f2fs: add f2fs_map_blocks
authorJaegeuk Kim <jaegeuk@kernel.org>
Tue, 7 Apr 2015 02:55:34 +0000 (19:55 -0700)
committerJaegeuk Kim <jaegeuk@kernel.org>
Thu, 28 May 2015 22:41:29 +0000 (15:41 -0700)
This patch introduces f2fs_map_blocks structure likewise ext4_map_blocks.
Now, f2fs uses f2fs_map_blocks when handling get_block.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
fs/f2fs/data.c
fs/f2fs/f2fs.h
include/trace/events/f2fs.h

index 1e1aae669fa86275d7b6c932aeda89ded129ce17..aa3c079ffd94d6b10a346e6bf2710fadd5d147de 100644 (file)
@@ -251,19 +251,6 @@ int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
        return err;
 }
 
-static void f2fs_map_bh(struct super_block *sb, pgoff_t pgofs,
-                       struct extent_info *ei, struct buffer_head *bh_result)
-{
-       unsigned int blkbits = sb->s_blocksize_bits;
-       size_t max_size = bh_result->b_size;
-       size_t mapped_size;
-
-       clear_buffer_new(bh_result);
-       map_bh(bh_result, sb, ei->blk + pgofs - ei->fofs);
-       mapped_size = (ei->fofs + ei->len - pgofs) << blkbits;
-       bh_result->b_size = min(max_size, mapped_size);
-}
-
 static bool lookup_extent_info(struct inode *inode, pgoff_t pgofs,
                                                        struct extent_info *ei)
 {
@@ -1208,18 +1195,18 @@ out:
 }
 
 /*
- * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh.
+ * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
+ * f2fs_map_blocks structure.
  * If original data blocks are allocated, then give them to blockdev.
  * Otherwise,
  *     a. preallocate requested block addresses
  *     b. do not use extent cache for better performance
  *     c. give the block addresses to blockdev
  */
-static int __get_data_block(struct inode *inode, sector_t iblock,
-                       struct buffer_head *bh_result, int create, bool fiemap)
+static int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
+                       int create, bool fiemap)
 {
-       unsigned int blkbits = inode->i_sb->s_blocksize_bits;
-       unsigned maxblocks = bh_result->b_size >> blkbits;
+       unsigned int maxblocks = map->m_len;
        struct dnode_of_data dn;
        int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA;
        pgoff_t pgofs, end_offset;
@@ -1227,11 +1214,16 @@ static int __get_data_block(struct inode *inode, sector_t iblock,
        struct extent_info ei;
        bool allocated = false;
 
-       /* Get the page offset from the block offset(iblock) */
-       pgofs = (pgoff_t)(iblock >> (PAGE_CACHE_SHIFT - blkbits));
+       map->m_len = 0;
+       map->m_flags = 0;
+
+       /* it only supports block size == page size */
+       pgofs = (pgoff_t)map->m_lblk;
 
        if (f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
-               f2fs_map_bh(inode->i_sb, pgofs, &ei, bh_result);
+               map->m_pblk = ei.blk + pgofs - ei.fofs;
+               map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
+               map->m_flags = F2FS_MAP_MAPPED;
                goto out;
        }
 
@@ -1250,21 +1242,21 @@ static int __get_data_block(struct inode *inode, sector_t iblock,
                goto put_out;
 
        if (dn.data_blkaddr != NULL_ADDR) {
-               clear_buffer_new(bh_result);
-               map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
+               map->m_flags = F2FS_MAP_MAPPED;
+               map->m_pblk = dn.data_blkaddr;
        } else if (create) {
                err = __allocate_data_block(&dn);
                if (err)
                        goto put_out;
                allocated = true;
-               set_buffer_new(bh_result);
-               map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
+               map->m_flags = F2FS_MAP_NEW | F2FS_MAP_MAPPED;
+               map->m_pblk = dn.data_blkaddr;
        } else {
                goto put_out;
        }
 
        end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
-       bh_result->b_size = (((size_t)1) << blkbits);
+       map->m_len = 1;
        dn.ofs_in_node++;
        pgofs++;
 
@@ -1288,22 +1280,22 @@ get_next:
                end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
        }
 
-       if (maxblocks > (bh_result->b_size >> blkbits)) {
+       if (maxblocks > map->m_len) {
                block_t blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
                if (blkaddr == NULL_ADDR && create) {
                        err = __allocate_data_block(&dn);
                        if (err)
                                goto sync_out;
                        allocated = true;
-                       set_buffer_new(bh_result);
+                       map->m_flags |= F2FS_MAP_NEW;
                        blkaddr = dn.data_blkaddr;
                }
                /* Give more consecutive addresses for the readahead */
-               if (blkaddr == (bh_result->b_blocknr + ofs)) {
+               if (map->m_pblk != NEW_ADDR && blkaddr == (map->m_pblk + ofs)) {
                        ofs++;
                        dn.ofs_in_node++;
                        pgofs++;
-                       bh_result->b_size += (((size_t)1) << blkbits);
+                       map->m_len++;
                        goto get_next;
                }
        }
@@ -1316,10 +1308,28 @@ unlock_out:
        if (create)
                f2fs_unlock_op(F2FS_I_SB(inode));
 out:
-       trace_f2fs_get_data_block(inode, iblock, bh_result, err);
+       trace_f2fs_map_blocks(inode, map, err);
        return err;
 }
 
