Merge tag 'f2fs-for-5.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs

Pull f2fs updates from Jaegeuk Kim:
 "In this round, we've added two small interfaces: (a) GC_URGENT_LOW
  mode for performance and (b) F2FS_IOC_SEC_TRIM_FILE ioctl for
  security.

  The new GC mode allows Android to run some lower priority GCs in
  background, while new ioctl discards user information without race
  condition when the account is removed.

  In addition, some patches were merged to address latency-related
  issues. We've fixed some compression-related bug fixes as well as edge
  race conditions.

  Enhancements:
   - add GC_URGENT_LOW mode in gc_urgent
   - introduce F2FS_IOC_SEC_TRIM_FILE ioctl
   - bypass racy readahead to improve read latencies
   - shrink node_write lock coverage to avoid long latency

  Bug fixes:
   - fix missing compression flag control, i_size, and mount option
   - fix deadlock between quota writes and checkpoint
   - remove inode eviction path in synchronous path to avoid deadlock
   - fix to wait GCed compressed page writeback
   - fix a kernel panic in f2fs_is_compressed_page
   - check page dirty status before writeback
   - wait page writeback before update in node page write flow
   - fix a race condition between f2fs_write_end_io and f2fs_del_fsync_node_entry

  We've added some minor sanity checks and refactored trivial code
  blocks for better readability and debugging information"

* tag 'f2fs-for-5.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (52 commits)
  f2fs: prepare a waiter before entering io_schedule
  f2fs: update_sit_entry: Make the judgment condition of f2fs_bug_on more intuitive
  f2fs: replace test_and_set/clear_bit() with set/clear_bit()
  f2fs: make file immutable even if releasing zero compression block
  f2fs: compress: disable compression mount option if compression is off
  f2fs: compress: add sanity check during compressed cluster read
  f2fs: use macro instead of f2fs verity version
  f2fs: fix deadlock between quota writes and checkpoint
  f2fs: correct comment of f2fs_exist_written_data
  f2fs: compress: delay temp page allocation
  f2fs: compress: fix to update isize when overwriting compressed file
  f2fs: space related cleanup
  f2fs: fix use-after-free issue
  f2fs: Change the type of f2fs_flush_inline_data() to void
  f2fs: add F2FS_IOC_SEC_TRIM_FILE ioctl
  f2fs: should avoid inode eviction in synchronous path
  f2fs: segment.h: delete a duplicated word
  f2fs: compress: fix to avoid memory leak on cc->cpages
  f2fs: use generic names for generic ioctls
  f2fs: don't keep meta inode pages used for compressed block migration
  ...

Author: torvalds

Documentation/ABI/testing/sysfs-fs-f2fs

@@ -229,7 +229,9 @@ Date:		August 2017
 Contact:	"Jaegeuk Kim" <jaegeuk@kernel.org>
 Description:	Do background GC agressively when set. When gc_urgent = 1,
 		background thread starts to do GC by given gc_urgent_sleep_time
-		interval. It is set to 0 by default.
+		interval. When gc_urgent = 2, F2FS will lower the bar of
+		checking idle in order to process outstanding discard commands
+		and GC a little bit aggressively. It is set to 0 by default.
 
 What:		/sys/fs/f2fs/<disk>/gc_urgent_sleep_time
 Date:		August 2017

Documentation/filesystems/f2fs.rst

@@ -258,6 +258,8 @@ compress_extension=%s	 Support adding specified extension, so that f2fs can enab
 			 on compression extension list and enable compression on
 			 these file by default rather than to enable it via ioctl.
 			 For other files, we can still enable compression via ioctl.
+			 Note that, there is one reserved special extension '*', it
+			 can be set to enable compression for all files.
 inlinecrypt		 When possible, encrypt/decrypt the contents of encrypted
 			 files using the blk-crypto framework rather than
 			 filesystem-layer encryption. This allows the use of
@@ -743,8 +745,8 @@ Compression implementation
 
 - In order to eliminate write amplification during overwrite, F2FS only
   support compression on write-once file, data can be compressed only when
-  all logical blocks in file are valid and cluster compress ratio is lower
-  than specified threshold.
+  all logical blocks in cluster contain valid data and compress ratio of
+  cluster data is lower than specified threshold.
 
