diff options
author | Martin J. Bligh <mbligh@mbligh.org> | 2007-10-16 23:30:46 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@woody.linux-foundation.org> | 2007-10-17 08:43:02 -0700 |
commit | a686cd898bd999fd026a51e90fb0a3410d258ddb (patch) | |
tree | 3f0a91a70fe1a3addf9e99f3babed9e9baba4b99 /fs/ext2/balloc.c | |
parent | 369f2389e7d03022abdd25e298bffb9613cd0e54 (diff) | |
download | kernel_goldelico_gta04-a686cd898bd999fd026a51e90fb0a3410d258ddb.zip kernel_goldelico_gta04-a686cd898bd999fd026a51e90fb0a3410d258ddb.tar.gz kernel_goldelico_gta04-a686cd898bd999fd026a51e90fb0a3410d258ddb.tar.bz2 |
ext2 reservations
Val's cross-port of the ext3 reservations code into ext2.
[mbligh@mbligh.org: Small type error for printk
[akpm@linux-foundation.org: fix types, sync with ext3]
[mbligh@mbligh.org: Bring ext2 reservations code in line with latest ext3]
[akpm@linux-foundation.org: kill noisy printk]
[akpm@linux-foundation.org: remember to dirty the gdp's block]
[akpm@linux-foundation.org: cross-port the missed 5dea5176e5c32ef9f0d1a41d28427b3bf6881b3a]
[akpm@linux-foundation.org: cross-port e6022603b9aa7d61d20b392e69edcdbbc1789969]
[akpm@linux-foundation.org: Port the omitted 08fb306fe63d98eb86e3b16f4cc21816fa47f18e]
[akpm@linux-foundation.org: Backport the missed 20acaa18d0c002fec180956f87adeb3f11f635a6]
[akpm@linux-foundation.org: fixes]
[cmm@us.ibm.com: fix reservation extension]
[bunk@stusta.de: make ext2_get_blocks() static]
[hugh@veritas.com: fix hang]
[hugh@veritas.com: ext2_new_blocks should reset the reservation window size]
[hugh@veritas.com: ext2 balloc: fix off-by-one against rsv_end]
[hugh@veritas.com: grp_goal 0 is a genuine goal (unlike -1), so ext2_try_to_allocate_with_rsv should treat it as such]
[hugh@veritas.com: rbtree usage cleanup]
[pbadari@us.ibm.com: Fix for ext2 reservation]
[bunk@kernel.org: remove fs/ext2/balloc.c:reserve_blocks()]
[hugh@veritas.com: ext2 balloc: use io_error label]
Cc: "Martin J. Bligh" <mbligh@mbligh.org>
Cc: Valerie Henson <val_henson@linux.intel.com>
Cc: Mingming Cao <cmm@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'fs/ext2/balloc.c')
-rw-r--r-- | fs/ext2/balloc.c | 1305 |
1 files changed, 1072 insertions, 233 deletions
diff --git a/fs/ext2/balloc.c b/fs/ext2/balloc.c index ffaa6d8..18a42de 100644 --- a/fs/ext2/balloc.c +++ b/fs/ext2/balloc.c @@ -133,41 +133,6 @@ error_out: return NULL; } -/* - * Set sb->s_dirt here because the superblock was "logically" altered. We - * need to recalculate its free blocks count and flush it out. - */ -static int reserve_blocks(struct super_block *sb, int count) -{ - struct ext2_sb_info *sbi = EXT2_SB(sb); - struct ext2_super_block *es = sbi->s_es; - unsigned free_blocks; - unsigned root_blocks; - - free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); - root_blocks = le32_to_cpu(es->s_r_blocks_count); - - if (free_blocks < count) - count = free_blocks; - - if (free_blocks < root_blocks + count && !capable(CAP_SYS_RESOURCE) && - sbi->s_resuid != current->fsuid && - (sbi->s_resgid == 0 || !in_group_p (sbi->s_resgid))) { - /* - * We are too close to reserve and we are not privileged. - * Can we allocate anything at all? - */ - if (free_blocks > root_blocks) - count = free_blocks - root_blocks; - else - return 0; - } - - percpu_counter_sub(&sbi->s_freeblocks_counter, count); - sb->s_dirt = 1; - return count; -} - static void release_blocks(struct super_block *sb, int count) { if (count) { @@ -178,25 +143,7 @@ static void release_blocks(struct super_block *sb, int count) } } -static int group_reserve_blocks(struct ext2_sb_info *sbi, int group_no, - struct ext2_group_desc *desc, struct buffer_head *bh, int count) -{ - unsigned free_blocks; - - if (!desc->bg_free_blocks_count) - return 0; - - spin_lock(sb_bgl_lock(sbi, group_no)); - free_blocks = le16_to_cpu(desc->bg_free_blocks_count); - if (free_blocks < count) - count = free_blocks; - desc->bg_free_blocks_count = cpu_to_le16(free_blocks - count); - spin_unlock(sb_bgl_lock(sbi, group_no)); - mark_buffer_dirty(bh); - return count; -} - -static void group_release_blocks(struct super_block *sb, int group_no, +static void group_adjust_blocks(struct super_block *sb, int group_no, struct ext2_group_desc *desc, struct buffer_head *bh, int count) { if (count) { @@ -212,7 +159,306 @@ static void group_release_blocks(struct super_block *sb, int group_no, } } -/* Free given blocks, update quota and i_blocks field */ +/* + * The reservation window structure operations + * -------------------------------------------- + * Operations include: + * dump, find, add, remove, is_empty, find_next_reservable_window, etc. + * + * We use a red-black tree to represent per-filesystem reservation + * windows. + * + */ + +/** + * __rsv_window_dump() -- Dump the filesystem block allocation reservation map + * @rb_root: root of per-filesystem reservation rb tree + * @verbose: verbose mode + * @fn: function which wishes to dump the reservation map + * + * If verbose is turned on, it will print the whole block reservation + * windows(start, end). Otherwise, it will only print out the "bad" windows, + * those windows that overlap with their immediate neighbors. + */ +#if 1 +static void __rsv_window_dump(struct rb_root *root, int verbose, + const char *fn) +{ + struct rb_node *n; + struct ext2_reserve_window_node *rsv, *prev; + int bad; + +restart: + n = rb_first(root); + bad = 0; + prev = NULL; + + printk("Block Allocation Reservation Windows Map (%s):\n", fn); + while (n) { + rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node); + if (verbose) + printk("reservation window 0x%p " + "start: %lu, end: %lu\n", + rsv, rsv->rsv_start, rsv->rsv_end); + if (rsv->rsv_start && rsv->rsv_start >= rsv->rsv_end) { + printk("Bad reservation %p (start >= end)\n", + rsv); + bad = 1; + } + if (prev && prev->rsv_end >= rsv->rsv_start) { + printk("Bad reservation %p (prev->end >= start)\n", + rsv); + bad = 1; + } + if (bad) { + if (!verbose) { + printk("Restarting reservation walk in verbose mode\n"); + verbose = 1; + goto restart; + } + } + n = rb_next(n); + prev = rsv; + } + printk("Window map complete.