From 0db18eec0d9a7ee525209e31e3ac2f673545b12f Mon Sep 17 00:00:00 2001 From: Mukesh Ojha Date: Thu, 27 Oct 2022 14:42:40 +0530 Subject: f2fs: fix the assign logic of iocb commit 18ae8d12991b ("f2fs: show more DIO information in tracepoint") introduces iocb field in 'f2fs_direct_IO_enter' trace event And it only assigns the pointer and later it accesses its field in trace print log. Unable to handle kernel paging request at virtual address ffffffc04cef3d30 Mem abort info: ESR = 0x96000007 EC = 0x25: DABT (current EL), IL = 32 bits pc : trace_raw_output_f2fs_direct_IO_enter+0x54/0xa4 lr : trace_raw_output_f2fs_direct_IO_enter+0x2c/0xa4 sp : ffffffc0443cbbd0 x29: ffffffc0443cbbf0 x28: ffffff8935b120d0 x27: ffffff8935b12108 x26: ffffff8935b120f0 x25: ffffff8935b12100 x24: ffffff8935b110c0 x23: ffffff8935b10000 x22: ffffff88859a936c x21: ffffff88859a936c x20: ffffff8935b110c0 x19: ffffff8935b10000 x18: ffffffc03b195060 x17: ffffff8935b11e76 x16: 00000000000000cc x15: ffffffef855c4f2c x14: 0000000000000001 x13: 000000000000004e x12: ffff0000ffffff00 x11: ffffffef86c350d0 x10: 00000000000010c0 x9 : 000000000fe0002c x8 : ffffffc04cef3d28 x7 : 7f7f7f7f7f7f7f7f x6 : 0000000002000000 x5 : ffffff8935b11e9a x4 : 0000000000006250 x3 : ffff0a00ffffff04 x2 : 0000000000000002 x1 : ffffffef86a0a31f x0 : ffffff8935b10000 Call trace: trace_raw_output_f2fs_direct_IO_enter+0x54/0xa4 print_trace_fmt+0x9c/0x138 print_trace_line+0x154/0x254 tracing_read_pipe+0x21c/0x380 vfs_read+0x108/0x3ac ksys_read+0x7c/0xec __arm64_sys_read+0x20/0x30 invoke_syscall+0x60/0x150 el0_svc_common.llvm.1237943816091755067+0xb8/0xf8 do_el0_svc+0x28/0xa0 Fix it by copying the required variables for printing and while at it fix the similar issue at some other places in the same file. Fixes: bd984c03097b ("f2fs: show more DIO information in tracepoint") Signed-off-by: Mukesh Ojha Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- include/trace/events/f2fs.h | 34 +++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-) (limited to 'include') diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index c6b372401c27..ff57e7f9914c 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -322,7 +322,7 @@ TRACE_EVENT(f2fs_unlink_enter, __field(ino_t, ino) __field(loff_t, size) __field(blkcnt_t, blocks) - __field(const char *, name) + __string(name, dentry->d_name.name) ), TP_fast_assign( @@ -330,7 +330,7 @@ TRACE_EVENT(f2fs_unlink_enter, __entry->ino = dir->i_ino; __entry->size = dir->i_size; __entry->blocks = dir->i_blocks; - __entry->name = dentry->d_name.name; + __assign_str(name, dentry->d_name.name); ), TP_printk("dev = (%d,%d), dir ino = %lu, i_size = %lld, " @@ -338,7 +338,7 @@ TRACE_EVENT(f2fs_unlink_enter, show_dev_ino(__entry), __entry->size, (unsigned long long)__entry->blocks, - __entry->name) + __get_str(name)) ); DEFINE_EVENT(f2fs__inode_exit, f2fs_unlink_exit, @@ -940,25 +940,29 @@ TRACE_EVENT(f2fs_direct_IO_enter, TP_STRUCT__entry( __field(dev_t, dev) __field(ino_t, ino) - __field(struct kiocb *, iocb) + __field(loff_t, ki_pos) + __field(int, ki_flags) + __field(u16, ki_ioprio) __field(unsigned long, len) __field(int, rw) ), TP_fast_assign( - __entry->dev = inode->i_sb->s_dev; - __entry->ino = inode->i_ino; - __entry->iocb = iocb; - __entry->len = len; - __entry->rw = rw; + __entry->dev = inode->i_sb->s_dev; + __entry->ino = inode->i_ino; + __entry->ki_pos = iocb->ki_pos; + __entry->ki_flags = iocb->ki_flags; + __entry->ki_ioprio = iocb->ki_ioprio; + __entry->len = len; + __entry->rw = rw; ), TP_printk("dev = (%d,%d), ino = %lu pos = %lld len = %lu ki_flags = %x ki_ioprio = %x rw = %d", show_dev_ino(__entry), - __entry->iocb->ki_pos, + __entry->ki_pos, __entry->len, - __entry->iocb->ki_flags, - __entry->iocb->ki_ioprio, + __entry->ki_flags, + __entry->ki_ioprio, __entry->rw) ); @@ -1407,19 +1411,19 @@ TRACE_EVENT(f2fs_write_checkpoint, TP_STRUCT__entry( __field(dev_t, dev) __field(int, reason) - __field(char *, msg) + __string(dest_msg, msg) ), TP_fast_assign( __entry->dev = sb->s_dev; __entry->reason = reason; - __entry->msg = msg; + __assign_str(dest_msg, msg); ), TP_printk("dev = (%d,%d), checkpoint for %s, state = %s", show_dev(__entry->dev), show_cpreason(__entry->reason), - __entry->msg) + __get_str(dest_msg)) ); DECLARE_EVENT_CLASS(f2fs_discard, -- cgit v1.2.3 From 195623f2d8e9361eaddec071ad298998ec0590ba Mon Sep 17 00:00:00 2001 From: Mukesh Ojha Date: Thu, 27 Oct 2022 14:42:41 +0530 Subject: f2fs: fix the msg data type Data type of msg in f2fs_write_checkpoint trace should be const char * instead of char *. Signed-off-by: Mukesh Ojha Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- include/trace/events/f2fs.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include') diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index ff57e7f9914c..7fbfce498472 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -1404,7 +1404,7 @@ TRACE_EVENT(f2fs_readpages, TRACE_EVENT(f2fs_write_checkpoint, - TP_PROTO(struct super_block *sb, int reason, char *msg), + TP_PROTO(struct super_block *sb, int reason, const char *msg), TP_ARGS(sb, reason, msg), -- cgit v1.2.3 From 41e8f85a75fc60e1543e4903428a1b481b672a17 Mon Sep 17 00:00:00 2001 From: Daeho Jeong Date: Fri, 11 Nov 2022 09:04:06 -0800 Subject: f2fs: introduce F2FS_IOC_START_ATOMIC_REPLACE introduce a new ioctl to replace the whole content of a file atomically, which means it induces truncate and content update at the same time. We can start it with F2FS_IOC_START_ATOMIC_REPLACE and complete it with F2FS_IOC_COMMIT_ATOMIC_WRITE. Or abort it with F2FS_IOC_ABORT_ATOMIC_WRITE. Signed-off-by: Daeho Jeong Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- include/uapi/linux/f2fs.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/uapi/linux/f2fs.h b/include/uapi/linux/f2fs.h index 3121d127d5aa..955d440be104 100644 --- a/include/uapi/linux/f2fs.h +++ b/include/uapi/linux/f2fs.h @@ -42,6 +42,7 @@ struct f2fs_comp_option) #define F2FS_IOC_DECOMPRESS_FILE _IO(F2FS_IOCTL_MAGIC, 23) #define F2FS_IOC_COMPRESS_FILE _IO(F2FS_IOCTL_MAGIC, 24) +#define F2FS_IOC_START_ATOMIC_REPLACE _IO(F2FS_IOCTL_MAGIC, 25) /* * should be same as XFS_IOC_GOINGDOWN. -- cgit v1.2.3 From e7547daccd6a37522f0af74ec4b5a3036f3dd328 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 30 Nov 2022 09:26:29 -0800 Subject: f2fs: refactor extent_cache to support for read and more This patch prepares extent_cache to be ready for addition. Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- include/trace/events/f2fs.h | 62 ++++++++++++++++++++++++++------------------- 1 file changed, 36 insertions(+), 26 deletions(-) (limited to 'include') diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index 7fbfce498472..2bb37892d2ba 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -48,6 +48,7 @@ TRACE_DEFINE_ENUM(CP_DISCARD); TRACE_DEFINE_ENUM(CP_TRIMMED); TRACE_DEFINE_ENUM(CP_PAUSE); TRACE_DEFINE_ENUM(CP_RESIZE); +TRACE_DEFINE_ENUM(EX_READ); #define show_block_type(type) \ __print_symbolic(type, \ @@ -1522,28 +1523,31 @@ TRACE_EVENT(f2fs_issue_flush, TRACE_EVENT(f2fs_lookup_extent_tree_start, - TP_PROTO(struct inode *inode, unsigned int pgofs), + TP_PROTO(struct inode *inode, unsigned int pgofs, enum extent_type type), - TP_ARGS(inode, pgofs), + TP_ARGS(inode, pgofs, type), TP_STRUCT__entry( __field(dev_t, dev) __field(ino_t, ino) __field(unsigned int, pgofs) + __field(enum extent_type, type) ), TP_fast_assign( __entry->dev = inode->i_sb->s_dev; __entry->ino = inode->i_ino; __entry->pgofs = pgofs; + __entry->type = type; ), - TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u", + TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, type = %s", show_dev_ino(__entry), - __entry->pgofs) + __entry->pgofs, + __entry->type == EX_READ ? "Read" : "N/A") ); -TRACE_EVENT_CONDITION(f2fs_lookup_extent_tree_end, +TRACE_EVENT_CONDITION(f2fs_lookup_read_extent_tree_end, TP_PROTO(struct inode *inode, unsigned int pgofs, struct extent_info *ei), @@ -1557,8 +1561,8 @@ TRACE_EVENT_CONDITION(f2fs_lookup_extent_tree_end, __field(ino_t, ino) __field(unsigned int, pgofs) __field(unsigned int, fofs) - __field(u32, blk) __field(unsigned int, len) + __field(u32, blk) ), TP_fast_assign( @@ -1566,26 +1570,26 @@ TRACE_EVENT_CONDITION(f2fs_lookup_extent_tree_end, __entry->ino = inode->i_ino; __entry->pgofs = pgofs; __entry->fofs = ei->fofs; - __entry->blk = ei->blk; __entry->len = ei->len; + __entry->blk = ei->blk; ), TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, " - "ext_info(fofs: %u, blk: %u, len: %u)", + "read_ext_info(fofs: %u, len: %u, blk: %u)", show_dev_ino(__entry), __entry->pgofs, __entry->fofs, - __entry->blk, - __entry->len) + __entry->len, + __entry->blk) ); -TRACE_EVENT(f2fs_update_extent_tree_range, +TRACE_EVENT(f2fs_update_read_extent_tree_range, - TP_PROTO(struct inode *inode, unsigned int pgofs, block_t blkaddr, - unsigned int len, + TP_PROTO(struct inode *inode, unsigned int pgofs, unsigned int len, + block_t blkaddr, unsigned int c_len), - TP_ARGS(inode, pgofs, blkaddr, len, c_len), + TP_ARGS(inode, pgofs, len, blkaddr, c_len), TP_STRUCT__entry( __field(dev_t, dev) @@ -1600,67 +1604,73 @@ TRACE_EVENT(f2fs_update_extent_tree_range, __entry->dev = inode->i_sb->s_dev; __entry->ino = inode->i_ino; __entry->pgofs = pgofs; - __entry->blk = blkaddr; __entry->len = len; + __entry->blk = blkaddr; __entry->c_len = c_len; ), TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, " - "blkaddr = %u, len = %u, " - "c_len = %u", + "len = %u, blkaddr = %u, c_len = %u", show_dev_ino(__entry), __entry->pgofs, - __entry->blk, __entry->len, + __entry->blk, __entry->c_len) ); TRACE_EVENT(f2fs_shrink_extent_tree, TP_PROTO(struct f2fs_sb_info *sbi, unsigned int node_cnt, - unsigned int tree_cnt), + unsigned int tree_cnt, enum extent_type type), - TP_ARGS(sbi, node_cnt, tree_cnt), + TP_ARGS(sbi, node_cnt, tree_cnt, type), TP_STRUCT__entry( __field(dev_t, dev) __field(unsigned int, node_cnt) __field(unsigned