diff options
author | Josef Bacik <jbacik@redhat.com> | 2009-02-20 11:00:09 -0500 |
---|---|---|
committer | Chris Mason <chris.mason@oracle.com> | 2009-02-20 11:00:09 -0500 |
commit | 6a63209fc02d5483371f07e4913ee8abad608051 (patch) | |
tree | 7595e0df452928b677b66a64baf0cb3b7ec53dfc /fs | |
parent | 2cfbd50b536c878e58ab3681c4e944fa3d99b415 (diff) |
Btrfs: add better -ENOSPC handling
This is a step in the direction of better -ENOSPC handling. Instead of
checking the global bytes counter we check the space_info bytes counters to
make sure we have enough space.
If we don't we go ahead and try to allocate a new chunk, and then if that fails
we return -ENOSPC. This patch adds two counters to btrfs_space_info,
bytes_delalloc and bytes_may_use.
bytes_delalloc account for extents we've actually setup for delalloc and will
be allocated at some point down the line.
bytes_may_use is to keep track of how many bytes we may use for delalloc at
some point. When we actually set the extent_bit for the delalloc bytes we
subtract the reserved bytes from the bytes_may_use counter. This keeps us from
not actually being able to allocate space for any delalloc bytes.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Diffstat (limited to 'fs')
-rw-r--r-- | fs/btrfs/btrfs_inode.h | 8 | ||||
-rw-r--r-- | fs/btrfs/ctree.h | 40 | ||||
-rw-r--r-- | fs/btrfs/extent-tree.c | 215 | ||||
-rw-r--r-- | fs/btrfs/file.c | 16 | ||||
-rw-r--r-- | fs/btrfs/inode.c | 62 | ||||
-rw-r--r-- | fs/btrfs/ioctl.c | 6 |
6 files changed, 271 insertions, 76 deletions
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index a8c9693b75ac..72677ce2b74f 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h | |||
@@ -66,6 +66,9 @@ struct btrfs_inode { | |||
66 | */ | 66 | */ |
67 | struct list_head delalloc_inodes; | 67 | struct list_head delalloc_inodes; |
68 | 68 | ||
69 | /* the space_info for where this inode's data allocations are done */ | ||
70 | struct btrfs_space_info *space_info; | ||
71 | |||
69 | /* full 64 bit generation number, struct vfs_inode doesn't have a big | 72 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
70 | * enough field for this. | 73 | * enough field for this. |
71 | */ | 74 | */ |
@@ -94,6 +97,11 @@ struct btrfs_inode { | |||
94 | */ | 97 | */ |
95 | u64 delalloc_bytes; | 98 | u64 delalloc_bytes; |
96 | 99 | ||
100 | /* total number of bytes that may be used for this inode for | ||
101 | * delalloc | ||
102 | */ | ||
103 | u64 reserved_bytes; | ||
104 | |||
97 | /* | 105 | /* |
98 | * the size of the file stored in the metadata on disk. data=ordered | 106 | * the size of the file stored in the metadata on disk. data=ordered |
99 | * means the in-memory i_size might be larger than the size on disk | 107 | * means the in-memory i_size might be larger than the size on disk |
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 766b31ae3186..82491ba8fa40 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h | |||
@@ -596,13 +596,27 @@ struct btrfs_block_group_item { | |||
596 | 596 | ||
597 | struct btrfs_space_info { | 597 | struct btrfs_space_info { |
598 | u64 flags; | 598 | u64 flags; |
599 | u64 total_bytes; | 599 | |
600 | u64 bytes_used; | 600 | u64 total_bytes; /* total bytes in the space */ |
601 | u64 bytes_pinned; | 601 | u64 bytes_used; /* total bytes used on disk */ |
602 | u64 bytes_reserved; | 602 | u64 bytes_pinned; /* total bytes pinned, will be freed when the |
603 | u64 bytes_readonly; | 603 | transaction finishes */ |
604 | int full; | 604 | u64 bytes_reserved; /* total bytes the allocator has reserved for |
605 | int force_alloc; | 605 | current allocations */ |
606 | u64 bytes_readonly; /* total bytes that are read only */ | ||
607 | |||
608 | /* delalloc accounting */ | ||
609 | u64 bytes_delalloc; /* number of bytes reserved for allocation, | ||
610 | this space is not necessarily reserved yet | ||
611 | by the allocator */ | ||
612 | u64 bytes_may_use; /* number of bytes that may be used for | ||
613 | delalloc */ | ||
614 | |||
615 | int full; /* indicates that we cannot allocate any more | ||
616 | chunks for this space */ | ||
