| 1 | # SPDX-License-Identifier: GPL-2.0-only |
| 2 | config SQUASHFS |
| 3 | tristate "SquashFS 4.0 - Squashed file system support" |
| 4 | depends on BLOCK |
| 5 | help |
| 6 | Saying Y here includes support for SquashFS 4.0 (a Compressed |
| 7 | Read-Only File System). Squashfs is a highly compressed read-only |
| 8 | filesystem for Linux. It uses zlib, lz4, lzo, xz or zstd compression |
| 9 | to compress both files, inodes and directories. Inodes in the system |
| 10 | are very small and all blocks are packed to minimise data overhead. |
| 11 | Block sizes greater than 4K are supported up to a maximum of 1 Mbytes |
| 12 | (default block size 128K). SquashFS 4.0 supports 64 bit filesystems |
| 13 | and files (larger than 4GB), full uid/gid information, hard links and |
| 14 | timestamps. |
| 15 | |
| 16 | Squashfs is intended for general read-only filesystem use, for |
| 17 | archival use (i.e. in cases where a .tar.gz file may be used), and in |
| 18 | embedded systems where low overhead is needed. Further information |
| 19 | and tools are available from github.com/plougher/squashfs-tools. |
| 20 | |
| 21 | If you want to compile this as a module ( = code which can be |
| 22 | inserted in and removed from the running kernel whenever you want), |
| 23 | say M here. The module will be called squashfs. Note that the root |
| 24 | file system (the one containing the directory /) cannot be compiled |
| 25 | as a module. |
| 26 | |
| 27 | If unsure, say N. |
| 28 | |
| 29 | choice |
| 30 | prompt "File decompression options" |
| 31 | depends on SQUASHFS |
| 32 | help |
| 33 | Squashfs now supports two options for decompressing file |
| 34 | data. Traditionally Squashfs has decompressed into an |
| 35 | intermediate buffer and then memcopied it into the page cache. |
| 36 | Squashfs now supports the ability to decompress directly into |
| 37 | the page cache. |
| 38 | |
| 39 | If unsure, select "Decompress file data into an intermediate buffer" |
| 40 | |
| 41 | config SQUASHFS_FILE_CACHE |
| 42 | bool "Decompress file data into an intermediate buffer" |
| 43 | help |
| 44 | Decompress file data into an intermediate buffer and then |
| 45 | memcopy it into the page cache. |
| 46 | |
| 47 | config SQUASHFS_FILE_DIRECT |
| 48 | bool "Decompress files directly into the page cache" |
| 49 | help |
| 50 | Directly decompress file data into the page cache. |
| 51 | Doing so can significantly improve performance because |
| 52 | it eliminates a memcpy and it also removes the lock contention |
| 53 | on the single buffer. |
| 54 | |
| 55 | endchoice |
| 56 | |
| 57 | config SQUASHFS_DECOMP_SINGLE |
| 58 | depends on SQUASHFS |
| 59 | def_bool n |
| 60 | |
| 61 | config SQUASHFS_DECOMP_MULTI |
| 62 | depends on SQUASHFS |
| 63 | def_bool n |
| 64 | |
| 65 | config SQUASHFS_DECOMP_MULTI_PERCPU |
| 66 | depends on SQUASHFS |
| 67 | def_bool n |
| 68 | |
| 69 | config SQUASHFS_CHOICE_DECOMP_BY_MOUNT |
| 70 | bool "Select the parallel decompression mode during mount" |
| 71 | depends on SQUASHFS |
| 72 | default n |
| 73 | select SQUASHFS_DECOMP_SINGLE |
| 74 | select SQUASHFS_DECOMP_MULTI |
| 75 | select SQUASHFS_DECOMP_MULTI_PERCPU |
| 76 | select SQUASHFS_MOUNT_DECOMP_THREADS |
| 77 | help |
| 78 | Compile all parallel decompression modes and specify the |
| 79 | decompression mode by setting "threads=" during mount. |
| 80 | default Decompressor parallelisation is SQUASHFS_DECOMP_SINGLE |
| 81 | |
| 82 | choice |
| 83 | prompt "Select decompression parallel mode at compile time" |
| 84 | depends on SQUASHFS |
| 85 | depends on !SQUASHFS_CHOICE_DECOMP_BY_MOUNT |
| 86 | help |
| 87 | Squashfs now supports three parallelisation options for |
| 88 | decompression. Each one exhibits various trade-offs between |
| 89 | decompression performance and CPU and memory usage. |
| 90 | |
| 91 | If in doubt, select "Single threaded compression" |
| 92 | |
| 93 | config SQUASHFS_COMPILE_DECOMP_SINGLE |
| 94 | bool "Single threaded compression" |
| 95 | select SQUASHFS_DECOMP_SINGLE |
| 96 | help |
| 97 | Traditionally Squashfs has used single-threaded decompression. |
| 98 | Only one block (data or metadata) can be decompressed at any |
| 99 | one time. This limits CPU and memory usage to a minimum. |
| 100 | |
| 101 | config SQUASHFS_COMPILE_DECOMP_MULTI |
| 102 | bool "Use multiple decompressors for parallel I/O" |
| 103 | select SQUASHFS_DECOMP_MULTI |
| 104 | help |
| 105 | By default Squashfs uses a single decompressor but it gives |
| 106 | poor performance on parallel I/O workloads when using multiple CPU |
| 107 | machines due to waiting on decompressor availability. |
| 108 | |
| 109 | If you have a parallel I/O workload and your system has enough memory, |
| 110 | using this option may improve overall I/O performance. |
| 111 | |
| 112 | This decompressor implementation uses up to two parallel |
| 113 | decompressors per core. It dynamically allocates decompressors |
| 114 | on a demand basis. |
| 115 | |
| 116 | config SQUASHFS_COMPILE_DECOMP_MULTI_PERCPU |
| 117 | bool "Use percpu multiple decompressors for parallel I/O" |
| 118 | select SQUASHFS_DECOMP_MULTI_PERCPU |
| 119 | help |
| 120 | By default Squashfs uses a single decompressor but it gives |
| 121 | poor performance on parallel I/O workloads when using multiple CPU |
| 122 | machines due to waiting on decompressor availability. |
| 123 | |
| 124 | This decompressor implementation uses a maximum of one |
| 125 | decompressor per core. It uses percpu variables to ensure |
| 126 | decompression is load-balanced across the cores. |
| 127 | endchoice |
| 128 | |
| 129 | config SQUASHFS_MOUNT_DECOMP_THREADS |
| 130 | bool "Add the mount parameter 'threads=' for squashfs" |
| 131 | depends on SQUASHFS |
| 132 | depends on SQUASHFS_DECOMP_MULTI |
| 133 | default n |
| 134 | help |
| 135 | Use threads= to set the decompression parallel mode and the number of threads. |
| 136 | If SQUASHFS_CHOICE_DECOMP_BY_MOUNT=y |
| 137 | threads=<single|multi|percpu|1|2|3|...> |
| 138 | else |
| 139 | threads=<2|3|...> |
| 140 | The upper limit is num_online_cpus() * 2. |
| 141 | |
| 142 | config SQUASHFS_XATTR |
| 143 | bool "Squashfs XATTR support" |
| 144 | depends on SQUASHFS |
| 145 | help |
| 146 | Saying Y here includes support for extended attributes (xattrs). |
| 147 | Xattrs are name:value pairs associated with inodes by |
| 148 | the kernel or by users (see the attr(5) manual page). |
| 149 | |
| 150 | If unsure, say N. |
| 151 | |
| 152 | config SQUASHFS_COMP_CACHE_FULL |
| 153 | bool "Enable full caching of compressed blocks" |
| 154 | depends on SQUASHFS |
| 155 | default n |
| 156 | help |
| 157 | This option enables caching of all compressed blocks, Without caching, |
| 158 | repeated reads of the same files trigger excessive disk I/O, significantly |
| 159 | reducinng performance in workloads like fio-based benchmarks. |
| 160 | |
| 161 | For example, fio tests (iodepth=1, numjobs=1, ioengine=psync) show: |
| 162 | With caching: IOPS=2223, BW=278MiB/s (291MB/s) |
| 163 | Without caching: IOPS=815, BW=102MiB/s (107MB/s) |
| 164 | |
| 165 | Enabling this option restores performance to pre-regression levels by |
| 166 | caching all compressed blocks in the page cache, reducing disk I/O for |
| 167 | repeated reads. However, this increases memory usage, which may be a |
| 168 | concern in memory-constrained environments. |
| 169 | |
| 170 | Enable this option if your workload involves frequent repeated reads and |
| 171 | memory usage is not a limiting factor. If unsure, say N. |
| 172 | |
| 173 | config SQUASHFS_ZLIB |
| 174 | bool "Include support for ZLIB compressed file systems" |
| 175 | depends on SQUASHFS |
| 176 | select ZLIB_INFLATE |
| 177 | default y |
| 178 | help |
| 179 | ZLIB compression is the standard compression used by Squashfs |
| 180 | file systems. It offers a good trade-off between compression |
| 181 | achieved and the amount of CPU time and memory necessary to |
| 182 | compress and decompress. |
| 183 | |
| 184 | If unsure, say Y. |
| 185 | |
| 186 | config SQUASHFS_LZ4 |
| 187 | bool "Include support for LZ4 compressed file systems" |
| 188 | depends on SQUASHFS |
| 189 | select LZ4_DECOMPRESS |
| 190 | help |
| 191 | Saying Y here includes support for reading Squashfs file systems |