int, tree_cnt) + __field(enum extent_type, type) ), TP_fast_assign( __entry->dev = sbi->sb->s_dev; __entry->node_cnt = node_cnt; __entry->tree_cnt = tree_cnt; + __entry->type = type; ), - TP_printk("dev = (%d,%d), shrunk: node_cnt = %u, tree_cnt = %u", + TP_printk("dev = (%d,%d), shrunk: node_cnt = %u, tree_cnt = %u, type = %s", show_dev(__entry->dev), __entry->node_cnt, - __entry->tree_cnt) + __entry->tree_cnt, + __entry->type == EX_READ ? "Read" : "N/A") ); TRACE_EVENT(f2fs_destroy_extent_tree, - TP_PROTO(struct inode *inode, unsigned int node_cnt), + TP_PROTO(struct inode *inode, unsigned int node_cnt, + enum extent_type type), - TP_ARGS(inode, node_cnt), + TP_ARGS(inode, node_cnt, type), TP_STRUCT__entry( __field(dev_t, dev) __field(ino_t, ino) __field(unsigned int, node_cnt) + __field(enum extent_type, type) ), TP_fast_assign( __entry->dev = inode->i_sb->s_dev; __entry->ino = inode->i_ino; __entry->node_cnt = node_cnt; + __entry->type = type; ), - TP_printk("dev = (%d,%d), ino = %lu, destroyed: node_cnt = %u", + TP_printk("dev = (%d,%d), ino = %lu, destroyed: node_cnt = %u, type = %s", show_dev_ino(__entry), - __entry->node_cnt) + __entry->node_cnt, + __entry->type == EX_READ ? "Read" : "N/A") ); DECLARE_EVENT_CLASS(f2fs_sync_dirty_inodes, -- cgit v1.2.3 From 71644dff481180ba024ac4f5cb1f068756357adf Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Thu, 1 Dec 2022 17:37:15 -0800 Subject: f2fs: add block_age-based extent cache This patch introduces a runtime hot/cold data separation method for f2fs, in order to improve the accuracy for data temperature classification, reduce the garbage collection overhead after long-term data updates. Enhanced hot/cold data separation can record data block update frequency as "age" of the extent per inode, and take use of the age info to indicate better temperature type for data block allocation: - It records total data blocks allocated since mount; - When file extent has been updated, it calculate the count of data blocks allocated since last update as the age of the extent; - Before the data block allocated, it searches for the age info and chooses the suitable segment for allocation. Test and result: - Prepare: create about 30000 files * 3% for cold files (with cold file extension like .apk, from 3M to 10M) * 50% for warm files (with random file extension like .FcDxq, from 1K to 4M) * 47% for hot files (with hot file extension like .db, from 1K to 256K) - create(5%)/random update(90%)/delete(5%) the files * total write amount is about 70G * fsync will be called for .db files, and buffered write will be used for other files The storage of test device is large enough(128G) so that it will not switch to SSR mode during the test. Benefit: dirty segment count increment reduce about 14% - before: Dirty +21110 - after: Dirty +18286 Signed-off-by: qixiaoyu1 Signed-off-by: xiongping1 Signed-off-by: Jaegeuk Kim --- include/trace/events/f2fs.h | 86 +++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 83 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index 2bb37892d2ba..31d994e6b4ca 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -49,6 +49,7 @@ TRACE_DEFINE_ENUM(CP_TRIMMED); TRACE_DEFINE_ENUM(CP_PAUSE); TRACE_DEFINE_ENUM(CP_RESIZE); TRACE_DEFINE_ENUM(EX_READ); +TRACE_DEFINE_ENUM(EX_BLOCK_AGE); #define show_block_type(type) \ __print_symbolic(type, \ @@ -155,6 +156,11 @@ TRACE_DEFINE_ENUM(EX_READ); { COMPRESS_ZSTD, "ZSTD" }, \ { COMPRESS_LZORLE, "LZO-RLE" }) +#define show_extent_type(type) \ + __print_symbolic(type, \ + { EX_READ, "Read" }, \ + { EX_BLOCK_AGE, "Block Age" }) + struct f2fs_sb_info; struct f2fs_io_info; struct extent_info; @@ -1544,7 +1550,7 @@ TRACE_EVENT(f2fs_lookup_extent_tree_start, TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, type = %s", show_dev_ino(__entry), __entry->pgofs, - __entry->type == EX_READ ? "Read" : "N/A") + show_extent_type(__entry->type)) ); TRACE_EVENT_CONDITION(f2fs_lookup_read_extent_tree_end, @@ -1583,6 +1589,45 @@ TRACE_EVENT_CONDITION(f2fs_lookup_read_extent_tree_end, __entry->blk) ); +TRACE_EVENT_CONDITION(f2fs_lookup_age_extent_tree_end, + + TP_PROTO(struct inode *inode, unsigned int pgofs, + struct extent_info *ei), + + TP_ARGS(inode, pgofs, ei), + + TP_CONDITION(ei), + + TP_STRUCT__entry( + __field(dev_t, dev) + __field(ino_t, ino) + __field(unsigned int, pgofs) + __field(unsigned int, fofs) + __field(unsigned int, len) + __field(unsigned long long, age) + __field(unsigned long long, blocks) + ), + + TP_fast_assign( + __entry->dev = inode->i_sb->s_dev; + __entry->ino = inode->i_ino; + __entry->pgofs = pgofs; + __entry->fofs = ei->fofs; + __entry->len = ei->len; + __entry->age = ei->age; + __entry->blocks = ei->last_blocks; + ), + + TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, " + "age_ext_info(fofs: %u, len: %u, age: %llu, blocks: %llu)", + show_dev_ino(__entry), + __entry->pgofs, + __entry->fofs, + __entry->len, + __entry->age, + __entry->blocks) +); + TRACE_EVENT(f2fs_update_read_extent_tree_range, TP_PROTO(struct inode *inode, unsigned int pgofs, unsigned int len, @@ -1618,6 +1663,41 @@ TRACE_EVENT(f2fs_update_read_extent_tree_range, __entry->c_len) ); +TRACE_EVENT(f2fs_update_age_extent_tree_range, + + TP_PROTO(struct inode *inode, unsigned int pgofs, unsigned int len, + unsigned long long age, + unsigned long long last_blks), + + TP_ARGS(inode, pgofs, len, age, last_blks), + + TP_STRUCT__entry( + __field(dev_t, dev) + __field(ino_t, ino) + __field(unsigned int, pgofs) + __field(unsigned int, len) + __field(unsigned long long, age) + __field(unsigned long long, blocks) + ), + + TP_fast_assign( + __entry->dev = inode->i_sb->s_dev; + __entry->ino = inode->i_ino; + __entry->pgofs = pgofs; + __entry->len = len; + __entry->age = age; + __entry->blocks = last_blks; + ), + + TP_printk("dev = (%d,%d), ino = %lu, pgofs = %u, " + "len = %u, age = %llu, blocks = %llu", + show_dev_ino(__entry), + __entry->pgofs, + __entry->len, + __entry->age, + __entry->blocks) +); + TRACE_EVENT(f2fs_shrink_extent_tree, TP_PROTO(struct f2fs_sb_info *sbi, unsigned int node_cnt, @@ -1643,7 +1723,7 @@ TRACE_EVENT(f2fs_shrink_extent_tree, show_dev(__entry->dev), __entry->node_cnt, __entry->tree_cnt, - __entry->type == EX_READ ? "Read" : "N/A") + show_extent_type(__entry->type)) ); TRACE_EVENT(f2fs_destroy_extent_tree, @@ -1670,7 +1750,7 @@ TRACE_EVENT(f2fs_destroy_extent_tree, TP_printk("dev = (%d,%d), ino = %lu, destroyed: node_cnt = %u, type = %s", show_dev_ino(__entry), __entry->node_cnt, - __entry->type == EX_READ ? "Read" : "N/A") + show_extent_type(__entry->type)) ); DECLARE_EVENT_CLASS(f2fs_sync_dirty_inodes, -- cgit v1.2.3