617 | int force_alloc; /* set if we need to force a chunk alloc for | ||
618 | this space */ | ||
619 | |||
606 | struct list_head list; | 620 | struct list_head list; |
607 | 621 | ||
608 | /* for block groups in our same type */ | 622 | /* for block groups in our same type */ |
@@ -1782,6 +1796,16 @@ int btrfs_add_dead_reloc_root(struct btrfs_root *root); | |||
1782 | int btrfs_cleanup_reloc_trees(struct btrfs_root *root); | 1796 | int btrfs_cleanup_reloc_trees(struct btrfs_root *root); |
1783 | int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len); | 1797 | int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len); |
1784 | u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags); | 1798 | u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags); |
1799 | void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde); | ||
1800 | int btrfs_check_metadata_free_space(struct btrfs_root *root); | ||
1801 | int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode, | ||
1802 | u64 bytes); | ||
1803 | void btrfs_free_reserved_data_space(struct btrfs_root *root, | ||
1804 | struct inode *inode, u64 bytes); | ||
1805 | void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode, | ||
1806 | u64 bytes); | ||
1807 | void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode, | ||
1808 | u64 bytes); | ||
1785 | /* ctree.c */ | 1809 | /* ctree.c */ |
1786 | int btrfs_previous_item(struct btrfs_root *root, | 1810 | int btrfs_previous_item(struct btrfs_root *root, |
1787 | struct btrfs_path *path, u64 min_objectid, | 1811 | struct btrfs_path *path, u64 min_objectid, |
@@ -2027,8 +2051,6 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, | |||
2027 | unsigned long btrfs_force_ra(struct address_space *mapping, | 2051 | unsigned long btrfs_force_ra(struct address_space *mapping, |
2028 | struct file_ra_state *ra, struct file *file, | 2052 | struct file_ra_state *ra, struct file *file, |
2029 | pgoff_t offset, pgoff_t last_index); | 2053 | pgoff_t offset, pgoff_t last_index); |
2030 | int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, | ||
2031 | int for_del); | ||
2032 | int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); | 2054 | int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); |
2033 | int btrfs_readpage(struct file *file, struct page *page); | 2055 | int btrfs_readpage(struct file *file, struct page *page); |
2034 | void btrfs_delete_inode(struct inode *inode); | 2056 | void btrfs_delete_inode(struct inode *inode); |
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 0a5d796c9f7e..e11875e97c2f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c | |||
@@ -60,6 +60,10 @@ static int update_block_group(struct btrfs_trans_handle *trans, | |||
60 | u64 bytenr, u64 num_bytes, int alloc, | 60 | u64 bytenr, u64 num_bytes, int alloc, |
61 | int mark_free); | 61 | int mark_free); |
62 | 62 | ||
63 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, | ||
64 | struct btrfs_root *extent_root, u64 alloc_bytes, | ||
65 | u64 flags, int force); | ||
66 | |||
63 | static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) | 67 | static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) |
64 | { | 68 | { |
65 | return (cache->flags & bits) == bits; | 69 | return (cache->flags & bits) == bits; |
@@ -1909,6 +1913,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, | |||
1909 | found->bytes_pinned = 0; | 1913 | found->bytes_pinned = 0; |
1910 | found->bytes_reserved = 0; | 1914 | found->bytes_reserved = 0; |
1911 | found->bytes_readonly = 0; | 1915 | found->bytes_readonly = 0; |
1916 | found->bytes_delalloc = 0; | ||
1912 | found->full = 0; | 1917 | found->full = 0; |
1913 | found->force_alloc = 0; | 1918 | found->force_alloc = 0; |
1914 | *space_info = found; | 1919 | *space_info = found; |
@@ -1972,6 +1977,196 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags) | |||
1972 | return flags; | 1977 | return flags; |
1973 | } | 1978 | } |
1974 | 1979 | ||
1980 | static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data) | ||
1981 | { | ||
1982 | struct btrfs_fs_info *info = root->fs_info; | ||
1983 | u64 alloc_profile; | ||
