os-linux.h revision 6d459ee7e90393af490a64709aec206d8a585ebd
1#ifndef FIO_OS_LINUX_H
2#define FIO_OS_LINUX_H
3
4#include <sys/ioctl.h>
5#include <sys/uio.h>
6#include <sys/syscall.h>
7#include <unistd.h>
8#include <fcntl.h>
9#include <linux/unistd.h>
10#include <linux/raw.h>
11#include <linux/major.h>
12
13#include "indirect.h"
14
15#define FIO_HAVE_LIBAIO
16#define FIO_HAVE_POSIXAIO
17#define FIO_HAVE_FADVISE
18#define FIO_HAVE_CPU_AFFINITY
19#define FIO_HAVE_DISK_UTIL
20#define FIO_HAVE_SGIO
21#define FIO_HAVE_IOPRIO
22#define FIO_HAVE_SPLICE
23#define FIO_HAVE_IOSCHED_SWITCH
24#define FIO_HAVE_ODIRECT
25#define FIO_HAVE_HUGETLB
26#define FIO_HAVE_RAWBIND
27#define FIO_HAVE_BLKTRACE
28#define FIO_HAVE_STRSEP
29#define FIO_HAVE_FALLOCATE
30#define FIO_HAVE_POSIXAIO_FSYNC
31
32#define OS_MAP_ANON		MAP_ANONYMOUS
33
34#ifndef CLOCK_MONOTONIC
35#define CLOCK_MONOTONIC 1
36#endif
37
38#ifdef FIO_HAVE_CPU_AFFINITY
39typedef cpu_set_t os_cpu_mask_t;
40#else
41typedef int os_cpu_mask_t;
42#endif
43typedef struct drand48_data os_random_state_t;
44
45/*
46 * we want fadvise64 really, but it's so tangled... later
47 */
48#ifdef FIO_HAVE_FADVISE
49#define fadvise(fd, off, len, advice)	\
50	posix_fadvise((fd), (off_t)(off), (len), (advice))
51#endif
52
53/*
54 * If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
55 * the affinity helpers to work.
56 */
57#ifndef GLIBC_2_3_2
58#define fio_setaffinity(td)		\
59	sched_setaffinity((td)->pid, sizeof((td)->o.cpumask), &(td)->o.cpumask)
60#define fio_getaffinity(pid, ptr)	\
61	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
62#else
63#define fio_setaffinity(td)		\
64	sched_setaffinity((td)->pid, &(td)->o.cpumask)
65#define fio_getaffinity(pid, ptr)	\
66	sched_getaffinity((pid), (ptr))
67#endif
68
69#define fio_cpu_clear(mask, cpu)	CPU_CLR((cpu), (mask))
70#define fio_cpu_set(mask, cpu)		CPU_SET((cpu), (mask))
71#define fio_cpuset_init(td)		CPU_ZERO(&(td)->o.cpumask)
72
73#define FIO_MAX_CPUS			CPU_SETSIZE
74
75static inline int ioprio_set(int which, int who, int ioprio)
76{
77	return syscall(__NR_ioprio_set, which, who, ioprio);
78}
79
80/*
81 * Just check for SPLICE_F_MOVE, if that isn't there, assume the others
82 * aren't either.
83 */
84#ifndef SPLICE_F_MOVE
85#define SPLICE_F_MOVE	(0x01)	/* move pages instead of copying */
86#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
87				 /* we may still block on the fd we splice */
88				 /* from/to, of course */
89#define SPLICE_F_MORE	(0x04)	/* expect more data */
90#define SPLICE_F_GIFT   (0x08)  /* pages passed in are a gift */
91
92static inline int splice(int fdin, loff_t *off_in, int fdout, loff_t *off_out,
93			 size_t len, unsigned int flags)
94{
95	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
96}
97
98static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
99{
100	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
101}
102
103static inline int vmsplice(int fd, const struct iovec *iov,
104			   unsigned long nr_segs, unsigned int flags)
105{
106	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
107}
108#endif
109
110#define SPLICE_DEF_SIZE	(64*1024)
111
112#ifdef FIO_HAVE_SYSLET
113
114struct syslet_uatom;
115struct async_head_user;
116
117/*
118 * syslet stuff
119 */
120static inline struct syslet_uatom *
121async_exec(struct syslet_uatom *atom, struct async_head_user *ahu)
122{
123	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
124}
125
126static inline long
127async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
128	   struct async_head_user *ahu)
129{
130	return syscall(__NR_async_wait, min_wait_events,
131			user_ring_idx, ahu);
132}
133
134static inline long async_thread(void *event, struct async_head_user *ahu)
135{
136	return syscall(__NR_async_thread, event, ahu);
137}
138
139static inline long umem_add(unsigned long *uptr, unsigned long inc)
140{
141	return syscall(__NR_umem_add, uptr, inc);
142}
143#endif /* FIO_HAVE_SYSLET */
144
145enum {
146	IOPRIO_CLASS_NONE,
147	IOPRIO_CLASS_RT,
148	IOPRIO_CLASS_BE,
149	IOPRIO_CLASS_IDLE,
150};
151
152enum {
153	IOPRIO_WHO_PROCESS = 1,
154	IOPRIO_WHO_PGRP,
155	IOPRIO_WHO_USER,
156};
157
158#define IOPRIO_BITS		16
159#define IOPRIO_CLASS_SHIFT	13
160
161#ifndef BLKGETSIZE64
162#define BLKGETSIZE64	_IOR(0x12,114,size_t)
163#endif
164
165#ifndef BLKFLSBUF
166#define BLKFLSBUF	_IO(0x12,97)
167#endif
168
169static inline int blockdev_invalidate_cache(int fd)
170{
171	return ioctl(fd, BLKFLSBUF);
172}
173
174static inline int blockdev_size(int fd, unsigned long long *bytes)
175{
176	if (!ioctl(fd, BLKGETSIZE64, bytes))
177		return 0;
178
179	return errno;
180}
181
182static inline unsigned long long os_phys_mem(void)
183{
184	long pagesize, pages;
185
186	pagesize = sysconf(_SC_PAGESIZE);
187	pages = sysconf(_SC_PHYS_PAGES);
188	if (pages == -1 || pagesize == -1)
189		return 0;
190
191	return (unsigned long long) pages * (unsigned long long) pagesize;
192}
193
194static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
195{
196	srand48_r(seed, rs);
197}
198
199static inline long os_random_long(os_random_state_t *rs)
200{
201	long val;
202
203	lrand48_r(rs, &val);
204	return val;
205}
206
207static inline int fio_lookup_raw(dev_t dev, int *majdev, int *mindev)
208{
209	struct raw_config_request rq;
210	int fd;
211
212	if (major(dev) != RAW_MAJOR)
213		return 1;
214
215	/*
216	 * we should be able to find /dev/rawctl or /dev/raw/rawctl
217	 */
218	fd = open("/dev/rawctl", O_RDONLY);
219	if (fd < 0) {
220		fd = open("/dev/raw/rawctl", O_RDONLY);
221		if (fd < 0)
222			return 1;
223	}
224
225	rq.raw_minor = minor(dev);
226	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
227		close(fd);
228		return 1;
229	}
230
231	close(fd);
232	*majdev = rq.block_major;
233	*mindev = rq.block_minor;
234	return 0;
235}
236
237#ifdef O_NOATIME
238#define FIO_O_NOATIME	O_NOATIME
239#else
240#define FIO_O_NOATIME	0
241#endif
242
243#endif
244