 - To enable compression on regular inode, there are three ways:
 

fs/f2fs/checkpoint.c

@@ -523,7 +523,7 @@ void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 	__remove_ino_entry(sbi, ino, type);
 }
 
-/* mode should be APPEND_INO or UPDATE_INO */
+/* mode should be APPEND_INO, UPDATE_INO or TRANS_DIR_INO */
 bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
 {
 	struct inode_management *im = &sbi->im[mode];
@@ -1258,8 +1258,6 @@ void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type)
 	DEFINE_WAIT(wait);
 
 	for (;;) {
-		prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE);
-
 		if (!get_pages(sbi, type))
 			break;
 
@@ -1269,6 +1267,10 @@ void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type)
 		if (type == F2FS_DIRTY_META)
 			f2fs_sync_meta_pages(sbi, META, LONG_MAX,
 							FS_CP_META_IO);
+		else if (type == F2FS_WB_CP_DATA)
+			f2fs_submit_merged_write(sbi, DATA);
+
+		prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE);
 		io_schedule_timeout(DEFAULT_IO_TIMEOUT);
 	}
 	finish_wait(&sbi->cp_wait, &wait);
@@ -1415,7 +1417,7 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
 	}
 
-	/* 2 cp  + n data seg summary + orphan inode blocks */
+	/* 2 cp + n data seg summary + orphan inode blocks */
 	data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false);
 	spin_lock_irqsave(&sbi->cp_lock, flags);
 	if (data_sum_blocks < NR_CURSEG_DATA_TYPE)
@@ -1515,9 +1517,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 
 	/*
 	 * invalidate intermediate page cache borrowed from meta inode which are
-	 * used for migration of encrypted or verity inode's blocks.
+	 * used for migration of encrypted, verity or compressed inode's blocks.
 	 */
-	if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi))
+	if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi) ||
+		f2fs_sb_has_compression(sbi))
 		invalidate_mapping_pages(META_MAPPING(sbi),
 				MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1);
 

fs/f2fs/compress.c

@@ -49,6 +49,13 @@ bool f2fs_is_compressed_page(struct page *page)
 		return false;
 	if (IS_ATOMIC_WRITTEN_PAGE(page) || IS_DUMMY_WRITTEN_PAGE(page))
 		return false;
+	/*
+	 * page->private may be set with pid.
+	 * pid_max is enough to check if it is traced.
+	 */
+	if (IS_IO_TRACED_PAGE(page))
+		return false;
+
 	f2fs_bug_on(F2FS_M_SB(page->mapping),
 		*((u32 *)page_private(page)) != F2FS_COMPRESSED_PAGE_MAGIC);
 	return true;
@@ -506,7 +513,7 @@ bool f2fs_is_compress_backend_ready(struct inode *inode)
 	return f2fs_cops[F2FS_I(inode)->i_compress_algorithm];
 }
 
-static mempool_t *compress_page_pool = NULL;
+static mempool_t *compress_page_pool;
 static int num_compress_pages = 512;
 module_param(num_compress_pages, uint, 0444);
 MODULE_PARM_DESC(num_compress_pages,
@@ -663,6 +670,7 @@ void f2fs_decompress_pages(struct bio *bio, struct page *page, bool verity)
 	const struct f2fs_compress_ops *cops =
 			f2fs_cops[fi->i_compress_algorithm];
 	int ret;
+	int i;
 
 	dec_page_count(sbi, F2FS_RD_DATA);
 
@@ -681,6 +689,26 @@ void f2fs_decompress_pages(struct bio *bio, struct page *page, bool verity)
 		goto out_free_dic;
 	}
 
+	dic->tpages = f2fs_kzalloc(sbi, sizeof(struct page *) *
+					dic->cluster_size, GFP_NOFS);
+	if (!dic->tpages) {
+		ret = -ENOMEM;
+		goto out_free_dic;
+	}
+
+	for (i = 0; i < dic->cluster_size; i++) {
+		if (dic->rpages[i]) {
+			dic->tpages[i] = dic->rpages[i];
+			continue;
+		}
+
+		dic->tpages[i] = f2fs_compress_alloc_page();
+		if (!dic->tpages[i]) {
+			ret = -ENOMEM;
+			goto out_free_dic;
+		}
+	}
+
 	if (cops->init_decompress_ctx) {
 		ret = cops->init_decompress_ctx(dic);
 		if (ret)
@@ -821,7 +849,7 @@ static int f2fs_compressed_blocks(struct compress_ctx *cc)
 }
 
 /* return # of valid blocks in compressed cluster */
-static int f2fs_cluster_blocks(struct compress_ctx *cc, bool compr)
+static int f2fs_cluster_blocks(struct compress_ctx *cc)
 {
 	return __f2fs_cluster_blocks(cc, false);
 }
@@ -835,7 +863,7 @@ int f2fs_is_compressed_cluster(struct inode *inode, pgoff_t index)
 		.cluster_idx = index >> F2FS_I(inode)->i_log_cluster_size,
 	};
 
