f2fs: support atomic writes
This patch introduces a very limited functionality for atomic write support.
In order to support atomic write, this patch adds two ioctls:
o F2FS_IOC_START_ATOMIC_WRITE
o F2FS_IOC_COMMIT_ATOMIC_WRITE
The database engine should be aware of the following sequence.
1. open
-> ioctl(F2FS_IOC_START_ATOMIC_WRITE);
2. writes
: all the written data will be treated as atomic pages.
3. commit
-> ioctl(F2FS_IOC_COMMIT_ATOMIC_WRITE);
: this flushes all the data blocks to the disk, which will be shown all or
nothing by f2fs recovery procedure.
4. repeat to #2.
The IO pattens should be:
,- START_ATOMIC_WRITE ,- COMMIT_ATOMIC_WRITE
CP | D D D D D D | FSYNC | D D D D | FSYNC ...
`- COMMIT_ATOMIC_WRITE
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 4d1c49a..923cb76 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -26,6 +26,7 @@
static struct kmem_cache *discard_entry_slab;
static struct kmem_cache *sit_entry_set_slab;
+static struct kmem_cache *inmem_entry_slab;
/*
* __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since
@@ -173,6 +174,60 @@
return result + __reverse_ffz(tmp);
}
+void register_inmem_page(struct inode *inode, struct page *page)
+{
+ struct f2fs_inode_info *fi = F2FS_I(inode);
+ struct inmem_pages *new;
+
+ new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
+
+ /* add atomic page indices to the list */
+ new->page = page;
+ INIT_LIST_HEAD(&new->list);
+
+ /* increase reference count with clean state */
+ mutex_lock(&fi->inmem_lock);
+ get_page(page);
+ list_add_tail(&new->list, &fi->inmem_pages);
+ mutex_unlock(&fi->inmem_lock);
+}
+
+void commit_inmem_pages(struct inode *inode, bool abort)
+{
+ struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+ struct f2fs_inode_info *fi = F2FS_I(inode);
+ struct inmem_pages *cur, *tmp;
+ bool submit_bio = false;
+ struct f2fs_io_info fio = {
+ .type = DATA,
+ .rw = WRITE_SYNC,
+ };
+
+ f2fs_balance_fs(sbi);
+ f2fs_lock_op(sbi);
+
+ mutex_lock(&fi->inmem_lock);
+ list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
+ lock_page(cur->page);
+ if (!abort && cur->page->mapping == inode->i_mapping) {
+ f2fs_wait_on_page_writeback(cur->page, DATA);
+ if (clear_page_dirty_for_io(cur->page))
+ inode_dec_dirty_pages(inode);
+ do_write_data_page(cur->page, &fio);
+ submit_bio = true;
+ }
+ f2fs_put_page(cur->page, 1);
+ list_del(&cur->list);
+ kmem_cache_free(inmem_entry_slab, cur);
+ }
+ if (submit_bio)
+ f2fs_submit_merged_bio(sbi, DATA, WRITE);
+ mutex_unlock(&fi->inmem_lock);
+
+ filemap_fdatawait_range(inode->i_mapping, 0, LLONG_MAX);
+ f2fs_unlock_op(sbi);
+}
+
/*
* This function balances dirty node and dentry pages.
* In addition, it controls garbage collection.
@@ -2148,8 +2203,15 @@
sizeof(struct nat_entry_set));
if (!sit_entry_set_slab)
goto destory_discard_entry;
+
+ inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
+ sizeof(struct inmem_pages));
+ if (!inmem_entry_slab)
+ goto destroy_sit_entry_set;
return 0;
+destroy_sit_entry_set:
+ kmem_cache_destroy(sit_entry_set_slab);
destory_discard_entry:
kmem_cache_destroy(discard_entry_slab);
fail:
@@ -2160,4 +2222,5 @@
{
kmem_cache_destroy(sit_entry_set_slab);
kmem_cache_destroy(discard_entry_slab);
+ kmem_cache_destroy(inmem_entry_slab);
}