->bi_partno);
2994
-	if (part)
2995
-		maxsector = part_nr_sects_read(part);
2996
-	else
2997
-		maxsector = get_capacity(bio->bi_disk);
2998
-	rcu_read_unlock();
2999
-
3000
-	if (!maxsector)
3001
-		return;
3002
-
3003
-	/*
3004
-	 * If the *whole* IO is past the end of the device,
3005
-	 * let it through, and the IO layer will turn it into
3006
-	 * an EIO.
3007
-	 */
3008
-	if (unlikely(bio->bi_iter.bi_sector >= maxsector))
3009
-		return;
3010
-
3011
-	maxsector -= bio->bi_iter.bi_sector;
3012
-	if (likely((bio->bi_iter.bi_size >> 9) <= maxsector))
3013
-		return;
3014
-
3015
-	/* Uhhuh. We've got a bio that straddles the device size! */
3016
-	truncated_bytes = bio->bi_iter.bi_size - (maxsector << 9);
3017
-
3018
-	/*
3019
-	 * The bio contains more than one segment which spans EOD, just return
3020
-	 * and let IO layer turn it into an EIO
3021
-	 */
3022
-	if (truncated_bytes > bvec->bv_len)
3023
-		return;
3024
-
3025
-	/* Truncate the bio.. */
3026
-	bio->bi_iter.bi_size -= truncated_bytes;
3027
-	bvec->bv_len -= truncated_bytes;
3028
-
3029
-	/* ..and clear the end of the buffer for reads */
3030
-	if (op == REQ_OP_READ) {
3031
-		zero_user(bvec->bv_page, bvec->bv_offset + bvec->bv_len,
3032
-				truncated_bytes);
3033
-	}
3034
-}
3035
-
30363037 static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
30373038 			 enum rw_hint write_hint, struct writeback_control *wbc)
30383039 {
....@@ -3050,18 +3051,9 @@
30503051 	if (test_set_buffer_req(bh) && (op == REQ_OP_WRITE))
30513052 		clear_buffer_write_io_error(bh);
30523053 
3053
-	/*
3054
-	 * from here on down, it's all bio -- do the initial mapping,
3055
-	 * submit_bio -> generic_make_request may further map this bio around
3056
-	 */
30573054 	bio = bio_alloc(GFP_NOIO, 1);
30583055 
30593056 	fscrypt_set_bio_crypt_ctx_bh(bio, bh, GFP_NOIO);
3060
-
3061
-	if (wbc) {
3062
-		wbc_init_bio(wbc, bio);
3063
-		wbc_account_io(wbc, bh->b_page, bh->b_size);
3064
-	}
30653057 
30663058 	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
30673059 	bio_set_dev(bio, bh->b_bdev);
....@@ -3073,14 +3065,19 @@
30733065 	bio->bi_end_io = end_bio_bh_io_sync;
30743066 	bio->bi_private = bh;
30753067 
3076
-	/* Take care of bh's that straddle the end of the device */
3077
-	guard_bio_eod(op, bio);
3078
-
30793068 	if (buffer_meta(bh))
30803069 		op_flags |= REQ_META;
30813070 	if (buffer_prio(bh))
30823071 		op_flags |= REQ_PRIO;
30833072 	bio_set_op_attrs(bio, op, op_flags);
3073
+
3074
+	/* Take care of bh's that straddle the end of the device */
3075
+	guard_bio_eod(bio);
3076
+
3077
+	if (wbc) {
3078
+		wbc_init_bio(wbc, bio);
3079
+		wbc_account_cgroup_owner(wbc, bh->b_page, bh->b_size);
3080
+	}
30843081 
30853082 	submit_bio(bio);
30863083 	return 0;
....@@ -3145,7 +3142,7 @@
31453142 		unlock_buffer(bh);
31463143 	}
31473144 }
3148
-EXPORT_SYMBOL(ll_rw_block);
3145
+EXPORT_SYMBOL_NS(ll_rw_block, ANDROID_GKI_VFS_EXPORT_ONLY);
31493146 
31503147 void write_dirty_buffer(struct buffer_head *bh, int op_flags)
31513148 {
....@@ -3198,7 +3195,7 @@
31983195 {
31993196 	return __sync_dirty_buffer(bh, REQ_SYNC);
32003197 }
3201
-EXPORT_SYMBOL(sync_dirty_buffer);
3198
+EXPORT_SYMBOL_NS(sync_dirty_buffer, ANDROID_GKI_VFS_EXPORT_ONLY);
32023199 
32033200 /*
32043201  * try_to_free_buffers() checks if all the buffers on this particular page
....@@ -3247,7 +3244,7 @@
32473244 		bh = next;
32483245 	} while (bh != head);
32493246 	*buffers_to_free = head;
3250
-	__clear_page_buffers(page);
3247
+	detach_page_private(page);
32513248 	return 1;
32523249 failed:
32533250 	return 0;
....@@ -3367,7 +3364,7 @@
33673364 	struct buffer_head *ret = kmem_cache_zalloc(bh_cachep, gfp_flags);
33683365 	if (ret) {
33693366 		INIT_LIST_HEAD(&ret->b_assoc_buffers);
3370
-		buffer_head_init_locks(ret);
3367
+		spin_lock_init(&ret->b_uptodate_lock);
33713368 		preempt_disable();
33723369 		__this_cpu_inc(bh_accounting.nr);
33733370 		recalc_bh_state();