#include <linux/module.h>
#include <linux/spinlock.h>
#include <linux/blkdev.h>
+#include <linux/swap.h>
+#include <linux/version.h>
+#include <linux/writeback.h>
#include "extent_map.h"
/* temporary define until extent_map moves out of btrfs */
static struct kmem_cache *extent_map_cache;
static struct kmem_cache *extent_state_cache;
+static struct kmem_cache *extent_buffer_cache;
+
+static LIST_HEAD(buffers);
+static LIST_HEAD(states);
+
+static spinlock_t state_lock = SPIN_LOCK_UNLOCKED;
+#define BUFFER_LRU_MAX 64
struct tree_entry {
u64 start;
struct rb_node rb_node;
};
+struct extent_page_data {
+ struct bio *bio;
+ struct extent_map_tree *tree;
+ get_extent_t *get_extent;
+};
+
void __init extent_map_init(void)
{
extent_map_cache = btrfs_cache_create("extent_map",
- sizeof(struct extent_map),
- SLAB_DESTROY_BY_RCU,
+ sizeof(struct extent_map), 0,
NULL);
extent_state_cache = btrfs_cache_create("extent_state",
- sizeof(struct extent_state),
- SLAB_DESTROY_BY_RCU,
+ sizeof(struct extent_state), 0,
+ NULL);
+ extent_buffer_cache = btrfs_cache_create("extent_buffers",
+ sizeof(struct extent_buffer), 0,
NULL);
}
void __exit extent_map_exit(void)
{
+ struct extent_state *state;
+
+ while (!list_empty(&states)) {
+ state = list_entry(states.next, struct extent_state, list);
+ printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs));
+ list_del(&state->list);
+ kmem_cache_free(extent_state_cache, state);
+
+ }
+
if (extent_map_cache)
kmem_cache_destroy(extent_map_cache);
if (extent_state_cache)
kmem_cache_destroy(extent_state_cache);
+ if (extent_buffer_cache)
+ kmem_cache_destroy(extent_buffer_cache);
}
void extent_map_tree_init(struct extent_map_tree *tree,
tree->state.rb_node = NULL;
tree->ops = NULL;
rwlock_init(&tree->lock);
+ spin_lock_init(&tree->lru_lock);
tree->mapping = mapping;
+ INIT_LIST_HEAD(&tree->buffer_lru);
+ tree->lru_size = 0;
}
EXPORT_SYMBOL(extent_map_tree_init);
+void extent_map_tree_empty_lru(struct extent_map_tree *tree)
+{
+ struct extent_buffer *eb;
+ while(!list_empty(&tree->buffer_lru)) {
+ eb = list_entry(tree->buffer_lru.next, struct extent_buffer,
+ lru);
+ list_del(&eb->lru);
+ free_extent_buffer(eb);
+ }
+}
+EXPORT_SYMBOL(extent_map_tree_empty_lru);
+
struct extent_map *alloc_extent_map(gfp_t mask)
{
struct extent_map *em;
struct extent_state *alloc_extent_state(gfp_t mask)
{
struct extent_state *state;
+ unsigned long flags;
+
state = kmem_cache_alloc(extent_state_cache, mask);
if (!state || IS_ERR(state))
return state;
state->state = 0;
state->in_tree = 0;
state->private = 0;
+
+ spin_lock_irqsave(&state_lock, flags);
+ list_add(&state->list, &states);
+ spin_unlock_irqrestore(&state_lock, flags);
+
atomic_set(&state->refs, 1);
init_waitqueue_head(&state->wq);
return state;
void free_extent_state(struct extent_state *state)
{
+ unsigned long flags;
if (!state)
return;
if (atomic_dec_and_test(&state->refs)) {
WARN_ON(state->in_tree);
+ spin_lock_irqsave(&state_lock, flags);
+ list_del(&state->list);
+ spin_unlock_irqrestore(&state_lock, flags);
kmem_cache_free(extent_state_cache, state);
}
}
if (prev && prev->end + 1 == em->start &&
((em->block_start == EXTENT_MAP_HOLE &&
prev->block_start == EXTENT_MAP_HOLE) ||
- (em->block_start == prev->block_end + 1))) {
+ (em->block_start == EXTENT_MAP_INLINE &&
+ prev->block_start == EXTENT_MAP_INLINE) ||
+ (em->block_start == EXTENT_MAP_DELALLOC &&
+ prev->block_start == EXTENT_MAP_DELALLOC) ||
+ (em->block_start < EXTENT_MAP_DELALLOC - 1 &&
+ em->block_start == prev->block_end + 1))) {
em->start = prev->start;
em->block_start = prev->block_start;
rb_erase(&prev->rb_node, &tree->map);
state->state |= bits;
state->start = start;
state->end = end;
- if ((end & 4095) == 0) {
- printk("insert state %Lu %Lu strange end\n", start, end);
- WARN_ON(1);
- }
node = tree_insert(&tree->state, end, &state->rb_node);
if (node) {
struct extent_state *found;
prealloc->end = split - 1;
prealloc->state = orig->state;
orig->start = split;
- if ((prealloc->end & 4095) == 0) {
- printk("insert state %Lu %Lu strange end\n", prealloc->start,
- prealloc->end);
- WARN_ON(1);
- }
+
node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
if (node) {
struct extent_state *found;
return set;
search_again:
- if (start >= end)
+ if (start > end)
goto out;
write_unlock_irqrestore(&tree->lock, flags);
if (mask & __GFP_WAIT)
}
EXPORT_SYMBOL(set_extent_dirty);
+int set_extent_bits(struct extent_map_tree *tree, u64 start, u64 end,
+ int bits, gfp_t mask)
+{
+ return set_extent_bit(tree, start, end, bits, 0, NULL,
+ mask);
+}
+EXPORT_SYMBOL(set_extent_bits);
+
+int clear_extent_bits(struct extent_map_tree *tree, u64 start, u64 end,
+ int bits, gfp_t mask)
+{
+ return clear_extent_bit(tree, start, end, bits, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_bits);
+
int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end,
gfp_t mask)
{
struct extent_state *state;
int ret = 1;
- write_lock_irq(&tree->lock);
+ read_lock_irq(&tree->lock);
/*
* this search will find all the extents that end after
* our range starts.
