[PATCH] ext4: 48bit physical block number support in extents
Signed-off-by: Alex Tomas <alex@clusterfs.com>
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index f67b2ef..4a13b56 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -44,6 +44,44 @@
#include <asm/uaccess.h>
+/* this macro combines low and hi parts of phys. blocknr into ext4_fsblk_t */
+static inline ext4_fsblk_t ext_pblock(struct ext4_extent *ex)
+{
+ ext4_fsblk_t block;
+
+ block = le32_to_cpu(ex->ee_start);
+ if (sizeof(ext4_fsblk_t) > 4)
+ block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
+ return block;
+}
+
+/* this macro combines low and hi parts of phys. blocknr into ext4_fsblk_t */
+static inline ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
+{
+ ext4_fsblk_t block;
+
+ block = le32_to_cpu(ix->ei_leaf);
+ if (sizeof(ext4_fsblk_t) > 4)
+ block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
+ return block;
+}
+
+/* the routine stores large phys. blocknr into extent breaking it into parts */
+static inline void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
+{
+ ex->ee_start = cpu_to_le32((unsigned long) (pb & 0xffffffff));
+ if (sizeof(ext4_fsblk_t) > 4)
+ ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
+}
+
+/* the routine stores large phys. blocknr into index breaking it into parts */
+static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix, ext4_fsblk_t pb)
+{
+ ix->ei_leaf = cpu_to_le32((unsigned long) (pb & 0xffffffff));
+ if (sizeof(ext4_fsblk_t) > 4)
+ ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
+}
+
static int ext4_ext_check_header(const char *function, struct inode *inode,
struct ext4_extent_header *eh)
{
@@ -124,13 +162,13 @@
return err;
}
-static int ext4_ext_find_goal(struct inode *inode,
+static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
struct ext4_ext_path *path,
- unsigned long block)
+ ext4_fsblk_t block)
{
struct ext4_inode_info *ei = EXT4_I(inode);
- unsigned long bg_start;
- unsigned long colour;
+ ext4_fsblk_t bg_start;
+ ext4_grpblk_t colour;
int depth;
if (path) {
@@ -139,8 +177,7 @@
/* try to predict block placement */
if ((ex = path[depth].p_ext))
- return le32_to_cpu(ex->ee_start)
- + (block - le32_to_cpu(ex->ee_block));
+ return ext_pblock(ex)+(block-le32_to_cpu(ex->ee_block));
/* it looks index is empty
* try to find starting from index itself */
@@ -156,12 +193,12 @@
return bg_start + colour + block;
}
-static int
+static ext4_fsblk_t
ext4_ext_new_block(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path,
struct ext4_extent *ex, int *err)
{
- int goal, newblock;
+ ext4_fsblk_t goal, newblock;
goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
newblock = ext4_new_block(handle, inode, goal, err);
@@ -230,13 +267,13 @@
ext_debug("path:");
for (k = 0; k <= l; k++, path++) {
if (path->p_idx) {
- ext_debug(" %d->%d", le32_to_cpu(path->p_idx->ei_block),
- le32_to_cpu(path->p_idx->ei_leaf));
+ ext_debug(" %d->"E3FSBLK, le32_to_cpu(path->p_idx->ei_block),
+ idx_pblock(path->p_idx));
} else if (path->p_ext) {
- ext_debug(" %d:%d:%d",
+ ext_debug(" %d:%d:"E3FSBLK" ",
le32_to_cpu(path->p_ext->ee_block),
le16_to_cpu(path->p_ext->ee_len),
- le32_to_cpu(path->p_ext->ee_start));
+ ext_pblock(path->p_ext));
} else
ext_debug(" []");
}
@@ -257,9 +294,8 @@
ex = EXT_FIRST_EXTENT(eh);