1984 | |||
1985 | if (data) { | ||
1986 | alloc_profile = info->avail_data_alloc_bits & | ||
1987 | info->data_alloc_profile; | ||
1988 | data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; | ||
1989 | } else if (root == root->fs_info->chunk_root) { | ||
1990 | alloc_profile = info->avail_system_alloc_bits & | ||
1991 | info->system_alloc_profile; | ||
1992 | data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; | ||
1993 | } else { | ||
1994 | alloc_profile = info->avail_metadata_alloc_bits & | ||
1995 | info->metadata_alloc_profile; | ||
1996 | data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; | ||
1997 | } | ||
1998 | |||
1999 | return btrfs_reduce_alloc_profile(root, data); | ||
2000 | } | ||
2001 | |||
2002 | void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode) | ||
2003 | { | ||
2004 | u64 alloc_target; | ||
2005 | |||
2006 | alloc_target = btrfs_get_alloc_profile(root, 1); | ||
2007 | BTRFS_I(inode)->space_info = __find_space_info(root->fs_info, | ||
2008 | alloc_target); | ||
2009 | } | ||
2010 | |||
2011 | /* | ||
2012 | * for now this just makes sure we have at least 5% of our metadata space free | ||
2013 | * for use. | ||
2014 | */ | ||
2015 | int btrfs_check_metadata_free_space(struct btrfs_root *root) | ||
2016 | { | ||
2017 | struct btrfs_fs_info *info = root->fs_info; | ||
2018 | struct btrfs_space_info *meta_sinfo; | ||
2019 | u64 alloc_target, thresh; | ||
2020 | |||
2021 | /* get the space info for where the metadata will live */ | ||
2022 | alloc_target = btrfs_get_alloc_profile(root, 0); | ||
2023 | meta_sinfo = __find_space_info(info, alloc_target); | ||
2024 | |||
2025 | /* | ||
2026 | * if the metadata area isn't maxed out then there is no sense in | ||
2027 | * checking how much is used, since we can always allocate a new chunk | ||
2028 | */ | ||
2029 | if (!meta_sinfo->full) | ||
2030 | return 0; | ||
2031 | |||
2032 | spin_lock(&meta_sinfo->lock); | ||
2033 | thresh = meta_sinfo->total_bytes * 95; | ||
2034 | |||
2035 | do_div(thresh, 100); | ||
2036 | |||
2037 | if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved + | ||
2038 | meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) { | ||
2039 | spin_unlock(&meta_sinfo->lock); | ||
2040 | return -ENOSPC; | ||
2041 | } | ||
2042 | spin_unlock(&meta_sinfo->lock); | ||
2043 | |||
2044 | return 0; | ||
2045 | } | ||
2046 | |||
2047 | /* | ||
2048 | * This will check the space that the inode allocates from to make sure we have | ||
2049 | * enough space for bytes. | ||
2050 | */ | ||
2051 | int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode, | ||
2052 | u64 bytes) | ||
2053 | { | ||
2054 | struct btrfs_space_info *data_sinfo; | ||
2055 | int ret = 0; | ||
2056 | |||
2057 | /* make sure bytes are sectorsize aligned */ | ||
2058 | bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | ||
2059 | |||
2060 | data_sinfo = BTRFS_I(inode)->space_info; | ||
2061 | again: | ||
2062 | /* make sure we have enough space to handle the data first */ | ||
2063 | spin_lock(&data_sinfo->lock); | ||
2064 | if (data_sinfo->total_bytes - data_sinfo->bytes_used - | ||
2065 | data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved - | ||
2066 | data_sinfo->bytes_pinned - data_sinfo->bytes_readonly - | ||
2067 | data_sinfo->bytes_may_use < bytes) { | ||
2068 | /* | ||
2069 | * if we don't have enough free bytes in this space then we need | ||
2070 | * to alloc a new chunk. | ||
2071 | */ | ||
2072 | if (!data_sinfo->full) { | ||
2073 | u64 alloc_target; | ||
2074 | struct btrfs_trans_handle *trans; | ||
2075 | |||
2076 | data_sinfo->force_alloc = 1; | ||
2077 | spin_unlock(&data_sinfo->lock); | ||
2078 | |||
2079 | alloc_target = btrfs_get_alloc_profile(root, 1); | ||
2080 | trans = btrfs_start_transaction(root, 1); | ||
2081 | if (!trans) | ||
2082 | return -ENOMEM; | ||
2083 | |||
2084 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, | ||
2085 | bytes + 2 * 1024 * 1024, | ||
2086 | alloc_target, 0); | ||
2087 | btrfs_end_transaction(trans, root); | ||
2088 | if (ret) | ||
2089 | return ret; | ||
2090 | goto again; | ||
2091 | } | ||
