[fio.git] / os / os-linux.h

#ifndef FIO_OS_LINUX_H
#define FIO_OS_LINUX_H

#include <sys/ioctl.h>
#include <sys/uio.h>
#include <sys/syscall.h>
#include <sys/vfs.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <sched.h>
#include <linux/unistd.h>
#include <linux/raw.h>
#include <linux/major.h>

#include "indirect.h"

#define FIO_HAVE_LIBAIO
#define FIO_HAVE_POSIXAIO
#define FIO_HAVE_FADVISE
#define FIO_HAVE_CPU_AFFINITY
#define FIO_HAVE_DISK_UTIL
#define FIO_HAVE_SGIO
#define FIO_HAVE_IOPRIO
#define FIO_HAVE_SPLICE
#define FIO_HAVE_IOSCHED_SWITCH
#define FIO_HAVE_ODIRECT
#define FIO_HAVE_HUGETLB
#define FIO_HAVE_RAWBIND
#define FIO_HAVE_BLKTRACE
#define FIO_HAVE_STRSEP
#define FIO_HAVE_FALLOCATE
#define FIO_HAVE_POSIXAIO_FSYNC
#define FIO_HAVE_PSHARED_MUTEX
#define FIO_HAVE_CL_SIZE
#define FIO_HAVE_CGROUPS
#define FIO_HAVE_FDATASYNC
#define FIO_HAVE_FS_STAT

#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
#define FIO_HAVE_SYNC_FILE_RANGE
#endif

#define OS_MAP_ANON		MAP_ANONYMOUS

#ifndef CLOCK_MONOTONIC
#define CLOCK_MONOTONIC 1
#endif

typedef cpu_set_t os_cpu_mask_t;

typedef struct drand48_data os_random_state_t;

/*
 * we want fadvise64 really, but it's so tangled... later
 */
#ifdef FIO_HAVE_FADVISE
#define fadvise(fd, off, len, advice)	\
	posix_fadvise((fd), (off_t)(off), (len), (advice))
#endif

/*
 * If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
 * the affinity helpers to work.
 */
#ifndef GLIBC_2_3_2
#define fio_setaffinity(pid, cpumask)		\
	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
#define fio_getaffinity(pid, ptr)	\
	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
#else
#define fio_setaffinity(pid, cpumask)	\
	sched_setaffinity((pid), &(cpumask))
#define fio_getaffinity(pid, ptr)	\
	sched_getaffinity((pid), (ptr))
#endif

#define fio_cpu_clear(mask, cpu)	CPU_CLR((cpu), (mask))
#define fio_cpu_set(mask, cpu)		CPU_SET((cpu), (mask))

static inline int fio_cpuset_init(os_cpu_mask_t *mask)
{
	CPU_ZERO(mask);
	return 0;
}

static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
{
	return 0;
}

#define FIO_MAX_CPUS			CPU_SETSIZE

static inline int ioprio_set(int which, int who, int ioprio)
{
	return syscall(__NR_ioprio_set, which, who, ioprio);
}

/*
 * Just check for SPLICE_F_MOVE, if that isn't there, assume the others
 * aren't either.
 */
#ifndef SPLICE_F_MOVE
#define SPLICE_F_MOVE	(0x01)	/* move pages instead of copying */
#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
				 /* we may still block on the fd we splice */
				 /* from/to, of course */
#define SPLICE_F_MORE	(0x04)	/* expect more data */
#define SPLICE_F_GIFT   (0x08)  /* pages passed in are a gift */

static inline int splice(int fdin, loff_t *off_in, int fdout, loff_t *off_out,
			 size_t len, unsigned int flags)
{
	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
}

static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
{
	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
}

static inline int vmsplice(int fd, const struct iovec *iov,
			   unsigned long nr_segs, unsigned int flags)
{
	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
}
#endif

#define SPLICE_DEF_SIZE	(64*1024)

#ifdef FIO_HAVE_SYSLET

struct syslet_uatom;
struct async_head_user;