\n"); + if (bad) + BUG(); +} +#define rsv_window_dump(root, verbose) \ + __rsv_window_dump((root), (verbose), __FUNCTION__) +#else +#define rsv_window_dump(root, verbose) do {} while (0) +#endif + +/** + * goal_in_my_reservation() + * @rsv: inode's reservation window + * @grp_goal: given goal block relative to the allocation block group + * @group: the current allocation block group + * @sb: filesystem super block + * + * Test if the given goal block (group relative) is within the file's + * own block reservation window range. + * + * If the reservation window is outside the goal allocation group, return 0; + * grp_goal (given goal block) could be -1, which means no specific + * goal block. In this case, always return 1. + * If the goal block is within the reservation window, return 1; + * otherwise, return 0; + */ +static int +goal_in_my_reservation(struct ext2_reserve_window *rsv, ext2_grpblk_t grp_goal, + unsigned int group, struct super_block * sb) +{ + ext2_fsblk_t group_first_block, group_last_block; + + group_first_block = ext2_group_first_block_no(sb, group); + group_last_block = group_first_block + EXT2_BLOCKS_PER_GROUP(sb) - 1; + + if ((rsv->_rsv_start > group_last_block) || + (rsv->_rsv_end < group_first_block)) + return 0; + if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start) + || (grp_goal + group_first_block > rsv->_rsv_end))) + return 0; + return 1; +} + +/** + * search_reserve_window() + * @rb_root: root of reservation tree + * @goal: target allocation block + * + * Find the reserved window which includes the goal, or the previous one + * if the goal is not in any window. + * Returns NULL if there are no windows or if all windows start after the goal. + */ +static struct ext2_reserve_window_node * +search_reserve_window(struct rb_root *root, ext2_fsblk_t goal) +{ + struct rb_node *n = root->rb_node; + struct ext2_reserve_window_node *rsv; + + if (!n) + return NULL; + + do { + rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node); + + if (goal < rsv->rsv_start) + n = n->rb_left; + else if (goal > rsv->rsv_end) + n = n->rb_right; + else + return rsv; + } while (n); + /* + * We've fallen off the end of the tree: the goal wasn't inside + * any particular node. OK, the previous node must be to one + * side of the interval containing the goal. If it's the RHS, + * we need to back up one. + */ + if (rsv->rsv_start > goal) { + n = rb_prev(&rsv->rsv_node); + rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node); + } + return rsv; +} + +/* + * ext2_rsv_window_add() -- Insert a window to the block reservation rb tree. + * @sb: super block + * @rsv: reservation window to add + * + * Must be called with rsv_lock held. + */ +void ext2_rsv_window_add(struct super_block *sb, + struct ext2_reserve_window_node *rsv) +{ + struct rb_root *root = &EXT2_SB(sb)->s_rsv_window_root; + struct rb_node *node = &rsv->rsv_node; + ext2_fsblk_t start = rsv->rsv_start; + + struct rb_node ** p = &root->rb_node; + struct rb_node * parent = NULL; + struct ext2_reserve_window_node *this; + + while (*p) + { + parent = *p; + this = rb_entry(parent, struct ext2_reserve_window_node, rsv_node); + + if (start < this->rsv_start) + p = &(*p)->rb_left; + else if (start > this->rsv_end) + p = &(*p)->rb_right; + else { + rsv_window_dump(root, 1); + BUG(); + } + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); +} + +/** + * rsv_window_remove() -- unlink a window from the reservation rb tree + * @sb: super block + * @rsv: reservation window to remove + * + * Mark the block reservation window as not allocated, and unlink it + * from the filesystem reservation window rb tree. Must be called with + * rsv_lock held. + */ +static void rsv_window_remove(struct super_block *sb, + struct ext2_reserve_window_node *rsv) +{ + rsv->rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED; + rsv->rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED; + rsv->rsv_alloc_hit = 0; + rb_erase(&rsv->rsv_node, &EXT2_SB(sb)->s_rsv_window_root); +} + +/* + * rsv_is_empty() -- Check if the reservation window is allocated. + * @rsv: given reservation window to check + * + * returns 1 if the end block is EXT2_RESERVE_WINDOW_NOT_ALLOCATED. + */ +static inline int rsv_is_empty(struct ext2_reserve_window *rsv) +{ + /* a valid reservation end block could not be 0 */ + return (rsv->_rsv_end == EXT2_RESERVE_WINDOW_NOT_ALLOCATED); +} + +/** + * ext2_init_block_alloc_info() + * @inode: file inode structure + * + * Allocate and initialize the reservation window structure, and + * link the window to the ext2 inode structure at last + * + * The reservation window structure is only dynamically allocated + * and linked to ext2 inode the first time the open file + * needs a new block. So, before every ext2_new_block(s) call, for + * regular files, we should check whether the reservation window + * structure exists or not. In the latter case, this function is called. + * Fail to do so will result in block reservation being turned off for that + * open file. + * + * This function is called from ext2_get_blocks_handle(), also called + * when setting the reservation window size through ioctl before the file + * is open for write (needs block allocation). + * + * Needs truncate_mutex protection prior to calling this function. + */ +void ext2_init_block_alloc_info(struct inode *inode) +{ + struct ext2_inode_info *ei = EXT2_I(inode); + struct ext2_block_alloc_info *block_i = ei->i_block_alloc_info; + struct super_block *sb = inode->i_sb; + + block_i = kmalloc(sizeof(*block_i), GFP_NOFS); + if (block_i) { + struct ext2_reserve_window_node *rsv = &block_i->rsv_window_node; + + rsv->rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED; + rsv->rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED; + + /* + * if filesystem is mounted with NORESERVATION, the goal + * reservation window size is set to zero to indicate + * block reservation is off + */ + if (!test_opt(sb, RESERVATION)) + rsv->rsv_goal_size = 0; + else + rsv->rsv_goal_size = EXT2_DEFAULT_RESERVE_BLOCKS; + rsv->rsv_alloc_hit = 0; + block_i->last_alloc_logical_block = 0; + block_i->last_alloc_physical_block = 0; + } + ei->i_block_alloc_info = block_i; +} + +/** + * ext2_discard_reservation() + * @inode: inode + * + * Discard(free) block reservation window on last file close, or truncate + * or at last iput(). + * + * It is being called in three cases: + * ext2_release_file(): last writer closes the file + * ext2_clear_inode(): last iput(), when nobody links to this file. + * ext2_truncate(): when the block indirect map is about to change. + */ +void ext2_discard_reservation(struct inode *inode) +{ + struct ext2_inode_info *ei = EXT2_I(inode); + struct ext2_block_alloc_info *block_i = ei->i_block_alloc_info; + struct ext2_reserve_window_node *rsv; + spinlock_t *rsv_lock = &EXT2_SB(inode->i_sb)->s_rsv_window_lock; + + if (!block_i) + return; + + rsv = &block_i->rsv_window_node; + if (!rsv_is_empty(&rsv->rsv_window)) { + spin_lock(rsv_lock); + if (!rsv_is_empty(&rsv->rsv_window)) + rsv_window_remove(inode->i_sb, rsv); + spin_unlock(rsv_lock); + } +} + +/** + * ext2_free_blocks_sb() -- Free given blocks and update quota and i_blocks + * @inode: inode + * @block: start physcial block to free + * @count: number of blocks to free + */ void ext2_free_blocks (struct inode * inode, unsigned long block, unsigned long count) { @@ -287,7 +533,7 @@ do_more: if (sb->s_flags & MS_SYNCHRONOUS) sync_dirty_buffer(bitmap_bh); - group_release_blocks(sb, block_group, desc, bh2, group_freed); + group_adjust_blocks(sb, block_group, desc, bh2, group_freed); freed += group_freed; if (overflow) { @@ -301,16 +547,46 @@ error_return: DQUOT_FREE_BLOCK(inode, freed); } -static int grab_block(spinlock_t *lock, char *map, unsigned size, int goal) +/** + * bitmap_search_next_usable_block() + * @start: the starting block (group relative) of the search + * @bh: bufferhead contains the block group bitmap + * @maxblocks: the ending block (group relative) of the reservation + * + * The bitmap search --- search forward through the actual bitmap on disk until + * we find a bit free. + */ +static ext2_grpblk_t +bitmap_search_next_usable_block(ext2_grpblk_t start, struct buffer_head *bh, + ext2_grpblk_t maxblocks) { - int k; - char *p, *r; + ext2_grpblk_t next; - if (!ext2_test_bit(goal, map)) - goto got_it; + next = ext2_find_next_zero_bit(bh->b_data, maxblocks, start); + if (next >= maxblocks) + return -1; + return next; +} -repeat: - if (goal) { +/** + * find_next_usable_block() + * @start: the starting block (group relative) to find next + * allocatable block in bitmap. + * @bh: bufferhead contains the block group bitmap + * @maxblocks: the ending block (group relative) for the search + * + * Find an allocatable block in a bitmap. We perform the "most + * appropriate allocation" algorithm of looking for a free block near + * the initial goal; then for a free byte somewhere in the bitmap; + * then for any free bit in the bitmap. + */ +static ext2_grpblk_t +find_next_usable_block(int start, struct buffer_head *bh, int maxblocks) +{ + ext2_grpblk_t here, next; + char *p, *r; + + if (start > 0) { /* * The goal was occupied; search forward for a free * block within the next XX blocks. @@ -319,244 +595,807 @@ repeat: * less than EXT2_BLOCKS_PER_GROUP. Aligning up to the * next 64-bit boundary is simple.. */ - k = (goal + 63) & ~63; - goal = ext2_find_next_zero_bit(map, k, goal); - if (goal < k) - goto got_it; + ext2_grpblk_t end_goal = (start + 63) & ~63; + if (end_goal > maxblocks) + end_goal = maxblocks; + here = ext2_find_next_zero_bit(bh->b_data, end_goal, start); + if (here < end_goal) + return here; + ext2_debug("Bit not found near goal\n"); + } + + here = start; + if (here < 0) + here = 0; + + p = ((char *)bh->b_data) + (here >> 3); + r = memscan(p, 0, ((maxblocks + 7) >> 3) - (here >> 3)); + next = (r - ((char *)bh->b_data)) << 3; + + if (next < maxblocks && next >= here) + return next; + + here = bitmap_search_next_usable_block(here, bh, maxblocks); + return here; +} + +/* + * ext2_try_to_allocate() + * @sb: superblock + * @handle: handle to this transaction + * @group: given allocation block group + * @bitmap_bh: bufferhead holds the block bitmap + * @grp_goal: given target block within the group + * @count: target number of blocks to allocate + * @my_rsv: reservation window + * + * Attempt to allocate blocks within a give range. Set the range of allocation + * first, then find the first free bit(s) from the bitmap (within the range), + * and at last, allocate the blocks by claiming the found free bit as allocated. + * + * To set the range of this allocation: + * if there is a reservation window, only try to allocate block(s) + * from the file's own reservation window; + * Otherwise, the allocation range