-	return f2fs_cluster_blocks(&cc, false);
+	return f2fs_cluster_blocks(&cc);
 }
 
 static bool cluster_may_compress(struct compress_ctx *cc)
@@ -886,7 +914,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 	bool prealloc;
 
 retry:
-	ret = f2fs_cluster_blocks(cc, false);
+	ret = f2fs_cluster_blocks(cc);
 	if (ret <= 0)
 		return ret;
 
@@ -949,7 +977,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 	}
 
 	if (prealloc) {
-		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
+		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
 
 		set_new_dnode(&dn, cc->inode, NULL, NULL, 0);
 
@@ -964,7 +992,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 				break;
 		}
 
-		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
+		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
 	}
 
 	if (likely(!ret)) {
@@ -1096,8 +1124,16 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 	loff_t psize;
 	int i, err;
 
-	if (!IS_NOQUOTA(inode) && !f2fs_trylock_op(sbi))
+	if (IS_NOQUOTA(inode)) {
+		/*
+		 * We need to wait for node_write to avoid block allocation during
+		 * checkpoint. This can only happen to quota writes which can cause
+		 * the below discard race condition.
+		 */
+		down_read(&sbi->node_write);
+	} else if (!f2fs_trylock_op(sbi)) {
 		return -EAGAIN;
+	}
 
 	set_new_dnode(&dn, cc->inode, NULL, NULL, 0);
 
@@ -1137,6 +1173,13 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 		f2fs_set_compressed_page(cc->cpages[i], inode,
 					cc->rpages[i + 1]->index, cic);
 		fio.compressed_page = cc->cpages[i];
+
+		fio.old_blkaddr = data_blkaddr(dn.inode, dn.node_page,
+						dn.ofs_in_node + i + 1);
+
+		/* wait for GCed page writeback via META_MAPPING */
+		f2fs_wait_on_block_writeback(inode, fio.old_blkaddr);
+
 		if (fio.encrypted) {
 			fio.page = cc->rpages[i + 1];
 			err = f2fs_encrypt_one_page(&fio);
@@ -1203,7 +1246,9 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 		set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
 
 	f2fs_put_dnode(&dn);
-	if (!IS_NOQUOTA(inode))
+	if (IS_NOQUOTA(inode))
+		up_read(&sbi->node_write);
+	else
 		f2fs_unlock_op(sbi);
 
 	spin_lock(&fi->i_size_lock);
@@ -1230,7 +1275,9 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 out_put_dnode:
 	f2fs_put_dnode(&dn);
 out_unlock_op:
-	if (!IS_NOQUOTA(inode))
+	if (IS_NOQUOTA(inode))
+		up_read(&sbi->node_write);
+	else
 		f2fs_unlock_op(sbi);
 	return -EAGAIN;
 }
@@ -1310,6 +1357,12 @@ static int f2fs_write_raw_pages(struct compress_ctx *cc,
 				congestion_wait(BLK_RW_ASYNC,
 						DEFAULT_IO_TIMEOUT);
 				lock_page(cc->rpages[i]);
+
+				if (!PageDirty(cc->rpages[i])) {
+					unlock_page(cc->rpages[i]);
+					continue;
+				}
+
 				clear_page_dirty_for_io(cc->rpages[i]);
 				goto retry_write;
 			}
@@ -1353,6 +1406,8 @@ int f2fs_write_multi_pages(struct compress_ctx *cc,
 		err = f2fs_write_compressed_pages(cc, submitted,
 							wbc, io_type);
 		cops->destroy_compress_ctx(cc);
+		kfree(cc->cpages);
+		cc->cpages = NULL;
 		if (!err)
 			return 0;
 		f2fs_bug_on(F2FS_I_SB(cc->inode), err != -EAGAIN);
@@ -1415,22 +1470,6 @@ struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc)
 		dic->cpages[i] = page;
 	}
 
-	dic->tpages = f2fs_kzalloc(sbi, sizeof(struct page *) *
-					dic->cluster_size, GFP_NOFS);
-	if (!dic->tpages)
-		goto out_free;
-
-	for (i = 0; i < dic->cluster_size; i++) {
-		if (cc->rpages[i]) {
-			dic->tpages[i] = cc->rpages[i];
-			continue;
-		}
-
-		dic->tpages[i] = f2fs_compress_alloc_page();
-		if (!dic->tpages[i])
-			goto out_free;
-	}
-
 	return dic;
 
 out_free: