1ALLOC_HUGEPAGES(2) Linux Programmer's Manual ALLOC_HUGEPAGES(2)
2
3
4
6 alloc_hugepages, free_hugepages - allocate or free huge pages
7
9 void *syscall(SYS_alloc_hugepages, int key, void *addr, size_t len,
10 int prot, int flag);
11 int syscall(SYS_free_hugepages, void *addr);
12
13 Note: glibc provides no wrappers for these system calls, necessitating
14 the use of syscall(2).
15
17 The system calls alloc_hugepages() and free_hugepages() were introduced
18 in Linux 2.5.36 and removed again in 2.5.54. They existed only on i386
19 and ia64 (when built with CONFIG_HUGETLB_PAGE). In Linux 2.4.20, the
20 syscall numbers exist, but the calls fail with the error ENOSYS.
21
22 On i386 the memory management hardware knows about ordinary pages
23 (4 KiB) and huge pages (2 or 4 MiB). Similarly ia64 knows about huge
24 pages of several sizes. These system calls serve to map huge pages
25 into the process's memory or to free them again. Huge pages are locked
26 into memory, and are not swapped.
27
28 The key argument is an identifier. When zero the pages are private,
29 and not inherited by children. When positive the pages are shared with
30 other applications using the same key, and inherited by child pro‐
31 cesses.
32
33 The addr argument of free_hugepages() tells which page is being freed:
34 it was the return value of a call to alloc_hugepages(). (The memory is
35 first actually freed when all users have released it.) The addr argu‐
36 ment of alloc_hugepages() is a hint, that the kernel may or may not
37 follow. Addresses must be properly aligned.
38
39 The len argument is the length of the required segment. It must be a
40 multiple of the huge page size.
41
42 The prot argument specifies the memory protection of the segment. It
43 is one of PROT_READ, PROT_WRITE, PROT_EXEC.
44
45 The flag argument is ignored, unless key is positive. In that case, if
46 flag is IPC_CREAT, then a new huge page segment is created when none
47 with the given key existed. If this flag is not set, then ENOENT is
48 returned when no segment with the given key exists.
49
51 On success, alloc_hugepages() returns the allocated virtual address,
52 and free_hugepages() returns zero. On error, -1 is returned, and errno
53 is set to indicate the error.
54
56 ENOSYS The system call is not supported on this kernel.
57
59 /proc/sys/vm/nr_hugepages
60 Number of configured hugetlb pages. This can be read and writ‐
61 ten.
62
63 /proc/meminfo
64 Gives info on the number of configured hugetlb pages and on
65 their size in the three variables HugePages_Total,
66 HugePages_Free, Hugepagesize.
67
69 These extinct system calls were specific to Linux on Intel processors.
70
72 These system calls are gone; they existed only in Linux 2.5.36 through
73 to 2.5.54. Now the hugetlbfs filesystem can be used instead. Memory
74 backed by huge pages (if the CPU supports them) is obtained by using
75 mmap(2) to map files in this virtual filesystem.
76
77 The maximal number of huge pages can be specified using the hugepages=
78 boot parameter.
79
81 This page is part of release 5.13 of the Linux man-pages project. A
82 description of the project, information about reporting bugs, and the
83 latest version of this page, can be found at
84 https://www.kernel.org/doc/man-pages/.
85
86
87
88Linux 2021-03-22 ALLOC_HUGEPAGES(2)