1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ============== 3 ==============
4 Fuse I/O Modes 4 Fuse I/O Modes
5 ============== 5 ==============
6 6
7 Fuse supports the following I/O modes: 7 Fuse supports the following I/O modes:
8 8
9 - direct-io 9 - direct-io
10 - cached 10 - cached
11 + write-through 11 + write-through
12 + writeback-cache 12 + writeback-cache
13 13
14 The direct-io mode can be selected with the FO 14 The direct-io mode can be selected with the FOPEN_DIRECT_IO flag in the
15 FUSE_OPEN reply. 15 FUSE_OPEN reply.
16 16
17 In direct-io mode the page cache is completely 17 In direct-io mode the page cache is completely bypassed for reads and writes.
18 No read-ahead takes place. Shared mmap is disa !! 18 No read-ahead takes place. Shared mmap is disabled.
19 mmap, the FUSE_DIRECT_IO_ALLOW_MMAP flag may b <<
20 19
21 In cached mode reads may be satisfied from the 20 In cached mode reads may be satisfied from the page cache, and data may be
22 read-ahead by the kernel to fill the cache. T 21 read-ahead by the kernel to fill the cache. The cache is always kept consistent
23 after any writes to the file. All mmap modes 22 after any writes to the file. All mmap modes are supported.
24 23
25 The cached mode has two sub modes controlling 24 The cached mode has two sub modes controlling how writes are handled. The
26 write-through mode is the default and is suppo 25 write-through mode is the default and is supported on all kernels. The
27 writeback-cache mode may be selected by the FU 26 writeback-cache mode may be selected by the FUSE_WRITEBACK_CACHE flag in the
28 FUSE_INIT reply. 27 FUSE_INIT reply.
29 28
30 In write-through mode each write is immediatel 29 In write-through mode each write is immediately sent to userspace as one or more
31 WRITE requests, as well as updating any cached 30 WRITE requests, as well as updating any cached pages (and caching previously
32 uncached, but fully written pages). No READ r 31 uncached, but fully written pages). No READ requests are ever sent for writes,
33 so when an uncached page is partially written, 32 so when an uncached page is partially written, the page is discarded.
34 33
35 In writeback-cache mode (enabled by the FUSE_W 34 In writeback-cache mode (enabled by the FUSE_WRITEBACK_CACHE flag) writes go to
36 the cache only, which means that the write(2) 35 the cache only, which means that the write(2) syscall can often complete very
37 fast. Dirty pages are written back implicitly 36 fast. Dirty pages are written back implicitly (background writeback or page
38 reclaim on memory pressure) or explicitly (inv 37 reclaim on memory pressure) or explicitly (invoked by close(2), fsync(2) and
39 when the last ref to the file is being release 38 when the last ref to the file is being released on munmap(2)). This mode
40 assumes that all changes to the filesystem go 39 assumes that all changes to the filesystem go through the FUSE kernel module
41 (size and atime/ctime/mtime attributes are kep 40 (size and atime/ctime/mtime attributes are kept up-to-date by the kernel), so
42 it's generally not suitable for network filesy 41 it's generally not suitable for network filesystems. If a partial page is
43 written, then the page needs to be first read 42 written, then the page needs to be first read from userspace. This means, that
44 even for files opened for O_WRONLY it is possi 43 even for files opened for O_WRONLY it is possible that READ requests will be
45 generated by the kernel. 44 generated by the kernel.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.