while(1) {
state = rb_entry(node, struct extent_state, rb_node);
- if (state->state & bits) {
+ if (state->end >= start && (state->state & bits)) {
*start_ret = state->start;
*end_ret = state->end;
ret = 0;
+ break;
}
node = rb_next(node);
if (!node)
break;
}
out:
- write_unlock_irq(&tree->lock);
+ read_unlock_irq(&tree->lock);
return ret;
}
EXPORT_SYMBOL(find_first_extent_bit);
if (state->state & EXTENT_LOCKED) {
DEFINE_WAIT(wait);
atomic_inc(&state->refs);
+ prepare_to_wait(&state->wq, &wait,
+ TASK_UNINTERRUPTIBLE);
write_unlock_irq(&tree->lock);
schedule();
write_lock_irq(&tree->lock);
node = rb_next(node);
if (!node)
break;
- total_bytes = state->end - state->start + 1;
+ total_bytes += state->end - state->start + 1;
if (total_bytes >= max_bytes)
break;
}
out:
write_unlock_irq(&tree->lock);
return ret;
-
}
int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private)
* has the bits set. Otherwise, 1 is returned if any bit in the
* range is found set.
*/
-static int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
- int bits, int filled)
+int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
+ int bits, int filled)
{
struct extent_state *state = NULL;
struct rb_node *node;
read_unlock_irq(&tree->lock);
return bitset;
}
+EXPORT_SYMBOL(test_range_bit);
/*
* helper function to set a given page up to date if all the
static int check_page_uptodate(struct extent_map_tree *tree,
struct page *page)
{
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 end = start + PAGE_CACHE_SIZE - 1;
if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1))
SetPageUptodate(page);
static int check_page_locked(struct extent_map_tree *tree,
struct page *page)
{
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 end = start + PAGE_CACHE_SIZE - 1;
if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0))
unlock_page(page);
static int check_page_writeback(struct extent_map_tree *tree,
struct page *page)
{
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 end = start + PAGE_CACHE_SIZE - 1;
if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
end_page_writeback(page);
* Scheduling is not allowed, so the extent state tree is expected
* to have one and only one object corresponding to this IO.
*/
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_bio_extent_writepage(struct bio *bio, int err)
+#else
static int end_bio_extent_writepage(struct bio *bio,
unsigned int bytes_done, int err)
+#endif
{
const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
u64 end;
int whole_page;
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
if (bio->bi_size)
return 1;
+#endif
do {
struct page *page = bvec->bv_page;
- start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+ start = ((u64)page->index << PAGE_CACHE_SHIFT) +
+ bvec->bv_offset;
end = start + bvec->bv_len - 1;
if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
} while (bvec >= bio->bi_io_vec);
bio_put(bio);
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
return 0;
+#endif
}
/*
* Scheduling is not allowed, so the extent state tree is expected
* to have one and only one object corresponding to this IO.
*/
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_bio_extent_readpage(struct bio *bio, int err)
+#else
static int end_bio_extent_readpage(struct bio *bio,
unsigned int bytes_done, int err)
+#endif
{
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
int whole_page;
int ret;
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
if (bio->bi_size)
return 1;
+#endif
do {
struct page *page = bvec->bv_page;
- start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+ start = ((u64)page->index << PAGE_CACHE_SHIFT) +
+ bvec->bv_offset;
end = start + bvec->bv_len - 1;
if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
} while (bvec >= bio->bi_io_vec);
bio_put(bio);
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
return 0;
+#endif
}
/*
* the structs in the extent tree when done, and set the uptodate bits
* as appropriate.
