From 632faca72938f9f63049e48a8c438913828ac7a9 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 23 Dec 2020 11:44:25 -0800 Subject: [PATCH 01/32] f2fs: handle unallocated section and zone on pinned/atgc If we have large section/zone, unallocated segment makes them corrupted. E.g., - Pinned file: -1 119304647 119304647 - ATGC data: -1 119304647 119304647 Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/segment.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h index e81eb0748e2a..229814b4f4a6 100644 --- a/fs/f2fs/segment.h +++ b/fs/f2fs/segment.h @@ -101,11 +101,11 @@ static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi, #define BLKS_PER_SEC(sbi) \ ((sbi)->segs_per_sec * (sbi)->blocks_per_seg) #define GET_SEC_FROM_SEG(sbi, segno) \ - ((segno) / (sbi)->segs_per_sec) + (((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec) #define GET_SEG_FROM_SEC(sbi, secno) \ ((secno) * (sbi)->segs_per_sec) #define GET_ZONE_FROM_SEC(sbi, secno) \ - ((secno) / (sbi)->secs_per_zone) + (((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone) #define GET_ZONE_FROM_SEG(sbi, segno) \ GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) From 36218b81f094648d929994399eb6eb5c97b991e5 Mon Sep 17 00:00:00 2001 From: Zheng Yongjun Date: Tue, 22 Dec 2020 21:34:15 +0800 Subject: [PATCH 02/32] f2fs: Replace expression with offsetof() Use the existing offsetof() macro instead of duplicating code. Signed-off-by: Zheng Yongjun Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/node.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c index 3a24423ac65f..5e3fabacefb5 100644 --- a/fs/f2fs/node.c +++ b/fs/f2fs/node.c @@ -2696,7 +2696,7 @@ retry: src = F2FS_INODE(page); dst = F2FS_INODE(ipage); - memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); + memcpy(dst, src, offsetof(struct f2fs_inode, i_ext)); dst->i_size = 0; dst->i_blocks = cpu_to_le64(1); dst->i_links = cpu_to_le32(1); From a28d9aa1a2c7c774c38f2da1a662434bc29cb98e Mon Sep 17 00:00:00 2001 From: Weichao Guo Date: Mon, 14 Dec 2020 11:54:53 +0800 Subject: [PATCH 03/32] f2fs: fix to set inode->i_mode correctly for posix_acl_update_mode We should update the ~S_IRWXUGO part of inode->i_mode in __setattr_copy, because posix_acl_update_mode updates mode based on inode->i_mode, which finally overwrites the ~S_IRWXUGO part of i_acl_mode with old i_mode. Testcase to reproduce this bug: 0. adduser abc 1. mkfs.f2fs /dev/sdd 2. mount -t f2fs /dev/sdd /mnt/f2fs 3. mkdir /mnt/f2fs/test 4. setfacl -m u:abc:r /mnt/f2fs/test 5. chmod +s /mnt/f2fs/test Signed-off-by: Weichao Guo Signed-off-by: Bin Shu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index f585545277d7..eced14882fc1 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -850,6 +850,7 @@ static void __setattr_copy(struct inode *inode, const struct iattr *attr) if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) mode &= ~S_ISGID; + inode->i_mode = (inode->i_mode & S_IRWXUGO) | (mode & ~S_IRWXUGO); set_acl_inode(inode, mode); } } From 17232e830afb800acdcc22ae8980bf9d330393ef Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Fri, 25 Dec 2020 16:52:27 +0800 Subject: [PATCH 04/32] f2fs: enhance to update i_mode and acl atomically in f2fs_setattr() Previously, in f2fs_setattr(), we don't update S_ISUID|S_ISGID|S_ISVTX bits with S_IRWXUGO bits and acl entries atomically, so in error path, chmod() may partially success, this patch enhances to make chmod() flow being atomical. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/acl.c | 23 ++++++++++++++++++++++- fs/f2fs/file.c | 7 ++++--- fs/f2fs/xattr.c | 15 +++++++++------ 3 files changed, 35 insertions(+), 10 deletions(-) diff --git a/fs/f2fs/acl.c b/fs/f2fs/acl.c index 1e5e9b1136ee..732ec10e7890 100644 --- a/fs/f2fs/acl.c +++ b/fs/f2fs/acl.c @@ -200,6 +200,27 @@ struct posix_acl *f2fs_get_acl(struct inode *inode, int type) return __f2fs_get_acl(inode, type, NULL); } +static int f2fs_acl_update_mode(struct inode *inode, umode_t *mode_p, + struct posix_acl **acl) +{ + umode_t mode = inode->i_mode; + int error; + + if (is_inode_flag_set(inode, FI_ACL_MODE)) + mode = F2FS_I(inode)->i_acl_mode; + + error = posix_acl_equiv_mode(*acl, &mode); + if (error < 0) + return error; + if (error == 0) + *acl = NULL; + if (!in_group_p(inode->i_gid) && + !capable_wrt_inode_uidgid(inode, CAP_FSETID)) + mode &= ~S_ISGID; + *mode_p = mode; + return 0; +} + static int __f2fs_set_acl(struct inode *inode, int type, struct posix_acl *acl, struct page *ipage) { @@ -213,7 +234,7 @@ static int __f2fs_set_acl(struct inode *inode, int type, case ACL_TYPE_ACCESS: name_index = F2FS_XATTR_INDEX_POSIX_ACL_ACCESS; if (acl && !ipage) { - error = posix_acl_update_mode(inode, &mode, &acl); + error = f2fs_acl_update_mode(inode, &mode, &acl); if (error) return error; set_acl_inode(inode, mode); diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index eced14882fc1..2ddc4baaf173 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -850,7 +850,6 @@ static void __setattr_copy(struct inode *inode, const struct iattr *attr) if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) mode &= ~S_ISGID; - inode->i_mode = (inode->i_mode & S_IRWXUGO) | (mode & ~S_IRWXUGO); set_acl_inode(inode, mode); } } @@ -950,8 +949,10 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr) if (attr->ia_valid & ATTR_MODE) { err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode)); - if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { - inode->i_mode = F2FS_I(inode)->i_acl_mode; + + if (is_inode_flag_set(inode, FI_ACL_MODE)) { + if (!err) + inode->i_mode = F2FS_I(inode)->i_acl_mode; clear_inode_flag(inode, FI_ACL_MODE); } } diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c index 65afcc3cc68a..2086bef6c154 100644 --- a/fs/f2fs/xattr.c +++ b/fs/f2fs/xattr.c @@ -673,7 +673,7 @@ static int __f2fs_setxattr(struct inode *inode, int index, } if (value && f2fs_xattr_value_same(here, value, size)) - goto exit; + goto same; } else if ((flags & XATTR_REPLACE)) { error = -ENODATA; goto exit; @@ -738,17 +738,20 @@ static int __f2fs_setxattr(struct inode *inode, int index, if (error) goto exit; - if (is_inode_flag_set(inode, FI_ACL_MODE)) { - inode->i_mode = F2FS_I(inode)->i_acl_mode; - inode->i_ctime = current_time(inode); - clear_inode_flag(inode, FI_ACL_MODE); - } if (index == F2FS_XATTR_INDEX_ENCRYPTION && !strcmp(name, F2FS_XATTR_NAME_ENCRYPTION_CONTEXT)) f2fs_set_encrypted_inode(inode); f2fs_mark_inode_dirty_sync(inode, true); if (!error && S_ISDIR(inode->i_mode)) set_sbi_flag(F2FS_I_SB(inode), SBI_NEED_CP); + +same: + if (is_inode_flag_set(inode, FI_ACL_MODE)) { + inode->i_mode = F2FS_I(inode)->i_acl_mode; + inode->i_ctime = current_time(inode); + clear_inode_flag(inode, FI_ACL_MODE); + } + exit: kfree(base_addr); return error; From e0fcd01510ad025c9bbce704c5c2579294056141 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Sat, 26 Dec 2020 18:07:01 +0800 Subject: [PATCH 05/32] f2fs: enforce the immutable flag on open files This patch ports commit 02b016ca7f99 ("ext4: enforce the immutable flag on open files") to f2fs. According to the chattr man page, "a file with the 'i' attribute cannot be modified..." Historically, this was only enforced when the file was opened, per the rest of the description, "... and the file can not be opened in write mode". There is general agreement that we should standardize all file systems to prevent modifications even for files that were opened at the time the immutable flag is set. Eventually, a change to enforce this at the VFS layer should be landing in mainline. Cc: stable@kernel.org Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 2ddc4baaf173..d57b54643918 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -60,6 +60,9 @@ static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) bool need_alloc = true; int err = 0; + if (unlikely(IS_IMMUTABLE(inode))) + return VM_FAULT_SIGBUS; + if (unlikely(f2fs_cp_error(sbi))) { err = -EIO; goto err; @@ -865,6 +868,14 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr) if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) return -EIO; + if (unlikely(IS_IMMUTABLE(inode))) + return -EPERM; + + if (unlikely(IS_APPEND(inode) && + (attr->ia_valid & (ATTR_MODE | ATTR_UID | + ATTR_GID | ATTR_TIMES_SET)))) + return -EPERM; + if ((attr->ia_valid & ATTR_SIZE) && !f2fs_is_compress_backend_ready(inode)) return -EOPNOTSUPP; @@ -4351,6 +4362,11 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) inode_lock(inode); } + if (unlikely(IS_IMMUTABLE(inode))) { + ret = -EPERM; + goto unlock; + } + ret = generic_write_checks(iocb, from); if (ret > 0) { bool preallocated = false; @@ -4415,6 +4431,7 @@ write: if (ret > 0) f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); } +unlock: inode_unlock(inode); out: trace_f2fs_file_write_iter(inode, iocb->ki_pos, From 0b979f1bded3e6808184842133e6afeba312a4ff Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Sat, 26 Dec 2020 18:07:41 +0800 Subject: [PATCH 06/32] f2fs: relocate f2fs_precache_extents() Relocate f2fs_precache_extents() in prior to check_swap_activate(), then extent cache can be enabled before its use. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index aa34d620bec9..57b9aab2b142 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -4108,12 +4108,13 @@ static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file, if (!f2fs_disable_compressed_file(inode)) return -EINVAL; + f2fs_precache_extents(inode); + ret = check_swap_activate(sis, file, span); if (ret < 0) return ret; set_inode_flag(inode, FI_PIN_FILE); - f2fs_precache_extents(inode); f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); return ret; } From 32be0e97c71366a19d11d1965e3f0957ea0be609 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Fri, 22 Jan 2021 17:40:13 +0800 Subject: [PATCH 07/32] f2fs: compress: deny setting unsupported compress algorithm If kernel doesn't support certain kinds of compress algorithm, deny to set them as compress algorithm of f2fs via 'compress_algorithm=%s' mount option. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/super.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index b4a07fe62d1a..a275bd312ae5 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -882,17 +882,33 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount) if (!name) return -ENOMEM; if (!strcmp(name, "lzo")) { +#ifdef CONFIG_F2FS_FS_LZO F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO; +#else + f2fs_info(sbi, "kernel doesn't support lzo compression"); +#endif } else if (!strcmp(name, "lz4")) { +#ifdef CONFIG_F2FS_FS_LZ4 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; +#else + f2fs_info(sbi, "kernel doesn't support lz4 compression"); +#endif } else if (!strcmp(name, "zstd")) { +#ifdef CONFIG_F2FS_FS_ZSTD F2FS_OPTION(sbi).compress_algorithm = COMPRESS_ZSTD; +#else + f2fs_info(sbi, "kernel doesn't support zstd compression"); +#endif } else if (!strcmp(name, "lzo-rle")) { +#ifdef CONFIG_F2FS_FS_LZORLE F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZORLE; +#else + f2fs_info(sbi, "kernel doesn't support lzorle compression"); +#endif } else { kfree(name); return -EINVAL; From 3fde13f817e23f05ce407d136325df4cbc913e67 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Fri, 22 Jan 2021 17:46:43 +0800 Subject: [PATCH 08/32] f2fs: compress: support compress level Expand 'compress_algorithm' mount option to accept parameter as format of :, by this way, it gives a way to allow user to do more specified config on lz4 and zstd compression level, then f2fs compression can provide higher compress ratio. In order to set compress level for lz4 algorithm, it needs to set CONFIG_LZ4HC_COMPRESS and CONFIG_F2FS_FS_LZ4HC config to enable lz4hc compress algorithm. CR and performance number on lz4/lz4hc algorithm: dd if=enwik9 of=compressed_file conv=fsync Original blocks: 244382 lz4 lz4hc-9 compressed blocks 170647 163270 compress ratio 69.8% 66.8% speed 16.4207 s, 60.9 MB/s 26.7299 s, 37.4 MB/s compress ratio = after / before Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.rst | 5 ++ fs/f2fs/Kconfig | 10 ++++ fs/f2fs/compress.c | 41 +++++++++++++- fs/f2fs/f2fs.h | 9 +++ fs/f2fs/super.c | 89 +++++++++++++++++++++++++++++- include/linux/f2fs_fs.h | 3 + 6 files changed, 152 insertions(+), 5 deletions(-) diff --git a/Documentation/filesystems/f2fs.rst b/Documentation/filesystems/f2fs.rst index dae15c96e659..5eff4009e77e 100644 --- a/Documentation/filesystems/f2fs.rst +++ b/Documentation/filesystems/f2fs.rst @@ -249,6 +249,11 @@ checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enabl This space is reclaimed once checkpoint=enable. compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo", "lz4", "zstd" and "lzo-rle" algorithm. +compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only + "lz4" and "zstd" support compress level config. + algorithm level range + lz4 3 - 16 + zstd 1 - 22 compress_log_size=%u Support configuring compress cluster size, the size will be 4KB * (1 << %u), 16KB is minimum size, also it's default size. diff --git a/fs/f2fs/Kconfig b/fs/f2fs/Kconfig index d13c5c6a9787..63c1fc1a0e3b 100644 --- a/fs/f2fs/Kconfig +++ b/fs/f2fs/Kconfig @@ -119,6 +119,16 @@ config F2FS_FS_LZ4 help Support LZ4 compress algorithm, if unsure, say Y. +config F2FS_FS_LZ4HC + bool "LZ4HC compression support" + depends on F2FS_FS_COMPRESSION + depends on F2FS_FS_LZ4 + select LZ4HC_COMPRESS + default y + help + Support LZ4HC compress algorithm, LZ4HC has compatible on-disk + layout with LZ4, if unsure, say Y. + config F2FS_FS_ZSTD bool "ZSTD compression support" depends on F2FS_FS_COMPRESSION diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c index 4bcbacfe3325..a345a41e2119 100644 --- a/fs/f2fs/compress.c +++ b/fs/f2fs/compress.c @@ -252,8 +252,14 @@ static const struct f2fs_compress_ops f2fs_lzo_ops = { #ifdef CONFIG_F2FS_FS_LZ4 static int lz4_init_compress_ctx(struct compress_ctx *cc) { - cc->private = f2fs_kvmalloc(F2FS_I_SB(cc->inode), - LZ4_MEM_COMPRESS, GFP_NOFS); + unsigned int size = LZ4_MEM_COMPRESS; + +#ifdef CONFIG_F2FS_FS_LZ4HC + if (F2FS_I(cc->inode)->i_compress_flag >> COMPRESS_LEVEL_OFFSET) + size = LZ4HC_MEM_COMPRESS; +#endif + + cc->private = f2fs_kvmalloc(F2FS_I_SB(cc->inode), size, GFP_NOFS); if (!cc->private) return -ENOMEM; @@ -272,10 +278,34 @@ static void lz4_destroy_compress_ctx(struct compress_ctx *cc) cc->private = NULL; } +#ifdef CONFIG_F2FS_FS_LZ4HC +static int lz4hc_compress_pages(struct compress_ctx *cc) +{ + unsigned char level = F2FS_I(cc->inode)->i_compress_flag >> + COMPRESS_LEVEL_OFFSET; + int len; + + if (level) + len = LZ4_compress_HC(cc->rbuf, cc->cbuf->cdata, cc->rlen, + cc->clen, level, cc->private); + else + len = LZ4_compress_default(cc->rbuf, cc->cbuf->cdata, cc->rlen, + cc->clen, cc->private); + if (!len) + return -EAGAIN; + + cc->clen = len; + return 0; +} +#endif + static int lz4_compress_pages(struct compress_ctx *cc) { int len; +#ifdef CONFIG_F2FS_FS_LZ4HC + return lz4hc_compress_pages(cc); +#endif len = LZ4_compress_default(cc->rbuf, cc->cbuf->cdata, cc->rlen, cc->clen, cc->private); if (!len) @@ -325,8 +355,13 @@ static int zstd_init_compress_ctx(struct compress_ctx *cc) ZSTD_CStream *stream; void *workspace; unsigned int workspace_size; + unsigned char level = F2FS_I(cc->inode)->i_compress_flag >> + COMPRESS_LEVEL_OFFSET; - params = ZSTD_getParams(F2FS_ZSTD_DEFAULT_CLEVEL, cc->rlen, 0); + if (!level) + level = F2FS_ZSTD_DEFAULT_CLEVEL; + + params = ZSTD_getParams(level, cc->rlen, 0); workspace_size = ZSTD_CStreamWorkspaceBound(params.cParams); workspace = f2fs_kvmalloc(F2FS_I_SB(cc->inode), diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index bb11759191dc..36012181c17f 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -146,6 +146,7 @@ struct f2fs_mount_info { /* For compression */ unsigned char compress_algorithm; /* algorithm type */ unsigned char compress_log_size; /* cluster log size */ + unsigned char compress_level; /* compress level */ bool compress_chksum; /* compressed data chksum */ unsigned char compress_ext_cnt; /* extension count */ int compress_mode; /* compression mode */ @@ -735,6 +736,7 @@ struct f2fs_inode_info { atomic_t i_compr_blocks; /* # of compressed blocks */ unsigned char i_compress_algorithm; /* algorithm type */ unsigned char i_log_cluster_size; /* log of cluster size */ + unsigned char i_compress_level; /* compress level (lz4hc,zstd) */ unsigned short i_compress_flag; /* compress flag */ unsigned int i_cluster_size; /* cluster size */ }; @@ -1310,6 +1312,8 @@ struct compress_data { #define F2FS_COMPRESSED_PAGE_MAGIC 0xF5F2C000 +#define COMPRESS_LEVEL_OFFSET 8 + /* compress context */ struct compress_ctx { struct inode *inode; /* inode the context belong to */ @@ -3934,6 +3938,11 @@ static inline void set_compress_context(struct inode *inode) 1 << COMPRESS_CHKSUM : 0; F2FS_I(inode)->i_cluster_size = 1 << F2FS_I(inode)->i_log_cluster_size; + if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_LZ4 && + F2FS_OPTION(sbi).compress_level) + F2FS_I(inode)->i_compress_flag |= + F2FS_OPTION(sbi).compress_level << + COMPRESS_LEVEL_OFFSET; F2FS_I(inode)->i_flags |= F2FS_COMPR_FL; set_inode_flag(inode, FI_COMPRESSED_FILE); stat_inc_compr_inode(inode); diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index a275bd312ae5..c8be27a9eed6 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -25,6 +25,8 @@ #include #include #include +#include +#include #include "f2fs.h" #include "node.h" @@ -464,6 +466,74 @@ static int f2fs_set_test_dummy_encryption(struct super_block *sb, return 0; } +#ifdef CONFIG_F2FS_FS_COMPRESSION +#ifdef CONFIG_F2FS_FS_LZ4 +static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) +{ +#ifdef CONFIG_F2FS_FS_LZ4HC + unsigned int level; +#endif + + if (strlen(str) == 3) { + F2FS_OPTION(sbi).compress_level = 0; + return 0; + } + +#ifdef CONFIG_F2FS_FS_LZ4HC + str += 3; + + if (str[0] != ':') { + f2fs_info(sbi, "wrong format, e.g. :"); + return -EINVAL; + } + if (kstrtouint(str + 1, 10, &level)) + return -EINVAL; + + if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { + f2fs_info(sbi, "invalid lz4hc compress level: %d", level); + return -EINVAL; + } + + F2FS_OPTION(sbi).compress_level = level; + return 0; +#else + f2fs_info(sbi, "kernel doesn't support lz4hc compression"); + return -EINVAL; +#endif +} +#endif + +#ifdef CONFIG_F2FS_FS_ZSTD +static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) +{ + unsigned int level; + int len = 4; + + if (strlen(str) == len) { + F2FS_OPTION(sbi).compress_level = 0; + return 0; + } + + str += len; + + if (str[0] != ':') { + f2fs_info(sbi, "wrong format, e.g. :"); + return -EINVAL; + } + if (kstrtouint(str + 1, 10, &level)) + return -EINVAL; + + if (!level || level > ZSTD_maxCLevel()) { + f2fs_info(sbi, "invalid zstd compress level: %d", level); + return -EINVAL; + } + + F2FS_OPTION(sbi).compress_level = level; + return 0; +} +#endif +#endif + static int parse_options(struct super_block *sb, char *options, bool is_remount) { struct f2fs_sb_info *sbi = F2FS_SB(sb); @@ -883,20 +953,31 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount) return -ENOMEM; if (!strcmp(name, "lzo")) { #ifdef CONFIG_F2FS_FS_LZO + F2FS_OPTION(sbi).compress_level = 0; F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO; #else f2fs_info(sbi, "kernel doesn't support lzo compression"); #endif - } else if (!strcmp(name, "lz4")) { + } else if (!strncmp(name, "lz4", 3)) { #ifdef CONFIG_F2FS_FS_LZ4 + ret = f2fs_set_lz4hc_level(sbi, name); + if (ret) { + kfree(name); + return -EINVAL; + } F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; #else f2fs_info(sbi, "kernel doesn't support lz4 compression"); #endif - } else if (!strcmp(name, "zstd")) { + } else if (!strncmp(name, "zstd", 4)) { #ifdef CONFIG_F2FS_FS_ZSTD + ret = f2fs_set_zstd_level(sbi, name); + if (ret) { + kfree(name); + return -EINVAL; + } F2FS_OPTION(sbi).compress_algorithm = COMPRESS_ZSTD; #else @@ -904,6 +985,7 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount) #endif } else if (!strcmp(name, "lzo-rle")) { #ifdef CONFIG_F2FS_FS_LZORLE + F2FS_OPTION(sbi).compress_level = 0; F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZORLE; #else @@ -1555,6 +1637,9 @@ static inline void f2fs_show_compress_options(struct seq_file *seq, } seq_printf(seq, ",compress_algorithm=%s", algtype); + if (F2FS_OPTION(sbi).compress_level) + seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); + seq_printf(seq, ",compress_log_size=%u", F2FS_OPTION(sbi).compress_log_size); diff --git a/include/linux/f2fs_fs.h b/include/linux/f2fs_fs.h index 7dc2a06cf19a..c6cc0a566ef5 100644 --- a/include/linux/f2fs_fs.h +++ b/include/linux/f2fs_fs.h @@ -274,6 +274,9 @@ struct f2fs_inode { __u8 i_compress_algorithm; /* compress algorithm */ __u8 i_log_cluster_size; /* log of cluster size */ __le16 i_compress_flag; /* compress flag */ + /* 0 bit: chksum flag + * [10,15] bits: compress level + */ __le32 i_extra_end[0]; /* for attribute size calculation */ } __packed; __le32 i_addr[DEF_ADDRS_PER_INODE]; /* Pointers to data blocks */ From 5d4daa579e56adc97fb77c7dfda6c1f747c9ef25 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Wed, 9 Dec 2020 16:43:27 +0800 Subject: [PATCH 09/32] f2fs: introduce a new per-sb directory in sysfs Add a new directory 'stat' in path of /sys/fs/f2fs//, later we can add new readonly stat sysfs file into this directory, it will make directory less mess. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/f2fs.h | 5 +++- fs/f2fs/sysfs.c | 69 +++++++++++++++++++++++++++++++++++++++++++++---- 2 files changed, 68 insertions(+), 6 deletions(-) diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 36012181c17f..ccbbf86d14e5 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1545,9 +1545,12 @@ struct f2fs_sb_info { unsigned int node_io_flag; /* For sysfs suppport */ - struct kobject s_kobj; + struct kobject s_kobj; /* /sys/fs/f2fs/ */ struct completion s_kobj_unregister; + struct kobject s_stat_kobj; /* /sys/fs/f2fs//stat */ + struct completion s_stat_kobj_unregister; + /* For shrinker support */ struct list_head s_list; int s_ndevs; /* number of devices */ diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c index 30bae57428d1..bd1174ed2e6f 100644 --- a/fs/f2fs/sysfs.c +++ b/fs/f2fs/sysfs.c @@ -702,6 +702,11 @@ static struct attribute *f2fs_feat_attrs[] = { }; ATTRIBUTE_GROUPS(f2fs_feat); +static struct attribute *f2fs_stat_attrs[] = { + NULL, +}; +ATTRIBUTE_GROUPS(f2fs_stat); + static const struct sysfs_ops f2fs_attr_ops = { .show = f2fs_attr_show, .store = f2fs_attr_store, @@ -730,6 +735,44 @@ static struct kobject f2fs_feat = { .kset = &f2fs_kset, }; +static ssize_t f2fs_stat_attr_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, + s_stat_kobj); + struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); + + return a->show ? a->show(a, sbi, buf) : 0; +} + +static ssize_t f2fs_stat_attr_store(struct kobject *kobj, struct attribute *attr, + const char *buf, size_t len) +{ + struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, + s_stat_kobj); + struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); + + return a->store ? a->store(a, sbi, buf, len) : 0; +} + +static void f2fs_stat_kobj_release(struct kobject *kobj) +{ + struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, + s_stat_kobj); + complete(&sbi->s_stat_kobj_unregister); +} + +static const struct sysfs_ops f2fs_stat_attr_ops = { + .show = f2fs_stat_attr_show, + .store = f2fs_stat_attr_store, +}; + +static struct kobj_type f2fs_stat_ktype = { + .default_groups = f2fs_stat_groups, + .sysfs_ops = &f2fs_stat_attr_ops, + .release = f2fs_stat_kobj_release, +}; + static int __maybe_unused segment_info_seq_show(struct seq_file *seq, void *offset) { @@ -936,11 +979,15 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi) init_completion(&sbi->s_kobj_unregister); err = kobject_init_and_add(&sbi->s_kobj, &f2fs_sb_ktype, NULL, "%s", sb->s_id); - if (err) { - kobject_put(&sbi->s_kobj); - wait_for_completion(&sbi->s_kobj_unregister); - return err; - } + if (err) + goto put_sb_kobj; + + sbi->s_stat_kobj.kset = &f2fs_kset; + init_completion(&sbi->s_stat_kobj_unregister); + err = kobject_init_and_add(&sbi->s_stat_kobj, &f2fs_stat_ktype, + &sbi->s_kobj, "stat"); + if (err) + goto put_stat_kobj; if (f2fs_proc_root) sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); @@ -956,6 +1003,13 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi) victim_bits_seq_show, sb); } return 0; +put_stat_kobj: + kobject_put(&sbi->s_stat_kobj); + wait_for_completion(&sbi->s_stat_kobj_unregister); +put_sb_kobj: + kobject_put(&sbi->s_kobj); + wait_for_completion(&sbi->s_kobj_unregister); + return err; } void f2fs_unregister_sysfs(struct f2fs_sb_info *sbi) @@ -967,6 +1021,11 @@ void f2fs_unregister_sysfs(struct f2fs_sb_info *sbi) remove_proc_entry("victim_bits", sbi->s_proc); remove_proc_entry(sbi->sb->s_id, f2fs_proc_root); } + + kobject_del(&sbi->s_stat_kobj); + kobject_put(&sbi->s_stat_kobj); + wait_for_completion(&sbi->s_stat_kobj_unregister); + kobject_del(&sbi->s_kobj); kobject_put(&sbi->s_kobj); wait_for_completion(&sbi->s_kobj_unregister); From 0953fe864c4d05f5a5cde626a630a76918cf4f9c Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Mon, 14 Dec 2020 17:20:57 +0800 Subject: [PATCH 10/32] f2fs: fix to tag FIEMAP_EXTENT_MERGED in f2fs_fiemap() f2fs does not natively support extents in metadata, 'extent' in f2fs is used as a virtual concept, so in f2fs_fiemap() interface, it needs to tag FIEMAP_EXTENT_MERGED flag to indicated the extent status is a result of merging. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 57b9aab2b142..547c9d4b430b 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -1964,6 +1964,7 @@ next: } if (size) { + flags |= FIEMAP_EXTENT_MERGED; if (IS_ENCRYPTED(inode)) flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; From 2562515f0ad7342bde6456602c491b64c63fe950 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Wed, 16 Dec 2020 17:15:23 +0800 Subject: [PATCH 11/32] f2fs: fix out-of-repair __setattr_copy() __setattr_copy() was copied from setattr_copy() in fs/attr.c, there is two missing patches doesn't cover this inner function, fix it. Commit 7fa294c8991c ("userns: Allow chown and setgid preservation") Commit 23adbe12ef7d ("fs,userns: Change inode_capable to capable_wrt_inode_uidgid") Fixes: fbfa2cc58d53 ("f2fs: add file operations") Cc: stable@vger.kernel.org Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index d57b54643918..4e6d4b9120a8 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -851,7 +851,8 @@ static void __setattr_copy(struct inode *inode, const struct iattr *attr) if (ia_valid & ATTR_MODE) { umode_t mode = attr->ia_mode; - if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) + if (!in_group_p(inode->i_gid) && + !capable_wrt_inode_uidgid(inode, CAP_FSETID)) mode &= ~S_ISGID; set_acl_inode(inode, mode); } From cf7404036019fada99d99ea01f49cb5c3142099d Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Wed, 30 Dec 2020 16:38:35 +0800 Subject: [PATCH 12/32] f2fs: trival cleanup in move_data_block() Trival cleanups: - relocate set_summary() before its use - relocate "allocate block address" to correct place - remove unneeded f2fs_wait_on_page_writeback() Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/gc.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index 3ef84e6ded41..39330ad3c44e 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -1169,8 +1169,6 @@ static int move_data_block(struct inode *inode, block_t bidx, if (err) goto put_out; - set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); - /* read page */ fio.page = page; fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; @@ -1207,6 +1205,9 @@ static int move_data_block(struct inode *inode, block_t bidx, } } + set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); + + /* allocate block address */ f2fs_allocate_data_block(fio.sbi, NULL, fio.old_blkaddr, &newaddr, &sum, type, NULL); @@ -1233,9 +1234,6 @@ static int move_data_block(struct inode *inode, block_t bidx, set_page_writeback(fio.encrypted_page); ClearPageError(page); - /* allocate block address */ - f2fs_wait_on_page_writeback(dn.node_page, NODE, true, true); - fio.op = REQ_OP_WRITE; fio.op_flags = REQ_SYNC; fio.new_blkaddr = newaddr; From 7f59b277f79e8aacaa2ec7e549be6c27985c27f2 Mon Sep 17 00:00:00 2001 From: Eric Biggers Date: Mon, 4 Jan 2021 22:33:02 -0800 Subject: [PATCH 13/32] f2fs: clean up post-read processing Rework the post-read processing logic to be much easier to understand. At least one bug is fixed by this: if an I/O error occurred when reading from disk, decryption and verity would be performed on the uninitialized data, causing misleading messages in the kernel log. Signed-off-by: Eric Biggers Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/compress.c | 149 +++++++++++++----- fs/f2fs/data.c | 375 ++++++++++++++++++--------------------------- fs/f2fs/f2fs.h | 55 ++++++- 3 files changed, 306 insertions(+), 273 deletions(-) diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c index a345a41e2119..1696f9183ff5 100644 --- a/fs/f2fs/compress.c +++ b/fs/f2fs/compress.c @@ -756,38 +756,27 @@ out: return ret; } -void f2fs_decompress_pages(struct bio *bio, struct page *page, bool verity) +static void f2fs_decompress_cluster(struct decompress_io_ctx *dic) { - struct decompress_io_ctx *dic = - (struct decompress_io_ctx *)page_private(page); struct f2fs_sb_info *sbi = F2FS_I_SB(dic->inode); - struct f2fs_inode_info *fi= F2FS_I(dic->inode); + struct f2fs_inode_info *fi = F2FS_I(dic->inode); const struct f2fs_compress_ops *cops = f2fs_cops[fi->i_compress_algorithm]; int ret; int i; - dec_page_count(sbi, F2FS_RD_DATA); - - if (bio->bi_status || PageError(page)) - dic->failed = true; - - if (atomic_dec_return(&dic->pending_pages)) - return; - trace_f2fs_decompress_pages_start(dic->inode, dic->cluster_idx, dic->cluster_size, fi->i_compress_algorithm); - /* submit partial compressed pages */ if (dic->failed) { ret = -EIO; - goto out_free_dic; + goto out_end_io; } dic->tpages = page_array_alloc(dic->inode, dic->cluster_size); if (!dic->tpages) { ret = -ENOMEM; - goto