On Aug 8, 2019, at 9:45 PM, Harshad Shirwadkar <harshadshirwadkar@xxxxxxxxx> wrote: > > Ext4's fast commit feature tracks changed files and maintains them in > a queue. We also remember for each file the logical block range that > needs to be committed. This patch adds these fields to ext4_inode_info > and ext4_sb_info and also adds initialization calls. > > Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@xxxxxxxxx> > > --- > > Changelog: > > V2: Converted s_fc_lock from mutex to spinlock to improve parallelism > performance. > --- > fs/ext4/ext4.h | 34 ++++++++++++++++++++++++++++++++++ > fs/ext4/ext4_jbd2.c | 13 +++++++++++++ > fs/ext4/ext4_jbd2.h | 2 ++ > fs/ext4/inode.c | 1 + > fs/ext4/super.c | 7 +++++++ > 5 files changed, 57 insertions(+) > > diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h > index becbda38b7db..0d15d4539dda 100644 > --- a/fs/ext4/ext4.h > +++ b/fs/ext4/ext4.h > @@ -921,6 +921,27 @@ enum { > I_DATA_SEM_QUOTA, > }; > > +/* > + * Ext4 fast commit inode specific information > + */ > +struct ext4_fast_commit_inode_info { > + /* TID / SUB-TID when old_i_size and i_size were recorded */ > + tid_t fc_tid; > + tid_t fc_subtid; > + > + /* > + * Start of logical block range that needs to be committed in this fast > + * commit > + */ > + loff_t fc_lblk_start; > + > + /* > + * End of logical block range that needs to be committed in this fast > + * commit > + */ > + loff_t fc_lblk_end; Since these are logical block numbers within the journal, they certainly don't need to be 64-bit values. loff_t is for byte offsets, this should use ext4_lblk_t, which will also reduce the size of the struct by 8 bytes. > +}; > + > > /* > * fourth extended file system inode data in memory > @@ -955,6 +976,9 @@ struct ext4_inode_info { > > struct list_head i_orphan; /* unlinked but open inodes */ > > + struct list_head i_fc_list; /* inodes that need fast commit */ This comment should document what lock is protecting this list, along with the other fields. > + struct ext4_fast_commit_inode_info i_fc; Since this increases the size of the inode, does it affect the number of inodes that can fit into one page of ext4_inode_cachep? > /* > * i_disksize keeps track of what the inode size is ON DISK, not > * in memory. During truncate, i_size is set to the new size by > @@ -1529,6 +1553,16 @@ struct ext4_sb_info { > /* Barrier between changing inodes' journal flags and writepages ops. */ > struct percpu_rw_semaphore s_journal_flag_rwsem; > struct dax_device *s_daxdev; > + > + /* Ext4 fast commit stuff */ > + bool fc_replay; /* Fast commit replay in progress */ > + struct list_head s_fc_q; /* Inodes that need fast commit. */ This comment should document what lock is protecting this list, along with the other fields. > + __u32 s_fc_q_cnt; /* Number of inodes in the fc queue */ > + bool s_fc_eligible; /* > + * Are changes after the last commit > + * eligible for fast commit? > + */ It is slightly more space efficient to put the bool values together rather than interleaving them between 64-bit values. > + spinlock_t s_fc_lock; > }; > > static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) > diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c > index 7c70b08d104c..75b6db808837 100644 > --- a/fs/ext4/ext4_jbd2.c > +++ b/fs/ext4/ext4_jbd2.c > @@ -330,3 +330,16 @@ int __ext4_handle_dirty_super(const char *where, unsigned int line, > mark_buffer_dirty(bh); > return err; > } > + > +void ext4_init_inode_fc_info(struct inode *inode) > +{ > + handle_t *handle = ext4_journal_current_handle(); > + struct ext4_inode_info *ei = EXT4_I(inode); > + > + memset(&ei->i_fc, 0, sizeof(ei->i_fc)); > + if (ext4_handle_valid(handle)) { > + ei->i_fc.fc_tid = handle->h_transaction->t_tid; > + ei->i_fc.fc_subtid = handle->h_transaction->t_journal->j_subtid; > + } > + INIT_LIST_HEAD(&ei->i_fc_list); > +} > diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h > index ef8fcf7d0d3b..2305c1acd415 100644 > --- a/fs/ext4/ext4_jbd2.h > +++ b/fs/ext4/ext4_jbd2.h > @@ -459,4 +459,6 @@ static inline int ext4_should_dioread_nolock(struct inode *inode) > return 1; > } > > +void ext4_init_inode_fc_info(struct inode *inode); > + > #endif /* _EXT4_JBD2_H */ > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index 420fe3deed39..f230a888eddd 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -4996,6 +4996,7 @@ struct inode *__ext4_iget(struct super_block *sb, unsigned long ino, > for (block = 0; block < EXT4_N_BLOCKS; block++) > ei->i_data[block] = raw_inode->i_block[block]; > INIT_LIST_HEAD(&ei->i_orphan); > + ext4_init_inode_fc_info(&ei->vfs_inode); > > /* > * Set transaction id's of transactions that have to be committed > diff --git a/fs/ext4/super.c b/fs/ext4/super.c > index 6bab59ae81f7..0b833e9b61c1 100644 > --- a/fs/ext4/super.c > +++ b/fs/ext4/super.c > @@ -1100,6 +1100,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb) > ei->i_datasync_tid = 0; > atomic_set(&ei->i_unwritten, 0); > INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); > + ext4_init_inode_fc_info(&ei->vfs_inode); > return &ei->vfs_inode; > } > > @@ -1139,6 +1140,7 @@ static void init_once(void *foo) > init_rwsem(&ei->i_data_sem); > init_rwsem(&ei->i_mmap_sem); > inode_init_once(&ei->vfs_inode); > + ext4_init_inode_fc_info(&ei->vfs_inode); > } > > static int __init init_inodecache(void) > @@ -4301,6 +4303,11 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent) > INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ > mutex_init(&sbi->s_orphan_lock); > > + INIT_LIST_HEAD(&sbi->s_fc_q); > + sbi->s_fc_q_cnt = 0; > + sbi->s_fc_eligible = true; > + spin_lock_init(&sbi->s_fc_lock); > + > sb->s_root = NULL; > > needs_recovery = (es->s_last_orphan != 0 || > -- > 2.23.0.rc1.153.gdeed80330f-goog > Cheers, Andreas
Attachment:
signature.asc
Description: Message signed with OpenPGP