for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
- ext_debug("%d:%d:%d ", le32_to_cpu(ex->ee_block),
- le16_to_cpu(ex->ee_len),
- le32_to_cpu(ex->ee_start));
+ ext_debug("%d:%d:"E3FSBLK" ", le32_to_cpu(ex->ee_block),
+ le16_to_cpu(ex->ee_len), ext_pblock(ex));
}
ext_debug("\n");
}
@@ -308,8 +344,8 @@
}
path->p_idx = l - 1;
- ext_debug(" -> %d->%d ", le32_to_cpu(path->p_idx->ei_block),
- le32_to_cpu(path->p_idx->ei_leaf));
+ ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block),
+ idx_block(path->p_idx));
#ifdef CHECK_BINSEARCH
{
@@ -374,10 +410,10 @@
}
path->p_ext = l - 1;
- ext_debug(" -> %d:%d:%d ",
+ ext_debug(" -> %d:"E3FSBLK":%d ",
le32_to_cpu(path->p_ext->ee_block),
- le32_to_cpu(path->p_ext->ee_start),
- le16_to_cpu(path->p_ext->ee_len));
+ ext_pblock(path->p_ext),
+ le16_to_cpu(path->p_ext->ee_len));
#ifdef CHECK_BINSEARCH
{
@@ -442,7 +478,7 @@
ext_debug("depth %d: num %d, max %d\n",
ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
ext4_ext_binsearch_idx(inode, path + ppos, block);
- path[ppos].p_block = le32_to_cpu(path[ppos].p_idx->ei_leaf);
+ path[ppos].p_block = idx_pblock(path[ppos].p_idx);
path[ppos].p_depth = i;
path[ppos].p_ext = NULL;
@@ -489,7 +525,7 @@
*/
static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
struct ext4_ext_path *curp,
- int logical, int ptr)
+ int logical, ext4_fsblk_t ptr)
{
struct ext4_extent_idx *ix;
int len, err;
@@ -524,7 +560,7 @@
}
ix->ei_block = cpu_to_le32(logical);
- ix->ei_leaf = cpu_to_le32(ptr);
+ ext4_idx_store_pblock(ix, ptr);
curp->p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(curp->p_hdr->eh_entries)+1);
BUG_ON(le16_to_cpu(curp->p_hdr->eh_entries)
@@ -556,9 +592,9 @@
struct ext4_extent_idx *fidx;
struct ext4_extent *ex;
int i = at, k, m, a;
- unsigned long newblock, oldblock;
+ ext4_fsblk_t newblock, oldblock;
__le32 border;
- int *ablocks = NULL; /* array of allocated blocks */
+ ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
int err = 0;
/* make decision: where to split? */
@@ -591,10 +627,10 @@
* we need this to handle errors and free blocks
* upon them
*/
- ablocks = kmalloc(sizeof(unsigned long) * depth, GFP_NOFS);
+ ablocks = kmalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS);
if (!ablocks)
return -ENOMEM;
- memset(ablocks, 0, sizeof(unsigned long) * depth);
+ memset(ablocks, 0, sizeof(ext4_fsblk_t) * depth);
/* allocate all needed blocks */
ext_debug("allocate %d blocks for indexes/leaf\n", depth - at);
@@ -633,9 +669,9 @@
path[depth].p_ext++;
while (path[depth].p_ext <=
EXT_MAX_EXTENT(path[depth].p_hdr)) {
- ext_debug("move %d:%d:%d in new leaf %lu\n",
+ ext_debug("move %d:"E3FSBLK":%d in new leaf "E3FSBLK"\n",
le32_to_cpu(path[depth].p_ext->ee_block),
- le32_to_cpu(path[depth].p_ext->ee_start),
+ ext_pblock(path[depth].p_ext),
le16_to_cpu(path[depth].p_ext->ee_len),
newblock);
/*memmove(ex++, path[depth].p_ext++,
@@ -679,7 +715,7 @@
while (k--) {
oldblock = newblock;
newblock = ablocks[--a];
- bh = sb_getblk(inode->i_sb, newblock);
+ bh = sb_getblk(inode->i_sb, (ext4_fsblk_t)newblock);
if (!bh) {
err = -EIO;
goto cleanup;
@@ -696,9 +732,9 @@
neh->eh_depth = cpu_to_le16(depth - i);
fidx = EXT_FIRST_INDEX(neh);
fidx->ei_block = border;
- fidx->ei_leaf = cpu_to_le32(oldblock);
+ ext4_idx_store_pblock(fidx, oldblock);