out_free_dic; + goto out_end_io; } for (i = 0; i < dic->cluster_size; i++) { @@ -799,20 +788,20 @@ void f2fs_decompress_pages(struct bio *bio, struct page *page, bool verity) dic->tpages[i] = f2fs_compress_alloc_page(); if (!dic->tpages[i]) { ret = -ENOMEM; - goto out_free_dic; + goto out_end_io; } } if (cops->init_decompress_ctx) { ret = cops->init_decompress_ctx(dic); if (ret) - goto out_free_dic; + goto out_end_io; } dic->rbuf = f2fs_vmap(dic->tpages, dic->cluster_size); if (!dic->rbuf) { ret = -ENOMEM; - goto destroy_decompress_ctx; + goto out_destroy_decompress_ctx; } dic->cbuf = f2fs_vmap(dic->cpages, dic->nr_cpages); @@ -851,18 +840,34 @@ out_vunmap_cbuf: vm_unmap_ram(dic->cbuf, dic->nr_cpages); out_vunmap_rbuf: vm_unmap_ram(dic->rbuf, dic->cluster_size); -destroy_decompress_ctx: +out_destroy_decompress_ctx: if (cops->destroy_decompress_ctx) cops->destroy_decompress_ctx(dic); -out_free_dic: - if (!verity) - f2fs_decompress_end_io(dic->rpages, dic->cluster_size, - ret, false); - +out_end_io: trace_f2fs_decompress_pages_end(dic->inode, dic->cluster_idx, dic->clen, ret); - if (!verity) - f2fs_free_dic(dic); + f2fs_decompress_end_io(dic, ret); +} + +/* + * This is called when a page of a compressed cluster has been read from disk + * (or failed to be read from disk). It checks whether this page was the last + * page being waited on in the cluster, and if so, it decompresses the cluster + * (or in the case of a failure, cleans up without actually decompressing). + */ +void f2fs_end_read_compressed_page(struct page *page, bool failed) +{ + struct decompress_io_ctx *dic = + (struct decompress_io_ctx *)page_private(page); + struct f2fs_sb_info *sbi = F2FS_I_SB(dic->inode); + + dec_page_count(sbi, F2FS_RD_DATA); + + if (failed) + WRITE_ONCE(dic->failed, true); + + if (atomic_dec_and_test(&dic->remaining_pages)) + f2fs_decompress_cluster(dic); } static bool is_page_in_cluster(struct compress_ctx *cc, pgoff_t index) @@ -1529,6 +1534,8 @@ destroy_out: return err; } +static void f2fs_free_dic(struct decompress_io_ctx *dic); + struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc) { struct decompress_io_ctx *dic; @@ -1547,12 +1554,14 @@ struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc) dic->magic = F2FS_COMPRESSED_PAGE_MAGIC; dic->inode = cc->inode; - atomic_set(&dic->pending_pages, cc->nr_cpages); + atomic_set(&dic->remaining_pages, cc->nr_cpages); dic->cluster_idx = cc->cluster_idx; dic->cluster_size = cc->cluster_size; dic->log_cluster_size = cc->log_cluster_size; dic->nr_cpages = cc->nr_cpages; + refcount_set(&dic->refcnt, 1); dic->failed = false; + dic->need_verity = f2fs_need_verity(cc->inode, start_idx); for (i = 0; i < dic->cluster_size; i++) dic->rpages[i] = cc->rpages[i]; @@ -1581,7 +1590,7 @@ out_free: return ERR_PTR(-ENOMEM); } -void f2fs_free_dic(struct decompress_io_ctx *dic) +static void f2fs_free_dic(struct decompress_io_ctx *dic) { int i; @@ -1609,30 +1618,88 @@ void f2fs_free_dic(struct decompress_io_ctx *dic) kmem_cache_free(dic_entry_slab, dic); } -void f2fs_decompress_end_io(struct page **rpages, - unsigned int cluster_size, bool err, bool verity) +static void f2fs_put_dic(struct decompress_io_ctx *dic) +{ + if (refcount_dec_and_test(&dic->refcnt)) + f2fs_free_dic(dic); +} + +/* + * Update and unlock the cluster's pagecache pages, and release the reference to + * the decompress_io_ctx that was being held for I/O completion. + */ +static void __f2fs_decompress_end_io(struct decompress_io_ctx *dic, bool failed) { int i; - for (i = 0; i < cluster_size; i++) { - struct page *rpage = rpages[i]; + for (i = 0; i < dic->cluster_size; i++) { + struct page *rpage = dic->rpages[i]; if (!rpage) continue; - if (err || PageError(rpage)) - goto clear_uptodate; - - if (!verity || fsverity_verify_page(rpage)) { + /* PG_error was set if verity failed. */ + if (failed || PageError(rpage)) { + ClearPageUptodate(rpage); + /* will re-read again later */ + ClearPageError(rpage); + } else { SetPageUptodate(rpage); - goto unlock; } -clear_uptodate: - ClearPageUptodate(rpage); - ClearPageError(rpage); -unlock: unlock_page(rpage); } + + f2fs_put_dic(dic); +} + +static void f2fs_verify_cluster(struct work_struct *work) +{ + struct decompress_io_ctx *dic = + container_of(work, struct decompress_io_ctx, verity_work); + int i; + + /* Verify the cluster's decompressed pages with fs-verity. */ + for (i = 0; i < dic->cluster_size; i++) { + struct page *rpage = dic->rpages[i]; + + if (rpage && !fsverity_verify_page(rpage)) + SetPageError(rpage); + } + + __f2fs_decompress_end_io(dic, false); +} + +/* + * This is called when a compressed cluster has been decompressed + * (or failed to be read and/or decompressed). + */ +void f2fs_decompress_end_io(struct decompress_io_ctx *dic, bool failed) +{ + if (!failed && dic->need_verity) { + /* + * Note that to avoid deadlocks, the verity work can't be done + * on the decompression workqueue. This is because verifying + * the data pages can involve reading metadata pages from the + * file, and these metadata pages may be compressed. + */ + INIT_WORK(&dic->verity_work, f2fs_verify_cluster); + fsverity_enqueue_verify_work(&dic->verity_work); + } else { + __f2fs_decompress_end_io(dic, failed); + } +} + +/* + * Put a reference to a compressed page's decompress_io_ctx. + * + * This is called when the page is no longer needed and can be freed. + */ +void f2fs_put_page_dic(struct page *page) +{ + struct decompress_io_ctx *dic = + (struct decompress_io_ctx *)page_private(page); + + f2fs_put_dic(dic); } int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 547c9d4b430b..4d80f00e5e40 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -115,10 +115,21 @@ static enum count_type __read_io_type(struct page *page) /* postprocessing steps for read bios */ enum bio_post_read_step { - STEP_DECRYPT, - STEP_DECOMPRESS_NOWQ, /* handle normal cluster data inplace */ - STEP_DECOMPRESS, /* handle compressed cluster data in workqueue */ - STEP_VERITY, +#ifdef CONFIG_FS_ENCRYPTION + STEP_DECRYPT = 1 << 0, +#else + STEP_DECRYPT = 0, /* compile out the decryption-related code */ +#endif +#ifdef CONFIG_F2FS_FS_COMPRESSION + STEP_DECOMPRESS = 1 << 1, +#else + STEP_DECOMPRESS = 0, /* compile out the decompression-related code */ +#endif +#ifdef CONFIG_FS_VERITY + STEP_VERITY = 1 << 2, +#else + STEP_VERITY = 0, /* compile out the verity-related code */ +#endif }; struct bio_post_read_ctx { @@ -128,25 +139,26 @@ struct bio_post_read_ctx { unsigned int enabled_steps; }; -static void __read_end_io(struct bio *bio, bool compr, bool verity) +static void f2fs_finish_read_bio(struct bio *bio) { - struct page *page; struct bio_vec *bv; struct bvec_iter_all iter_all; + /* + * Update and unlock the bio's pagecache pages, and put the + * decompression context for any compressed pages. + */ bio_for_each_segment_all(bv, bio, iter_all) { - page = bv->bv_page; + struct page *page = bv->bv_page; -#ifdef CONFIG_F2FS_FS_COMPRESSION - if (compr && f2fs_is_compressed_page(page)) { - f2fs_decompress_pages(bio, page, verity); + if (f2fs_is_compressed_page(page)) { + if (bio->bi_status) + f2fs_end_read_compressed_page(page, true); + f2fs_put_page_dic(page); continue; } - if (verity) - continue; -#endif - /* PG_error was set if any post_read step failed */ + /* PG_error was set if decryption or verity failed. */ if (bio->bi_status || PageError(page)) { ClearPageUptodate(page); /* will re-read again later */ @@ -157,106 +169,104 @@ static void __read_end_io(struct bio *bio, bool compr, bool verity) dec_page_count(F2FS_P_SB(page), __read_io_type(page)); unlock_page(page); } + + if (bio->bi_private) + mempool_free(bio->bi_private, bio_post_read_ctx_pool); + bio_put(bio); } -static void f2fs_release_read_bio(struct bio *bio); -static void __f2fs_read_end_io(struct bio *bio, bool compr, bool verity) -{ - if (!compr) - __read_end_io(bio, false, verity); - f2fs_release_read_bio(bio); -} - -static void f2fs_decompress_bio(struct bio *bio, bool verity) -{ - __read_end_io(bio, true, verity); -} - -static void bio_post_read_processing(struct bio_post_read_ctx *ctx); - -static void f2fs_decrypt_work(struct bio_post_read_ctx *ctx) -{ - fscrypt_decrypt_bio(ctx->bio); -} - -static void f2fs_decompress_work(struct bio_post_read_ctx *ctx) -{ - f2fs_decompress_bio(ctx->bio, ctx->enabled_steps & (1 << STEP_VERITY)); -} - -#ifdef CONFIG_F2FS_FS_COMPRESSION -static void f2fs_verify_pages(struct page **rpages, unsigned int cluster_size) -{ - f2fs_decompress_end_io(rpages, cluster_size, false, true); -} - -static void f2fs_verify_bio(struct bio *bio) -{ - struct bio_vec *bv; - struct bvec_iter_all iter_all; - - bio_for_each_segment_all(bv, bio, iter_all) { - struct page *page = bv->bv_page; - struct decompress_io_ctx *dic; - - dic = (struct decompress_io_ctx *)page_private(page); - - if (dic) { - if (atomic_dec_return(&dic->verity_pages)) - continue; - f2fs_verify_pages(dic->rpages, - dic->cluster_size); - f2fs_free_dic(dic); - continue; - } - - if (bio->bi_status || PageError(page)) - goto clear_uptodate; - - if (fsverity_verify_page(page)) { - SetPageUptodate(page); - goto unlock; - } -clear_uptodate: - ClearPageUptodate(page); - ClearPageError(page); -unlock: - dec_page_count(F2FS_P_SB(page), __read_io_type(page)); - unlock_page(page); - } -} -#endif - -static void f2fs_verity_work(struct work_struct *work) +static void f2fs_verify_bio(struct work_struct *work) { struct bio_post_read_ctx *ctx = container_of(work, struct bio_post_read_ctx, work); struct bio *bio = ctx->bio; -#ifdef CONFIG_F2FS_FS_COMPRESSION - unsigned int enabled_steps = ctx->enabled_steps; -#endif + bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS); /* * fsverity_verify_bio() may call readpages() again, and while verity - * will be disabled for this, decryption may still be needed, resulting - * in another bio_post_read_ctx being allocated. So to prevent - * deadlocks we need to release the current ctx to the mempool first. - * This assumes that verity is the last post-read step. + * will be disabled for this, decryption and/or decompression may still + * be needed, resulting in another bio_post_read_ctx being allocated. + * So to prevent deadlocks we need to release the current ctx to the + * mempool first. This assumes that verity is the last post-read step. */ mempool_free(ctx, bio_post_read_ctx_pool); bio->bi_private = NULL; -#ifdef CONFIG_F2FS_FS_COMPRESSION - /* previous step is decompression */ - if (enabled_steps & (1 << STEP_DECOMPRESS)) { - f2fs_verify_bio(bio); - f2fs_release_read_bio(bio); - return; - } -#endif + /* + * Verify the bio's pages with fs-verity. Exclude compressed pages, + * as those were handled separately by f2fs_end_read_compressed_page(). + */ + if (may_have_compressed_pages) { + struct bio_vec *bv; + struct bvec_iter_all iter_all; - fsverity_verify_bio(bio); - __f2fs_read_end_io(bio, false, false); + bio_for_each_segment_all(bv, bio, iter_all) { + struct page *page = bv->bv_page; + + if (!f2fs_is_compressed_page(page) && + !PageError(page) && !fsverity_verify_page(page)) + SetPageError(page); + } + } else { + fsverity_verify_bio(bio); + } + + f2fs_finish_read_bio(bio); +} + +/* + * If the bio's data needs to be verified with fs-verity, then enqueue the + * verity work for the bio. Otherwise finish the bio now. + * + * Note that to avoid deadlocks, the verity work can't be done on the + * decryption/decompression workqueue. This is because verifying the data pages + * can involve reading verity metadata pages from the file, and these verity + * metadata pages may be encrypted and/or compressed. + */ +static void f2fs_verify_and_finish_bio(struct bio *bio) +{ + struct bio_post_read_ctx *ctx = bio->bi_private; + + if (ctx && (ctx->enabled_steps & STEP_VERITY)) { + INIT_WORK(&ctx->work, f2fs_verify_bio); + fsverity_enqueue_verify_work(&ctx->work); + } else { + f2fs_finish_read_bio(bio); + } +} + +/* + * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last + * remaining page was read by @ctx->bio. + * + * Note that a bio may span clusters (even a mix of compressed and uncompressed + * clusters) or be for just part of a cluster. STEP_DECOMPRESS just indicates + * that the bio includes at least one compressed page. The actual decompression + * is done on a per-cluster basis, not a per-bio basis. + */ +static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx) +{ + struct bio_vec *bv; + struct bvec_iter_all iter_all; + bool all_compressed = true; + + bio_for_each_segment_all(bv, ctx->bio, iter_all) { + struct page *page = bv->bv_page; + + /* PG_error was set if decryption failed. */ + if (f2fs_is_compressed_page(page)) + f2fs_end_read_compressed_page(page, PageError(page)); + else + all_compressed = false; + } + + /* + * Optimization: if all the bio's pages are compressed, then scheduling + * the per-bio verity work is unnecessary, as verity will be fully + * handled at the compression cluster level. + */ + if (all_compressed) + ctx->enabled_steps &= ~STEP_VERITY; } static void f2fs_post_read_work(struct work_struct *work) @@ -264,74 +274,36 @@ static void f2fs_post_read_work(struct work_struct *work) struct bio_post_read_ctx *ctx = container_of(work, struct bio_post_read_ctx, work); - if (ctx->enabled_steps & (1 << STEP_DECRYPT)) - f2fs_decrypt_work(ctx); + if (ctx->enabled_steps & STEP_DECRYPT) + fscrypt_decrypt_bio(ctx->bio); - if (ctx->enabled_steps & (1 << STEP_DECOMPRESS)) - f2fs_decompress_work(ctx); + if (ctx->enabled_steps & STEP_DECOMPRESS) + f2fs_handle_step_decompress(ctx); - if (ctx->enabled_steps & (1 << STEP_VERITY)) { - INIT_WORK(&ctx->work, f2fs_verity_work); - fsverity_enqueue_verify_work(&ctx->work); - return; - } - - __f2fs_read_end_io(ctx->bio, - ctx->enabled_steps & (1 << STEP_DECOMPRESS), false); -} - -static void f2fs_enqueue_post_read_work(struct f2fs_sb_info *sbi, - struct work_struct *work) -{ - queue_work(sbi->post_read_wq, work); -} - -static void bio_post_read_processing(struct bio_post_read_ctx *ctx) -{ - /* - * We use different work queues for decryption and for verity because - * verity may require reading metadata pages that need decryption, and - * we shouldn't recurse to the same workqueue. - */ - - if (ctx->enabled_steps & (1 << STEP_DECRYPT) || - ctx->enabled_steps & (1 << STEP_DECOMPRESS)) { - INIT_WORK(&ctx->work, f2fs_post_read_work); - f2fs_enqueue_post_read_work(ctx->sbi, &ctx->work); - return; - } - - if (ctx->enabled_steps & (1 << STEP_VERITY)) { - INIT_WORK(&ctx->work, f2fs_verity_work); - fsverity_enqueue_verify_work(&ctx->work); - return; - } - - __f2fs_read_end_io(ctx->bio, false, false); -} - -static bool f2fs_bio_post_read_required(struct bio *bio) -{ - return bio->bi_private; + f2fs_verify_and_finish_bio(ctx->bio); } static void f2fs_read_end_io(struct bio *bio) { struct f2fs_sb_info *sbi = F2FS_P_SB(bio_first_page_all(bio)); + struct bio_post_read_ctx *ctx = bio->bi_private; if (time_to_inject(sbi, FAULT_READ_IO)) { f2fs_show_injection_info(sbi, FAULT_READ_IO); bio->bi_status = BLK_STS_IOERR; } - if (f2fs_bio_post_read_required(bio)) { - struct