| Message ID | 20230811061905.301124-2-dongyangli@ddn.com |
|---|---|
| State | New |
| Headers | show |
| Series | [1/2] ext4: introduce EXT4_BG_TRIMMED to optimize fstrim | expand |
On Aug 11, 2023, at 12:19 AM, Li Dongyang <dongyangli@ddn.com> wrote: > > This adds EXT2_FLAG_BG_TRIMMED, which is used on block group > descriptors during mke2fs after discard is done. > The EXT2_FLAG_BG_TRIMMED flag is cleared on the block group when > we free blocks. > > Introduce EXT2_FLAGS_TRACK_TRIM, which is a new super block flag, > to indicate whether we should honour the EXT2_FLAG_BG_TRIMMED > set on each block group. > EXT2_FLAGS_TRACK_TRIM itself can be turned on/off via tune2fs. > > Make dumpe2fs aware of the new flags. Ted, I think this patch is uncontroversial, and could be landed as-is. Reviewed-by: Andreas Dilger <adilger@dilger.ca> > > Cc: Shuichi Ihara <sihara@ddn.com> > Cc: Andreas Dilger <adilger@dilger.ca> > Cc: Wang Shilong <wangshilong1991@gmail.com> > Signed-off-by: Wang Shilong <wshilong@ddn.com> > Signed-off-by: Li Dongyang <dongyangli@ddn.com> > --- > lib/e2p/ls.c | 4 ++++ > lib/ext2fs/alloc_stats.c | 8 ++++++-- > lib/ext2fs/ext2_fs.h | 2 ++ > misc/dumpe2fs.c | 2 ++ > misc/mke2fs.c | 9 +++++++++ > misc/tune2fs.8.in | 8 ++++++++ > misc/tune2fs.c | 10 ++++++++++ > 7 files changed, 41 insertions(+), 2 deletions(-) > > diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c > index 0b74aea2b..4b356eca6 100644 > --- a/lib/e2p/ls.c > +++ b/lib/e2p/ls.c > @@ -162,6 +162,10 @@ static void print_super_flags(struct ext2_super_block * s, FILE *f) > fputs("test_filesystem ", f); > flags_found++; > } > + if (s->s_flags & EXT2_FLAGS_TRACK_TRIM) { > + fputs("track_trim ", f); > + flags_found++; > + } > if (flags_found) > fputs("\n", f); > else > diff --git a/lib/ext2fs/alloc_stats.c b/lib/ext2fs/alloc_stats.c > index 6f98bcc7c..4e03f92a4 100644 > --- a/lib/ext2fs/alloc_stats.c > +++ b/lib/ext2fs/alloc_stats.c > @@ -70,10 +70,12 @@ void ext2fs_block_alloc_stats2(ext2_filsys fs, blk64_t blk, int inuse) > #endif > return; > } > - if (inuse > 0) > + if (inuse > 0) { > ext2fs_mark_block_bitmap2(fs->block_map, blk); > - else > + } else { > ext2fs_unmark_block_bitmap2(fs->block_map, blk); > + ext2fs_bg_flags_clear(fs, group, EXT2_BG_TRIMMED); > + } > ext2fs_bg_free_blocks_count_set(fs, group, ext2fs_bg_free_blocks_count(fs, group) - inuse); > ext2fs_bg_flags_clear(fs, group, EXT2_BG_BLOCK_UNINIT); > ext2fs_group_desc_csum_set(fs, group); > @@ -139,6 +141,8 @@ void ext2fs_block_alloc_stats_range(ext2_filsys fs, blk64_t blk, > ext2fs_bg_free_blocks_count(fs, group) - > inuse*n/EXT2FS_CLUSTER_RATIO(fs)); > ext2fs_bg_flags_clear(fs, group, EXT2_BG_BLOCK_UNINIT); > + if (inuse < 0) > + ext2fs_bg_flags_clear(fs, group, EXT2_BG_TRIMMED); > ext2fs_group_desc_csum_set(fs, group); > ext2fs_free_blocks_count_add(fs->super, -inuse * (blk64_t) n); > blk += n; > diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h > index 0fc9c09a5..88e1114c9 100644 > --- a/lib/ext2fs/ext2_fs.h > +++ b/lib/ext2fs/ext2_fs.h > @@ -223,6 +223,7 @@ struct ext4_group_desc > #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */ > #define EXT2_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not initialized */ > #define EXT2_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */ > +#define EXT2_BG_TRIMMED 0x0008 /* Block group was trimmed */ > > /* > * Data structures used by the directory indexing feature > @@ -563,6 +564,7 @@ struct ext2_inode *EXT2_INODE(struct ext2_inode_large *large_inode) > #define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */ > #define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */ > #define EXT2_FLAGS_TEST_FILESYS 0x0004 /* OK for use on development code */ > +#define EXT2_FLAGS_TRACK_TRIM 0x0008 /* Track trim status in each bg */ > #define EXT2_FLAGS_IS_SNAPSHOT 0x0010 /* This is a snapshot image */ > #define EXT2_FLAGS_FIX_SNAPSHOT 0x0020 /* Snapshot inodes corrupted */ > #define EXT2_FLAGS_FIX_EXCLUDE 0x0040 /* Exclude bitmaps corrupted */ > diff --git a/misc/dumpe2fs.c b/misc/dumpe2fs.c > index 7c080ed9f..afe569dff 100644 > --- a/misc/dumpe2fs.c > +++ b/misc/dumpe2fs.c > @@ -131,6 +131,8 @@ static void print_bg_opts(ext2_filsys fs, dgrp_t i) > &first); > print_bg_opt(bg_flags, EXT2_BG_INODE_ZEROED, "ITABLE_ZEROED", > &first); > + print_bg_opt(bg_flags, EXT2_BG_TRIMMED, "TRIMMED", > + &first); > if (!first) > fputc(']', stdout); > fputc('\n', stdout); > diff --git a/misc/mke2fs.c b/misc/mke2fs.c > index 4a9c1b092..bbfcde478 100644 > --- a/misc/mke2fs.c > +++ b/misc/mke2fs.c > @@ -3154,6 +3154,15 @@ int main (int argc, char *argv[]) > /* Can't undo discard ... */ > if (!noaction && discard && dev_size && (io_ptr != undo_io_manager)) { > retval = mke2fs_discard_device(fs); > + if (!retval) { > + dgrp_t i; > + > + fs->super->s_flags |= EXT2_FLAGS_TRACK_TRIM; > + for (i = 0; i < fs->group_desc_count; i++) { > + ext2fs_bg_flags_set(fs, i, EXT2_BG_TRIMMED); > + ext2fs_group_desc_csum_set(fs, i); > + } > + } > if (!retval && io_channel_discard_zeroes_data(fs->io)) { > if (verbose) > printf("%s", > diff --git a/misc/tune2fs.8.in b/misc/tune2fs.8.in > index dcf108c1f..2eb7e88ed 100644 > --- a/misc/tune2fs.8.in > +++ b/misc/tune2fs.8.in > @@ -273,6 +273,14 @@ mounted using experimental kernel code, such as the ext4dev file system. > .B ^test_fs > Clear the test_fs flag, indicating the file system should only be mounted > using production-level file system code. > +.TP > +.B track_trim > +Set a flag in the file system superblock to make fstrim save the trim status > +in each block group and skip the block groups already been trimmed. > +.TP > +.B ^track_trim > +Clear the track_trim flag to make fstrim ignore the trim status saved in > +each block group, and trim every block group. > .RE > .TP > .B \-f > diff --git a/misc/tune2fs.c b/misc/tune2fs.c > index 458f7cf6a..dd9e8eab0 100644 > --- a/misc/tune2fs.c > +++ b/misc/tune2fs.c > @@ -2312,6 +2312,14 @@ static int parse_extended_opts(ext2_filsys