xref: /linux/Documentation/mm/remap_file_pages.rst (revision 4359a011e259a4608afc7fb3635370c9d4ba5943)
1.. _remap_file_pages:
2
3==============================
4remap_file_pages() system call
5==============================
6
7The remap_file_pages() system call is used to create a nonlinear mapping,
8that is, a mapping in which the pages of the file are mapped into a
9nonsequential order in memory. The advantage of using remap_file_pages()
10over using repeated calls to mmap(2) is that the former approach does not
11require the kernel to create additional VMA (Virtual Memory Area) data
12structures.
13
14Supporting of nonlinear mapping requires significant amount of non-trivial
15code in kernel virtual memory subsystem including hot paths. Also to get
16nonlinear mapping work kernel need a way to distinguish normal page table
17entries from entries with file offset (pte_file). Kernel reserves flag in
18PTE for this purpose. PTE flags are scarce resource especially on some CPU
19architectures. It would be nice to free up the flag for other usage.
20
21Fortunately, there are not many users of remap_file_pages() in the wild.
22It's only known that one enterprise RDBMS implementation uses the syscall
23on 32-bit systems to map files bigger than can linearly fit into 32-bit
24virtual address space. This use-case is not critical anymore since 64-bit
25systems are widely available.
26
27The syscall is deprecated and replaced it with an emulation now. The
28emulation creates new VMAs instead of nonlinear mappings. It's going to
29work slower for rare users of remap_file_pages() but ABI is preserved.
30
31One side effect of emulation (apart from performance) is that user can hit
32vm.max_map_count limit more easily due to additional VMAs. See comment for
33DEFAULT_MAX_MAP_COUNT for more details on the limit.
34