bio_post_read_ctx *ctx = bio->bi_private; - - bio_post_read_processing(ctx); + if (bio->bi_status) { + f2fs_finish_read_bio(bio); return; } - __f2fs_read_end_io(bio, false, false); + if (ctx && (ctx->enabled_steps & (STEP_DECRYPT | STEP_DECOMPRESS))) { + INIT_WORK(&ctx->work, f2fs_post_read_work); + queue_work(ctx->sbi->post_read_wq, &ctx->work); + } else { + f2fs_verify_and_finish_bio(bio); + } } static void f2fs_write_end_io(struct bio *bio) @@ -1022,16 +994,9 @@ out: up_write(&io->io_rwsem); } -static inline bool f2fs_need_verity(const struct inode *inode, pgoff_t idx) -{ - return fsverity_active(inode) && - idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE); -} - static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr, unsigned nr_pages, unsigned op_flag, - pgoff_t first_idx, bool for_write, - bool for_verity) + pgoff_t first_idx, bool for_write) { struct f2fs_sb_info *sbi = F2FS_I_SB(inode); struct bio *bio; @@ -1050,13 +1015,19 @@ static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr, bio_set_op_attrs(bio, REQ_OP_READ, op_flag); if (fscrypt_inode_uses_fs_layer_crypto(inode)) - post_read_steps |= 1 << STEP_DECRYPT; - if (f2fs_compressed_file(inode)) - post_read_steps |= 1 << STEP_DECOMPRESS_NOWQ; - if (for_verity && f2fs_need_verity(inode, first_idx)) - post_read_steps |= 1 << STEP_VERITY; + post_read_steps |= STEP_DECRYPT; - if (post_read_steps) { + if (f2fs_need_verity(inode, first_idx)) + post_read_steps |= STEP_VERITY; + + /* + * STEP_DECOMPRESS is handled specially, since a compressed file might + * contain both compressed and uncompressed clusters. We'll allocate a + * bio_post_read_ctx if the file is compressed, but the caller is + * responsible for enabling STEP_DECOMPRESS if it's actually needed. + */ + + if (post_read_steps || f2fs_compressed_file(inode)) { /* Due to the mempool, this never fails. */ ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS); ctx->bio = bio; @@ -1068,13 +1039,6 @@ static struct bio *f2fs_grab_read_bio(struct inode *inode, block_t blkaddr, return bio; } -static void f2fs_release_read_bio(struct bio *bio) -{ - if (bio->bi_private) - mempool_free(bio->bi_private, bio_post_read_ctx_pool); - bio_put(bio); -} - /* This can handle encryption stuffs */ static int f2fs_submit_page_read(struct inode *inode, struct page *page, block_t blkaddr, int op_flags, bool for_write) @@ -1083,7 +1047,7 @@ static int f2fs_submit_page_read(struct inode *inode, struct page *page, struct bio *bio; bio = f2fs_grab_read_bio(inode, blkaddr, 1, op_flags, - page->index, for_write, true); + page->index, for_write); if (IS_ERR(bio)) return PTR_ERR(bio); @@ -2122,7 +2086,7 @@ submit_and_realloc: if (bio == NULL) { bio = f2fs_grab_read_bio(inode, block_nr, nr_pages, is_readahead ? REQ_RAHEAD : 0, page->index, - false, true); + false); if (IS_ERR(bio)) { ret = PTR_ERR(bio); bio = NULL; @@ -2168,8 +2132,6 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret, sector_t last_block_in_file; const unsigned blocksize = blks_to_bytes(inode, 1); struct decompress_io_ctx *dic = NULL; - struct bio_post_read_ctx *ctx; - bool for_verity = false; int i; int ret = 0; @@ -2235,29 +2197,10 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret, goto out_put_dnode; } - /* - * It's possible to enable fsverity on the fly when handling a cluster, - * which requires complicated error handling. Instead of adding more - * complexity, let's give a rule where end_io post-processes fsverity - * per cluster. In order to do that, we need to submit bio, if previous - * bio sets a different post-process policy. - */ - if (fsverity_active(cc->inode)) { - atomic_set(&dic->verity_pages, cc->nr_cpages); - for_verity = true; - - if (bio) { - ctx = bio->bi_private; - if (!(ctx->enabled_steps & (1 << STEP_VERITY))) { - __submit_bio(sbi, bio, DATA); - bio = NULL; - } - } - } - for (i = 0; i < dic->nr_cpages; i++) { struct page *page = dic->cpages[i]; block_t blkaddr; + struct bio_post_read_ctx *ctx; blkaddr = data_blkaddr(dn.inode, dn.node_page, dn.ofs_in_node + i + 1); @@ -2273,31 +2216,10 @@ submit_and_realloc: if (!bio) { bio = f2fs_grab_read_bio(inode, blkaddr, nr_pages, is_readahead ? REQ_RAHEAD : 0, - page->index, for_write, for_verity); + page->index, for_write); if (IS_ERR(bio)) { - unsigned int remained = dic->nr_cpages - i; - bool release = false; - ret = PTR_ERR(bio); - dic->failed = true; - - if (for_verity) { - if (!atomic_sub_return(remained, - &dic->verity_pages)) - release = true; - } else { - if (!atomic_sub_return(remained, - &dic->pending_pages)) - release = true; - } - - if (release) { - f2fs_decompress_end_io(dic->rpages, - cc->cluster_size, true, - false); - f2fs_free_dic(dic); - } - + f2fs_decompress_end_io(dic, ret); f2fs_put_dnode(&dn); *bio_ret = NULL; return ret; @@ -2309,10 +2231,9 @@ submit_and_realloc: if (bio_add_page(bio, page, blocksize, 0) < blocksize) goto submit_and_realloc; - /* tag STEP_DECOMPRESS to handle IO in wq */ ctx = bio->bi_private; - if (!(ctx->enabled_steps & (1 << STEP_DECOMPRESS))) - ctx->enabled_steps |= 1 << STEP_DECOMPRESS; + ctx->enabled_steps |= STEP_DECOMPRESS; + refcount_inc(&dic->refcnt); inc_page_count(sbi, F2FS_RD_DATA); f2fs_update_iostat(sbi, FS_DATA_READ_IO, F2FS_BLKSIZE); @@ -2329,7 +2250,13 @@ submit_and_realloc: out_put_dnode: f2fs_put_dnode(&dn); out: - f2fs_decompress_end_io(cc->rpages, cc->cluster_size, true, false); + for (i = 0; i < cc->cluster_size; i++) { + if (cc->rpages[i]) { + ClearPageUptodate(cc->rpages[i]); + ClearPageError(cc->rpages[i]); + unlock_page(cc->rpages[i]); + } + } *bio_ret = bio; return ret; } diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index ccbbf86d14e5..980e061f7968 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1341,7 +1341,7 @@ struct compress_io_ctx { atomic_t pending_pages; /* in-flight compressed page count */ }; -/* decompress io context for read IO path */ +/* Context for decompressing one cluster on the read IO path */ struct decompress_io_ctx { u32 magic; /* magic number to indicate page is compressed */ struct inode *inode; /* inode the context belong to */ @@ -1357,11 +1357,37 @@ struct decompress_io_ctx { struct compress_data *cbuf; /* virtual mapped address on cpages */ size_t rlen; /* valid data length in rbuf */ size_t clen; /* valid data length in cbuf */ - atomic_t pending_pages; /* in-flight compressed page count */ - atomic_t verity_pages; /* in-flight page count for verity */ - bool failed; /* indicate IO error during decompression */ + + /* + * The number of compressed pages remaining to be read in this cluster. + * This is initially nr_cpages. It is decremented by 1 each time a page + * has been read (or failed to be read). When it reaches 0, the cluster + * is decompressed (or an error is reported). + * + * If an error occurs before all the pages have been submitted for I/O, + * then this will never reach 0. In this case the I/O submitter is + * responsible for calling f2fs_decompress_end_io() instead. + */ + atomic_t remaining_pages; + + /* + * Number of references to this decompress_io_ctx. + * + * One reference is held for I/O completion. This reference is dropped + * after the pagecache pages are updated and unlocked -- either after + * decompression (and verity if enabled), or after an error. + * + * In addition, each compressed page holds a reference while it is in a + * bio. These references are necessary prevent compressed pages from + * being freed while they are still in a bio. + */ + refcount_t refcnt; + + bool failed; /* IO error occurred before decompression? */ + bool need_verity; /* need fs-verity verification after decompression? */ void *private; /* payload buffer for specified decompression algorithm */ void *private2; /* extra payload buffer */ + struct work_struct verity_work; /* work to verify the decompressed pages */ }; #define NULL_CLUSTER ((unsigned int)(~0)) @@ -3883,7 +3909,7 @@ void f2fs_compress_write_end_io(struct bio *bio, struct page *page); bool f2fs_is_compress_backend_ready(struct inode *inode); int f2fs_init_compress_mempool(void); void f2fs_destroy_compress_mempool(void); -void f2fs_decompress_pages(struct bio *bio, struct page *page, bool verity); +void f2fs_end_read_compressed_page(struct page *page, bool failed); bool f2fs_cluster_is_empty(struct compress_ctx *cc); bool f2fs_cluster_can_merge_page(struct compress_ctx *cc, pgoff_t index); void f2fs_compress_ctx_add_page(struct compress_ctx *cc, struct page *page); @@ -3896,9 +3922,8 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret, unsigned nr_pages, sector_t *last_block_in_bio, bool is_readahead, bool for_write); struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc); -void f2fs_free_dic(struct decompress_io_ctx *dic); -void f2fs_decompress_end_io(struct page **rpages, - unsigned int cluster_size, bool err, bool verity); +void f2fs_decompress_end_io(struct decompress_io_ctx *dic, bool failed); +void f2fs_put_page_dic(struct page *page); int f2fs_init_compress_ctx(struct compress_ctx *cc); void f2fs_destroy_compress_ctx(struct compress_ctx *cc); void f2fs_init_compress_info(struct f2fs_sb_info *sbi); @@ -3922,6 +3947,14 @@ static inline struct page *f2fs_compress_control_page(struct page *page) } static inline int f2fs_init_compress_mempool(void) { return 0; } static inline void f2fs_destroy_compress_mempool(void) { } +static inline void f2fs_end_read_compressed_page(struct page *page, bool failed) +{ + WARN_ON_ONCE(1); +} +static inline void f2fs_put_page_dic(struct page *page) +{ + WARN_ON_ONCE(1); +} static inline int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi) { return 0; } static inline void f2fs_destroy_page_array_cache(struct f2fs_sb_info *sbi) { } static inline int __init f2fs_init_compress_cache(void) { return 0; } @@ -4126,6 +4159,12 @@ static inline bool f2fs_force_buffered_io(struct inode *inode, return false; } +static inline bool f2fs_need_verity(const struct inode *inode, pgoff_t idx) +{ + return fsverity_active(inode) && + idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE); +} + #ifdef CONFIG_F2FS_FAULT_INJECTION extern void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, unsigned int type); From df0736d70c4fa6ed711ba103b61880fe72bb4777 Mon Sep 17 00:00:00 2001 From: Daeho Jeong Date: Wed, 6 Jan 2021 08:49:28 +0900 Subject: [PATCH 14/32] f2fs: fix null page reference in redirty_blocks By Colin's static analysis, we found out there is a null page reference under low memory situation in redirty_blocks. I've made the page finding loop stop immediately and return an error not to cause further memory pressure when we run into a failure to find a page under low memory condition. Signed-off-by: Daeho Jeong Reported-by: Colin Ian King Fixes: 5fdb322ff2c2 ("f2fs: add F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE") Reviewed-by: Colin Ian King Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 4e6d4b9120a8..e3a5b620b50a 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -4057,8 +4057,10 @@ static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) for (i = 0; i < page_len; i++, redirty_idx++) { page = find_lock_page(mapping, redirty_idx); - if (!page) - ret = -ENOENT; + if (!page) { + ret = -ENOMEM; + break; + } set_page_dirty(page); f2fs_put_page(page, 1); f2fs_put_page(page, 0); From 46085f37fc9e12d5c3539fb768b5ad7951e72acf Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Tue, 12 Jan 2021 09:55:09 +0800 Subject: [PATCH 15/32] f2fs: fix to set/clear I_LINKABLE under i_lock fsstress + fault injection test case reports a warning message as below: WARNING: CPU: 13 PID: 6226 at fs/inode.c:361 inc_nlink+0x32/0x40 Call Trace: f2fs_init_inode_metadata+0x25c/0x4a0 [f2fs] f2fs_add_inline_entry+0x153/0x3b0 [f2fs] f2fs_add_dentry+0x75/0x80 [f2fs] f2fs_do_add_link+0x108/0x160 [f2fs] f2fs_rename2+0x6ab/0x14f0 [f2fs] vfs_rename+0x70c/0x940 do_renameat2+0x4d8/0x4f0 __x64_sys_renameat2+0x4b/0x60 do_syscall_64+0x33/0x80 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Following race case can cause this: Thread A Kworker - f2fs_rename - f2fs_create_whiteout - __f2fs_tmpfile - f2fs_i_links_write - f2fs_mark_inode_dirty_sync - mark_inode_dirty_sync - writeback_single_inode - __writeback_single_inode - spin_lock(&inode->i_lock) - inode->i_state |= I_LINKABLE - inode->i_state &= ~dirty - spin_unlock(&inode->i_lock) - f2fs_add_link - f2fs_do_add_link - f2fs_add_dentry - f2fs_add_inline_entry - f2fs_init_inode_metadata - f2fs_i_links_write - inc_nlink - WARN_ON(!(inode->i_state & I_LINKABLE)) Fix to add i_lock to avoid i_state update race condition. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/namei.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c index 6edb1ab579a1..887804968576 100644 --- a/fs/f2fs/namei.c +++ b/fs/f2fs/namei.c @@ -855,7 +855,11 @@ static int __f2fs_tmpfile(struct inode *dir, struct dentry *dentry, if (whiteout) { f2fs_i_links_write(inode, false); + + spin_lock(&inode->i_lock); inode->i_state |= I_LINKABLE; + spin_unlock(&inode->i_lock); + *whiteout = inode; } else { d_tmpfile(dentry, inode); @@ -1041,7 +1045,11 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, err = f2fs_add_link(old_dentry, whiteout); if (err) goto put_out_dir; + + spin_lock(&whiteout->i_lock); whiteout->i_state &= ~I_LINKABLE; + spin_unlock(&whiteout->i_lock); + iput(whiteout); } From 794c43f716845e2d48ce195ed5c4179a4e05ce5f Mon Sep 17 00:00:00 2001 From: Eric Biggers Date: Mon, 28 Dec 2020 15:25:29 -0800 Subject: [PATCH 16/32] libfs: unexport generic_ci_d_compare() and generic_ci_d_hash() Now that generic_set_encrypted_ci_d_ops() has been added and ext4 and f2fs are using it, it's no longer necessary to export generic_ci_d_compare() and generic_ci_d_hash() to filesystems. Reviewed-by: Gabriel Krisman Bertazi Signed-off-by: Eric Biggers Signed-off-by: Jaegeuk Kim --- fs/libfs.c | 8 +++----- include/linux/fs.h | 5 ----- 2 files changed, 3 insertions(+), 10 deletions(-) diff --git a/fs/libfs.c b/fs/libfs.c index d1c3bade9f30..79721571e014 100644 --- a/fs/libfs.c +++ b/fs/libfs.c @@ -1388,8 +1388,8 @@ static bool needs_casefold(const struct inode *dir) * * Return: 0 if names match, 1 if mismatch, or -ERRNO */ -int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, - const char *str, const struct qstr *name) +static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, + const char *str, const struct qstr *name) { const struct dentry *parent = READ_ONCE(dentry->d_parent); const struct inode *dir = READ_ONCE(parent->d_inode); @@ -1426,7 +1426,6 @@ fallback: return 1; return !!memcmp(str, name->name, len); } -EXPORT_SYMBOL(generic_ci_d_compare); /** * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems @@ -1435,7 +1434,7 @@ EXPORT_SYMBOL(generic_ci_d_compare); * * Return: 0 if hash was successful or unchanged, and -EINVAL on error */ -int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) +static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) { const struct inode *dir = READ_ONCE(dentry->d_inode); struct super_block *sb = dentry->d_sb; @@ -1450,7 +1449,6 @@ int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) return -EINVAL; return 0; } -EXPORT_SYMBOL(generic_ci_d_hash); static const struct dentry_operations generic_ci_dentry_ops = { .d_hash = generic_ci_d_hash, diff --git a/include/linux/fs.h b/include/linux/fs.h index fd47deea7c17..6d8b1e7337e4 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -3192,11 +3192,6 @@ extern int generic_file_fsync(struct file *, loff_t, loff_t, int); extern int generic_check_addressable(unsigned, u64); -#ifdef CONFIG_UNICODE -extern int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str); -extern int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, - const char *str, const struct qstr *name); -#endif extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry); #ifdef CONFIG_MIGRATION From 3afae09ffea5e08f523823be99a784675995d6bb Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Mon, 11 Jan 2021 17:42:53 +0800 Subject: [PATCH 17/32] f2fs: compress: fix potential deadlock generic/269 reports a hangtask issue, the root cause is ABBA deadlock described as below: Thread A Thread B - down_write(&sbi->gc_lock) -- A - f2fs_write_data_pages - lock all pages in cluster -- B - f2fs_write_multi_pages - f2fs_write_raw_pages - f2fs_write_single_data_page - f2fs_balance_fs - down_write(&sbi->gc_lock) -- A - f2fs_gc - do_garbage_collect - ra_data_block - pagecache_get_page -- B To fix this, it needs to avoid calling f2fs_balance_fs() if there is still cluster pages been locked in context of cluster writeback, so instead, let's call f2fs_balance_fs() in the end of f2fs_write_raw_pages() when all cluster pages were unlocked. Fixes: 4c8ff7095bef ("f2fs: support data compression") Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/compress.c | 5 ++++- fs/f2fs/data.c | 10 ++++++---- fs/f2fs/f2fs.h | 2 +- 3 files changed, 11 insertions(+), 6 deletions(-) diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c index 1696f9183ff5..77fa342de38f 100644 --- a/fs/f2fs/compress.c +++ b/fs/f2fs/compress.c @@ -1455,7 +1455,7 @@ retry_write: ret = f2fs_write_single_data_page(cc->rpages[i], &_submitted, NULL, NULL, wbc, io_type, - compr_blocks); + compr_blocks, false); if (ret) { if (ret == AOP_WRITEPAGE_ACTIVATE) { unlock_page(cc->rpages[i]); @@ -1490,6 +1490,9 @@ retry_write: *submitted += _submitted; } + + f2fs_balance_fs(F2FS_M_SB(mapping), true); + return 0; out_err: for (++i; i < cc->cluster_size; i++) { diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 4d80f00e5e40..c7bb07dd9a20 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -2671,7 +2671,8 @@ int f2fs_write_single_data_page(struct page *page, int *submitted, sector_t *last_block, struct writeback_control *wbc, enum iostat_type io_type, - int compr_blocks) + int compr_blocks, + bool allow_balance) { struct inode *inode = page->mapping->host; struct f2fs_sb_info *sbi = F2FS_I_SB(inode); @@ -2809,7 +2810,7 @@ out: } unlock_page(page); if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) && - !F2FS_I(inode)->cp_task) + !F2FS_I(inode)->cp_task && allow_balance) f2fs_balance_fs(sbi, need_balance_fs); if (unlikely(f2fs_cp_error(sbi))) { @@ -2856,7 +2857,7 @@ out: #endif return f2fs_write_single_data_page(page, NULL, NULL, NULL, - wbc, FS_DATA_IO, 0); + wbc, FS_DATA_IO, 0, true); } /* @@ -3024,7 +3025,8 @@ continue_unlock: } #endif ret = f2fs_write_single_data_page(page, &submitted, - &bio, &last_block, wbc, io_type, 0); + &bio, &last_block, wbc, io_type, + 0, true); if (ret == AOP_WRITEPAGE_ACTIVATE) unlock_page(page); #ifdef CONFIG_F2FS_FS_COMPRESSION diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 980e061f7968..63852404151e 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -3502,7 +3502,7 @@ int f2fs_write_single_data_page(struct page *page, int *submitted, struct bio **bio, sector_t *last_block, struct writeback_control *wbc, enum iostat_type io_type, - int compr_blocks); + int compr_blocks, bool allow_balance); void f2fs_invalidate_page(struct page *page, unsigned int offset, unsigned int length); int f2fs_release_page(struct page *page, gfp_t wait); From 6d1451bf7f84ea45035553ae566b3c91661d902b Mon Sep 17 00:00:00 2001 From: Chengguang Xu Date: Wed, 13 Jan 2021 13:21:54 +0800 Subject: [PATCH 18/32] f2fs: fix to use per-inode maxbytes F2FS inode may have different max size, e.g. compressed file have less blkaddr entries in all its direct-node blocks, result in being with less max filesize. So change to use per-inode maxbytes. Suggested-by: Chao Yu Signed-off-by: Chengguang Xu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 2 +- fs/f2fs/f2fs.h | 2 +- fs/f2fs/file.c | 9 ++++++--- fs/f2fs/super.c | 12 ++++++++---- 4 files changed, 16 insertions(+), 9 deletions(-) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index c7bb07dd9a20..9aa458c01101 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -3761,7 +3761,7 @@ static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) filemap_write_and_wait(mapping); /* Block number less than F2FS MAX BLOCKS */ - if (unlikely(block >= F2FS_I_SB(inode)->max_file_blocks)) + if (unlikely(block >= max_file_blocks(inode))) goto out; if (f2fs_compressed_file(inode)) { diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 63852404151e..ca5f1ff14dab 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1474,7 +1474,6 @@ struct f2fs_sb_info { unsigned int total_sections; /* total section count */ unsigned int total_node_count; /* total node block count */ unsigned int total_valid_node_count; /* valid node block count */ - loff_t max_file_blocks; /* max block index of file */ int dir_level; /* directory level */ int readdir_ra; /* readahead inode in readdir */ u64 max_io_bytes; /* max io bytes to merge IOs */ @@ -3265,6 +3264,7 @@ int f2fs_inode_dirtied(struct inode *inode, bool sync); void f2fs_inode_synced(struct inode *inode); int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly); int f2fs_quota_sync(struct super_block *sb, int type); +loff_t max_file_blocks(struct inode *inode); void f2fs_quota_off_umount(struct super_block *sb); int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover); int f2fs_sync_fs(struct super_block *sb, int sync); diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index e3a5b620b50a..e768371c6575 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -486,6 +486,9 @@ static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) struct inode *inode = file->f_mapping->host; loff_t maxbytes = inode->i_sb->s_maxbytes; + if (f2fs_compressed_file(inode)) + maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; + switch (whence) { case SEEK_SET: case SEEK_CUR: @@ -670,7 +673,7 @@ int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) free_from = (pgoff_t)F2FS_BLK_ALIGN(from); - if (free_from >= sbi->max_file_blocks) + if (free_from >= max_file_blocks(inode)) goto free_partial; if (lock) @@ -2744,7 +2747,7 @@ static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) return -EINVAL; if (unlikely((range.start + range.len) >> PAGE_SHIFT > - sbi->max_file_blocks)) + max_file_blocks(inode))) return -EINVAL; err = mnt_want_write_file(filp); @@ -3307,7 +3310,7 @@ int f2fs_precache_extents(struct inode *inode) map.m_next_extent = &m_next_extent; map.m_seg_type = NO_CHECK_TYPE; map.m_may_create = false; - end = F2FS_I_SB(inode)->max_file_blocks; + end = max_file_blocks(inode); while (map.m_lblk < end) { map.m_len = end - map.m_lblk; diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index c8be27a9eed6..9749c9ad374f 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -2739,10 +2739,10 @@ static const struct export_operations f2fs_export_ops = { .get_parent = f2fs_get_parent, }; -static loff_t max_file_blocks(void) +loff_t max_file_blocks(struct inode *inode) { loff_t result = 0; - loff_t leaf_count = DEF_ADDRS_PER_BLOCK; + loff_t leaf_count; /* * note: previously, result is equal to (DEF_ADDRS_PER_INODE - @@ -2751,6 +2751,11 @@ static loff_t max_file_blocks(void) * result as zero. */ + if (inode && f2fs_compressed_file(inode)) + leaf_count = ADDRS_PER_BLOCK(inode); + else + leaf_count = DEF_ADDRS_PER_BLOCK; + /* two direct node blocks */ result += (leaf_count * 2); @@ -3634,8 +3639,7 @@ try_onemore: if (err) goto free_options; - sbi->max_file_blocks = max_file_blocks(); - sb->s_maxbytes = sbi->max_file_blocks << + sb->s_maxbytes = max_file_blocks(NULL) << le32_to_cpu(raw_super->log_blocksize); sb->s_max_links = F2FS_LINK_MAX; From 0bfe9f790448012ef38abf4e78feb2e691e2d366 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 14 Jan 2021 09:41:27 +0800 Subject: [PATCH 19/32] f2fs: introduce sb_status sysfs node Introduce /sys/fs/f2fs//stat/sb_status to show superblock status in real time as a hexadecimal value. value sb status macro description 0x1 SBI_IS_DIRTY, /* dirty flag for checkpoint */ 0x2 SBI_IS_CLOSE, /* specify unmounting */ 0x4 SBI_NEED_FSCK, /* need fsck.f2fs to fix */ 0x8 SBI_POR_DOING, /* recovery is doing or not */ 0x10 SBI_NEED_SB_WRITE, /* need to recover superblock */ 0x20 SBI_NEED_CP, /* need to checkpoint */ 0x40 SBI_IS_SHUTDOWN, /* shutdown by ioctl */ 0x80 SBI_IS_RECOVERED, /* recovered orphan/data */ 0x100 SBI_CP_DISABLED, /* CP was disabled last mount */ 0x200 SBI_CP_DISABLED_QUICK, /* CP was disabled quickly */ 0x400 SBI_QUOTA_NEED_FLUSH, /* need to flush quota info in CP */ 0x800 SBI_QUOTA_SKIP_FLUSH, /* skip flushing quota in current CP */ 0x1000 SBI_QUOTA_NEED_REPAIR, /* quota file may be corrupted */ 0x2000 SBI_IS_RESIZEFS, /* resizefs is in process */ Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/ABI/testing/sysfs-fs-f2fs | 23 +++++++++++++++++++++++ fs/f2fs/sysfs.c | 8 ++++++++ 2 files changed, 31 insertions(+) diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fs index 3dfee94e0618..362803901614 100644 --- a/Documentation/ABI/testing/sysfs-fs-f2fs +++ b/Documentation/ABI/testing/sysfs-fs-f2fs @@ -377,3 +377,26 @@ Description: This gives a control to limit the bio size in f2fs. Default is zero, which will follow underlying block layer limit, whereas, if it has a certain bytes value, f2fs won't submit a bio larger than that size. + +What: /sys/fs/f2fs//stat/sb_status +Date: December 2020 +Contact: "Chao Yu" +Description: Show status of f2fs superblock in real time. + + ====== ===================== ================================= + value sb status macro description + 0x1 SBI_IS_DIRTY dirty flag for checkpoint + 0x2 SBI_IS_CLOSE specify unmounting + 0x4 SBI_NEED_FSCK need fsck.f2fs to fix + 0x8 SBI_POR_DOING recovery is doing or not + 0x10 SBI_NEED_SB_WRITE need to recover superblock + 0x20 SBI_NEED_CP need to checkpoint + 0x40 SBI_IS_SHUTDOWN shutdown by ioctl + 0x80 SBI_IS_RECOVERED recovered orphan/data + 0x100 SBI_CP_DISABLED CP was disabled last mount + 0x200 SBI_CP_DISABLED_QUICK CP was disabled quickly + 0x400 SBI_QUOTA_NEED_FLUSH need to flush quota info in CP + 0x800 SBI_QUOTA_SKIP_FLUSH skip flushing quota in current CP + 0x1000 SBI_QUOTA_NEED_REPAIR quota file may be corrupted + 0x2000 SBI_IS_RESIZEFS resizefs is in process + ====== ===================== ================================= diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c index bd1174ed2e6f..f39874d512ea 100644 --- a/fs/f2fs/sysfs.c +++ b/fs/f2fs/sysfs.c @@ -96,6 +96,12 @@ static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a, sbi->sectors_written_start) >> 1))); } +static ssize_t sb_status_show(struct f2fs_attr *a, + struct f2fs_sb_info *sbi, char *buf) +{ + return sprintf(buf, "%lx\n", sbi->s_flag); +} + static ssize_t features_show(struct f2fs_attr *a, struct f2fs_sb_info *sbi, char *buf) { @@ -702,7 +708,9 @@ static struct attribute *f2fs_feat_attrs[] = { }; ATTRIBUTE_GROUPS(f2fs_feat); +F2FS_GENERAL_RO_ATTR(sb_status); static struct attribute *f2fs_stat_attrs[] = { + ATTR_LIST(sb_status), NULL, }; ATTRIBUTE_GROUPS(f2fs_stat); From deaa965fb01173478a1234f4305c71fffa4b5dc4 Mon Sep 17 00:00:00 2001 From: Jack Qiu Date: Wed, 13 Jan 2021 17:58:53 +0800 Subject: [PATCH 20/32] f2fs: remove unused stat_{inc, dec}_atomic_write Just clean code, no logical change. Signed-off-by: Jack Qiu Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/f2fs.h | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index ca5f1ff14dab..d37883ce9002 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -3748,8 +3748,6 @@ void f2fs_update_sit_info(struct f2fs_sb_info *sbi); #define stat_dec_compr_inode(inode) do { } while (0) #define stat_add_compr_blocks(inode, blocks) do { } while (0) #define stat_sub_compr_blocks(inode, blocks) do { } while (0) -#define stat_inc_atomic_write(inode) do { } while (0) -#define stat_dec_atomic_write(inode) do { } while (0) #define stat_update_max_atomic_write(inode) do { } while (0) #define stat_inc_volatile_write(inode) do { } while (0) #define stat_dec_volatile_write(inode) do { } while (0) From 12699fb781574d50871ec6a4d96ac5e0f0ede03e Mon Sep 17 00:00:00 2001 From: "Matthew Wilcox (Oracle)" Date: Thu, 14 Jan 2021 19:00:51 +0000 Subject: [PATCH 21/32] f2fs: Remove readahead collision detection With the new ->readahead operation, locked pages are added to the page cache, preventing two threads from racing with each other to read the same chunk of file, so this is dead code. Signed-off-by: Matthew Wilcox (Oracle) Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 25 ------------------------- fs/f2fs/f2fs.h | 1 - fs/f2fs/super.c | 2 -- 3 files changed, 28 deletions(-) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 9aa458c01101..d9a063d8a63d 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -2265,11 +2265,6 @@ out: /* * This function was originally taken from fs/mpage.c, and customized for f2fs. * Major change was from block_size == page_size in f2fs by default. - * - * Note that the aops->readpages() function is ONLY used for read-ahead. If - * this function ever deviates from doing just read-ahead, it should either - * use ->readpage() or do the necessary surgery to decouple ->readpages() - * from read-ahead. */ static int f2fs_mpage_readpages(struct inode *inode, struct readahead_control *rac, struct page *page) @@ -2292,7 +2287,6 @@ static int f2fs_mpage_readpages(struct inode *inode, unsigned nr_pages = rac ? readahead_count(rac) : 1; unsigned max_nr_pages = nr_pages; int ret = 0; - bool drop_ra = false; map.m_pblk = 0; map.m_lblk = 0; @@ -2303,26 +2297,10 @@ static int f2fs_mpage_readpages(struct inode *inode, map.m_seg_type = NO_CHECK_TYPE; map.m_may_create = false; - /* - * Two readahead threads for same address range can cause race condition - * which fragments sequential read IOs. So let's avoid each other. - */ - if (rac && readahead_count(rac)) { - if (READ_ONCE(F2FS_I(inode)->ra_offset) == readahead_index(rac)) - drop_ra = true; - else - WRITE_ONCE(F2FS_I(inode)->ra_offset, - readahead_index(rac)); - } - for (; nr_pages; nr_pages--) { if (rac) { page = readahead_page(rac); prefetchw(&page->flags); - if (drop_ra) { - f2fs_put_page(page, 1); - continue; - } } #ifdef CONFIG_F2FS_FS_COMPRESSION @@ -2385,9 +2363,6 @@ next_page: } if (bio) __submit_bio(F2FS_I_SB(inode), bio, DATA); - - if (rac && readahead_count(rac) && !drop_ra) - WRITE_ONCE(F2FS_I(inode)->ra_offset, -1); return