- ext_debug("int.index at %d (block %lu): %lu -> %lu\n", i,
+ ext_debug("int.index at %d (block "E3FSBLK"): %lu -> "E3FSBLK"\n", i,
newblock, (unsigned long) le32_to_cpu(border),
oldblock);
/* copy indexes */
@@ -710,9 +746,9 @@
BUG_ON(EXT_MAX_INDEX(path[i].p_hdr) !=
EXT_LAST_INDEX(path[i].p_hdr));
while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) {
- ext_debug("%d: move %d:%d in new index %lu\n", i,
+ ext_debug("%d: move %d:%d in new index "E3FSBLK"\n", i,
le32_to_cpu(path[i].p_idx->ei_block),
- le32_to_cpu(path[i].p_idx->ei_leaf),
+ idx_pblock(path[i].p_idx),
newblock);
/*memmove(++fidx, path[i].p_idx++,
sizeof(struct ext4_extent_idx));
@@ -791,7 +827,7 @@
struct ext4_extent_header *neh;
struct ext4_extent_idx *fidx;
struct buffer_head *bh;
- unsigned long newblock;
+ ext4_fsblk_t newblock;
int err = 0;
newblock = ext4_ext_new_block(handle, inode, path, newext, &err);
@@ -839,13 +875,13 @@
curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr);
/* FIXME: it works, but actually path[0] can be index */
curp->p_idx->ei_block = EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block;
- curp->p_idx->ei_leaf = cpu_to_le32(newblock);
+ ext4_idx_store_pblock(curp->p_idx, newblock);
neh = ext_inode_hdr(inode);
fidx = EXT_FIRST_INDEX(neh);
- ext_debug("new root: num %d(%d), lblock %d, ptr %d\n",
+ ext_debug("new root: num %d(%d), lblock %d, ptr "E3FSBLK"\n",
le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
- le32_to_cpu(fidx->ei_block), le32_to_cpu(fidx->ei_leaf));
+ le32_to_cpu(fidx->ei_block), idx_pblock(fidx));
neh->eh_depth = cpu_to_le16(path->p_depth + 1);
err = ext4_ext_dirty(handle, inode, curp);
@@ -1042,7 +1078,6 @@
ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
struct ext4_extent *ex2)
{
- /* FIXME: 48bit support */
if (le32_to_cpu(ex1->ee_block) + le16_to_cpu(ex1->ee_len)
!= le32_to_cpu(ex2->ee_block))
return 0;
@@ -1052,8 +1087,7 @@
return 0;
#endif
- if (le32_to_cpu(ex1->ee_start) + le16_to_cpu(ex1->ee_len)
- == le32_to_cpu(ex2->ee_start))
+ if (ext_pblock(ex1) + le16_to_cpu(ex1->ee_len) == ext_pblock(ex2))
return 1;
return 0;
}
@@ -1080,11 +1114,10 @@
/* try to insert block into found extent and return */
if (ex && ext4_can_extents_be_merged(inode, ex, newext)) {
- ext_debug("append %d block to %d:%d (from %d)\n",
+ ext_debug("append %d block to %d:%d (from "E3FSBLK")\n",
le16_to_cpu(newext->ee_len),
le32_to_cpu(ex->ee_block),
- le16_to_cpu(ex->ee_len),
- le32_to_cpu(ex->ee_start));
+ le16_to_cpu(ex->ee_len), ext_pblock(ex));
if ((err = ext4_ext_get_access(handle, inode, path + depth)))
return err;
ex->ee_len = cpu_to_le16(le16_to_cpu(ex->ee_len)
@@ -1140,9 +1173,9 @@
if (!nearex) {
/* there is no extent in this leaf, create first one */
- ext_debug("first extent in the leaf: %d:%d:%d\n",
+ ext_debug("first extent in the leaf: %d:"E3FSBLK":%d\n",
le32_to_cpu(newext->ee_block),
- le32_to_cpu(newext->ee_start),
+ ext_pblock(newext),
le16_to_cpu(newext->ee_len));
path[depth].p_ext = EXT_FIRST_EXTENT(eh);
} else if (le32_to_cpu(newext->ee_block)
@@ -1152,10 +1185,10 @@
len = EXT_MAX_EXTENT(eh) - nearex;
len = (len - 1) * sizeof(struct ext4_extent);
len = len < 0 ? 0 : len;
- ext_debug("insert %d:%d:%d after: nearest 0x%p, "
+ ext_debug("insert %d:"E3FSBLK":%d after: nearest 0x%p, "
