os-linux.h revision 495ee9b632c5629e79a46d1074cacf4eda965e13 
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#define FIO_HAVE_LIBAIO
14#define FIO_HAVE_POSIXAIO
15#define FIO_HAVE_FADVISE
16#define FIO_HAVE_CPU_AFFINITY
17#define FIO_HAVE_DISK_UTIL
18#define FIO_HAVE_SGIO
19#define FIO_HAVE_IOPRIO
20#define FIO_HAVE_SPLICE
21#define FIO_HAVE_IOSCHED_SWITCH
22#define FIO_HAVE_ODIRECT
23#define FIO_HAVE_HUGETLB
24#define FIO_HAVE_RAWBIND
25#define FIO_HAVE_BLKTRACE
26
27#define OS_MAP_ANON		(MAP_ANONYMOUS)
28
29typedef cpu_set_t os_cpu_mask_t;
30typedef struct drand48_data os_random_state_t;
31
32/*
33 * we want fadvise64 really, but it's so tangled... later
34 */
35#define fadvise(fd, off, len, advice)	\
36	posix_fadvise((fd), (off_t)(off), (len), (advice))
37
38/*
39 * If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
40 * the affinity helpers to work.
41 */
42#ifndef GLIBC_2_3_2
43#define fio_setaffinity(td)		\
44	sched_setaffinity((td)->pid, sizeof((td)->o.cpumask), &(td)->o.cpumask)
45#define fio_getaffinity(pid, ptr)	\
46	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
47#else
48#define fio_setaffinity(td)		\
49	sched_setaffinity((td)->pid, &(td)->o.cpumask)
50#define fio_getaffinity(pid, ptr)	\
51	sched_getaffinity((pid), (ptr))
52#endif
53
54static inline int ioprio_set(int which, int who, int ioprio)
55{
56	return syscall(__NR_ioprio_set, which, who, ioprio);
57}
58
59/*
60 * Just check for SPLICE_F_MOVE, if that isn't there, assume the others
61 * aren't either.
62 */
63#ifndef SPLICE_F_MOVE
64#define SPLICE_F_MOVE	(0x01)	/* move pages instead of copying */
65#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
66				 /* we may still block on the fd we splice */
67				 /* from/to, of course */
68#define SPLICE_F_MORE	(0x04)	/* expect more data */
69#define SPLICE_F_GIFT   (0x08)  /* pages passed in are a gift */
70
71static inline int splice(int fdin, loff_t *off_in, int fdout, loff_t *off_out,
72			 size_t len, unsigned int flags)
73{
74	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
75}
76
77static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
78{
79	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
80}
81
82static inline int vmsplice(int fd, const struct iovec *iov,
83			   unsigned long nr_segs, unsigned int flags)
84{
85	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
86}
87#endif
88
89#define SPLICE_DEF_SIZE	(64*1024)
90
91#ifdef FIO_HAVE_SYSLET
92
93struct syslet_uatom;
94struct async_head_user;
95
96/*
97 * syslet stuff
98 */
99static inline struct syslet_uatom *
100async_exec(struct syslet_uatom *atom, struct async_head_user *ahu)
101{
102	return (void *) syscall(__NR_async_exec, atom, ahu);
103}
104
105static inline long
106async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
107	   struct async_head_user *ahu)
108{
109	return syscall(__NR_async_wait, min_wait_events,
110			user_ring_idx, ahu);
111}
112
113static inline long async_thread(void *event, struct async_head_user *ahu)
114{
115	return syscall(__NR_async_thread, event, ahu);
116}
117
118static inline long umem_add(unsigned long *uptr, unsigned long inc)
119{
120	return syscall(__NR_umem_add, uptr, inc);
121}
122#endif /* FIO_HAVE_SYSLET */
123
124enum {
125	IOPRIO_CLASS_NONE,
126	IOPRIO_CLASS_RT,
127	IOPRIO_CLASS_BE,
128	IOPRIO_CLASS_IDLE,
129};
130
131enum {
132	IOPRIO_WHO_PROCESS = 1,
133	IOPRIO_WHO_PGRP,
134	IOPRIO_WHO_USER,
135};
136
137#define IOPRIO_BITS		16
138#define IOPRIO_CLASS_SHIFT	13
139
140#ifndef BLKGETSIZE64
141#define BLKGETSIZE64	_IOR(0x12,114,size_t)
142#endif
143
144#ifndef BLKFLSBUF
145#define BLKFLSBUF	_IO(0x12,97)
146#endif
147
148static inline int blockdev_invalidate_cache(int fd)
149{
150	return ioctl(fd, BLKFLSBUF);
151}
152
153static inline int blockdev_size(int fd, unsigned long long *bytes)
154{
155	if (!ioctl(fd, BLKGETSIZE64, bytes))
156		return 0;
157
158	return errno;
159}
160
161static inline unsigned long long os_phys_mem(void)
162{
163	long pagesize, pages;
164
165	pagesize = sysconf(_SC_PAGESIZE);
166	pages = sysconf(_SC_PHYS_PAGES);
167	if (pages == -1 || pagesize == -1)
168		return 0;
169
170	return (unsigned long long) pages * (unsigned long long) pagesize;
171}
172
173static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
174{
175	srand48_r(seed, rs);
176}
177
178static inline long os_random_long(os_random_state_t *rs)
179{
180	long val;
181
182	lrand48_r(rs, &val);
183	return val;
184}
185
186static inline double os_random_double(os_random_state_t *rs)
187{
188	double val;
189
190	drand48_r(rs, &val);
191	return val;
192}
193
194static inline int fio_lookup_raw(dev_t dev, int *majdev, int *mindev)
195{
196	struct raw_config_request rq;
197	int fd;
198
199	if (major(dev) != RAW_MAJOR)
200		return 1;
201
202	/*
203	 * we should be able to find /dev/rawctl or /dev/raw/rawctl
204	 */
205	fd = open("/dev/rawctl", O_RDONLY);
206	if (fd < 0) {
207		fd = open("/dev/raw/rawctl", O_RDONLY);
208		if (fd < 0)
209			return 1;
210	}
211
212	rq.raw_minor = minor(dev);
213	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
214		close(fd);
215		return 1;
216	}
217
218	close(fd);
219	*majdev = rq.block_major;
220	*mindev = rq.block_minor;
221	return 0;
222}
223
224#endif
225