2092 | spin_unlock(&data_sinfo->lock); | ||
2093 | printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes" | ||
2094 | ", %llu bytes_used, %llu bytes_reserved, " | ||
2095 | "%llu bytes_pinned, %llu bytes_readonly, %llu may use" | ||
2096 | "%llu total\n", bytes, data_sinfo->bytes_delalloc, | ||
2097 | data_sinfo->bytes_used, data_sinfo->bytes_reserved, | ||
2098 | data_sinfo->bytes_pinned, data_sinfo->bytes_readonly, | ||
2099 | data_sinfo->bytes_may_use, data_sinfo->total_bytes); | ||
2100 | return -ENOSPC; | ||
2101 | } | ||
2102 | data_sinfo->bytes_may_use += bytes; | ||
2103 | BTRFS_I(inode)->reserved_bytes += bytes; | ||
2104 | spin_unlock(&data_sinfo->lock); | ||
2105 | |||
2106 | return btrfs_check_metadata_free_space(root); | ||
2107 | } | ||
2108 | |||
2109 | /* | ||
2110 | * if there was an error for whatever reason after calling | ||
2111 | * btrfs_check_data_free_space, call this so we can cleanup the counters. | ||
2112 | */ | ||
2113 | void btrfs_free_reserved_data_space(struct btrfs_root *root, | ||
2114 | struct inode *inode, u64 bytes) | ||
2115 | { | ||
2116 | struct btrfs_space_info *data_sinfo; | ||
2117 | |||
2118 | /* make sure bytes are sectorsize aligned */ | ||
2119 | bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | ||
2120 | |||
2121 | data_sinfo = BTRFS_I(inode)->space_info; | ||
2122 | spin_lock(&data_sinfo->lock); | ||
2123 | data_sinfo->bytes_may_use -= bytes; | ||
2124 | BTRFS_I(inode)->reserved_bytes -= bytes; | ||
2125 | spin_unlock(&data_sinfo->lock); | ||
2126 | } | ||
2127 | |||
2128 | /* called when we are adding a delalloc extent to the inode's io_tree */ | ||
2129 | void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode, | ||
2130 | u64 bytes) | ||
2131 | { | ||
2132 | struct btrfs_space_info *data_sinfo; | ||
2133 | |||
2134 | /* get the space info for where this inode will be storing its data */ | ||
2135 | data_sinfo = BTRFS_I(inode)->space_info; | ||
2136 | |||
2137 | /* make sure we have enough space to handle the data first */ | ||
2138 | spin_lock(&data_sinfo->lock); | ||
2139 | data_sinfo->bytes_delalloc += bytes; | ||
2140 | |||
2141 | /* | ||
2142 | * we are adding a delalloc extent without calling | ||
2143 | * btrfs_check_data_free_space first. This happens on a weird | ||
2144 | * writepage condition, but shouldn't hurt our accounting | ||
2145 | */ | ||
2146 | if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) { | ||
2147 | data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes; | ||
2148 | BTRFS_I(inode)->reserved_bytes = 0; | ||
2149 | } else { | ||
2150 | data_sinfo->bytes_may_use -= bytes; | ||
2151 | BTRFS_I(inode)->reserved_bytes -= bytes; | ||
2152 | } | ||
2153 | |||
2154 | spin_unlock(&data_sinfo->lock); | ||
2155 | } | ||
2156 | |||
2157 | /* called when we are clearing an delalloc extent from the inode's io_tree */ | ||
2158 | void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode, | ||
2159 | u64 bytes) | ||
2160 | { | ||
2161 | struct btrfs_space_info *info; | ||
2162 | |||
2163 | info = BTRFS_I(inode)->space_info; | ||
2164 | |||
2165 | spin_lock(&info->lock); | ||
2166 | info->bytes_delalloc -= bytes; | ||
2167 | spin_unlock(&info->lock); | ||
2168 | } | ||
2169 | |||
1975 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, | 2170 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, |
1976 | struct btrfs_root *extent_root, u64 alloc_bytes, | 2171 | struct btrfs_root *extent_root, u64 alloc_bytes, |
1977 | u64 flags, int force) | 2172 | u64 flags, int force) |
@@ -3105,6 +3300,10 @@ static void dump_space_info(struct btrfs_space_info *info, u64 bytes) | |||
3105 | (unsigned long long)(info->total_bytes - info->bytes_used - | 3300 | (unsigned long long)(info->total_bytes - info->bytes_used - |
3106 | info->bytes_pinned - info->bytes_reserved), | 3301 | info->bytes_pinned - info->bytes_reserved), |
3107 | (info->full) ? "" : "not "); | 3302 | (info->full) ? "" : "not "); |
3303 | printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu," | ||
3304 | " may_use=%llu, used=%llu\n", info->total_bytes, | ||