*/
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_bio_extent_preparewrite(struct bio *bio, int err)
+#else
static int end_bio_extent_preparewrite(struct bio *bio,
unsigned int bytes_done, int err)
+#endif
{
const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
u64 start;
u64 end;
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
if (bio->bi_size)
return 1;
+#endif
do {
struct page *page = bvec->bv_page;
- start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+ start = ((u64)page->index << PAGE_CACHE_SHIFT) +
+ bvec->bv_offset;
end = start + bvec->bv_len - 1;
if (--bvec >= bio->bi_io_vec)
} while (bvec >= bio->bi_io_vec);
bio_put(bio);
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
return 0;
+#endif
}
-static int submit_extent_page(int rw, struct extent_map_tree *tree,
- struct page *page, sector_t sector,
- size_t size, unsigned long offset,
- struct block_device *bdev,
- bio_end_io_t end_io_func)
+static struct bio *
+extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
+ gfp_t gfp_flags)
{
struct bio *bio;
- int ret = 0;
- bio = bio_alloc(GFP_NOIO, 1);
+ bio = bio_alloc(gfp_flags, nr_vecs);
- bio->bi_sector = sector;
- bio->bi_bdev = bdev;
- bio->bi_io_vec[0].bv_page = page;
- bio->bi_io_vec[0].bv_len = size;
- bio->bi_io_vec[0].bv_offset = offset;
-
- bio->bi_vcnt = 1;
- bio->bi_idx = 0;
- bio->bi_size = size;
+ if (bio == NULL && (current->flags & PF_MEMALLOC)) {
+ while (!bio && (nr_vecs /= 2))
+ bio = bio_alloc(gfp_flags, nr_vecs);
+ }
- bio->bi_end_io = end_io_func;
- bio->bi_private = tree;
+ if (bio) {
+ bio->bi_bdev = bdev;
+ bio->bi_sector = first_sector;
+ }
+ return bio;
+}
+static int submit_one_bio(int rw, struct bio *bio)
+{
+ int ret = 0;
bio_get(bio);
submit_bio(rw, bio);
-
if (bio_flagged(bio, BIO_EOPNOTSUPP))
ret = -EOPNOTSUPP;
-
bio_put(bio);
return ret;
}
+static int submit_extent_page(int rw, struct extent_map_tree *tree,
+ struct page *page, sector_t sector,
+ size_t size, unsigned long offset,
+ struct block_device *bdev,
+ struct bio **bio_ret,
+ int max_pages,
+ bio_end_io_t end_io_func)
+{
+ int ret = 0;
+ struct bio *bio;
+ int nr;
+
+ if (bio_ret && *bio_ret) {
+ bio = *bio_ret;
+ if (bio->bi_sector + (bio->bi_size >> 9) != sector ||
+ bio_add_page(bio, page, size, offset) < size) {
+ ret = submit_one_bio(rw, bio);
+ bio = NULL;
+ } else {
+ return 0;
+ }
+ }
+ nr = min(max_pages, bio_get_nr_vecs(bdev));
+ bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
+ if (!bio) {
+ printk("failed to allocate bio nr %d\n", nr);
+ }
+ bio_add_page(bio, page, size, offset);
+ bio->bi_end_io = end_io_func;
+ bio->bi_private = tree;
+ if (bio_ret) {
+ *bio_ret = bio;
+ } else {
+ ret = submit_one_bio(rw, bio);
+ }
+
+ return ret;
+}
+
void set_page_extent_mapped(struct page *page)
{
if (!PagePrivate(page)) {
SetPagePrivate(page);
WARN_ON(!page->mapping->a_ops->invalidatepage);
- set_page_private(page, 1);
+ set_page_private(page, EXTENT_PAGE_PRIVATE);
page_cache_get(page);
}
}
get_extent_t *get_extent)
{
struct inode *inode = page->mapping->host;
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 page_end = start + PAGE_CACHE_SIZE - 1;
u64 end;
u64 cur = start;
if (!ret) {
ret = submit_extent_page(READ, tree, page,
sector, iosize, page_offset,
- bdev, end_bio_extent_readpage);
+ bdev, NULL, 1,
+ end_bio_extent_readpage);
}
if (ret)
SetPageError(page);
* are found, they are marked writeback. Then the lock bits are removed
* and the end_io handler clears the writeback ranges
*/
-int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
- get_extent_t *get_extent,
- struct writeback_control *wbc)
+static int __extent_writepage(struct page *page, struct writeback_control *wbc,
+ void *data)
{
struct inode *inode = page->mapping->host;
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ struct extent_page_data *epd = data;
+ struct extent_map_tree *tree = epd->tree;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 page_end = start + PAGE_CACHE_SIZE - 1;
u64 end;
u64 cur = start;
u64 extent_offset;
u64 last_byte = i_size_read(inode);
u64 block_start;
+ u64 iosize;
sector_t sector;
struct extent_map *em;
struct block_device *bdev;
int ret;
int nr = 0;
size_t page_offset = 0;
- size_t iosize;
size_t blocksize;
loff_t i_size = i_size_read(inode);
unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
break;
}
- em = get_extent(inode, page, page_offset, cur, end, 0);
+ em = epd->get_extent(inode, page, page_offset, cur, end, 1);
if (IS_ERR(em) || !em) {
SetPageError(page);
break;
if (ret)
SetPageError(page);
else {
+ unsigned long nr = end_index + 1;
set_range_writeback(tree, cur, cur + iosize - 1);
+
ret = submit_extent_page(WRITE, tree, page, sector,
iosize, page_offset, bdev,
+ &epd->bio, nr,
end_bio_extent_writepage);
if (ret)
SetPageError(page);
unlock_page(page);
return 0;
}
+
+int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
+ get_extent_t *get_extent,
+ struct writeback_control *wbc)
+{
+ int ret;
+ struct extent_page_data epd = {
+ .bio = NULL,
+ .tree = tree,
+ .get_extent = get_extent,
+ };
+
+ ret = __extent_writepage(page, wbc, &epd);
+ if (epd.bio)
+ submit_one_bio(WRITE, epd.bio);
+ return ret;
+}
EXPORT_SYMBOL(extent_write_full_page);
+int extent_writepages(struct extent_map_tree *tree,
+ struct address_space *mapping,
+ get_extent_t *get_extent,
+ struct writeback_control *wbc)
+{
+ int ret;
+ struct extent_page_data epd = {
+ .bio = NULL,
+ .tree = tree,
+ .get_extent = get_extent,
+ };
+
+ ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd);
+ if (epd.bio)
+ submit_one_bio(WRITE, epd.bio);
+ return ret;
+}
+EXPORT_SYMBOL(extent_writepages);
+
/*
* basic invalidatepage code, this waits on any locked or writeback
* ranges corresponding to the page, and then deletes any extent state
int extent_invalidatepage(struct extent_map_tree *tree,
struct page *page, unsigned long offset)
{
- u64 start = (page->index << PAGE_CACHE_SHIFT);
+ u64 start = ((u64)page->index << PAGE_CACHE_SHIFT);
u64 end = start + PAGE_CACHE_SIZE - 1;
size_t blocksize = page->mapping->host->i_sb->s_blocksize;
struct inode *inode, struct page *page,
unsigned from, unsigned to, get_extent_t *get_extent)
{
- u64 page_start = page->index << PAGE_CACHE_SHIFT;
+ u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
u64 block_start;
u64 orig_block_start;
EXTENT_LOCKED, 0, NULL, GFP_NOFS);
ret = submit_extent_page(READ, tree, page,
sector, iosize, page_offset, em->bdev,
+ NULL, 1,
end_bio_extent_preparewrite);
iocount++;
block_start = block_start + iosize;
int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page)
{
struct extent_map *em;
- u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 end = start + PAGE_CACHE_SIZE - 1;
u64 orig_start = start;
int ret = 1;
struct inode *inode = mapping->host;
u64 start = iblock << inode->i_blkbits;
u64 end = start + (1 << inode->i_blkbits) - 1;
+ sector_t sector = 0;
struct extent_map *em;
em = get_extent(inode, NULL, 0, start, end, 0);
if (!em || IS_ERR(em))
return 0;
- // XXX(hch): block 0 is valid in some cases, e.g. XFS RT device
if (em->block_start == EXTENT_MAP_INLINE ||
em->block_start == EXTENT_MAP_HOLE)
- return 0;
+ goto out;
+
+ sector = (em->block_start + start - em->start) >> inode->i_blkbits;
+out:
+ free_extent_map(em);
+ return sector;
+}
+
+static int add_lru(struct extent_map_tree *tree, struct extent_buffer *eb)
+{
+ if (list_empty(&eb->lru)) {
+ extent_buffer_get(eb);
+ list_add(&eb->lru, &tree->buffer_lru);
+ tree->lru_size++;
+ if (tree->lru_size >= BUFFER_LRU_MAX) {
+ struct extent_buffer *rm;
+ rm = list_entry(tree->buffer_lru.prev,
+ struct extent_buffer, lru);
+ tree->lru_size--;
+ list_del(&rm->lru);
+ free_extent_buffer(rm);
+ }
+ } else
+ list_move(&eb->lru, &tree->buffer_lru);
+ return 0;
+}
+static struct extent_buffer *find_lru(struct extent_map_tree *tree,
+ u64 start, unsigned long len)
+{
+ struct list_head *lru = &tree->buffer_lru;
+ struct list_head *cur = lru->next;
+ struct extent_buffer *eb;
+
+ if (list_empty(lru))
+ return NULL;
+
+ do {
+ eb = list_entry(cur, struct extent_buffer, lru);
+ if (eb->start == start && eb->len == len) {
+ extent_buffer_get(eb);
+ return eb;
+ }
+ cur = cur->next;
+ } while (cur != lru);
+ return NULL;
+}
+
+static inline unsigned long num_extent_pages(u64 start, u64 len)
+{
+ return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
+ (start >> PAGE_CACHE_SHIFT);
+}
+
+static inline struct page *extent_buffer_page(struct extent_buffer *eb,
+ unsigned long i)
+{
+ struct page *p;
+ struct address_space *mapping;
+
+ if (i == 0)