/*
 * syslet stuff
 */
static inline struct syslet_uatom *
async_exec(struct syslet_uatom *atom, struct async_head_user *ahu)
{
	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
}

static inline long
async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	   struct async_head_user *ahu)
{
	return syscall(__NR_async_wait, min_wait_events,
			user_ring_idx, ahu);
}

static inline long async_thread(void *event, struct async_head_user *ahu)
{
	return syscall(__NR_async_thread, event, ahu);
}

static inline long umem_add(unsigned long *uptr, unsigned long inc)
{
	return syscall(__NR_umem_add, uptr, inc);
}
#endif /* FIO_HAVE_SYSLET */

enum {
	IOPRIO_CLASS_NONE,
	IOPRIO_CLASS_RT,
	IOPRIO_CLASS_BE,
	IOPRIO_CLASS_IDLE,
};

enum {
	IOPRIO_WHO_PROCESS = 1,
	IOPRIO_WHO_PGRP,
	IOPRIO_WHO_USER,
};

#define IOPRIO_BITS		16
#define IOPRIO_CLASS_SHIFT	13

#ifndef BLKGETSIZE64
#define BLKGETSIZE64	_IOR(0x12,114,size_t)
#endif

#ifndef BLKFLSBUF
#define BLKFLSBUF	_IO(0x12,97)
#endif

static inline int blockdev_invalidate_cache(int fd)
{
	return ioctl(fd, BLKFLSBUF);
}

static inline int blockdev_size(int fd, unsigned long long *bytes)
{
	if (!ioctl(fd, BLKGETSIZE64, bytes))
		return 0;

	return errno;
}

static inline unsigned long long os_phys_mem(void)
{
	long pagesize, pages;

	pagesize = sysconf(_SC_PAGESIZE);
	pages = sysconf(_SC_PHYS_PAGES);
	if (pages == -1 || pagesize == -1)
		return 0;

	return (unsigned long long) pages * (unsigned long long) pagesize;
}

static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
{
	srand48_r(seed, rs);
}

static inline long os_random_long(os_random_state_t *rs)
{
	long val;

	lrand48_r(rs, &val);
	return val;
}

static inline int fio_lookup_raw(dev_t dev, int *majdev, int *mindev)
{
	struct raw_config_request rq;
	int fd;

	if (major(dev) != RAW_MAJOR)
		return 1;

	/*
	 * we should be able to find /dev/rawctl or /dev/raw/rawctl
	 */
	fd = open("/dev/rawctl", O_RDONLY);
	if (fd < 0) {
		fd = open("/dev/raw/rawctl", O_RDONLY);
		if (fd < 0)
			return 1;
	}

	rq.raw_minor = minor(dev);
	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
		close(fd);
		return 1;
	}

	close(fd);
	*majdev = rq.block_major;
	*mindev = rq.block_minor;
	return 0;
}

#ifdef O_NOATIME
#define FIO_O_NOATIME	O_NOATIME
#else
#define FIO_O_NOATIME	0
#endif

#ifdef MADV_REMOVE
#define FIO_MADV_FREE	MADV_REMOVE
#endif

#define CACHE_LINE_FILE	\
	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"

static inline int arch_cache_line_size(void)
{
	char size[32];
	int fd, ret;

	fd = open(CACHE_LINE_FILE, O_RDONLY);
	if (fd < 0)
		return -1;

	ret = read(fd, size, sizeof(size));

	close(fd);

	if (ret <= 0)
		return -1;
	else
		return atoi(size);
}

static inline unsigned long long get_fs_size(const char *path)
{
	unsigned long long ret;
	struct statfs s;

	if (statfs(path, &s) < 0)
		return -1ULL;