ret; } diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index d37883ce9002..a2e520a13630 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -718,7 +718,6 @@ struct f2fs_inode_info { struct list_head inmem_pages; /* inmemory pages managed by f2fs */ struct task_struct *inmem_task; /* store inmemory task */ struct mutex inmem_lock; /* lock for inmemory pages */ - pgoff_t ra_offset; /* ongoing readahead offset */ struct extent_tree *extent_tree; /* cached extent_tree entry */ /* avoid racing between foreground op and gc */ diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 9749c9ad374f..6a30876ff374 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -1174,8 +1174,6 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb) /* Will be used by directory only */ fi->i_dir_level = F2FS_SB(sb)->dir_level; - fi->ra_offset = -1; - return &fi->vfs_inode; } From d5f7bc0064e0541164bd3deeafad16bbb5992433 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Thu, 14 Jan 2021 13:59:09 -0800 Subject: [PATCH 22/32] f2fs: deprecate f2fs_trace_io This patch deprecates f2fs_trace_io, since f2fs uses page->private more broadly, resulting in more buggy cases. Acked-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/Kconfig | 10 --- fs/f2fs/Makefile | 1 - fs/f2fs/checkpoint.c | 3 - fs/f2fs/data.c | 4 -- fs/f2fs/file.c | 2 - fs/f2fs/node.c | 2 - fs/f2fs/segment.c | 3 - fs/f2fs/super.c | 6 -- fs/f2fs/trace.c | 165 ------------------------------------------- fs/f2fs/trace.h | 43 ----------- 10 files changed, 239 deletions(-) delete mode 100644 fs/f2fs/trace.c delete mode 100644 fs/f2fs/trace.h diff --git a/fs/f2fs/Kconfig b/fs/f2fs/Kconfig index 63c1fc1a0e3b..62e638a49bbf 100644 --- a/fs/f2fs/Kconfig +++ b/fs/f2fs/Kconfig @@ -76,16 +76,6 @@ config F2FS_CHECK_FS If you want to improve the performance, say N. -config F2FS_IO_TRACE - bool "F2FS IO tracer" - depends on F2FS_FS - depends on FUNCTION_TRACER - help - F2FS IO trace is based on a function trace, which gathers process - information and block IO patterns in the filesystem level. - - If unsure, say N. - config F2FS_FAULT_INJECTION bool "F2FS fault injection facility" depends on F2FS_FS diff --git a/fs/f2fs/Makefile b/fs/f2fs/Makefile index ee7316b42f69..e5295746208b 100644 --- a/fs/f2fs/Makefile +++ b/fs/f2fs/Makefile @@ -7,6 +7,5 @@ f2fs-y += shrinker.o extent_cache.o sysfs.o f2fs-$(CONFIG_F2FS_STAT_FS) += debug.o f2fs-$(CONFIG_F2FS_FS_XATTR) += xattr.o f2fs-$(CONFIG_F2FS_FS_POSIX_ACL) += acl.o -f2fs-$(CONFIG_F2FS_IO_TRACE) += trace.o f2fs-$(CONFIG_FS_VERITY) += verity.o f2fs-$(CONFIG_F2FS_FS_COMPRESSION) += compress.o diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 897edb7c951a..8c79ba0566b1 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -17,7 +17,6 @@ #include "f2fs.h" #include "node.h" #include "segment.h" -#include "trace.h" #include static struct kmem_cache *ino_entry_slab; @@ -443,7 +442,6 @@ static int f2fs_set_meta_page_dirty(struct page *page) __set_page_dirty_nobuffers(page); inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_META); f2fs_set_page_private(page, 0); - f2fs_trace_pid(page); return 1; } return 0; @@ -1017,7 +1015,6 @@ void f2fs_update_dirty_page(struct inode *inode, struct page *page) spin_unlock(&sbi->inode_lock[type]); f2fs_set_page_private(page, 0); - f2fs_trace_pid(page); } void f2fs_remove_dirty_inode(struct inode *inode) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index d9a063d8a63d..38476d0d3916 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -25,7 +25,6 @@ #include "f2fs.h" #include "node.h" #include "segment.h" -#include "trace.h" #include #define NUM_PREALLOC_POST_READ_CTXS 128 @@ -679,7 +678,6 @@ int f2fs_submit_page_bio(struct f2fs_io_info *fio) return -EFSCORRUPTED; trace_f2fs_submit_page_bio(page, fio); - f2fs_trace_ios(fio, 0); /* Allocate a new bio */ bio = __bio_alloc(fio, 1); @@ -884,7 +882,6 @@ int f2fs_merge_page_bio(struct f2fs_io_info *fio) return -EFSCORRUPTED; trace_f2fs_submit_page_bio(page, fio); - f2fs_trace_ios(fio, 0); if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block, fio->new_blkaddr)) @@ -981,7 +978,6 @@ alloc_new: wbc_account_cgroup_owner(fio->io_wbc, bio_page, PAGE_SIZE); io->last_block_in_bio = fio->new_blkaddr; - f2fs_trace_ios(fio, 0); trace_f2fs_submit_page_write(fio->page, fio); skip: diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index e768371c6575..7db27c81d034 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -29,7 +29,6 @@ #include "xattr.h" #include "acl.h" #include "gc.h" -#include "trace.h" #include #include @@ -369,7 +368,6 @@ flush_out: f2fs_update_time(sbi, REQ_TIME); out: trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); - f2fs_trace_ios(NULL, 1); return ret; } diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c index 5e3fabacefb5..a8a0fb890e8d 100644 --- a/fs/f2fs/node.c +++ b/fs/f2fs/node.c @@ -17,7 +17,6 @@ #include "node.h" #include "segment.h" #include "xattr.h" -#include "trace.h" #include #define on_f2fs_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock) @@ -2089,7 +2088,6 @@ static int f2fs_set_node_page_dirty(struct page *page) __set_page_dirty_nobuffers(page); inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); f2fs_set_page_private(page, 0); - f2fs_trace_pid(page); return 1; } return 0; diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index deca74cb17df..7d34f1cacdee 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -20,7 +20,6 @@ #include "segment.h" #include "node.h" #include "gc.h" -#include "trace.h" #include #define __reverse_ffz(x) __reverse_ffs(~(x)) @@ -187,8 +186,6 @@ void f2fs_register_inmem_page(struct inode *inode, struct page *page) { struct inmem_pages *new; - f2fs_trace_pid(page); - f2fs_set_page_private(page, ATOMIC_WRITTEN_PAGE); new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS); diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 6a30876ff374..fc343243799c 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -33,7 +33,6 @@ #include "segment.h" #include "xattr.h" #include "gc.h" -#include "trace.h" #define CREATE_TRACE_POINTS #include @@ -1448,8 +1447,6 @@ int f2fs_sync_fs(struct super_block *sb, int sync) err = f2fs_write_checkpoint(sbi, &cpc); up_write(&sbi->gc_lock); } - f2fs_trace_ios(NULL, 1); - return err; } @@ -4127,8 +4124,6 @@ static int __init init_f2fs_fs(void) return -EINVAL; } - f2fs_build_trace_ios(); - err = init_inodecache(); if (err) goto fail; @@ -4221,7 +4216,6 @@ static void __exit exit_f2fs_fs(void) f2fs_destroy_segment_manager_caches(); f2fs_destroy_node_manager_caches(); destroy_inodecache(); - f2fs_destroy_trace_ios(); } module_init(init_f2fs_fs) diff --git a/fs/f2fs/trace.c b/fs/f2fs/trace.c deleted file mode 100644 index d0ab533a9ce8..000000000000 --- a/fs/f2fs/trace.c +++ /dev/null @@ -1,165 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/* - * f2fs IO tracer - * - * Copyright (c) 2014 Motorola Mobility - * Copyright (c) 2014 Jaegeuk Kim - */ -#include -#include -#include -#include - -#include "f2fs.h" -#include "trace.h" - -static RADIX_TREE(pids, GFP_ATOMIC); -static spinlock_t pids_lock; -static struct last_io_info last_io; - -static inline void __print_last_io(void) -{ - if (!last_io.len) - return; - - trace_printk("%3x:%3x %4x %-16s %2x %5x %5x %12x %4x\n", - last_io.major, last_io.minor, - last_io.pid, "----------------", - last_io.type, - last_io.fio.op, last_io.fio.op_flags, - last_io.fio.new_blkaddr, - last_io.len); - memset(&last_io, 0, sizeof(last_io)); -} - -static int __file_type(struct inode *inode, pid_t pid) -{ - if (f2fs_is_atomic_file(inode)) - return __ATOMIC_FILE; - else if (f2fs_is_volatile_file(inode)) - return __VOLATILE_FILE; - else if (S_ISDIR(inode->i_mode)) - return __DIR_FILE; - else if (inode->i_ino == F2FS_NODE_INO(F2FS_I_SB(inode))) - return __NODE_FILE; - else if (inode->i_ino == F2FS_META_INO(F2FS_I_SB(inode))) - return __META_FILE; - else if (pid) - return __NORMAL_FILE; - else - return __MISC_FILE; -} - -void f2fs_trace_pid(struct page *page) -{ - struct inode *inode = page->mapping->host; - pid_t pid = task_pid_nr(current); - void *p; - - set_page_private(page, (unsigned long)pid); - -retry: - if (radix_tree_preload(GFP_NOFS)) - return; - - spin_lock(&pids_lock); - p = radix_tree_lookup(&pids, pid); - if (p == current) - goto out; - if (p) - radix_tree_delete(&pids, pid); - - if (radix_tree_insert(&pids, pid, current)) { - spin_unlock(&pids_lock); - radix_tree_preload_end(); - cond_resched(); - goto retry; - } - - trace_printk("%3x:%3x %4x %-16s\n", - MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), - pid, current->comm); -out: - spin_unlock(&pids_lock); - radix_tree_preload_end(); -} - -void f2fs_trace_ios(struct f2fs_io_info *fio, int flush) -{ - struct inode *inode; - pid_t pid; - int major, minor; - - if (flush) { - __print_last_io(); - return; - } - - inode = fio->page->mapping->host; - pid = page_private(fio->page); - - major = MAJOR(inode->i_sb->s_dev); - minor = MINOR(inode->i_sb->s_dev); - - if (last_io.major == major && last_io.minor == minor && - last_io.pid == pid && - last_io.type == __file_type(inode, pid) && - last_io.fio.op == fio->op && - last_io.fio.op_flags == fio->op_flags && - last_io.fio.new_blkaddr + last_io.len == - fio->new_blkaddr) { - last_io.len++; - return; - } - - __print_last_io(); - - last_io.major = major; - last_io.minor = minor; - last_io.pid = pid; - last_io.type = __file_type(inode, pid); - last_io.fio = *fio; - last_io.len = 1; - return; -} - -void f2fs_build_trace_ios(void) -{ - spin_lock_init(&pids_lock); -} - -#define PIDVEC_SIZE 128 -static unsigned int gang_lookup_pids(pid_t *results, unsigned long first_index, - unsigned int max_items) -{ - struct radix_tree_iter iter; - void **slot; - unsigned int ret = 0; - - if (unlikely(!max_items)) - return 0; - - radix_tree_for_each_slot(slot, &pids, &iter, first_index) { - results[ret] = iter.index; - if (++ret == max_items) - break; - } - return ret; -} - -void f2fs_destroy_trace_ios(void) -{ - pid_t pid[PIDVEC_SIZE]; - pid_t next_pid = 0; - unsigned int found; - - spin_lock(&pids_lock); - while ((found = gang_lookup_pids(pid, next_pid, PIDVEC_SIZE))) { - unsigned idx; - - next_pid = pid[found - 1] + 1; - for (idx = 0; idx < found; idx++) - radix_tree_delete(&pids, pid[idx]); - } - spin_unlock(&pids_lock); -} diff --git a/fs/f2fs/trace.h b/fs/f2fs/trace.h deleted file mode 100644 index 789f6aa727fc..000000000000 --- a/fs/f2fs/trace.h +++ /dev/null @@ -1,43 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 */ -/* - * f2fs IO tracer - * - * Copyright (c) 2014 Motorola Mobility - * Copyright (c) 2014 Jaegeuk Kim - */ -#ifndef __F2FS_TRACE_H__ -#define __F2FS_TRACE_H__ - -#ifdef CONFIG_F2FS_IO_TRACE -#include - -enum file_type { - __NORMAL_FILE, - __DIR_FILE, - __NODE_FILE, - __META_FILE, - __ATOMIC_FILE, - __VOLATILE_FILE, - __MISC_FILE, -}; - -struct last_io_info { - int major, minor; - pid_t pid; - enum file_type type; - struct f2fs_io_info fio; - block_t len; -}; - -extern void f2fs_trace_pid(struct page *); -extern void f2fs_trace_ios(struct f2fs_io_info *, int); -extern void f2fs_build_trace_ios(void); -extern void f2fs_destroy_trace_ios(void); -#else -#define f2fs_trace_pid(p) -#define f2fs_trace_ios(i, n) -#define f2fs_build_trace_ios() -#define f2fs_destroy_trace_ios() - -#endif -#endif /* __F2FS_TRACE_H__ */ From b0ff4fe746fd028eef920ddc8c7b0361c1ede6ec Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Tue, 26 Jan 2021 17:00:42 -0800 Subject: [PATCH 23/32] f2fs: flush data when enabling checkpoint back During checkpoint=disable period, f2fs bypasses all the synchronous IOs such as sync and fsync. So, when enabling it back, we must flush all of them in order to keep the data persistent. Otherwise, suddern power-cut right after enabling checkpoint will cause data loss. Fixes: 4354994f097d ("f2fs: checkpoint disabling") Cc: stable@vger.kernel.org Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/super.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index fc343243799c..429bc00af440 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -1892,6 +1892,9 @@ restore_flag: static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) { + /* we should flush all the data to keep data consistency */ + sync_inodes_sb(sbi->sb); + down_write(&sbi->gc_lock); f2fs_dirty_to_prefree(sbi); From 25fb04dbce6a0e165d28fd1fa8a1d7018c637fe8 Mon Sep 17 00:00:00 2001 From: Yi Chen Date: Thu, 28 Jan 2021 17:02:56 +0800 Subject: [PATCH 24/32] f2fs: fix to avoid inconsistent quota data Occasionally, quota data may be corrupted detected by fsck: Info: checkpoint state = 45 : crc compacted_summary unmount [QUOTA WARNING] Usage inconsistent for ID 0:actual (1543036928, 762) != expected (1543032832, 762) [ASSERT] (fsck_chk_quota_files:1986) --> Quota file is missing or invalid quota file content found. [QUOTA WARNING] Usage inconsistent for ID 0:actual (1352478720, 344) != expected (1352474624, 344) [ASSERT] (fsck_chk_quota_files:1986) --> Quota file is missing or invalid quota file content found. [FSCK] Unreachable nat entries [Ok..] [0x0] [FSCK] SIT valid block bitmap checking [Ok..] [FSCK] Hard link checking for regular file [Ok..] [0x0] [FSCK] valid_block_count matching with CP [Ok..] [0xdf299] [FSCK] valid_node_count matcing with CP (de lookup) [Ok..] [0x2b01] [FSCK] valid_node_count matcing with CP (nat lookup) [Ok..] [0x2b01] [FSCK] valid_inode_count matched with CP [Ok..] [0x2665] [FSCK] free segment_count matched with CP [Ok..] [0xcb04] [FSCK] next block offset is free [Ok..] [FSCK] fixing SIT types [FSCK] other corrupted bugs [Fail] The root cause is: If we open file w/ readonly flag, disk quota info won't be initialized for this file, however, following mmap() will force to convert inline inode via f2fs_convert_inline_inode(), which may increase block usage for this inode w/o updating quota data, it causes inconsistent disk quota info. The issue will happen in following stack: open(file, O_RDONLY) mmap(file) - f2fs_convert_inline_inode - f2fs_convert_inline_page - f2fs_reserve_block - f2fs_reserve_new_block - f2fs_reserve_new_blocks - f2fs_i_blocks_write - dquot_claim_block inode->i_blocks increase, but the dqb_curspace keep the size for the dquots is NULL. To fix this issue, let's call dquot_initialize() anyway in both f2fs_truncate() and f2fs_convert_inline_inode() functions to avoid potential inconsistent quota data issue. Fixes: 0abd675e97e6 ("f2fs: support plain user/group quota") Signed-off-by: Daiyue Zhang Signed-off-by: Dehe Gu Signed-off-by: Junchao Jiang Signed-off-by: Ge Qiu Signed-off-by: Yi Chen Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 4 ++++ fs/f2fs/inline.c | 4 ++++ 2 files changed, 8 insertions(+) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 7db27c81d034..00b2ce47fa37 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -771,6 +771,10 @@ int f2fs_truncate(struct inode *inode) return -EIO; } + err = dquot_initialize(inode); + if (err) + return err; + /* we should check inline_data size */ if (!f2fs_may_inline_data(inode)) { err = f2fs_convert_inline_inode(inode); diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c index 806ebabf5870..993caefcd2bb 100644 --- a/fs/f2fs/inline.c +++ b/fs/f2fs/inline.c @@ -192,6 +192,10 @@ int f2fs_convert_inline_inode(struct inode *inode) f2fs_hw_is_readonly(sbi) || f2fs_readonly(sbi->sb)) return 0; + err = dquot_initialize(inode); + if (err) + return err; + page = f2fs_grab_cache_page(inode->i_mapping, 0, false); if (!page) return -ENOMEM; From 2e0cd472a0dd9b9a35699502570015af15d7c70f Mon Sep 17 00:00:00 2001 From: Liu Song Date: Sun, 31 Jan 2021 20:26:05 +0800 Subject: [PATCH 25/32] f2fs: remove unnecessary initialization in xattr.c These variables will be explicitly assigned before use, so there is no need to initialize. Signed-off-by: Liu Song Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/xattr.