+ return eb->first_page;
+ i += eb->start >> PAGE_CACHE_SHIFT;
+ mapping = eb->first_page->mapping;
+ read_lock_irq(&mapping->tree_lock);
+ p = radix_tree_lookup(&mapping->page_tree, i);
+ read_unlock_irq(&mapping->tree_lock);
+ return p;
+}
+
+static struct extent_buffer *__alloc_extent_buffer(struct extent_map_tree *tree,
+ u64 start,
+ unsigned long len,
+ gfp_t mask)
+{
+ struct extent_buffer *eb = NULL;
+
+ spin_lock(&tree->lru_lock);
+ eb = find_lru(tree, start, len);
+ if (eb) {
+ goto lru_add;
+ }
+ spin_unlock(&tree->lru_lock);
+
+ if (eb) {
+ memset(eb, 0, sizeof(*eb));
+ } else {
+ eb = kmem_cache_zalloc(extent_buffer_cache, mask);
+ }
+ INIT_LIST_HEAD(&eb->lru);
+ eb->start = start;
+ eb->len = len;
+ atomic_set(&eb->refs, 1);
+
+ spin_lock(&tree->lru_lock);
+lru_add:
+ add_lru(tree, eb);
+ spin_unlock(&tree->lru_lock);
+ return eb;
+}
- return (em->block_start + start - em->start) >> inode->i_blkbits;
+static void __free_extent_buffer(struct extent_buffer *eb)
+{
+ kmem_cache_free(extent_buffer_cache, eb);
}
struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
u64 start, unsigned long len,
+ struct page *page0,
gfp_t mask)
{
- unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
- (start >> PAGE_CACHE_SHIFT) + 1;
+ unsigned long num_pages = num_extent_pages(start, len);
unsigned long i;
unsigned long index = start >> PAGE_CACHE_SHIFT;
struct extent_buffer *eb;
struct page *p;
struct address_space *mapping = tree->mapping;
- int uptodate = 0;
+ int uptodate = 1;
- eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+ eb = __alloc_extent_buffer(tree, start, len, mask);
if (!eb || IS_ERR(eb))
return NULL;
- eb->start = start;
- eb->len = len;
- atomic_set(&eb->refs, 1);
+ if (eb->flags & EXTENT_BUFFER_FILLED)
+ return eb;
- for (i = 0; i < num_pages; i++, index++) {
+ if (page0) {
+ eb->first_page = page0;
+ i = 1;
+ index++;
+ page_cache_get(page0);
+ mark_page_accessed(page0);
+ set_page_extent_mapped(page0);
+ set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ i = 0;
+ }
+ for (; i < num_pages; i++, index++) {
p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
- if (!p)
+ if (!p) {
+ WARN_ON(1);
+ /* make sure the free only frees the pages we've
+ * grabbed a reference on
+ */
+ eb->len = i << PAGE_CACHE_SHIFT;
+ eb->start &= ~((u64)PAGE_CACHE_SIZE - 1);
goto fail;
- eb->pages[i] = p;
+ }
+ set_page_extent_mapped(p);
+ mark_page_accessed(p);
+ if (i == 0) {
+ eb->first_page = p;
+ set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ set_page_private(p, EXTENT_PAGE_PRIVATE);
+ }
if (!PageUptodate(p))
uptodate = 0;
unlock_page(p);
}
if (uptodate)
eb->flags |= EXTENT_UPTODATE;
+ eb->flags |= EXTENT_BUFFER_FILLED;
return eb;
fail:
free_extent_buffer(eb);
u64 start, unsigned long len,
gfp_t mask)
{
- unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
- (start >> PAGE_CACHE_SHIFT) + 1;
- unsigned long i;
- unsigned long index = start >> PAGE_CACHE_SHIFT;
+ unsigned long num_pages = num_extent_pages(start, len);
+ unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT;
struct extent_buffer *eb;
struct page *p;
struct address_space *mapping = tree->mapping;
+ int uptodate = 1;
- eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+ eb = __alloc_extent_buffer(tree, start, len, mask);
if (!eb || IS_ERR(eb))
return NULL;
- eb->start = start;
- eb->len = len;
- atomic_set(&eb->refs, 1);
+ if (eb->flags & EXTENT_BUFFER_FILLED)
+ return eb;
for (i = 0; i < num_pages; i++, index++) {
- p = find_get_page(mapping, index);
- if (!p)
+ p = find_lock_page(mapping, index);
+ if (!p) {
+ /* make sure the free only frees the pages we've
+ * grabbed a reference on
+ */
+ eb->len = i << PAGE_CACHE_SHIFT;
+ eb->start &= ~((u64)PAGE_CACHE_SIZE - 1);
goto fail;
- eb->pages[i] = p;
+ }
+ set_page_extent_mapped(p);
+ mark_page_accessed(p);
+
+ if (i == 0) {
+ eb->first_page = p;
+ set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ set_page_private(p, EXTENT_PAGE_PRIVATE);
+ }
+
+ if (!PageUptodate(p))
+ uptodate = 0;
+ unlock_page(p);
}
+ if (uptodate)
+ eb->flags |= EXTENT_UPTODATE;
+ eb->flags |= EXTENT_BUFFER_FILLED;
return eb;
fail:
free_extent_buffer(eb);
if (!atomic_dec_and_test(&eb->refs))
return;
- num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