| 192 | compressed with LZ4 compression. LZ4 compression is mainly |
| 193 | aimed at embedded systems with slower CPUs where the overheads |
| 194 | of zlib are too high. |
| 195 | |
| 196 | LZ4 is not the standard compression used in Squashfs and so most |
| 197 | file systems will be readable without selecting this option. |
| 198 | |
| 199 | If unsure, say N. |
| 200 | |
| 201 | config SQUASHFS_LZO |
| 202 | bool "Include support for LZO compressed file systems" |
| 203 | depends on SQUASHFS |
| 204 | select LZO_DECOMPRESS |
| 205 | help |
| 206 | Saying Y here includes support for reading Squashfs file systems |
| 207 | compressed with LZO compression. LZO compression is mainly |
| 208 | aimed at embedded systems with slower CPUs where the overheads |
| 209 | of zlib are too high. |
| 210 | |
| 211 | LZO is not the standard compression used in Squashfs and so most |
| 212 | file systems will be readable without selecting this option. |
| 213 | |
| 214 | If unsure, say N. |
| 215 | |
| 216 | config SQUASHFS_XZ |
| 217 | bool "Include support for XZ compressed file systems" |
| 218 | depends on SQUASHFS |
| 219 | select XZ_DEC |
| 220 | help |
| 221 | Saying Y here includes support for reading Squashfs file systems |
| 222 | compressed with XZ compression. XZ gives better compression than |
| 223 | the default zlib compression, at the expense of greater CPU and |
| 224 | memory overhead. |
| 225 | |
| 226 | XZ is not the standard compression used in Squashfs and so most |
| 227 | file systems will be readable without selecting this option. |
| 228 | |
| 229 | If unsure, say N. |
| 230 | |
| 231 | config SQUASHFS_ZSTD |
| 232 | bool "Include support for ZSTD compressed file systems" |
| 233 | depends on SQUASHFS |
| 234 | select ZSTD_DECOMPRESS |
| 235 | help |
| 236 | Saying Y here includes support for reading Squashfs file systems |
| 237 | compressed with ZSTD compression. ZSTD gives better compression than |
| 238 | the default ZLIB compression, while using less CPU. |
| 239 | |
| 240 | ZSTD is not the standard compression used in Squashfs and so most |
| 241 | file systems will be readable without selecting this option. |
| 242 | |
| 243 | If unsure, say N. |
| 244 | |
| 245 | config SQUASHFS_4K_DEVBLK_SIZE |
| 246 | bool "Use 4K device block size?" |
| 247 | depends on SQUASHFS |
| 248 | help |
| 249 | By default Squashfs sets the dev block size (sb_min_blocksize) |
| 250 | to 1K or the smallest block size supported by the block device |
| 251 | (if larger). This, because blocks are packed together and |
| 252 | unaligned in Squashfs, should reduce latency. |
| 253 | |
| 254 | This, however, gives poor performance on MTD NAND devices where |
| 255 | the optimal I/O size is 4K (even though the devices can support |
| 256 | smaller block sizes). |
| 257 | |
| 258 | Using a 4K device block size may also improve overall I/O |
| 259 | performance for some file access patterns (e.g. sequential |
| 260 | accesses of files in filesystem order) on all media. |
| 261 | |
| 262 | Setting this option will force Squashfs to use a 4K device block |
| 263 | size by default. |
| 264 | |
| 265 | If unsure, say N. |
| 266 | |
| 267 | config SQUASHFS_EMBEDDED |
| 268 | bool "Additional option for memory-constrained systems" |
| 269 | depends on SQUASHFS |
| 270 | help |
| 271 | Saying Y here allows you to specify cache size. |
| 272 | |
| 273 | If unsure, say N. |
| 274 | |
| 275 | config SQUASHFS_FRAGMENT_CACHE_SIZE |
| 276 | int "Number of fragments cached" if SQUASHFS_EMBEDDED |
| 277 | depends on SQUASHFS |
| 278 | default "3" |
| 279 | help |
| 280 | By default SquashFS caches the last 3 fragments read from |
| 281 | the filesystem. Increasing this amount may mean SquashFS |
| 282 | has to re-read fragments less often from disk, at the expense |
| 283 | of extra system memory. Decreasing this amount will mean |
| 284 | SquashFS uses less memory at the expense of extra reads from disk. |
| 285 | |
| 286 | Note there must be at least one cached fragment. Anything |
| 287 | much more than three will probably not make much difference. |