3305 | info->bytes_pinned, info->bytes_delalloc, info->bytes_may_use, | ||
3306 | info->bytes_used); | ||
3108 | 3307 | ||
3109 | down_read(&info->groups_sem); | 3308 | down_read(&info->groups_sem); |
3110 | list_for_each_entry(cache, &info->block_groups, list) { | 3309 | list_for_each_entry(cache, &info->block_groups, list) { |
@@ -3131,24 +3330,10 @@ static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans, | |||
3131 | { | 3330 | { |
3132 | int ret; | 3331 | int ret; |
3133 | u64 search_start = 0; | 3332 | u64 search_start = 0; |
3134 | u64 alloc_profile; | ||
3135 | struct btrfs_fs_info *info = root->fs_info; | 3333 | struct btrfs_fs_info *info = root->fs_info; |
3136 | 3334 | ||
3137 | if (data) { | 3335 | data = btrfs_get_alloc_profile(root, data); |
3138 | alloc_profile = info->avail_data_alloc_bits & | ||
3139 | info->data_alloc_profile; | ||
3140 | data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; | ||
3141 | } else if (root == root->fs_info->chunk_root) { | ||
3142 | alloc_profile = info->avail_system_alloc_bits & | ||
3143 | info->system_alloc_profile; | ||
3144 | data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; | ||
3145 | } else { | ||
3146 | alloc_profile = info->avail_metadata_alloc_bits & | ||
3147 | info->metadata_alloc_profile; | ||
3148 | data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; | ||
3149 | } | ||
3150 | again: | 3336 | again: |
3151 | data = btrfs_reduce_alloc_profile(root, data); | ||
3152 | /* | 3337 | /* |
3153 | * the only place that sets empty_size is btrfs_realloc_node, which | 3338 | * the only place that sets empty_size is btrfs_realloc_node, which |
3154 | * is not called recursively on allocations | 3339 | * is not called recursively on allocations |
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 872f104576e5..dc78954861b3 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c | |||
@@ -1091,19 +1091,24 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |||
1091 | WARN_ON(num_pages > nrptrs); | 1091 | WARN_ON(num_pages > nrptrs); |
1092 | memset(pages, 0, sizeof(struct page *) * nrptrs); | 1092 | memset(pages, 0, sizeof(struct page *) * nrptrs); |
1093 | 1093 | ||
1094 | ret = btrfs_check_free_space(root, write_bytes, 0); | 1094 | ret = btrfs_check_data_free_space(root, inode, write_bytes); |
1095 | if (ret) | 1095 | if (ret) |
1096 | goto out; | 1096 | goto out; |
1097 | 1097 | ||
1098 | ret = prepare_pages(root, file, pages, num_pages, | 1098 | ret = prepare_pages(root, file, pages, num_pages, |
1099 | pos, first_index, last_index, | 1099 | pos, first_index, last_index, |
1100 | write_bytes); | 1100 | write_bytes); |
1101 | if (ret) | 1101 | if (ret) { |
1102 | btrfs_free_reserved_data_space(root, inode, | ||
1103 | write_bytes); | ||
1102 | goto out; | 1104 | goto out; |
1105 | } | ||
1103 | 1106 | ||
1104 | ret = btrfs_copy_from_user(pos, num_pages, | 1107 | ret = btrfs_copy_from_user(pos, num_pages, |
1105 | write_bytes, pages, buf); | 1108 | write_bytes, pages, buf); |
1106 | if (ret) { | 1109 | if (ret) { |
1110 | btrfs_free_reserved_data_space(root, inode, | ||
1111 | write_bytes); | ||
1107 | btrfs_drop_pages(pages, num_pages); | 1112 | btrfs_drop_pages(pages, num_pages); |
1108 | goto out; | 1113 | goto out; |
1109 | } | 1114 | } |
@@ -1111,8 +1116,11 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |||
1111 | ret = dirty_and_release_pages(NULL, root, file, pages, | 1116 | ret = dirty_and_release_pages(NULL, root, file, pages, |
1112 | num_pages, pos, write_bytes); | 1117 | num_pages, pos, write_bytes); |
1113 | btrfs_drop_pages(pages, num_pages); | 1118 | btrfs_drop_pages(pages, num_pages); |
1114 | if (ret) | 1119 | if (ret) { |
1120 | btrfs_free_reserved_data_space(root, inode, | ||
1121 | write_bytes); | ||
1115 | goto out; | 1122 | goto out; |
1123 | } | ||
1116 | 1124 | ||
1117 | if (will_write) { | 1125 | if (will_write) { |
1118 | btrfs_fdatawrite_range(inode->i_mapping, pos, | 1126 | btrfs_fdatawrite_range(inode->i_mapping, pos, |
@@ -1136,6 +1144,8 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |||
1136 | } | 1144 | } |
1137 | out: | 1145 | out: |