starts from the give goal block, + * ends at the block group's last block. + * + * If we failed to allocate the desired block then we may end up crossing to a + * new bitmap. + */ +static int +ext2_try_to_allocate(struct super_block *sb, int group, + struct buffer_head *bitmap_bh, ext2_grpblk_t grp_goal, + unsigned long *count, + struct ext2_reserve_window *my_rsv) +{ + ext2_fsblk_t group_first_block; + ext2_grpblk_t start, end; + unsigned long num = 0; + + /* we do allocation within the reservation window if we have a window */ + if (my_rsv) { + group_first_block = ext2_group_first_block_no(sb, group); + if (my_rsv->_rsv_start >= group_first_block) + start = my_rsv->_rsv_start - group_first_block; + else + /* reservation window cross group boundary */ + start = 0; + end = my_rsv->_rsv_end - group_first_block + 1; + if (end > EXT2_BLOCKS_PER_GROUP(sb)) + /* reservation window crosses group boundary */ + end = EXT2_BLOCKS_PER_GROUP(sb); + if ((start <= grp_goal) && (grp_goal < end)) + start = grp_goal; + else + grp_goal = -1; + } else { + if (grp_goal > 0) + start = grp_goal; + else + start = 0; + end = EXT2_BLOCKS_PER_GROUP(sb); + } + + BUG_ON(start > EXT2_BLOCKS_PER_GROUP(sb)); + +repeat: + if (grp_goal < 0) { + grp_goal = find_next_usable_block(start, bitmap_bh, end); + if (grp_goal < 0) + goto fail_access; + if (!my_rsv) { + int i; + + for (i = 0; i < 7 && grp_goal > start && + !ext2_test_bit(grp_goal - 1, + bitmap_bh->b_data); + i++, grp_goal--) + ; + } + } + start = grp_goal; + + if (ext2_set_bit_atomic(sb_bgl_lock(EXT2_SB(sb), group), grp_goal, + bitmap_bh->b_data)) { + /* + * The block was allocated by another thread, or it was + * allocated and then freed by another thread + */ + start++; + grp_goal++; + if (start >= end) + goto fail_access; + goto repeat; + } + num++; + grp_goal++; + while (num < *count && grp_goal < end + && !ext2_set_bit_atomic(sb_bgl_lock(EXT2_SB(sb), group), + grp_goal, bitmap_bh->b_data)) { + num++; + grp_goal++; + } + *count = num; + return grp_goal - num; +fail_access: + *count = num; + return -1; +} + +/** + * find_next_reservable_window(): + * find a reservable space within the given range. + * It does not allocate the reservation window for now: + * alloc_new_reservation() will do the work later. + * + * @search_head: the head of the searching list; + * This is not necessarily the list head of the whole filesystem + * + * We have both head and start_block to assist the search + * for the reservable space. The list starts from head, + * but we will shift to the place where start_block is, + * then start from there, when looking for a reservable space. + * + * @size: the target new reservation window size + * + * @group_first_block: the first block we consider to start + * the real search from + * + * @last_block: + * the maximum block number that our goal reservable space + * could start from. This is normally the last block in this + * group. The search will end when we found the start of next + * possible reservable space is out of this boundary. + * This could handle the cross boundary reservation window + * request. + * + * basically we search from the given range, rather than the whole + * reservation double linked list, (start_block, last_block) + * to find a free region that is of my size and has not + * been reserved. + * + */ +static int find_next_reservable_window( + struct ext2_reserve_window_node *search_head, + struct ext2_reserve_window_node *my_rsv, + struct super_block * sb, + ext2_fsblk_t start_block, + ext2_fsblk_t last_block) +{ + struct rb_node *next; + struct ext2_reserve_window_node *rsv, *prev; + ext2_fsblk_t cur; + int size = my_rsv->rsv_goal_size; + + /* TODO: make the start of the reservation window byte-aligned */ + /* cur = *start_block & ~7;*/ + cur = start_block; + rsv = search_head; + if (!rsv) + return -1; + + while (1) { + if (cur <= rsv->rsv_end) + cur = rsv->rsv_end + 1; + + /* TODO? + * in the case we could not find a reservable space + * that is what is expected, during the re-search, we could + * remember what's the largest reservable space we could have + * and return that one. + * + * For now it will fail if we could not find the reservable + * space with expected-size (or more)... + */ + if (cur > last_block) + return -1; /* fail */ + + prev = rsv; + next = rb_next(&rsv->rsv_node); + rsv = rb_entry(next,struct ext2_reserve_window_node,rsv_node); + /* - * Search in the remainder of the current group. + * Reached the last reservation, we can just append to the + * previous one. */ + if (!next) + break; + + if (cur + size <= rsv->rsv_start) { + /* + * Found a reserveable space big enough. We could + * have a reservation across the group boundary here + */ + break; + } } + /* + * we come here either : + * when we reach the end of the whole list, + * and there is empty reservable space after last entry in the list. + * append it to the end of the list. + * + * or we found one reservable space in the middle of the list, + * return the reservation window that we could append to. + * succeed. + */ - p = map + (goal >> 3); - r = memscan(p, 0, (size - goal + 7) >> 3); - k = (r - map) << 3; - if (k < size) { - /* - * We have succeeded in finding a free byte in the block - * bitmap. Now search backwards to find the start of this - * group of free blocks - won't take more than 7 iterations. + if ((prev != my_rsv) && (!rsv_is_empty(&my_rsv->rsv_window))) + rsv_window_remove(sb, my_rsv); + + /* + * Let's book the whole avaliable window for now. We will check the + * disk bitmap later and then, if there are free blocks then we adjust + * the window size if it's larger