	ret = s.f_bsize;
	ret *= (unsigned long long) s.f_bfree;
	return ret;
}

#endif
Commit	Line	Data
ebac4655 JA	1	#ifndef FIO_OS_LINUX_H
	2	#define FIO_OS_LINUX_H
	3
	4	#include <sys/ioctl.h>
8756e4d4	5	#include <sys/uio.h>
3c39a379	6	#include <sys/syscall.h>
2e3bd4c2	7	#include <sys/vfs.h>
3c39a379	8	#include <unistd.h>
97fbdfa1	9	#include <fcntl.h>
3c2d93ed	10	#include <errno.h>
d09913c2	11	#include <sched.h>
ea421790	12	#include <linux/unistd.h>
07e5b264 JA	13	#include <linux/raw.h>
07e5b264 JA	14	#include <linux/major.h>
ebac4655	15
f3de88a7 JA	16	#include "indirect.h"
f3de88a7 JA	17
ebac4655 JA	18	#define FIO_HAVE_LIBAIO
	19	#define FIO_HAVE_POSIXAIO
	20	#define FIO_HAVE_FADVISE
	21	#define FIO_HAVE_CPU_AFFINITY
	22	#define FIO_HAVE_DISK_UTIL
	23	#define FIO_HAVE_SGIO
ba4f8923	24	#define FIO_HAVE_IOPRIO
8756e4d4	25	#define FIO_HAVE_SPLICE
22f78b32	26	#define FIO_HAVE_IOSCHED_SWITCH
2c0ecd28	27	#define FIO_HAVE_ODIRECT
74b025b0	28	#define FIO_HAVE_HUGETLB
07e5b264	29	#define FIO_HAVE_RAWBIND
5e62c22a	30	#define FIO_HAVE_BLKTRACE
5921e80c	31	#define FIO_HAVE_STRSEP
fffca02d	32	#define FIO_HAVE_FALLOCATE
207cb0f0	33	#define FIO_HAVE_POSIXAIO_FSYNC
f356d01d	34	#define FIO_HAVE_PSHARED_MUTEX
eb7ccf38	35	#define FIO_HAVE_CL_SIZE
a696fa2a	36	#define FIO_HAVE_CGROUPS
c36d16f5	37	#define FIO_HAVE_FDATASYNC
2e3bd4c2	38	#define FIO_HAVE_FS_STAT
b6959b55	39
093729f7 JA	40	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	41	#define FIO_HAVE_SYNC_FILE_RANGE
	42	#endif
	43
dc873b6f	44	#define OS_MAP_ANON MAP_ANONYMOUS
ebac4655	45
4d8947de SL	46	#ifndef CLOCK_MONOTONIC
	47	#define CLOCK_MONOTONIC 1
	48	#endif
	49
ebac4655	50	typedef cpu_set_t os_cpu_mask_t;
4e78e405	51
6dfd46b9	52	typedef struct drand48_data os_random_state_t;
ebac4655 JA	53
	54	/*
	55	* we want fadvise64 really, but it's so tangled... later
	56	*/
4d8947de	57	#ifdef FIO_HAVE_FADVISE
ebac4655 JA	58	#define fadvise(fd, off, len, advice) \
ebac4655 JA	59	posix_fadvise((fd), (off_t)(off), (len), (advice))
4d8947de	60	#endif
ebac4655	61
c8f025f8 JA	62	/*
	63	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	64	* the affinity helpers to work.
	65	*/
	66	#ifndef GLIBC_2_3_2
e8462bd8 JA	67	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	68	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
ebac4655 JA	69	#define fio_getaffinity(pid, ptr) \
ebac4655 JA	70	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
c8f025f8	71	#else
e8462bd8 JA	72	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	73	sched_setaffinity((pid), &(cpumask))
c8f025f8 JA	74	#define fio_getaffinity(pid, ptr) \
	75	sched_getaffinity((pid), (ptr))
	76	#endif
be4ecfdf JA	77
be4ecfdf JA	78	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
6d459ee7	79	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
d2ce18b5 JA	80
	81	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	82	{
	83	CPU_ZERO(mask);
	84	return 0;
	85	}
	86
	87	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	88	{
	89	return 0;
	90	}
6d459ee7 JA	91
6d459ee7 JA	92	#define FIO_MAX_CPUS CPU_SETSIZE
4d8947de	93
ebac4655 JA	94	static inline int ioprio_set(int which, int who, int ioprio)