1138 | mutex_unlock(&inode->i_mutex); | 1146 | mutex_unlock(&inode->i_mutex); |
1147 | if (ret) | ||
1148 | err = ret; | ||
1139 | 1149 | ||
1140 | out_nolock: | 1150 | out_nolock: |
1141 | kfree(pages); | 1151 | kfree(pages); |
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3cee77ae03c8..7d4f948bc22a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c | |||
@@ -102,34 +102,6 @@ static int btrfs_init_inode_security(struct inode *inode, struct inode *dir) | |||
102 | } | 102 | } |
103 | 103 | ||
104 | /* | 104 | /* |
105 | * a very lame attempt at stopping writes when the FS is 85% full. There | ||
106 | * are countless ways this is incorrect, but it is better than nothing. | ||
107 | */ | ||
108 | int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, | ||
109 | int for_del) | ||
110 | { | ||
111 | u64 total; | ||
112 | u64 used; | ||
113 | u64 thresh; | ||
114 | int ret = 0; | ||
115 | |||
116 | spin_lock(&root->fs_info->delalloc_lock); | ||
117 | total = btrfs_super_total_bytes(&root->fs_info->super_copy); | ||
118 | used = btrfs_super_bytes_used(&root->fs_info->super_copy); | ||
119 | if (for_del) | ||
120 | thresh = total * 90; | ||
121 | else | ||
122 | thresh = total * 85; | ||
123 | |||
124 | do_div(thresh, 100); | ||
125 | |||
126 | if (used + root->fs_info->delalloc_bytes + num_required > thresh) | ||
127 | ret = -ENOSPC; | ||
128 | spin_unlock(&root->fs_info->delalloc_lock); | ||
129 | return ret; | ||
130 | } | ||
131 | |||
132 | /* | ||
133 | * this does all the hard work for inserting an inline extent into | 105 | * this does all the hard work for inserting an inline extent into |
134 | * the btree. The caller should have done a btrfs_drop_extents so that | 106 | * the btree. The caller should have done a btrfs_drop_extents so that |
135 | * no overlapping inline items exist in the btree | 107 | * no overlapping inline items exist in the btree |
@@ -1190,6 +1162,7 @@ static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, | |||
1190 | */ | 1162 | */ |
1191 | if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { | 1163 | if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { |
1192 | struct btrfs_root *root = BTRFS_I(inode)->root; | 1164 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1165 | btrfs_delalloc_reserve_space(root, inode, end - start + 1); | ||
1193 | spin_lock(&root->fs_info->delalloc_lock); | 1166 | spin_lock(&root->fs_info->delalloc_lock); |
1194 | BTRFS_I(inode)->delalloc_bytes += end - start + 1; | 1167 | BTRFS_I(inode)->delalloc_bytes += end - start + 1; |
1195 | root->fs_info->delalloc_bytes += end - start + 1; | 1168 | root->fs_info->delalloc_bytes += end - start + 1; |
@@ -1223,9 +1196,12 @@ static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, | |||
1223 | (unsigned long long)end - start + 1, | 1196 | (unsigned long long)end - start + 1, |
1224 | (unsigned long long) | 1197 | (unsigned long long) |
1225 | root->fs_info->delalloc_bytes); | 1198 | root->fs_info->delalloc_bytes); |
1199 | btrfs_delalloc_free_space(root, inode, (u64)-1); | ||
1226 | root->fs_info->delalloc_bytes = 0; | 1200 | root->fs_info->delalloc_bytes = 0; |
1227 | BTRFS_I(inode)->delalloc_bytes = 0; | 1201 | BTRFS_I(inode)->delalloc_bytes = 0; |
1228 | } else { | 1202 | } else { |
1203 | btrfs_delalloc_free_space(root, inode, | ||
1204 | end - start + 1); | ||
1229 | root->fs_info->delalloc_bytes -= end - start + 1; | 1205 | root->fs_info->delalloc_bytes -= end - start + 1; |
1230 | BTRFS_I(inode)->delalloc_bytes -= end - start + 1; | 1206 | BTRFS_I(inode)->delalloc_bytes -= end - start + 1; |
1231 | } | 1207 | } |
@@ -2245,10 +2221,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) | |||
2245 | 2221 | ||
2246 | root = BTRFS_I(dir)->root; | 2222 | root = BTRFS_I(dir)->root; |
2247 | 2223 | ||
2248 | ret = btrfs_check_free_space(root, 1, 1); | ||
2249 | if (ret) | ||
2250 | goto fail; | ||
2251 | |||
2252 | trans = btrfs_start_transaction(root, 1); | 2224 | trans = btrfs_start_transaction(root, 1); |
2253 | 2225 | ||
2254 | btrfs_set_trans_block_group(trans, dir); | 2226 | btrfs_set_trans_block_group(trans, dir); |