than requested. + * Otherwise, we will remove this node from the tree next time + * call find_next_reservable_window. + */ + my_rsv->rsv_start = cur; + my_rsv->rsv_end = cur + size - 1; + my_rsv->rsv_alloc_hit = 0; + + if (prev != my_rsv) + ext2_rsv_window_add(sb, my_rsv); + + return 0; +} + +/** + * alloc_new_reservation()--allocate a new reservation window + * + * To make a new reservation, we search part of the filesystem + * reservation list (the list that inside the group). We try to + * allocate a new reservation window near the allocation goal, + * or the beginning of the group, if there is no goal. + * + * We first find a reservable space after the goal, then from + * there, we check the bitmap for the first free block after + * it. If there is no free block until the end of group, then the + * whole group is full, we failed. Otherwise, check if the free + * block is inside the expected reservable space, if so, we + * succeed. + * If the first free block is outside the reservable space, then + * start from the first free block, we search for next available + * space, and go on. + * + * on succeed, a new reservation will be found and inserted into the list + * It contains at least one free block, and it does not overlap with other + * reservation windows. + * + * failed: we failed to find a reservation window in this group + * + * @rsv: the reservation + * + * @grp_goal: The goal (group-relative). It is where the search for a + * free reservable space should start from. + * if we have a goal(goal >0 ), then start from there, + * no goal(goal = -1), we start from the first block + * of the group. + * + * @sb: the super block + * @group: the group we are trying to allocate in + * @bitmap_bh: the block group block bitmap + * + */ +static int alloc_new_reservation(struct ext2_reserve_window_node *my_rsv, + ext2_grpblk_t grp_goal, struct super_block *sb, + unsigned int group, struct buffer_head *bitmap_bh) +{ + struct ext2_reserve_window_node *search_head; + ext2_fsblk_t group_first_block, group_end_block, start_block; + ext2_grpblk_t first_free_block; + struct rb_root *fs_rsv_root = &EXT2_SB(sb)->s_rsv_window_root; + unsigned long size; + int ret; + spinlock_t *rsv_lock = &EXT2_SB(sb)->s_rsv_window_lock; + + group_first_block = ext2_group_first_block_no(sb, group); + group_end_block = group_first_block + (EXT2_BLOCKS_PER_GROUP(sb) - 1); + + if (grp_goal < 0) + start_block = group_first_block; + else + start_block = grp_goal + group_first_block; + + size = my_rsv->rsv_goal_size; + + if (!rsv_is_empty(&my_rsv->rsv_window)) { + /* + * if the old reservation is cross group boundary + * and if the goal is inside the old reservation window, + * we will come here when we just failed to allocate from + * the first part of the window. We still have another part + * that belongs to the next group. In this case, there is no + * point to discard our window and try to allocate a new one + * in this group(which will fail). we should + * keep the reservation window, just simply move on. + * + * Maybe we could shift the start block of the reservation + * window to the first block of next group. */ - for (goal = k; goal && !ext2_test_bit (goal - 1, map); goal--) - ; - goto got_it; + + if ((my_rsv->rsv_start <= group_end_block) && + (my_rsv->rsv_end > group_end_block) && + (start_block >= my_rsv->rsv_start)) + return -1; + + if ((my_rsv->rsv_alloc_hit > + (my_rsv->rsv_end - my_rsv->rsv_start + 1) / 2)) { + /* + * if the previously allocation hit ratio is + * greater than 1/2, then we double the size of + * the reservation window the next time, + * otherwise we keep the same size window + */ + size = size * 2; + if (size > EXT2_MAX_RESERVE_BLOCKS) + size = EXT2_MAX_RESERVE_BLOCKS; + my_rsv->rsv_goal_size= size; + } } - k = ext2_find_next_zero_bit ((u32 *)map, size, goal); - if (k < size) { - goal = k; - goto got_it; + spin_lock(rsv_lock); + /* + * shift the search start to the window near the goal block + */ + search_head = search_reserve_window(fs_rsv_root, start_block); + + /* + * find_next_reservable_window() simply finds a reservable window + * inside the given range(start_block, group_end_block). + * + * To make sure the reservation window has a free bit inside it, we + * need to check the bitmap after we found a reservable window. + */ +retry: + ret = find_next_reservable_window(search_head, my_rsv, sb, + start_block, group_end_block); + + if (ret == -1) { + if (!rsv_is_empty(&my_rsv->rsv_window)) + rsv_window_remove(sb, my_rsv); + spin_unlock(rsv_lock); + return -1; } - return -1; -got_it: - if (ext2_set_bit_atomic(lock, goal, (void *) map)) - goto repeat; - return goal; + + /* + * On success, find_next_reservable_window() returns the + * reservation window where there is a reservable space after it. + * Before we reserve this reservable space, we need + * to make sure there is at least a free block inside this region. + * + * Search the first free bit on the block bitmap. Search starts from + * the start block of the reservable space we just found. + */ + spin_unlock(rsv_lock); + first_free_block = bitmap_search_next_usable_block( + my_rsv->rsv_start - group_first_block, + bitmap_bh, group_end_block - group_first_block + 1); + + if (first_free_block < 0) { + /* + * no free block left on the bitmap, no point + * to reserve the space. return failed. + */ + spin_lock(rsv_lock); + if (!rsv_is_empty(&my_rsv->rsv_window)) + rsv_window_remove(sb, my_rsv); + spin_unlock(rsv_lock); + return -1; /* failed */ + } + + start_block = first_free_block + group_first_block; + /* + * check if the first free block is within the + * free space we just reserved + */ + if (start_block >= my_rsv->rsv_start && start_block <= my_rsv->rsv_end) + return 0; /* success */ + /* + * if the first free bit we found is out of the reservable space + * continue search for next reservable space, + * start from where the free block is, + * we also shift the list head to where we stopped last time + */ + search_head = my_rsv; + spin_lock(rsv_lock); + goto retry; +} + +/** + * try_to_extend_reservation() + * @my_rsv: given reservation window + * @sb: super block + * @size: the delta to extend + * + * Attempt to expand the reservation window large enough to have + * required number of free blocks + * + * Since ext2_try_to_allocate() will always allocate blocks within + * the reservation window range, if the window size is too small, + * multiple blocks allocation has to stop at the end of the reservation + * window. To make this more efficient, given the total number of + * blocks needed and the current size of the window, we try to + * expand the reservation window size if necessary on a best-effort + * basis before ext2_new_blocks() tries to allocate blocks. + */ +static void try_to_extend_reservation(struct ext2_reserve_window_node *my_rsv, + struct super_block *sb, int size) +{ + struct ext2_reserve_window_node *next_rsv; + struct rb_node *next; + spinlock_t *rsv_lock = &EXT2_SB(sb)->s_rsv_window_lock; + + if (!spin_trylock(rsv_lock)) + return; + + next = rb_next(&my_rsv->rsv_node); + + if (!next) + my_rsv->rsv_end += size; + else { + next_rsv = rb_entry(next, struct ext2_reserve_window_node, rsv_node); + + if ((next_rsv->rsv_start - my_rsv->rsv_end - 1) >= size) + my_rsv->rsv_end += size; + else + my_rsv->rsv_end = next_rsv->rsv_start - 1; + } + spin_unlock(rsv_lock); +} + +/** + * ext2_try_to_allocate_with_rsv() + * @sb: superblock + * @group: given allocation block group + * @bitmap_bh: bufferhead holds the block bitmap + * @grp_goal: given target block within the group + * @count: target number of blocks to allocate + * @my_rsv: reservation window + * + * This is the main function used to allocate a new block and its reservation + * window. + * + * Each time when a new block allocation is need, first try to allocate from + * its own reservation. If it does not have a reservation window, instead of + * looking for a free bit on bitmap first, then look up the reservation list to + * see if it is inside somebody else's reservation window, we try to allocate a + * reservation window for it starting from the goal first. Then do the block + * allocation within the reservation window. + * + * This will avoid keeping on searching the reservation list again and + * again when somebody is looking for a free block (without + * reservation), and there are lots of free blocks, but they are all + * being reserved. + * + * We use a red-black tree for the per-filesystem reservation list. + */ +static ext2_grpblk_t +ext2_try_to_allocate_with_rsv(struct super_block *sb, unsigned int group, + struct buffer_head *bitmap_bh, ext2_grpblk_t grp_goal, + struct ext2_reserve_window_node * my_rsv, + unsigned long *count) +{ + ext2_fsblk_t group_first_block, group_last_block; + ext2_grpblk_t ret = 0; + unsigned long num = *count; + + /* + * we don't deal with reservation when + * filesystem is mounted without reservation + * or the file is not a regular file + * or last attempt to allocate a block with reservation turned on failed + */ + if (my_rsv == NULL) { + return ext2_try_to_allocate(sb, group, bitmap_bh, + grp_goal, count, NULL); + } + /* + * grp_goal is a group relative block number (if there is a goal) + * 0 <= grp_goal < EXT2_BLOCKS_PER_GROUP(sb) + * first block is a filesystem wide block number + * first block is the block number of the first block in this group + */ + group_first_block = ext2_group_first_block_no(sb, group); + group_last_block = group_first_block + (EXT2_BLOCKS_PER_GROUP(sb) - 1); + + /* + * Basically we will allocate a new block from inode's reservation + * window. + * + * We need to allocate a new reservation window, if: + * a) inode does not have a reservation window; or + * b) last attempt to allocate a block from existing reservation + * failed; or + * c) we come here with a goal and with a reservation window + * + * We do not need to allocate a new reservation window if we come here + * at the beginning with a goal and the goal is inside the window, or + * we don't have a goal but already have a reservation window. + * then we could go to allocate from the reservation window directly. + */ + while (1) { + if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) || + !goal_in_my_reservation(&my_rsv->rsv_window, + grp_goal, group, sb)) { + if (my_rsv->rsv_goal_size < *count) + my_rsv->rsv_goal_size = *count; + ret = alloc_new_reservation(my_rsv, grp_goal, sb, + group, bitmap_bh); + if (ret < 0) + break; /* failed */ + + if (!goal_in_my_reservation(&my_rsv->rsv_window, + grp_goal, group, sb)) + grp_goal = -1; + } else if (grp_goal >= 0) { + int curr = my_rsv->rsv_end - + (grp_goal + group_first_block) + 1; + + if (curr < *count) + try_to_extend_reservation(my_rsv, sb, + *count - curr); + } + + if ((my_rsv->rsv_start > group_last_block) || + (my_rsv->rsv_end < group_first_block)) { + rsv_window_dump(&EXT2_SB(sb)->s_rsv_window_root, 1); + BUG(); + } + ret = ext2_try_to_allocate(sb, group, bitmap_bh, grp_goal, + &num, &my_rsv->rsv_window); + if (ret >= 0) { + my_rsv->rsv_alloc_hit += num; + *count = num; + break; /* succeed */ + } + num = *count; + } + return ret; +} + +/** + * ext2_has_free_blocks() + * @sbi: in-core super block structure. + * + * Check if filesystem has at least 1 free block