- (eb->start >> PAGE_CACHE_SHIFT) + 1;
+ num_pages = num_extent_pages(eb->start, eb->len);
for (i = 0; i < num_pages; i++) {
- if (eb->pages[i])
- page_cache_release(eb->pages[i]);
+ page_cache_release(extent_buffer_page(eb, i));
}
- kfree(eb);
+ __free_extent_buffer(eb);
}
EXPORT_SYMBOL(free_extent_buffer);
u64 end = start + eb->len - 1;
set = clear_extent_dirty(tree, start, end, GFP_NOFS);
- num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
- (eb->start >> PAGE_CACHE_SHIFT) + 1;
+ num_pages = num_extent_pages(eb->start, eb->len);
for (i = 0; i < num_pages; i++) {
- page = eb->pages[i];
+ page = extent_buffer_page(eb, i);
lock_page(page);
/*
* if we're on the last page or the first page and the
*/
if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
((i == num_pages - 1) &&
- ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
- start = page->index << PAGE_CACHE_SHIFT;
+ ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
+ start = (u64)page->index << PAGE_CACHE_SHIFT;
end = start + PAGE_CACHE_SIZE - 1;
if (test_range_bit(tree, start, end,
EXTENT_DIRTY, 0)) {
int set_extent_buffer_dirty(struct extent_map_tree *tree,
struct extent_buffer *eb)
{
- return set_range_dirty(tree, eb->start, eb->start + eb->len - 1);
+ unsigned long i;
+ unsigned long num_pages;
+
+ num_pages = num_extent_pages(eb->start, eb->len);
+ for (i = 0; i < num_pages; i++) {
+ struct page *page = extent_buffer_page(eb, i);
+ /* writepage may need to do something special for the
+ * first page, we have to make sure page->private is
+ * properly set. releasepage may drop page->private
+ * on us if the page isn't already dirty.
+ */
+ if (i == 0) {
+ lock_page(page);
+ set_page_private(page,
+ EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ eb->len << 2);
+ }
+ __set_page_dirty_nobuffers(extent_buffer_page(eb, i));
+ if (i == 0)
+ unlock_page(page);
+ }
+ return set_extent_dirty(tree, eb->start,
+ eb->start + eb->len - 1, GFP_NOFS);
}
EXPORT_SYMBOL(set_extent_buffer_dirty);
struct page *page;
unsigned long num_pages;
- num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
- (eb->start >> PAGE_CACHE_SHIFT) + 1;
+ num_pages = num_extent_pages(eb->start, eb->len);
set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
GFP_NOFS);
for (i = 0; i < num_pages; i++) {
- page = eb->pages[i];
+ page = extent_buffer_page(eb, i);
if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
((i == num_pages - 1) &&
- ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
+ ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
check_page_uptodate(tree, page);
continue;
}
EXPORT_SYMBOL(extent_buffer_uptodate);
int read_extent_buffer_pages(struct extent_map_tree *tree,
- struct extent_buffer *eb, int wait)
+ struct extent_buffer *eb,
+ u64 start,
+ int wait)
{
unsigned long i;
+ unsigned long start_i;
struct page *page;
int err;
int ret = 0;
if (eb->flags & EXTENT_UPTODATE)
return 0;
- if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+ if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1,
EXTENT_UPTODATE, 1)) {
return 0;
}
+ if (start) {
+ WARN_ON(start < eb->start);
+ start_i = (start >> PAGE_CACHE_SHIFT) -
+ (eb->start >> PAGE_CACHE_SHIFT);
+ } else {
+ start_i = 0;
+ }
- num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
- (eb->start >> PAGE_CACHE_SHIFT) + 1;
- for (i = 0; i < num_pages; i++) {
- page = eb->pages[i];
+ num_pages = num_extent_pages(eb->start, eb->len);
+ for (i = start_i; i < num_pages; i++) {
+ page = extent_buffer_page(eb, i);
if (PageUptodate(page)) {
continue;
}
return ret;
}
- for (i = 0; i < num_pages; i++) {
- page = eb->pages[i];
+ for (i = start_i; i < num_pages; i++) {
+ page = extent_buffer_page(eb, i);
wait_on_page_locked(page);
if (!PageUptodate(page)) {
ret = -EIO;
}
}
- eb->flags |= EXTENT_UPTODATE;
+ if (!ret)
+ eb->flags |= EXTENT_UPTODATE;
return ret;
}
EXPORT_SYMBOL(read_extent_buffer_pages);
char *dst = (char *)dstv;
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+ unsigned long num_pages = num_extent_pages(eb->start, eb->len);
WARN_ON(start > eb->len);
WARN_ON(start + len > eb->start + eb->len);
- page = eb->pages[i];
- offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
- if (i == 0)
- offset += start_offset;
+ offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