@@ -2261,7 +2233,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) | |||
2261 | nr = trans->blocks_used; | 2233 | nr = trans->blocks_used; |
2262 | 2234 | ||
2263 | btrfs_end_transaction_throttle(trans, root); | 2235 | btrfs_end_transaction_throttle(trans, root); |
2264 | fail: | ||
2265 | btrfs_btree_balance_dirty(root, nr); | 2236 | btrfs_btree_balance_dirty(root, nr); |
2266 | return ret; | 2237 | return ret; |
2267 | } | 2238 | } |
@@ -2284,10 +2255,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) | |||
2284 | return -ENOTEMPTY; | 2255 | return -ENOTEMPTY; |
2285 | } | 2256 | } |
2286 | 2257 | ||
2287 | ret = btrfs_check_free_space(root, 1, 1); | ||
2288 | if (ret) | ||
2289 | goto fail; | ||
2290 | |||
2291 | trans = btrfs_start_transaction(root, 1); | 2258 | trans = btrfs_start_transaction(root, 1); |
2292 | btrfs_set_trans_block_group(trans, dir); | 2259 | btrfs_set_trans_block_group(trans, dir); |
2293 | 2260 | ||
@@ -2304,7 +2271,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) | |||
2304 | fail_trans: | 2271 | fail_trans: |
2305 | nr = trans->blocks_used; | 2272 | nr = trans->blocks_used; |
2306 | ret = btrfs_end_transaction_throttle(trans, root); | 2273 | ret = btrfs_end_transaction_throttle(trans, root); |
2307 | fail: | ||
2308 | btrfs_btree_balance_dirty(root, nr); | 2274 | btrfs_btree_balance_dirty(root, nr); |
2309 | 2275 | ||
2310 | if (ret && !err) | 2276 | if (ret && !err) |
@@ -2818,7 +2784,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t size) | |||
2818 | if (size <= hole_start) | 2784 | if (size <= hole_start) |
2819 | return 0; | 2785 | return 0; |
2820 | 2786 | ||
2821 | err = btrfs_check_free_space(root, 1, 0); | 2787 | err = btrfs_check_metadata_free_space(root); |
2822 | if (err) | 2788 | if (err) |
2823 | return err; | 2789 | return err; |
2824 | 2790 | ||
@@ -3014,6 +2980,7 @@ static noinline void init_btrfs_i(struct inode *inode) | |||
3014 | bi->last_trans = 0; | 2980 | bi->last_trans = 0; |
3015 | bi->logged_trans = 0; | 2981 | bi->logged_trans = 0; |
3016 | bi->delalloc_bytes = 0; | 2982 | bi->delalloc_bytes = 0; |
2983 | bi->reserved_bytes = 0; | ||
3017 | bi->disk_i_size = 0; | 2984 | bi->disk_i_size = 0; |
3018 | bi->flags = 0; | 2985 | bi->flags = 0; |
3019 | bi->index_cnt = (u64)-1; | 2986 | bi->index_cnt = (u64)-1; |
@@ -3035,6 +3002,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) | |||
3035 | inode->i_ino = args->ino; | 3002 | inode->i_ino = args->ino; |
3036 | init_btrfs_i(inode); | 3003 | init_btrfs_i(inode); |
3037 | BTRFS_I(inode)->root = args->root; | 3004 | BTRFS_I(inode)->root = args->root; |
3005 | btrfs_set_inode_space_info(args->root, inode); | ||
3038 | return 0; | 3006 | return 0; |
3039 | } | 3007 | } |
3040 | 3008 | ||
@@ -3455,6 +3423,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, | |||
3455 | BTRFS_I(inode)->index_cnt = 2; | 3423 | BTRFS_I(inode)->index_cnt = 2; |
3456 | BTRFS_I(inode)->root = root; | 3424 | BTRFS_I(inode)->root = root; |
3457 | BTRFS_I(inode)->generation = trans->transid; | 3425 | BTRFS_I(inode)->generation = trans->transid; |
3426 | btrfs_set_inode_space_info(root, inode); | ||
3458 | 3427 | ||
3459 | if (mode & S_IFDIR) | 3428 | if (mode & S_IFDIR) |
3460 | owner = 0; | 3429 | owner = 0; |
@@ -3602,7 +3571,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, | |||
3602 | if (!new_valid_dev(rdev)) | 3571 | if (!new_valid_dev(rdev)) |
3603 | return -EINVAL; | 3572 | return -EINVAL; |
3604 | 3573 | ||
3605 | err = btrfs_check_free_space(root, 1, 0); | 3574 | err = btrfs_check_metadata_free_space(root); |
3606 | if (err) | 3575 | if (err) |
3607 | goto fail; | 3576 | goto fail; |
3608 | 3577 | ||
@@ -3665,7 +3634,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, | |||
3665 | u64 objectid; | 3634 | u64 objectid; |
3666 | u64 index = 0; | 3635 | u64 index = 0; |
3667 | 3636 | ||
3668 | err = btrfs_check_free_space(root, 1, 0); | 3637 | err = btrfs_check_metadata_free_space(root); |
3669 | if (err) | 3638 | if (err) |
3670 | goto fail; | 3639 | goto fail; |
3671 | trans = btrfs_start_transaction(root, 1); | 3640 | trans = btrfs_start_transaction(root, 1); |