+static int __get_data_block(struct inode *inode, sector_t iblock,
+                       struct buffer_head *bh, int create, bool fiemap)
+{
+       struct f2fs_map_blocks map;
+       int ret;
+
+       map.m_lblk = iblock;
+       map.m_len = bh->b_size >> inode->i_blkbits;
+
+       ret = f2fs_map_blocks(inode, &map, create, fiemap);
+       if (!ret) {
+               map_bh(bh, inode->i_sb, map.m_pblk);
+               bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags;
+               bh->b_size = map.m_len << inode->i_blkbits;
+       }
+       return ret;
+}
+
 static int get_data_block(struct inode *inode, sector_t iblock,
                        struct buffer_head *bh_result, int create)
 {
index e1dd986262cb373c92a959ee80e7aacf81f231a0..c98454d011ff1cfe8db90f9a3203f85d05043996 100644 (file)
@@ -321,6 +321,22 @@ struct extent_tree {
        unsigned int count;             /* # of extent node in rb-tree*/
 };
 
+/*
+ * This structure is taken from ext4_map_blocks.
+ *
+ * Note that, however, f2fs uses NEW and MAPPED flags for f2fs_map_blocks().
+ */
+#define F2FS_MAP_NEW           (1 << BH_New)
+#define F2FS_MAP_MAPPED                (1 << BH_Mapped)
+#define F2FS_MAP_FLAGS         (F2FS_MAP_NEW | F2FS_MAP_MAPPED)
+
+struct f2fs_map_blocks {
+       block_t m_pblk;
+       block_t m_lblk;
+       unsigned int m_len;
+       unsigned int m_flags;
+};
+
 /*
  * i_advise uses FADVISE_XXX_BIT. We can add additional hints later.
  */
index 7a61ccc90a9cbdda4b3ec5214d8dceaf20509f28..04856a2d8c82d7d8579ba65125942b02a7820359 100644 (file)
@@ -117,6 +117,7 @@ TRACE_DEFINE_ENUM(CP_DISCARD);
                { CP_DISCARD,   "Discard" })
 
 struct victim_sel_policy;
+struct f2fs_map_blocks;
 
 DECLARE_EVENT_CLASS(f2fs__inode,
 
@@ -481,36 +482,35 @@ TRACE_EVENT(f2fs_truncate_partial_nodes,
                __entry->err)
 );
 
-TRACE_EVENT(f2fs_get_data_block,
-       TP_PROTO(struct inode *inode, sector_t iblock,
-                               struct buffer_head *bh, int ret),
+TRACE_EVENT(f2fs_map_blocks,
+       TP_PROTO(struct inode *inode, struct f2fs_map_blocks *map, int ret),
 
-       TP_ARGS(inode, iblock, bh, ret),
+       TP_ARGS(inode, map, ret),
 
        TP_STRUCT__entry(
                __field(dev_t,  dev)
                __field(ino_t,  ino)
-               __field(sector_t,       iblock)
-               __field(sector_t,       bh_start)
-               __field(size_t, bh_size)
+               __field(block_t,        m_lblk)
+               __field(block_t,        m_pblk)
+               __field(unsigned int,   m_len)
                __field(int,    ret)
        ),
 
        TP_fast_assign(
                __entry->dev            = inode->i_sb->s_dev;
                __entry->ino            = inode->i_ino;
-               __entry->iblock         = iblock;
-               __entry->bh_start       = bh->b_blocknr;
-               __entry->bh_size        = bh->b_size;
+               __entry->m_lblk         = map->m_lblk;
+               __entry->m_pblk         = map->m_pblk;
+               __entry->m_len          = map->m_len;
                __entry->ret            = ret;
        ),
 
        TP_printk("dev = (%d,%d), ino = %lu, file offset = %llu, "
-               "start blkaddr = 0x%llx, len = 0x%llx bytes, err = %d",
+               "start blkaddr = 0x%llx, len = 0x%llx, err = %d",
                show_dev_ino(__entry),
-               (unsigned long long)__entry->iblock,
-               (unsigned long long)__entry->bh_start,
-               (unsigned long long)__entry->bh_size,
+               (unsigned long long)__entry->m_lblk,
+               (unsigned long long)__entry->m_pblk,
+               (unsigned long long)__entry->m_len,
                __entry->ret)
 );