0001 .. SPDX-License-Identifier: GPL-2.0
0002
0003 Bigalloc
0004 --------
0005
0006 At the moment, the default size of a block is 4KiB, which is a commonly
0007 supported page size on most MMU-capable hardware. This is fortunate, as
0008 ext4 code is not prepared to handle the case where the block size
0009 exceeds the page size. However, for a filesystem of mostly huge files,
0010 it is desirable to be able to allocate disk blocks in units of multiple
0011 blocks to reduce both fragmentation and metadata overhead. The
0012 bigalloc feature provides exactly this ability.
0013
0014 The bigalloc feature (EXT4_FEATURE_RO_COMPAT_BIGALLOC) changes ext4 to
0015 use clustered allocation, so that each bit in the ext4 block allocation
0016 bitmap addresses a power of two number of blocks. For example, if the
0017 file system is mainly going to be storing large files in the 4-32
0018 megabyte range, it might make sense to set a cluster size of 1 megabyte.
0019 This means that each bit in the block allocation bitmap now addresses
0020 256 4k blocks. This shrinks the total size of the block allocation
0021 bitmaps for a 2T file system from 64 megabytes to 256 kilobytes. It also
0022 means that a block group addresses 32 gigabytes instead of 128 megabytes,
0023 also shrinking the amount of file system overhead for metadata.
0024
0025 The administrator can set a block cluster size at mkfs time (which is
0026 stored in the s_log_cluster_size field in the superblock); from then
0027 on, the block bitmaps track clusters, not individual blocks. This means
0028 that block groups can be several gigabytes in size (instead of just
0029 128MiB); however, the minimum allocation unit becomes a cluster, not a
0030 block, even for directories. TaoBao had a patchset to extend the “use
0031 units of clusters instead of blocks” to the extent tree, though it is
0032 not clear where those patches went-- they eventually morphed into
0033 “extent tree v2” but that code has not landed as of May 2015.
0034