available for allocation. + */ +static int ext2_has_free_blocks(struct ext2_sb_info *sbi) +{ + ext2_fsblk_t free_blocks, root_blocks; + + free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); + root_blocks = le32_to_cpu(sbi->s_es->s_r_blocks_count); + if (free_blocks < root_blocks + 1 && !capable(CAP_SYS_RESOURCE) && + sbi->s_resuid != current->fsuid && + (sbi->s_resgid == 0 || !in_group_p (sbi->s_resgid))) { + return 0; + } + return 1; } /* - * ext2_new_block uses a goal block to assist allocation. If the goal is + * ext2_new_blocks() -- core block(s) allocation function + * @inode: file inode + * @goal: given target block(filesystem wide) + * @count: target number of blocks to allocate + * @errp: error code + * + * ext2_new_blocks uses a goal block to assist allocation. If the goal is * free, or there is a free block within 32 blocks of the goal, that block * is allocated. Otherwise a forward search is made for a free block; within * each block group the search first looks for an entire free byte in the block * bitmap, and then for any free bit if that fails. * This function also updates quota and i_blocks field. */ -int ext2_new_block(struct inode *inode, unsigned long goal, - u32 *prealloc_count, u32 *prealloc_block, int *err) +ext2_fsblk_t ext2_new_blocks(struct inode *inode, ext2_fsblk_t goal, + unsigned long *count, int *errp) { struct buffer_head *bitmap_bh = NULL; - struct buffer_head *gdp_bh; /* bh2 */ - struct ext2_group_desc *desc; - int group_no; /* i */ - int ret_block; /* j */ - int group_idx; /* k */ - int target_block; /* tmp */ - int block = 0; - struct super_block *sb = inode->i_sb; - struct ext2_sb_info *sbi = EXT2_SB(sb); - struct ext2_super_block *es = sbi->s_es; - unsigned group_size = EXT2_BLOCKS_PER_GROUP(sb); - unsigned prealloc_goal = es->s_prealloc_blocks; - unsigned group_alloc = 0, es_alloc, dq_alloc; - int nr_scanned_groups; - - if (!prealloc_goal--) - prealloc_goal = EXT2_DEFAULT_PREALLOC_BLOCKS - 1; - if (!prealloc_count || *prealloc_count) - prealloc_goal = 0; - - if (DQUOT_ALLOC_BLOCK(inode, 1)) { - *err = -EDQUOT; - goto out; + struct buffer_head *gdp_bh; + int group_no; + int goal_group; + ext2_grpblk_t grp_target_blk; /* blockgroup relative goal block */ + ext2_grpblk_t grp_alloc_blk; /* blockgroup-relative allocated block*/ + ext2_fsblk_t ret_block; /* filesyetem-wide allocated block */ + int bgi; /* blockgroup iteration index */ + int performed_allocation = 0; + ext2_grpblk_t free_blocks; /* number of free blocks in a group */ + struct super_block *sb; + struct ext2_group_desc *gdp; + struct ext2_super_block *es; + struct ext2_sb_info *sbi; + struct ext2_reserve_window_node *my_rsv = NULL; + struct ext2_block_alloc_info *block_i; + unsigned short windowsz = 0; + unsigned long ngroups; + unsigned long num = *count; + + *errp = -ENOSPC; + sb = inode->i_sb; + if (!sb) { + printk("ext2_new_blocks: nonexistent device"); + return 0; } - while (prealloc_goal && DQUOT_PREALLOC_BLOCK(inode, prealloc_goal)) - prealloc_goal--; + /* + * Check quota for allocation of this block. + */ + if (DQUOT_ALLOC_BLOCK(inode, num)) { + *errp = -EDQUOT; + return 0; + } - dq_alloc = prealloc_goal + 1; - es_alloc = reserve_blocks(sb, dq_alloc); - if (!es_alloc) { - *err = -ENOSPC; - goto out_dquot; + sbi = EXT2_SB(sb); + es = EXT2_SB(sb)->s_es; + ext2_debug("goal=%lu.\n", goal); + /* + * Allocate a block from reservation only when + * filesystem is mounted with reservation(default,-o reservation), and + * it's a regular file, and + * the desired window size is greater than 0 (One could use ioctl + * command EXT2_IOC_SETRSVSZ to set the window size to 0 to turn off + * reservation on that particular file) + */ + block_i = EXT2_I(inode)->i_block_alloc_info; + if (block_i) { + windowsz = block_i->rsv_window_node.rsv_goal_size; + if (windowsz > 0) + my_rsv = &block_i->rsv_window_node; } - ext2_debug ("goal=%lu.\n", goal); + if (!ext2_has_free_blocks(sbi)) { + *errp = -ENOSPC; + goto out; + } + /* + * First, test whether the goal block is free. + */ if (goal < le32_to_cpu(es->s_first_data_block) || goal >= le32_to_cpu(es->s_blocks_count)) goal = le32_to_cpu(es->s_first_data_block); - group_no = (goal - le32_to_cpu(es->s_first_data_block)) / group_size; - desc = ext2_get_group_desc (sb, group_no, &gdp_bh); - if (!desc) { - /* - * gdp_bh may still be uninitialised. But group_release_blocks - * will not touch it because group_alloc is zero. - */ + group_no = (goal - le32_to_cpu(es->s_first_data_block)) / + EXT2_BLOCKS_PER_GROUP(sb); + goal_group = group_no; +retry_alloc: + gdp = ext2_get_group_desc(sb, group_no, &gdp_bh); + if (!gdp) goto io_error; - } - group_alloc = group_reserve_blocks(sbi, group_no, desc, - gdp_bh, es_alloc); - if (group_alloc) { - ret_block = ((goal - le32_to_cpu(es->s_first_data_block)) % - group_size); - brelse(bitmap_bh); + free_blocks = le16_to_cpu(gdp->bg_free_blocks_count); + /* + * if there is not enough free blocks to make a new resevation + * turn off reservation for this allocation + */ + if (my_rsv && (free_blocks < windowsz) + && (rsv_is_empty(&my_rsv->rsv_window))) + my_rsv = NULL; + + if (free_blocks > 0) { + grp_target_blk = ((goal - le32_to_cpu(es->s_first_data_block)) % + EXT2_BLOCKS_PER_GROUP(sb)); bitmap_bh = read_block_bitmap(sb, group_no); if (!bitmap_bh) goto io_error; - - ext2_debug("goal is at %d:%d.\n", group_no, ret_block); - - ret_block = grab_block(sb_bgl_lock(sbi, group_no), - bitmap_bh->b_data, group_size, ret_block); - if (ret_block >= 0) - goto got_block; - group_release_blocks(sb, group_no, desc, gdp_bh, group_alloc); - group_alloc = 0; + grp_alloc_blk = ext2_try_to_allocate_with_rsv(sb, group_no, + bitmap_bh, grp_target_blk, + my_rsv, &num); + if (grp_alloc_blk >= 0) + goto allocated; } - ext2_debug ("Bit not found in block group %d.