@@ -3733,7 +3702,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, | |||
3733 | return -ENOENT; | 3702 | return -ENOENT; |
3734 | 3703 | ||
3735 | btrfs_inc_nlink(inode); | 3704 | btrfs_inc_nlink(inode); |
3736 | err = btrfs_check_free_space(root, 1, 0); | 3705 | err = btrfs_check_metadata_free_space(root); |
3737 | if (err) | 3706 | if (err) |
3738 | goto fail; | 3707 | goto fail; |
3739 | err = btrfs_set_inode_index(dir, &index); | 3708 | err = btrfs_set_inode_index(dir, &index); |
@@ -3779,7 +3748,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) | |||
3779 | u64 index = 0; | 3748 | u64 index = 0; |
3780 | unsigned long nr = 1; | 3749 | unsigned long nr = 1; |
3781 | 3750 | ||
3782 | err = btrfs_check_free_space(root, 1, 0); | 3751 | err = btrfs_check_metadata_free_space(root); |
3783 | if (err) | 3752 | if (err) |
3784 | goto out_unlock; | 3753 | goto out_unlock; |
3785 | 3754 | ||
@@ -4336,7 +4305,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) | |||
4336 | u64 page_start; | 4305 | u64 page_start; |
4337 | u64 page_end; | 4306 | u64 page_end; |
4338 | 4307 | ||
4339 | ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); | 4308 | ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE); |
4340 | if (ret) | 4309 | if (ret) |
4341 | goto out; | 4310 | goto out; |
4342 | 4311 | ||
@@ -4349,6 +4318,7 @@ again: | |||
4349 | 4318 | ||
4350 | if ((page->mapping != inode->i_mapping) || | 4319 | if ((page->mapping != inode->i_mapping) || |
4351 | (page_start >= size)) { | 4320 | (page_start >= size)) { |
4321 | btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE); | ||
4352 | /* page got truncated out from underneath us */ | 4322 | /* page got truncated out from underneath us */ |
4353 | goto out_unlock; | 4323 | goto out_unlock; |
4354 | } | 4324 | } |
@@ -4631,7 +4601,7 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |||
4631 | if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) | 4601 | if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) |
4632 | return -EXDEV; | 4602 | return -EXDEV; |
4633 | 4603 | ||
4634 | ret = btrfs_check_free_space(root, 1, 0); | 4604 | ret = btrfs_check_metadata_free_space(root); |
4635 | if (ret) | 4605 | if (ret) |
4636 | goto out_unlock; | 4606 | goto out_unlock; |
4637 | 4607 | ||
@@ -4749,7 +4719,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, | |||
4749 | if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) | 4719 | if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) |
4750 | return -ENAMETOOLONG; | 4720 | return -ENAMETOOLONG; |
4751 | 4721 | ||
4752 | err = btrfs_check_free_space(root, 1, 0); | 4722 | err = btrfs_check_metadata_free_space(root); |
4753 | if (err) | 4723 | if (err) |
4754 | goto out_fail; | 4724 | goto out_fail; |
4755 | 4725 | ||
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 988fdc8b49eb..bca729fc80c8 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c | |||
@@ -70,7 +70,7 @@ static noinline int create_subvol(struct btrfs_root *root, | |||
70 | u64 index = 0; | 70 | u64 index = 0; |
71 | unsigned long nr = 1; | 71 | unsigned long nr = 1; |
72 | 72 | ||
73 | ret = btrfs_check_free_space(root, 1, 0); | 73 | ret = btrfs_check_metadata_free_space(root); |
74 | if (ret) | 74 | if (ret) |
75 | goto fail_commit; | 75 | goto fail_commit; |
76 | 76 | ||
@@ -203,7 +203,7 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, | |||
203 | if (!root->ref_cows) | 203 | if (!root->ref_cows) |
204 | return -EINVAL; | 204 | return -EINVAL; |
205 | 205 | ||
206 | ret = btrfs_check_free_space(root, 1, 0); | 206 | ret = btrfs_check_metadata_free_space(root); |
207 | if (ret) | 207 | if (ret) |
208 | goto fail_unlock; | 208 | goto fail_unlock; |
209 | 209 | ||
@@ -374,7 +374,7 @@ static int btrfs_defrag_file(struct file *file) | |||
374 | unsigned long i; | 374 | unsigned long i; |
375 | int ret; | 375 | int ret; |
376 | 376 | ||
377 | ret = btrfs_check_free_space(root, inode->i_size, 0); | 377 | ret = btrfs_check_data_free_space(root, inode, inode->i_size); |
378 | if (ret) | 378 | if (ret) |
379 | return -ENOSPC; | 379 | return -ENOSPC; |
380 | 380 | ||