while(len > 0) {
+ page = extent_buffer_page(eb, i);
+ if (!PageUptodate(page)) {
+ printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len);
+ WARN_ON(1);
+ }
WARN_ON(!PageUptodate(page));
cur = min(len, (PAGE_CACHE_SIZE - offset));
- // kaddr = kmap_atomic(page, KM_USER0);
- kaddr = page_address(page);
+ kaddr = kmap_atomic(page, KM_USER1);
memcpy(dst, kaddr + offset, cur);
- // kunmap_atomic(kaddr, KM_USER0);
+ kunmap_atomic(kaddr, KM_USER1);
dst += cur;
len -= cur;
offset = 0;
i++;
- page = eb->pages[i];
}
}
EXPORT_SYMBOL(read_extent_buffer);
-int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
- unsigned long min_len,
- char **token, char **map,
- unsigned long *map_start,
- unsigned long *map_len, int km)
+int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
+ unsigned long min_len, char **token, char **map,
+ unsigned long *map_start,
+ unsigned long *map_len, int km)
{
size_t offset = start & (PAGE_CACHE_SIZE - 1);
char *kaddr;
+ struct page *p;
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
- unsigned long end_i = (start_offset + start + min_len) >>
- PAGE_CACHE_SHIFT;
+ unsigned long end_i = (start_offset + start + min_len - 1) >>
+ PAGE_CACHE_SHIFT;
if (i != end_i)
return -EINVAL;
- WARN_ON(start > eb->len);
-
if (i == 0) {
offset = start_offset;
*map_start = 0;
} else {
+ offset = 0;
*map_start = (i << PAGE_CACHE_SHIFT) - start_offset;
}
+ if (start + min_len > eb->len) {
+printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len);
+ WARN_ON(1);
+ }
- // kaddr = kmap_atomic(eb->pages[i], km);
- kaddr = page_address(eb->pages[i]);
+ p = extent_buffer_page(eb, i);
+ WARN_ON(!PageUptodate(p));
+ kaddr = kmap_atomic(p, km);
*token = kaddr;
*map = kaddr + offset;
*map_len = PAGE_CACHE_SIZE - offset;
return 0;
}
+EXPORT_SYMBOL(map_private_extent_buffer);
+
+int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+ unsigned long min_len,
+ char **token, char **map,
+ unsigned long *map_start,
+ unsigned long *map_len, int km)
+{
+ int err;
+ int save = 0;
+ if (eb->map_token) {
+ unmap_extent_buffer(eb, eb->map_token, km);
+ eb->map_token = NULL;
+ save = 1;
+ }
+ err = map_private_extent_buffer(eb, start, min_len, token, map,
+ map_start, map_len, km);
+ if (!err && save) {
+ eb->map_token = *token;
+ eb->kaddr = *map;
+ eb->map_start = *map_start;
+ eb->map_len = *map_len;
+ }
+ return err;
+}
EXPORT_SYMBOL(map_extent_buffer);
void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
{
- // kunmap_atomic(token, km);
+ kunmap_atomic(token, km);
}
EXPORT_SYMBOL(unmap_extent_buffer);
WARN_ON(start > eb->len);
WARN_ON(start + len > eb->start + eb->len);
- page = eb->pages[i];
- offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
- if (i == 0)
- offset += start_offset;
+ offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
while(len > 0) {
+ page = extent_buffer_page(eb, i);
WARN_ON(!PageUptodate(page));
cur = min(len, (PAGE_CACHE_SIZE - offset));
- // kaddr = kmap_atomic(page, KM_USER0);
- kaddr = page_address(page);
+ kaddr = kmap_atomic(page, KM_USER0);
ret = memcmp(ptr, kaddr + offset, cur);
- // kunmap_atomic(kaddr, KM_USER0);
+ kunmap_atomic(kaddr, KM_USER0);
if (ret)
break;
len -= cur;
offset = 0;
i++;
- page = eb->pages[i];
}
return ret;
}
WARN_ON(start > eb->len);
WARN_ON(start + len > eb->start + eb->len);
- page = eb->pages[i];
- offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
- if (i == 0)
- offset += start_offset;
+ offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
while(len > 0) {
+ page = extent_buffer_page(eb, i);
WARN_ON(!PageUptodate(page));
cur = min(len, PAGE_CACHE_SIZE - offset);
- // kaddr = kmap_atomic(page, KM_USER0);
- kaddr = page_address(page);
+ kaddr = kmap_atomic(page, KM_USER1);
memcpy(kaddr + offset, src, cur);
- // kunmap_atomic(kaddr, KM_USER0);
+ kunmap_atomic(kaddr, KM_USER1);
src += cur;
len -= cur;
offset = 0;
i++;
- page = eb->pages[i];
}
}
EXPORT_SYMBOL(write_extent_buffer);
WARN_ON(start > eb->len);
WARN_ON(start + len > eb->start + eb->len);
- page = eb->pages[i];
- offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
- if (i == 0)
- offset += start_offset;
+ offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
while(len > 0) {
+ page = extent_buffer_page(eb, i);
WARN_ON(!PageUptodate(page));