fs, const char *opts) > sb->s_flags &= ~EXT2_FLAGS_TEST_FILESYS; > printf("Clearing test filesystem flag\n"); > ext2fs_mark_super_dirty(fs); > + } else if (!strcmp(token, "track_trim")) { > + sb->s_flags |= EXT2_FLAGS_TRACK_TRIM; > + printf("Setting track_trim flag\n"); > + ext2fs_mark_super_dirty(fs); > + } else if (!strcmp(token, "^track_trim")) { > + sb->s_flags &= ~EXT2_FLAGS_TRACK_TRIM; > + printf("Clearing track_trim flag\n"); > + ext2fs_mark_super_dirty(fs); > } else if (strcmp(token, "stride") == 0) { > if (!arg) { > r_usage++; > @@ -2458,6 +2466,8 @@ static int parse_extended_opts(ext2_filsys fs, const char *opts) > "\tforce_fsck\n" > "\ttest_fs\n" > "\t^test_fs\n" > + "\ttrack_trim\n" > + "\t^track_trim\n" > "\tencoding=<encoding>\n" > "\tencoding_flags=<flags>\n")); > free(buf); > -- > 2.41.0 > Cheers, Andreas
diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c index 0b74aea2b..4b356eca6 100644 --- a/lib/e2p/ls.c +++ b/lib/e2p/ls.c @@ -162,6 +162,10 @@ static void print_super_flags(struct ext2_super_block * s, FILE *f) fputs("test_filesystem ", f); flags_found++; } + if (s->s_flags & EXT2_FLAGS_TRACK_TRIM) { + fputs("track_trim ", f); + flags_found++; + } if (flags_found) fputs("\n", f); else diff --git a/lib/ext2fs/alloc_stats.c b/lib/ext2fs/alloc_stats.c index 6f98bcc7c..4e03f92a4 100644 --- a/lib/ext2fs/alloc_stats.c +++ b/lib/ext2fs/alloc_stats.c @@ -70,10 +70,12 @@ void ext2fs_block_alloc_stats2(ext2_filsys fs, blk64_t blk, int inuse) #endif return; } - if (inuse > 0) + if (inuse > 0) { ext2fs_mark_block_bitmap2(fs->block_map, blk); - else + } else { ext2fs_unmark_block_bitmap2(fs->block_map, blk); + ext2fs_bg_flags_clear(fs, group, EXT2_BG_TRIMMED); + } ext2fs_bg_free_blocks_count_set(fs, group, ext2fs_bg_free_blocks_count(fs, group) - inuse); ext2fs_bg_flags_clear(fs, group, EXT2_BG_BLOCK_UNINIT); ext2fs_group_desc_csum_set(fs, group); @@ -139,6 +141,8 @@ void ext2fs_block_alloc_stats_range(ext2_filsys fs, blk64_t blk, ext2fs_bg_free_blocks_count(fs, group) - inuse*n/EXT2FS_CLUSTER_RATIO(fs)); ext2fs_bg_flags_clear(fs, group, EXT2_BG_BLOCK_UNINIT); + if (inuse < 0) + ext2fs_bg_flags_clear(fs, group, EXT2_BG_TRIMMED); ext2fs_group_desc_csum_set(fs, group); ext2fs_free_blocks_count_add(fs->super, -inuse * (blk64_t) n); blk += n; diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h index 0fc9c09a5..88e1114c9 100644 --- a/lib/ext2fs/ext2_fs.h +++ b/lib/ext2fs/ext2_fs.h @@ -223,6 +223,7 @@ struct ext4_group_desc #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */ #define EXT2_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not initialized */ #define EXT2_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */ +#define EXT2_BG_TRIMMED 0x0008 /* Block group was trimmed */ /* * Data structures used by the directory indexing feature @@ -563,6 +564,7 @@ struct ext2_inode *EXT2_INODE(struct ext2_inode_large *large_inode) #define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */ #define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */ #define EXT2_FLAGS_TEST_FILESYS 0x0004 /* OK for use on development code */ +#define EXT2_FLAGS_TRACK_TRIM 0x0008 /* Track trim status in each bg */ #define EXT2_FLAGS_IS_SNAPSHOT 0x0010 /* This is a snapshot image */ #define EXT2_FLAGS_FIX_SNAPSHOT 0x0020 /* Snapshot inodes corrupted */ #define EXT2_FLAGS_FIX_EXCLUDE 0x0040 /* Exclude bitmaps corrupted */ diff --git a/misc/dumpe2fs.c b/misc/dumpe2fs.c index 7c080ed9f..afe569dff 100644 --- a/misc/dumpe2fs.c +++ b/misc/dumpe2fs.c @@ -131,6 +131,8 @@ static void print_bg_opts(ext2_filsys fs, dgrp_t i) &first); print_bg_opt(bg_flags, EXT2_BG_INODE_ZEROED, "ITABLE_ZEROED", &first); + print_bg_opt(bg_flags, EXT2_BG_TRIMMED, "TRIMMED", + &first); if (!first) fputc(']', stdout); fputc('\n', stdout); diff --git a/misc/mke2fs.c b/misc/mke2fs.c index 4a9c1b092..bbfcde478 100644 --- a/misc/mke2fs.c +++ b/misc/mke2fs.c @@ -3154,6 +3154,15 @@ int main (int argc, char *argv[]) /* Can't undo discard ... */ if (!noaction && discard && dev_size && (io_ptr != undo_io_manager)) { retval = mke2fs_discard_device(fs); + if (!retval) { + dgrp_t i; + + fs->super->s_flags |= EXT2_FLAGS_TRACK_TRIM; + for (i = 0; i < fs->group_desc_count; i++) { + ext2fs_bg_flags_set(fs, i, EXT2_BG_TRIMMED); + ext2fs_group_desc_csum_set(fs, i); + } + } if (!retval && io_channel_discard_zeroes_data(fs->io)) { if (verbose) printf("%s", diff --git a/misc/tune2fs.8.in b/misc/tune2fs.8.in index dcf108c1f..2eb7e88ed 100644 --- a/misc/tune2fs.8.in +++ b/misc/tune2fs.8.in @@ -273,6 +273,14 @@ mounted using experimental kernel code, such as the ext4dev file system. .B ^test_fs Clear the test_fs flag, indicating the file system should only be mounted using production-level file system code. +.TP +.B track_trim +Set a flag in the file system superblock to make fstrim save the trim status +in each block group and skip the block groups already been trimmed. +.TP +.B ^track_trim +Clear the track_trim flag to make fstrim ignore the trim status saved in +each block group, and trim every block group. .RE .TP .B \-f diff --git a/misc/tune2fs.c b/misc/tune2fs.c index 458f7cf6a..dd9e8eab0 100644 --- a/misc/tune2fs.c +++ b/misc/tune2fs.c @@ -2312,6 +2312,14 @@ static int parse_extended_opts(ext2_filsys fs, const char *opts) sb->s_flags &= ~EXT2_FLAGS_TEST_FILESYS; printf("Clearing test filesystem flag\n"); ext2fs_mark_super_dirty(fs); + } else if (!strcmp(token, "track_trim")) { + sb->s_flags |= EXT2_FLAGS_TRACK_TRIM; + printf("Setting track_trim flag\n"); + ext2fs_mark_super_dirty(fs); + } else if (!strcmp(token, "^track_trim")) { + sb->s_flags &= ~EXT2_FLAGS_TRACK_TRIM; + printf("Clearing track_trim flag\n"); + ext2fs_mark_super_dirty(fs); } else if (strcmp(token, "stride") == 0) { if (!arg) { r_usage++; @@ -2458,6 +2466,8 @@ static int parse_extended_opts(ext2_filsys fs, const char *opts) "\tforce_fsck\n" "\ttest_fs\n" "\t^test_fs\n" + "\ttrack_trim\n" + "\t^track_trim\n" "\tencoding=<encoding>\n" "\tencoding_flags=<flags>\n")); free(buf);