"move %d from 0x%p to 0x%p\n",
le32_to_cpu(newext->ee_block),
- le32_to_cpu(newext->ee_start),
+ ext_pblock(newext),
le16_to_cpu(newext->ee_len),
nearex, len, nearex + 1, nearex + 2);
memmove(nearex + 2, nearex + 1, len);
@@ -1165,10 +1198,10 @@
BUG_ON(newext->ee_block == nearex->ee_block);
len = (EXT_MAX_EXTENT(eh) - nearex) * sizeof(struct ext4_extent);
len = len < 0 ? 0 : len;
- ext_debug("insert %d:%d:%d before: nearest 0x%p, "
+ ext_debug("insert %d:"E3FSBLK":%d before: nearest 0x%p, "
"move %d from 0x%p to 0x%p\n",
le32_to_cpu(newext->ee_block),
- le32_to_cpu(newext->ee_start),
+ ext_pblock(newext),
le16_to_cpu(newext->ee_len),
nearex, len, nearex + 1, nearex + 2);
memmove(nearex + 1, nearex, len);
@@ -1179,9 +1212,8 @@
nearex = path[depth].p_ext;
nearex->ee_block = newext->ee_block;
nearex->ee_start = newext->ee_start;
+ nearex->ee_start_hi = newext->ee_start_hi;
nearex->ee_len = newext->ee_len;
- /* FIXME: support for large fs */
- nearex->ee_start_hi = 0;
merge:
/* try to merge extents to the right */
@@ -1290,7 +1322,7 @@
} else {
cbex.ec_block = le32_to_cpu(ex->ee_block);
cbex.ec_len = le16_to_cpu(ex->ee_len);
- cbex.ec_start = le32_to_cpu(ex->ee_start);
+ cbex.ec_start = ext_pblock(ex);
cbex.ec_type = EXT4_EXT_CACHE_EXTENT;
}
@@ -1398,13 +1430,13 @@
cex->ec_type != EXT4_EXT_CACHE_EXTENT);
if (block >= cex->ec_block && block < cex->ec_block + cex->ec_len) {
ex->ee_block = cpu_to_le32(cex->ec_block);
- ex->ee_start = cpu_to_le32(cex->ec_start);
+ ext4_ext_store_pblock(ex, cex->ec_start);
ex->ee_len = cpu_to_le16(cex->ec_len);
- ext_debug("%lu cached by %lu:%lu:%lu\n",
+ ext_debug("%lu cached by %lu:%lu:"E3FSBLK"\n",
(unsigned long) block,
(unsigned long) cex->ec_block,
(unsigned long) cex->ec_len,
- (unsigned long) cex->ec_start);
+ cex->ec_start);
return cex->ec_type;
}
@@ -1422,18 +1454,18 @@
{
struct buffer_head *bh;
int err;
- unsigned long leaf;
+ ext4_fsblk_t leaf;
/* free index block */
path--;
- leaf = le32_to_cpu(path->p_idx->ei_leaf);
+ leaf = idx_pblock(path->p_idx);
BUG_ON(path->p_hdr->eh_entries == 0);
if ((err = ext4_ext_get_access(handle, inode, path)))
return err;
path->p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(path->p_hdr->eh_entries)-1);
if ((err = ext4_ext_dirty(handle, inode, path)))
return err;
- ext_debug("index is empty, remove it, free block %lu\n", leaf);
+ ext_debug("index is empty, remove it, free block "E3FSBLK"\n", leaf);
bh = sb_find_get_block(inode->i_sb, leaf);
ext4_forget(handle, 1, inode, bh, leaf);
ext4_free_blocks(handle, inode, leaf, 1);
@@ -1515,10 +1547,11 @@
if (from >= le32_to_cpu(ex->ee_block)
&& to == le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len) - 1) {
/* tail removal */
- unsigned long num, start;
+ unsigned long num;
+ ext4_fsblk_t start;
num = le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len) - from;
- start = le32_to_cpu(ex->ee_start) + le16_to_cpu(ex->ee_len) - num;
- ext_debug("free last %lu blocks starting %lu\n", num, start);
+ start = ext_pblock(ex) + le16_to_cpu(ex->ee_len) - num;
+ ext_debug("free last %lu blocks starting "E3FSBLK"\n", num, start);
for (i = 0; i < num; i++) {
bh = sb_find_get_block(inode->i_sb, start + i);
ext4_forget(handle, 0, inode, bh, start + i);
@@ -1621,7 +1654,7 @@
if (num == 0) {
/* this extent is removed entirely mark slot unused */