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c index 2086bef6c154..8159fae74b9a 100644 --- a/fs/f2fs/xattr.c +++ b/fs/f2fs/xattr.c @@ -327,7 +327,7 @@ static int lookup_all_xattrs(struct inode *inode, struct page *ipage, void *last_addr = NULL; nid_t xnid = F2FS_I(inode)->i_xattr_nid; unsigned int inline_size = inline_xattr_size(inode); - int err = 0; + int err; if (!xnid && !inline_size) return -ENODATA; @@ -515,7 +515,7 @@ int f2fs_getxattr(struct inode *inode, int index, const char *name, void *buffer, size_t buffer_size, struct page *ipage) { struct f2fs_xattr_entry *entry = NULL; - int error = 0; + int error; unsigned int size, len; void *base_addr = NULL; int base_size; @@ -562,7 +562,7 @@ ssize_t f2fs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size) struct inode *inode = d_inode(dentry); struct f2fs_xattr_entry *entry; void *base_addr, *last_base_addr; - int error = 0; + int error; size_t rest = buffer_size; down_read(&F2FS_I(inode)->i_xattr_sem); @@ -632,7 +632,7 @@ static int __f2fs_setxattr(struct inode *inode, int index, int found, newsize; size_t len; __u32 new_hsize; - int error = 0; + int error; if (name == NULL) return -EINVAL; From 39f71b7e40e21805d6b15fc7750bdd9cab6a5010 Mon Sep 17 00:00:00 2001 From: Dehe Gu Date: Tue, 2 Feb 2021 17:39:22 +0800 Subject: [PATCH 26/32] f2fs: fix a wrong condition in __submit_bio We should use !F2FS_IO_ALIGNED() to check and submit_io directly. Fixes: 8223ecc456d0 ("f2fs: fix to add missing F2FS_IO_ALIGNED() condition") Reviewed-by: Chao Yu Signed-off-by: Dehe Gu Signed-off-by: Ge Qiu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 38476d0d3916..1ee966a63df9 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -470,7 +470,7 @@ static inline void __submit_bio(struct f2fs_sb_info *sbi, if (f2fs_lfs_mode(sbi) && current->plug) blk_finish_plug(current->plug); - if (F2FS_IO_ALIGNED(sbi)) + if (!F2FS_IO_ALIGNED(sbi)) goto submit_io; start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS; From c8e43d55b1aa05d175daac25d228c7c1c71c7b11 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Tue, 2 Feb 2021 16:03:11 +0800 Subject: [PATCH 27/32] f2fs: relocate inline conversion from mmap() to mkwrite() If there is page fault only for read case on inline inode, we don't need to convert inline inode, instead, let's do conversion for write case. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/file.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 00b2ce47fa37..8e53f8898688 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -72,6 +72,10 @@ static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) goto err; } + err = f2fs_convert_inline_inode(inode); + if (err) + goto err; + #ifdef CONFIG_F2FS_FS_COMPRESSION if (f2fs_compressed_file(inode)) { int ret = f2fs_is_compressed_cluster(inode, page->index); @@ -506,7 +510,6 @@ static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) { struct inode *inode = file_inode(file); - int err; if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) return -EIO; @@ -514,11 +517,6 @@ static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) if (!f2fs_is_compress_backend_ready(inode)) return -EOPNOTSUPP; - /* we don't need to use inline_data strictly */ - err = f2fs_convert_inline_inode(inode); - if (err) - return err; - file_accessed(file); vma->vm_ops = &f2fs_file_vm_ops; set_inode_flag(inode, FI_MMAP_FILE); From 261eeb9c1585de4515a770b48a3c89672c08ae7f Mon Sep 17 00:00:00 2001 From: Daeho Jeong Date: Tue, 19 Jan 2021 09:00:42 +0900 Subject: [PATCH 28/32] f2fs: introduce checkpoint_merge mount option We've added a new mount options, "checkpoint_merge" and "nocheckpoint_merge", which creates a kernel daemon and makes it to merge concurrent checkpoint requests as much as possible to eliminate redundant checkpoint issues. Plus, we can eliminate the sluggish issue caused by slow checkpoint operation when the checkpoint is done in a process context in a cgroup having low i/o budget and cpu shares. To make this do better, we set the default i/o priority of the kernel daemon to "3", to give one higher priority than other kernel threads. The below verification result explains this. The basic idea has come from https://opensource.samsung.com. [Verification] Android Pixel Device(ARM64, 7GB RAM, 256GB UFS) Create two I/O cgroups (fg w/ weight 100, bg w/ wight 20) Set "strict_guarantees" to "1" in BFQ tunables In "fg" cgroup, - thread A => trigger 1000 checkpoint operations "for i in `seq 1 1000`; do touch test_dir1/file; fsync test_dir1; done" - thread B => gererating async. I/O "fio --rw=write --numjobs=1 --bs=128k --runtime=3600 --time_based=1 --filename=test_img --name=test" In "bg" cgroup, - thread C => trigger repeated checkpoint operations "echo $$ > /dev/blkio/bg/tasks; while true; do touch test_dir2/file; fsync test_dir2; done" We've measured thread A's execution time. [ w/o patch ] Elapsed Time: Avg. 68 seconds [ w/ patch ] Elapsed Time: Avg. 48 seconds Reported-by: kernel test robot Reported-by: Dan Carpenter [Jaegeuk Kim: fix the return value in f2fs_start_ckpt_thread, reported by Dan] Signed-off-by: Daeho Jeong Signed-off-by: Sungjong Seo Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.rst | 11 ++ fs/f2fs/checkpoint.c | 177 +++++++++++++++++++++++++++++ fs/f2fs/debug.c | 12 ++ fs/f2fs/f2fs.h | 27 +++++ fs/f2fs/super.c | 58 ++++++++-- 5 files changed, 277 insertions(+), 8 deletions(-) diff --git a/Documentation/filesystems/f2fs.rst b/Documentation/filesystems/f2fs.rst index 5eff4009e77e..f75ec244762f 100644 --- a/Documentation/filesystems/f2fs.rst +++ b/Documentation/filesystems/f2fs.rst @@ -247,6 +247,17 @@ checkpoint=%s[:%u[%]] Set to "disable" to turn off checkpointing. Set to "enabl hide up to all remaining free space. The actual space that would be unusable can be viewed at /sys/fs/f2fs//unusable This space is reclaimed once checkpoint=enable. +checkpoint_merge When checkpoint is enabled, this can be used to create a kernel + daemon and make it to merge concurrent checkpoint requests as + much as possible to eliminate redundant checkpoint issues. Plus, + we can eliminate the sluggish issue caused by slow checkpoint + operation when the checkpoint is done in a process context in + a cgroup having low i/o budget and cpu shares. To make this + do better, we set the default i/o priority of the kernel daemon + to "3", to give one higher priority than other kernel threads. + This is the same way to give a I/O priority to the jbd2 + journaling thread of ext4 filesystem. +nocheckpoint_merge Disable checkpoint merge feature. compress_algorithm=%s Control compress algorithm, currently f2fs supports "lzo", "lz4", "zstd" and "lzo-rle" algorithm. compress_algorithm=%s:%d Control compress algorithm and its compress level, now, only diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 8c79ba0566b1..6f6ecba8f920 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -13,12 +13,15 @@ #include #include #include +#include #include "f2fs.h" #include "node.h" #include "segment.h" #include +#define DEFAULT_CHECKPOINT_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) + static struct kmem_cache *ino_entry_slab; struct kmem_cache *f2fs_inode_entry_slab; @@ -1704,3 +1707,177 @@ void f2fs_destroy_checkpoint_caches(void) kmem_cache_destroy(ino_entry_slab); kmem_cache_destroy(f2fs_inode_entry_slab); } + +static int __write_checkpoint_sync(struct f2fs_sb_info *sbi) +{ + struct cp_control cpc = { .reason = CP_SYNC, }; + int err; + + down_write(&sbi->gc_lock); + err = f2fs_write_checkpoint(sbi, &cpc); + up_write(&sbi->gc_lock); + + return err; +} + +static void __checkpoint_and_complete_reqs(struct f2fs_sb_info *sbi) +{ + struct ckpt_req_control *cprc = &sbi->cprc_info; + struct ckpt_req *req, *next; + struct llist_node *dispatch_list; + u64 sum_diff = 0, diff, count = 0; + int ret; + + dispatch_list = llist_del_all(&cprc->issue_list); + if (!dispatch_list) + return; + dispatch_list = llist_reverse_order(dispatch_list); + + ret = __write_checkpoint_sync(sbi); + atomic_inc(&cprc->issued_ckpt); + + llist_for_each_entry_safe(req, next, dispatch_list, llnode) { + diff = (u64)ktime_ms_delta(ktime_get(), req->queue_time); + req->ret = ret; + complete(&req->wait); + + sum_diff += diff; + count++; + } + atomic_sub(count, &cprc->queued_ckpt); + atomic_add(count, &cprc->total_ckpt); + + spin_lock(&cprc->stat_lock); + cprc->cur_time = (unsigned int)div64_u64(sum_diff, count); + if (cprc->peak_time < cprc->cur_time) + cprc->peak_time = cprc->cur_time; + spin_unlock(&cprc->stat_lock); +} + +static int issue_checkpoint_thread(void *data) +{ + struct f2fs_sb_info *sbi = data; + struct ckpt_req_control *cprc = &sbi->cprc_info; + wait_queue_head_t *q = &cprc->ckpt_wait_queue; +repeat: + if (kthread_should_stop()) + return 0; + + sb_start_intwrite(sbi->sb); + + if (!llist_empty(&cprc->issue_list)) + __checkpoint_and_complete_reqs(sbi); + + sb_end_intwrite(sbi->sb); + + wait_event_interruptible(*q, + kthread_should_stop() || !llist_empty(&cprc->issue_list)); + goto repeat; +} + +static void flush_remained_ckpt_reqs(struct f2fs_sb_info *sbi, + struct ckpt_req *wait_req) +{ + struct ckpt_req_control *cprc = &sbi->cprc_info; + + if (!llist_empty(&cprc->issue_list)) { + __checkpoint_and_complete_reqs(sbi); + } else { + /* already dispatched by issue_checkpoint_thread */ + if (wait_req) + wait_for_completion(&wait_req->wait); + } +} + +static void init_ckpt_req(struct ckpt_req *req) +{ + memset(req, 0, sizeof(struct ckpt_req)); + + init_completion(&req->wait); + req->queue_time = ktime_get(); +} + +int f2fs_issue_checkpoint(struct f2fs_sb_info *sbi) +{ + struct ckpt_req_control *cprc = &sbi->cprc_info; + struct ckpt_req req; + struct cp_control cpc; + + cpc.reason = __get_cp_reason(sbi); + if (!test_opt(sbi, MERGE_CHECKPOINT) || cpc.reason != CP_SYNC) { + int ret; + + down_write(&sbi->gc_lock); + ret = f2fs_write_checkpoint(sbi, &cpc); + up_write(&sbi->gc_lock); + + return ret; + } + + if (!cprc->f2fs_issue_ckpt) + return __write_checkpoint_sync(sbi); + + init_ckpt_req(&req); + + llist_add(&req.llnode, &cprc->issue_list); + atomic_inc(&cprc->queued_ckpt); + + /* update issue_list before we wake up issue_checkpoint thread */ + smp_mb(); + + if (waitqueue_active(&cprc->ckpt_wait_queue)) + wake_up(&cprc->ckpt_wait_queue); + + if (cprc->f2fs_issue_ckpt) + wait_for_completion(&req.wait); + else + flush_remained_ckpt_reqs(sbi, &req); + + return req.ret; +} + +int f2fs_start_ckpt_thread(struct f2fs_sb_info *sbi) +{ + dev_t dev = sbi->sb->s_bdev->bd_dev; + struct ckpt_req_control *cprc = &sbi->cprc_info; + + if (cprc->f2fs_issue_ckpt) + return 0; + + cprc->f2fs_issue_ckpt = kthread_run(issue_checkpoint_thread, sbi, + "f2fs_ckpt-%u:%u", MAJOR(dev), MINOR(dev)); + if (IS_ERR(cprc->f2fs_issue_ckpt)) { + cprc->f2fs_issue_ckpt = NULL; + return -ENOMEM; + } + + set_task_ioprio(cprc->f2fs_issue_ckpt, DEFAULT_CHECKPOINT_IOPRIO); + + return 0; +} + +void f2fs_stop_ckpt_thread(struct f2fs_sb_info *sbi) +{ + struct ckpt_req_control *cprc = &sbi->cprc_info; + + if (cprc->f2fs_issue_ckpt) { + struct task_struct *ckpt_task = cprc->f2fs_issue_ckpt; + + cprc->f2fs_issue_ckpt = NULL; + kthread_stop(ckpt_task); + + flush_remained_ckpt_reqs(sbi, NULL); + } +} + +void f2fs_init_ckpt_req_control(struct f2fs_sb_info *sbi) +{ + struct ckpt_req_control *cprc = &sbi->cprc_info; + + atomic_set(&cprc->issued_ckpt, 0); + atomic_set(&cprc->total_ckpt, 0); + atomic_set(&cprc->queued_ckpt, 0); + init_waitqueue_head(&cprc->ckpt_wait_queue); + init_llist_head(&cprc->issue_list); + spin_lock_init(&cprc->stat_lock); +} diff --git a/fs/f2fs/debug.c b/fs/f2fs/debug.c index 197c914119da..91855d5721cd 100644 --- a/fs/f2fs/debug.c +++ b/fs/f2fs/debug.c @@ -120,6 +120,13 @@ static void update_general_status(struct f2fs_sb_info *sbi) atomic_read(&SM_I(sbi)->dcc_info->discard_cmd_cnt); si->undiscard_blks = SM_I(sbi)->dcc_info->undiscard_blks; } + si->nr_issued_ckpt = atomic_read(&sbi->cprc_info.issued_ckpt); + si->nr_total_ckpt = atomic_read(&sbi->cprc_info.total_ckpt); + si->nr_queued_ckpt = atomic_read(&sbi->cprc_info.queued_ckpt); + spin_lock(&sbi->cprc_info.stat_lock); + si->cur_ckpt_time = sbi->cprc_info.cur_time; + si->peak_ckpt_time = sbi->cprc_info.peak_time; + spin_unlock(&sbi->cprc_info.stat_lock); si->total_count = (int)sbi->user_block_count / sbi->blocks_per_seg; si->rsvd_segs = reserved_segments(sbi); si->overp_segs = overprovision_segments(sbi); @@ -417,6 +424,11 @@ static int stat_show(struct seq_file *s, void *v) si->meta_count[META_NAT]); seq_printf(s, " - ssa blocks : %u\n", si->meta_count[META_SSA]); + seq_printf(s, "CP merge (Queued: %4d, Issued: %4d, Total: %4d, " + "Cur time: %4d(ms), Peak time: %4d(ms))\n", + si->nr_queued_ckpt, si->nr_issued_ckpt, + si->nr_total_ckpt, si->cur_ckpt_time, + si->peak_ckpt_time); seq_printf(s, "GC calls: %d (BG: %d)\n", si->call_count, si->bg_gc); seq_printf(s, " - data segments : %d (%d)\n", diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index a2e520a13630..f7536aca8a31 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -97,6 +97,7 @@ extern const char *f2fs_fault_name[FAULT_MAX]; #define F2FS_MOUNT_DISABLE_CHECKPOINT 0x02000000 #define F2FS_MOUNT_NORECOVERY 0x04000000 #define F2FS_MOUNT_ATGC 0x08000000 +#define F2FS_MOUNT_MERGE_CHECKPOINT 0x10000000 #define F2FS_OPTION(sbi) ((sbi)->mount_opt) #define clear_opt(sbi, option) (F2FS_OPTION(sbi).opt &= ~F2FS_MOUNT_##option) @@ -267,6 +268,25 @@ struct fsync_node_entry { unsigned int seq_id; /* sequence id */ }; +struct ckpt_req { + struct completion wait; /* completion for checkpoint done */ + struct llist_node llnode; /* llist_node to be linked in wait queue */ + int ret; /* return code of checkpoint */ + ktime_t queue_time; /* request queued time */ +}; + +struct ckpt_req_control { + struct task_struct *f2fs_issue_ckpt; /* checkpoint task */ + wait_queue_head_t ckpt_wait_queue; /* waiting queue for wake-up */ + atomic_t issued_ckpt; /* # of actually issued ckpts */ + atomic_t total_ckpt; /* # of total ckpts */ + atomic_t queued_ckpt; /* # of queued ckpts */ + struct llist_head issue_list; /* list for command issue */ + spinlock_t stat_lock; /* lock for below checkpoint time stats */ + unsigned int