\n", group_no); + ngroups = EXT2_SB(sb)->s_groups_count; + smp_rmb(); /* * Now search the rest of the groups. We assume that - * i and desc correctly point to the last group visited. + * i and gdp correctly point to the last group visited. */ - nr_scanned_groups = 0; -retry: - for (group_idx = 0; !group_alloc && - group_idx < sbi->s_groups_count; group_idx++) { + for (bgi = 0; bgi < ngroups; bgi++) { group_no++; - if (group_no >= sbi->s_groups_count) + if (group_no >= ngroups) group_no = 0; - desc = ext2_get_group_desc(sb, group_no, &gdp_bh); - if (!desc) + gdp = ext2_get_group_desc(sb, group_no, &gdp_bh); + if (!gdp) goto io_error; - group_alloc = group_reserve_blocks(sbi, group_no, desc, - gdp_bh, es_alloc); - } - if (!group_alloc) { - *err = -ENOSPC; - goto out_release; - } - brelse(bitmap_bh); - bitmap_bh = read_block_bitmap(sb, group_no); - if (!bitmap_bh) - goto io_error; - ret_block = grab_block(sb_bgl_lock(sbi, group_no), bitmap_bh->b_data, - group_size, 0); - if (ret_block < 0) { + free_blocks = le16_to_cpu(gdp->bg_free_blocks_count); /* - * If a free block counter is corrupted we can loop inifintely. - * Detect that here. + * skip this group if the number of + * free blocks is less than half of the reservation + * window size. */ - nr_scanned_groups++; - if (nr_scanned_groups > 2 * sbi->s_groups_count) { - ext2_error(sb, "ext2_new_block", - "corrupted free blocks counters"); + if (free_blocks <= (windowsz/2)) + continue; + + brelse(bitmap_bh); + bitmap_bh = read_block_bitmap(sb, group_no); + if (!bitmap_bh) goto io_error; - } /* - * Someone else grabbed the last free block in this blockgroup - * before us. Retry the scan. + * try to allocate block(s) from this group, without a goal(-1). */ - group_release_blocks(sb, group_no, desc, gdp_bh, group_alloc); - group_alloc = 0; - goto retry; + grp_alloc_blk = ext2_try_to_allocate_with_rsv(sb, group_no, + bitmap_bh, -1, my_rsv, &num); + if (grp_alloc_blk >= 0) + goto allocated; + } + /* + * We may end up a bogus ealier ENOSPC error due to + * filesystem is "full" of reservations, but + * there maybe indeed free blocks avaliable on disk + * In this case, we just forget about the reservations + * just do block allocation as without reservations. + */ + if (my_rsv) { + my_rsv = NULL; + windowsz = 0; + group_no = goal_group; + goto retry_alloc; } + /* No space left on the device */ + *errp = -ENOSPC; + goto out; + +allocated: -got_block: ext2_debug("using block group %d(%d)\n", - group_no, desc->bg_free_blocks_count); + group_no, gdp->bg_free_blocks_count); - target_block = ret_block + group_no * group_size + - le32_to_cpu(es->s_first_data_block); + ret_block = grp_alloc_blk + ext2_group_first_block_no(sb, group_no); - if (target_block == le32_to_cpu(desc->bg_block_bitmap) || - target_block == le32_to_cpu(desc->bg_inode_bitmap) || - in_range(target_block, le32_to_cpu(desc->bg_inode_table), - sbi->s_itb_per_group)) - ext2_error (sb, "ext2_new_block", + if (in_range(le32_to_cpu(gdp->bg_block_bitmap), ret_block, num) || + in_range(le32_to_cpu(gdp->bg_inode_bitmap), ret_block, num) || + in_range(ret_block, le32_to_cpu(gdp->bg_inode_table), + EXT2_SB(sb)->s_itb_per_group) || + in_range(ret_block + num - 1, le32_to_cpu(gdp->bg_inode_table), + EXT2_SB(sb)->s_itb_per_group)) + ext2_error(sb, "ext2_new_blocks", "Allocating block in system zone - " - "block = %u", target_block); + "blocks from "E2FSBLK", length %lu", + ret_block, num); - if (target_block >= le32_to_cpu(es->s_blocks_count)) { - ext2_error (sb, "ext2_new_block", - "block(%d) >= blocks count(%d) - " + performed_allocation = 1; + + if (ret_block + num - 1 >= le32_to_cpu(es->s_blocks_count)) { + ext2_error(sb, "ext2_new_blocks", + "block("E2FSBLK") >= blocks count(%d) - " "block_group = %d, es == %p ", ret_block, le32_to_cpu(es->s_blocks_count), group_no, es); - goto io_error; + goto out; } - block = target_block; - - /* OK, we _had_ allocated something */ - ext2_debug("found bit %d\n", ret_block); - - dq_alloc--; - es_alloc--; - group_alloc--; - /* - * Do block preallocation now if required. - */ - write_lock(&EXT2_I(inode)->i_meta_lock); - if (group_alloc && !*prealloc_count) { - unsigned n; - - for (n = 0; n < group_alloc && ++ret_block < group_size; n++) { - if (ext2_set_bit_atomic(sb_bgl_lock(sbi, group_no), - ret_block, - (void*) bitmap_bh->b_data)) - break; - } - *prealloc_block = block + 1; - *prealloc_count = n; - es_alloc -= n; - dq_alloc -= n; - group_alloc -= n; - } - write_unlock(&EXT2_I(inode)->i_meta_lock); + group_adjust_blocks(sb, group_no, gdp, gdp_bh, -num); + percpu_counter_sub(&sbi->s_freeblocks_counter, num); mark_buffer_dirty(bitmap_bh); if (sb->s_flags & MS_SYNCHRONOUS) sync_dirty_buffer(bitmap_bh); - ext2_debug ("allocating block %d. ", block); + *errp = 0; + brelse(bitmap_bh); + DQUOT_FREE_BLOCK(inode, *count-num); + *count = num; + return ret_block; - *err = 0; -out_release: - group_release_blocks(sb, group_no, desc, gdp_bh, group_alloc); - release_blocks(sb, es_alloc); -out_dquot: - DQUOT_FREE_BLOCK(inode, dq_alloc); +io_error: + *errp = -EIO; out: + /* + * Undo the block allocation + */ + if (!performed_allocation) + DQUOT_FREE_BLOCK(inode, *count); brelse(bitmap_bh); - return block; + return 0; +} -io_error: - *err = -EIO; - goto out_release; +ext2_fsblk_t ext2_new_block(struct inode *inode, unsigned long goal, int *errp) +{ + unsigned long count = 1; + + return ext2_new_blocks(inode, goal, &count, errp); } #ifdef EXT2FS_DEBUG |