- ex->ee_start = 0;
+ ext4_ext_store_pblock(ex, 0);
eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries)-1);
}
@@ -1632,8 +1665,8 @@
if (err)
goto out;
- ext_debug("new extent: %u:%u:%u\n", block, num,
- le32_to_cpu(ex->ee_start));
+ ext_debug("new extent: %u:%u:"E3FSBLK"\n", block, num,
+ ext_pblock(ex));
ex--;
ex_ee_block = le32_to_cpu(ex->ee_block);
ex_ee_len = le16_to_cpu(ex->ee_len);
@@ -1748,11 +1781,11 @@
path[i].p_idx);
if (ext4_ext_more_to_rm(path + i)) {
/* go to the next level */
- ext_debug("move to level %d (block %d)\n",
- i + 1, le32_to_cpu(path[i].p_idx->ei_leaf));
+ ext_debug("move to level %d (block "E3FSBLK")\n",
+ i + 1, idx_pblock(path[i].p_idx));
memset(path + i + 1, 0, sizeof(*path));
path[i+1].p_bh =
- sb_bread(sb, le32_to_cpu(path[i].p_idx->ei_leaf));
+ sb_bread(sb, idx_pblock(path[i].p_idx));
if (!path[i+1].p_bh) {
/* should we reset i_size? */
err = -EIO;
@@ -1851,13 +1884,15 @@
#endif
}
-int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, sector_t iblock,
+int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
+ ext4_fsblk_t iblock,
unsigned long max_blocks, struct buffer_head *bh_result,
int create, int extend_disksize)
{
struct ext4_ext_path *path = NULL;
struct ext4_extent newex, *ex;
- int goal, newblock, err = 0, depth;
+ ext4_fsblk_t goal, newblock;
+ int err = 0, depth;
unsigned long allocated = 0;
__clear_bit(BH_New, &bh_result->b_state);
@@ -1878,7 +1913,7 @@
/* block is already allocated */
newblock = iblock
- le32_to_cpu(newex.ee_block)
- + le32_to_cpu(newex.ee_start);
+ + ext_pblock(&newex);
/* number of remain blocks in the extent */
allocated = le16_to_cpu(newex.ee_len) -
(iblock - le32_to_cpu(newex.ee_block));
@@ -1907,14 +1942,14 @@
if ((ex = path[depth].p_ext)) {
unsigned long ee_block = le32_to_cpu(ex->ee_block);
- unsigned long ee_start = le32_to_cpu(ex->ee_start);
+ ext4_fsblk_t ee_start = ext_pblock(ex);
unsigned short ee_len = le16_to_cpu(ex->ee_len);
/* if found exent covers block, simple return it */
if (iblock >= ee_block && iblock < ee_block + ee_len) {
newblock = iblock - ee_block + ee_start;
/* number of remain blocks in the extent */
allocated = ee_len - (iblock - ee_block);
- ext_debug("%d fit into %lu:%d -> %d\n", (int) iblock,
+ ext_debug("%d fit into %lu:%d -> "E3FSBLK"\n", (int) iblock,
ee_block, ee_len, newblock);
ext4_ext_put_in_cache(inode, ee_block, ee_len,
ee_start, EXT4_EXT_CACHE_EXTENT);
@@ -1944,12 +1979,12 @@
newblock = ext4_new_blocks(handle, inode, goal, &allocated, &err);
if (!newblock)
goto out2;
- ext_debug("allocate new block: goal %d, found %d/%lu\n",
+ ext_debug("allocate new block: goal "E3FSBLK", found "E3FSBLK"/%lu\n",
goal, newblock, allocated);
/* try to insert new extent into found leaf and return */
newex.ee_block = cpu_to_le32(iblock);
- newex.ee_start = cpu_to_le32(newblock);
+ ext4_ext_store_pblock(&newex, newblock);
newex.ee_len = cpu_to_le16(allocated);
err = ext4_ext_insert_extent(handle, inode, path, &newex);
if (err)
@@ -1959,7 +1994,7 @@
EXT4_I(inode)->i_disksize = inode->i_size;
/* previous routine could use block we allocated */
- newblock = le32_to_cpu(newex.ee_start);
+ newblock = ext_pblock(&newex);
__set_bit(BH_New, &bh_result->b_state);
ext4_ext_put_in_cache(inode, iblock, allocated, newblock,