cur = min(len, PAGE_CACHE_SIZE - offset);
- // kaddr = kmap_atomic(page, KM_USER0);
- kaddr = page_address(page);
+ kaddr = kmap_atomic(page, KM_USER0);
memset(kaddr + offset, c, cur);
- // kunmap_atomic(kaddr, KM_USER0);
+ kunmap_atomic(kaddr, KM_USER0);
len -= cur;
offset = 0;
i++;
- page = eb->pages[i];
}
}
EXPORT_SYMBOL(memset_extent_buffer);
WARN_ON(src->len != dst_len);
- offset = dst_offset & ((unsigned long)PAGE_CACHE_SIZE - 1);
- if (i == 0)
- offset += start_offset;
+ offset = (start_offset + dst_offset) &
+ ((unsigned long)PAGE_CACHE_SIZE - 1);
while(len > 0) {
- page = dst->pages[i];
+ page = extent_buffer_page(dst, i);
WARN_ON(!PageUptodate(page));
cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
- // kaddr = kmap_atomic(page, KM_USER1);
- kaddr = page_address(page);
+ kaddr = kmap_atomic(page, KM_USER0);
read_extent_buffer(src, kaddr + offset, src_offset, cur);
- // kunmap_atomic(kaddr, KM_USER1);
+ kunmap_atomic(kaddr, KM_USER0);
src_offset += cur;
len -= cur;
unsigned long dst_off, unsigned long src_off,
unsigned long len)
{
- // char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
- char *dst_kaddr = page_address(dst_page);
+ char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
if (dst_page == src_page) {
memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
} else {
- // char *src_kaddr = kmap_atomic(src_page, KM_USER1);
- char *src_kaddr = page_address(src_page);
+ char *src_kaddr = kmap_atomic(src_page, KM_USER1);
char *p = dst_kaddr + dst_off + len;
char *s = src_kaddr + src_off + len;
while (len--)
*--p = *--s;
- // kunmap_atomic(src_kaddr, KM_USER1);
+ kunmap_atomic(src_kaddr, KM_USER1);
}
- // kunmap_atomic(dst_kaddr, KM_USER0);
+ kunmap_atomic(dst_kaddr, KM_USER0);
}
static void copy_pages(struct page *dst_page, struct page *src_page,
unsigned long dst_off, unsigned long src_off,
unsigned long len)
{
- //kmap_atomic(dst_page, KM_USER0);
- char *dst_kaddr = page_address(dst_page);
+ char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
char *src_kaddr;
if (dst_page != src_page)
- src_kaddr = page_address(src_page); // kmap_atomic(src_page, KM_USER1);
+ src_kaddr = kmap_atomic(src_page, KM_USER1);
else
src_kaddr = dst_kaddr;
memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
- /*
kunmap_atomic(dst_kaddr, KM_USER0);
if (dst_page != src_page)
kunmap_atomic(src_kaddr, KM_USER1);
- */
}
void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
}
while(len > 0) {
- dst_off_in_page = dst_offset &
+ dst_off_in_page = (start_offset + dst_offset) &
((unsigned long)PAGE_CACHE_SIZE - 1);
- src_off_in_page = src_offset &
+ src_off_in_page = (start_offset + src_offset) &
((unsigned long)PAGE_CACHE_SIZE - 1);
dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
- if (src_i == 0)
- src_off_in_page += start_offset;
- if (dst_i == 0)
- dst_off_in_page += start_offset;
-
cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
src_off_in_page));
- cur = min(cur, (unsigned long)(PAGE_CACHE_SIZE -
- dst_off_in_page));
+ cur = min_t(unsigned long, cur,
+ (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page));
- copy_pages(dst->pages[dst_i], dst->pages[src_i],
+ copy_pages(extent_buffer_page(dst, dst_i),
+ extent_buffer_page(dst, src_i),
dst_off_in_page, src_off_in_page, cur);
src_offset += cur;
dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
- dst_off_in_page = dst_end &
+ dst_off_in_page = (start_offset + dst_end) &
((unsigned long)PAGE_CACHE_SIZE - 1);
- src_off_in_page = src_end &
+ src_off_in_page = (start_offset + src_end) &
((unsigned long)PAGE_CACHE_SIZE - 1);
- if (src_i == 0)
- src_off_in_page += start_offset;
- if (dst_i == 0)
- dst_off_in_page += start_offset;
-
- cur = min(len, src_off_in_page + 1);
+ cur = min_t(unsigned long, len, src_off_in_page + 1);
cur = min(cur, dst_off_in_page + 1);
-// printk("move pages orig dst %lu src %lu len %lu, this %lu %lu %lu\n", dst_offset, src_offset, len, dst_off_in_page - cur + 1, src_off_in_page - cur + 1, cur);
- move_pages(dst->pages[dst_i], dst->pages[src_i],
+ move_pages(extent_buffer_page(dst, dst_i),
+ extent_buffer_page(dst, src_i),
dst_off_in_page - cur + 1,
src_off_in_page - cur + 1, cur);
- dst_end -= cur - 1;
- src_end -= cur - 1;
+ dst_end -= cur;
+ src_end -= cur;
len -= cur;
}
}