cur_time; /* cur wait time in msec for currently issued checkpoint */ + unsigned int peak_time; /* peak wait time in msec until now */ +}; + /* for the bitmap indicate blocks to be discarded */ struct discard_entry { struct list_head list; /* list head */ @@ -1433,6 +1453,7 @@ struct f2fs_sb_info { wait_queue_head_t cp_wait; unsigned long last_time[MAX_TIME]; /* to store time in jiffies */ long interval_time[MAX_TIME]; /* to store thresholds */ + struct ckpt_req_control cprc_info; /* for checkpoint request control */ struct inode_management im[MAX_INO_ENTRY]; /* manage inode cache */ @@ -3450,6 +3471,10 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc); void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi); int __init f2fs_create_checkpoint_caches(void); void f2fs_destroy_checkpoint_caches(void); +int f2fs_issue_checkpoint(struct f2fs_sb_info *sbi); +int f2fs_start_ckpt_thread(struct f2fs_sb_info *sbi); +void f2fs_stop_ckpt_thread(struct f2fs_sb_info *sbi); +void f2fs_init_ckpt_req_control(struct f2fs_sb_info *sbi); /* * data.c @@ -3562,6 +3587,8 @@ struct f2fs_stat_info { int nr_discarding, nr_discarded; int nr_discard_cmd; unsigned int undiscard_blks; + int nr_issued_ckpt, nr_total_ckpt, nr_queued_ckpt; + unsigned int cur_ckpt_time, peak_ckpt_time; int inline_xattr, inline_inode, inline_dir, append, update, orphans; int compr_inode; unsigned long long compr_blocks; diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 429bc00af440..1000d21120ca 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -144,6 +144,8 @@ enum { Opt_checkpoint_disable_cap, Opt_checkpoint_disable_cap_perc, Opt_checkpoint_enable, + Opt_checkpoint_merge, + Opt_nocheckpoint_merge, Opt_compress_algorithm, Opt_compress_log_size, Opt_compress_extension, @@ -214,6 +216,8 @@ static match_table_t f2fs_tokens = { {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, {Opt_checkpoint_enable, "checkpoint=enable"}, + {Opt_checkpoint_merge, "checkpoint_merge"}, + {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, {Opt_compress_algorithm, "compress_algorithm=%s"}, {Opt_compress_log_size, "compress_log_size=%u"}, {Opt_compress_extension, "compress_extension=%s"}, @@ -941,6 +945,12 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount) case Opt_checkpoint_enable: clear_opt(sbi, DISABLE_CHECKPOINT); break; + case Opt_checkpoint_merge: + set_opt(sbi, MERGE_CHECKPOINT); + break; + case Opt_nocheckpoint_merge: + clear_opt(sbi, MERGE_CHECKPOINT); + break; #ifdef CONFIG_F2FS_FS_COMPRESSION case Opt_compress_algorithm: if (!f2fs_sb_has_compression(sbi)) { @@ -1340,6 +1350,12 @@ static void f2fs_put_super(struct super_block *sb) /* prevent remaining shrinker jobs */ mutex_lock(&sbi->umount_mutex); + /* + * flush all issued checkpoints and stop checkpoint issue thread. + * after then, all checkpoints should be done by each process context. + */ + f2fs_stop_ckpt_thread(sbi); + /* * We don't need to do checkpoint when superblock is clean. * But, the previous checkpoint was not done by umount, it needs to do @@ -1438,15 +1454,9 @@ int f2fs_sync_fs(struct super_block *sb, int sync) if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) return -EAGAIN; - if (sync) { - struct cp_control cpc; + if (sync) + err = f2fs_issue_checkpoint(sbi); - cpc.reason = __get_cp_reason(sbi); - - down_write(&sbi->gc_lock); - err = f2fs_write_checkpoint(sbi, &cpc); - up_write(&sbi->gc_lock); - } return err; } @@ -1770,6 +1780,10 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root) if (test_opt(sbi, DISABLE_CHECKPOINT)) seq_printf(seq, ",checkpoint=disable:%u", F2FS_OPTION(sbi).unusable_cap); + if (test_opt(sbi, MERGE_CHECKPOINT)) + seq_puts(seq, ",checkpoint_merge"); + else + seq_puts(seq, ",nocheckpoint_merge"); if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) seq_printf(seq, ",fsync_mode=%s", "posix"); else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) @@ -2053,6 +2067,19 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data) } } + if (!test_opt(sbi, DISABLE_CHECKPOINT) && + test_opt(sbi, MERGE_CHECKPOINT)) { + err = f2fs_start_ckpt_thread(sbi); + if (err) { + f2fs_err(sbi, + "Failed to start F2FS issue_checkpoint_thread (%d)", + err); + goto restore_gc; + } + } else { + f2fs_stop_ckpt_thread(sbi); + } + /* * We stop issue flush thread if FS is mounted as RO * or if flush_merge is not passed in mount option. @@ -3804,6 +3831,19 @@ try_onemore: f2fs_init_fsync_node_info(sbi); + /* setup checkpoint request control and start checkpoint issue thread */ + f2fs_init_ckpt_req_control(sbi); + if (!test_opt(sbi, DISABLE_CHECKPOINT) && + test_opt(sbi, MERGE_CHECKPOINT)) { + err = f2fs_start_ckpt_thread(sbi); + if (err) { + f2fs_err(sbi, + "Failed to start F2FS issue_checkpoint_thread (%d)", + err); + goto stop_ckpt_thread; + } + } + /* setup f2fs internal modules */ err = f2fs_build_segment_manager(sbi); if (err) { @@ -4013,6 +4053,8 @@ free_nm: free_sm: f2fs_destroy_segment_manager(sbi); f2fs_destroy_post_read_wq(sbi); +stop_ckpt_thread: + f2fs_stop_ckpt_thread(sbi); free_devices: destroy_device_list(sbi); kvfree(sbi->ckpt); From e65920661708b7c0f3db45c9cd5d0095034ee37f Mon Sep 17 00:00:00 2001 From: Daeho Jeong Date: Thu, 21 Jan 2021 22:45:29 +0900 Subject: [PATCH 29/32] f2fs: add ckpt_thread_ioprio sysfs node Added "ckpt_thread_ioprio" sysfs node to give a way to change checkpoint merge daemon's io priority. Its default value is "be,3", which means "BE" I/O class and I/O priority "3". We can select the class between "rt" and "be", and set the I/O priority within valid range of it. "," delimiter is necessary in between I/O class and priority number. Signed-off-by: Daeho Jeong Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/ABI/testing/sysfs-fs-f2fs | 9 ++++ fs/f2fs/checkpoint.c | 3 +- fs/f2fs/f2fs.h | 1 + fs/f2fs/sysfs.c | 55 +++++++++++++++++++++++++ 4 files changed, 67 insertions(+), 1 deletion(-) diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fs index 362803901614..cbeac1bebe2f 100644 --- a/Documentation/ABI/testing/sysfs-fs-f2fs +++ b/Documentation/ABI/testing/sysfs-fs-f2fs @@ -400,3 +400,12 @@ Description: Show status of f2fs superblock in real time. 0x1000 SBI_QUOTA_NEED_REPAIR quota file may be corrupted 0x2000 SBI_IS_RESIZEFS resizefs is in process ====== ===================== ================================= + +What: /sys/fs/f2fs//ckpt_thread_ioprio +Date: January 2021 +Contact: "Daeho Jeong" +Description: Give a way to change checkpoint merge daemon's io priority. + Its default value is "be,3", which means "BE" I/O class and + I/O priority "3". We can select the class between "rt" and "be", + and set the I/O priority within valid range of it. "," delimiter + is necessary in between I/O class and priority number. diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 6f6ecba8f920..579b9c3603cc 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -1851,7 +1851,7 @@ int f2fs_start_ckpt_thread(struct f2fs_sb_info *sbi) return -ENOMEM; } - set_task_ioprio(cprc->f2fs_issue_ckpt, DEFAULT_CHECKPOINT_IOPRIO); + set_task_ioprio(cprc->f2fs_issue_ckpt, cprc->ckpt_thread_ioprio); return 0; } @@ -1877,6 +1877,7 @@ void f2fs_init_ckpt_req_control(struct f2fs_sb_info *sbi) atomic_set(&cprc->issued_ckpt, 0); atomic_set(&cprc->total_ckpt, 0); atomic_set(&cprc->queued_ckpt, 0); + cprc->ckpt_thread_ioprio = DEFAULT_CHECKPOINT_IOPRIO; init_waitqueue_head(&cprc->ckpt_wait_queue); init_llist_head(&cprc->issue_list); spin_lock_init(&cprc->stat_lock); diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index f7536aca8a31..2860003a09ed 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -277,6 +277,7 @@ struct ckpt_req { struct ckpt_req_control { struct task_struct *f2fs_issue_ckpt; /* checkpoint task */ + int ckpt_thread_ioprio; /* checkpoint merge thread ioprio */ wait_queue_head_t ckpt_wait_queue; /* waiting queue for wake-up */ atomic_t issued_ckpt; /* # of actually issued ckpts */ atomic_t total_ckpt; /* # of total ckpts */ diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c index f39874d512ea..e38a7f6921dd 100644 --- a/fs/f2fs/sysfs.c +++ b/fs/f2fs/sysfs.c @@ -11,6 +11,7 @@ #include #include #include +#include #include "f2fs.h" #include "segment.h" @@ -34,6 +35,7 @@ enum { FAULT_INFO_TYPE, /* struct f2fs_fault_info */ #endif RESERVED_BLOCKS, /* struct f2fs_sb_info */ + CPRC_INFO, /* struct ckpt_req_control */ }; struct f2fs_attr { @@ -70,6 +72,8 @@ static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type) else if (struct_type == STAT_INFO) return (unsigned char *)F2FS_STAT(sbi); #endif + else if (struct_type == CPRC_INFO) + return (unsigned char *)&sbi->cprc_info; return NULL; } @@ -261,6 +265,23 @@ static ssize_t f2fs_sbi_show(struct f2fs_attr *a, return len; } + if (!strcmp(a->attr.name, "ckpt_thread_ioprio")) { + struct ckpt_req_control *cprc = &sbi->cprc_info; + int len = 0; + int class = IOPRIO_PRIO_CLASS(cprc->ckpt_thread_ioprio); + int data = IOPRIO_PRIO_DATA(cprc->ckpt_thread_ioprio); + + if (class == IOPRIO_CLASS_RT) + len += scnprintf(buf + len, PAGE_SIZE - len, "rt,"); + else if (class == IOPRIO_CLASS_BE) + len += scnprintf(buf + len, PAGE_SIZE - len, "be,"); + else + return -EINVAL; + + len += scnprintf(buf + len, PAGE_SIZE - len, "%d\n", data); + return len; + } + ui = (unsigned int *)(ptr + a->offset); return sprintf(buf, "%u\n", *ui); @@ -314,6 +335,38 @@ out: return ret ? ret : count; } + if (!strcmp(a->attr.name, "ckpt_thread_ioprio")) { + const char *name = strim((char *)buf); + struct ckpt_req_control *cprc = &sbi->cprc_info; + int class; + long data; + int ret; + + if (!strncmp(name, "rt,", 3)) + class = IOPRIO_CLASS_RT; + else if (!strncmp(name, "be,", 3)) + class = IOPRIO_CLASS_BE; + else + return -EINVAL; + + name += 3; + ret = kstrtol(name, 10, &data); + if (ret) + return ret; + if (data >= IOPRIO_BE_NR || data < 0) + return -EINVAL; + + cprc->ckpt_thread_ioprio = IOPRIO_PRIO_VALUE(class, data); + if (test_opt(sbi, MERGE_CHECKPOINT)) { + ret = set_task_ioprio(cprc->f2fs_issue_ckpt, + cprc->ckpt_thread_ioprio); + if (ret) + return ret; + } + + return count; + } + ui = (unsigned int *)(ptr + a->offset); ret = kstrtoul(skip_spaces(buf), 0, &t); @@ -573,6 +626,7 @@ F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type); #endif F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, data_io_flag, data_io_flag); F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, node_io_flag, node_io_flag); +F2FS_RW_ATTR(CPRC_INFO, ckpt_req_control, ckpt_thread_ioprio, ckpt_thread_ioprio); F2FS_GENERAL_RO_ATTR(dirty_segments); F2FS_GENERAL_RO_ATTR(free_segments); F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes); @@ -658,6 +712,7 @@ static struct attribute *f2fs_attrs[] = { #endif ATTR_LIST(data_io_flag), ATTR_LIST(node_io_flag), + ATTR_LIST(ckpt_thread_ioprio), ATTR_LIST(dirty_segments), ATTR_LIST(free_segments), ATTR_LIST(unusable), From d50dfc0c7df7bf037442045fbe63952ae0c4ce46 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Mon, 8 Feb 2021 13:42:21 -0800 Subject: [PATCH 30/32] f2fs: don't grab superblock freeze for flush/ckpt thread There are controlled by f2fs_freeze(). This fixes xfstests/generic/068 which is stuck at task:f2fs_ckpt-252:3 state:D stack: 0 pid: 5761 ppid: 2 flags:0x00004000 Call Trace: __schedule+0x44c/0x8a0 schedule+0x4f/0xc0 percpu_rwsem_wait+0xd8/0x140 ? percpu_down_write+0xf0/0xf0 __percpu_down_read+0x56/0x70 issue_checkpoint_thread+0x12c/0x160 [f2fs] ? wait_woken+0x80/0x80 kthread+0x114/0x150 ? __checkpoint_and_complete_reqs+0x110/0x110 [f2fs] ? kthread_park+0x90/0x90 ret_from_fork+0x22/0x30 Signed-off-by: Jaegeuk Kim --- fs/f2fs/checkpoint.c | 4 ---- fs/f2fs/segment.c | 4 ---- fs/f2fs/super.c | 4 ++++ 3 files changed, 4 insertions(+), 8 deletions(-) diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 579b9c3603cc..174a0819ad96 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -1763,13 +1763,9 @@ repeat: if (kthread_should_stop()) return 0; - sb_start_intwrite(sbi->sb); - if (!llist_empty(&cprc->issue_list)) __checkpoint_and_complete_reqs(sbi); - sb_end_intwrite(sbi->sb); - wait_event_interruptible(*q, kthread_should_stop() || !llist_empty(&cprc->issue_list)); goto repeat; diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 7d34f1cacdee..440634dfaa56 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -607,8 +607,6 @@ repeat: if (kthread_should_stop()) return 0; - sb_start_intwrite(sbi->sb); - if (!llist_empty(&fcc->issue_list)) { struct flush_cmd *cmd, *next; int ret; @@ -629,8 +627,6 @@ repeat: fcc->dispatch_list = NULL; } - sb_end_intwrite(sbi->sb); - wait_event_interruptible(*q, kthread_should_stop() || !llist_empty(&fcc->issue_list)); goto repeat; diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 1000d21120ca..4aa533cb4340 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -1472,6 +1472,10 @@ static int f2fs_freeze(struct super_block *sb) /* must be clean, since sync_filesystem() was already called */ if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) return -EINVAL; + + /* ensure no checkpoint required */ + if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) + return -EINVAL; return 0; } From 938a184265d75ea474f1c6fe1da96a5196163789 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Fri, 12 Feb 2021 14:09:54 -0800 Subject: [PATCH 31/32] f2fs: give a warning only for readonly partition Let's allow mounting readonly partition. We're able to recovery later once we have it as read-write back. Signed-off-by: Jaegeuk Kim --- fs/f2fs/super.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 4aa533cb4340..30d5abef4361 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -3933,12 +3933,10 @@ try_onemore: * previous checkpoint was not done by clean system shutdown. */ if (f2fs_hw_is_readonly(sbi)) { - if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { - err = -EROFS; + if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); - goto free_meta; - } - f2fs_info(sbi, "write access unavailable, skipping recovery"); + else + f2fs_info(sbi, "write access unavailable, skipping recovery"); goto reset_checkpoint; } From 092af2eb180062f5bafe02a75da9856676eb4f89 Mon Sep 17 00:00:00 2001 From: Ed Tsai Date: Thu, 4 Feb 2021 21:25:56 +0800 Subject: [PATCH 32/32] Documentation: f2fs: fix typo s/automaic/automatic Fix typo in f2fs documentation. Signed-off-by: Ed Tsai Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Documentation/filesystems/f2fs.rst b/Documentation/filesystems/f2fs.rst index f75ec244762f..81c05baa8312 100644 --- a/Documentation/filesystems/f2fs.rst +++ b/Documentation/filesystems/f2fs.rst @@ -847,7 +847,7 @@ This is the default option. f2fs does automatic compression in the writeback of compression enabled files. 2) compress_mode=user -This disables the automaic compression and gives the user discretion of choosing the +This disables the automatic compression and gives the user discretion of choosing the target file and the timing. The user can do manual compression/decompression on the compression enabled files using F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE ioctls like the below.