	95	{
	96	return syscall(__NR_ioprio_set, which, who, ioprio);
	97	}
	98
97fbdfa1 JA	99	/*
	100	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	101	* aren't either.
	102	*/
	103	#ifndef SPLICE_F_MOVE
	104	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	105	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	106	/* we may still block on the fd we splice */
	107	/* from/to, of course */
	108	#define SPLICE_F_MORE (0x04) /* expect more data */
	109	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
8756e4d4 JA	110
8756e4d4 JA	111	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
495ee9b6	112	size_t len, unsigned int flags)
8756e4d4	113	{
97fbdfa1	114	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
8756e4d4 JA	115	}
	116
	117	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	118	{
97fbdfa1	119	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
8756e4d4 JA	120	}
	121
	122	static inline int vmsplice(int fd, const struct iovec *iov,
	123	unsigned long nr_segs, unsigned int flags)
	124	{
97fbdfa1	125	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
8756e4d4	126	}
97fbdfa1	127	#endif
8756e4d4	128
3feedc60 JA	129	#define SPLICE_DEF_SIZE (64*1024)
3feedc60 JA	130
b6959b55	131	#ifdef FIO_HAVE_SYSLET
bf0dc8fa IM	132
	133	struct syslet_uatom;
	134	struct async_head_user;
	135
a4f4fdd7 JA	136	/*
	137	* syslet stuff
	138	*/
bf0dc8fa IM	139	static inline struct syslet_uatom *
bf0dc8fa IM	140	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
a4f4fdd7	141	{
2b52511f	142	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
a4f4fdd7 JA	143	}
a4f4fdd7 JA	144
bf0dc8fa IM	145	static inline long
	146	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	147	struct async_head_user *ahu)
a4f4fdd7	148	{
bf0dc8fa IM	149	return syscall(__NR_async_wait, min_wait_events,
bf0dc8fa IM	150	user_ring_idx, ahu);
a4f4fdd7 JA	151	}
a4f4fdd7 JA	152
7756b0d0	153	static inline long async_thread(void event, struct async_head_user ahu)
a4f4fdd7	154	{
7756b0d0	155	return syscall(__NR_async_thread, event, ahu);
a4f4fdd7 JA	156	}
	157
	158	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	159	{
	160	return syscall(__NR_umem_add, uptr, inc);
	161	}
b6959b55	162	#endif /* FIO_HAVE_SYSLET */
a4f4fdd7	163
6cefbe33 JA	164	enum {
	165	IOPRIO_CLASS_NONE,
	166	IOPRIO_CLASS_RT,
	167	IOPRIO_CLASS_BE,
	168	IOPRIO_CLASS_IDLE,
	169	};
	170
ebac4655 JA	171	enum {
	172	IOPRIO_WHO_PROCESS = 1,
	173	IOPRIO_WHO_PGRP,
	174	IOPRIO_WHO_USER,
	175	};
	176
6cefbe33	177	#define IOPRIO_BITS 16
ebac4655 JA	178	#define IOPRIO_CLASS_SHIFT 13
	179
	180	#ifndef BLKGETSIZE64
	181	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	182	#endif
	183
e5b401d4 JA	184	#ifndef BLKFLSBUF
	185	#define BLKFLSBUF _IO(0x12,97)
	186	#endif
	187
	188	static inline int blockdev_invalidate_cache(int fd)
	189	{
a16211bb	190	return ioctl(fd, BLKFLSBUF);
e5b401d4 JA	191	}
e5b401d4 JA	192
9104f874	193	static inline int blockdev_size(int fd, unsigned long long *bytes)
ebac4655 JA	194	{
	195	if (!ioctl(fd, BLKGETSIZE64, bytes))
	196	return 0;
	197
	198	return errno;
	199	}
	200
32cd46a0 JA	201	static inline unsigned long long os_phys_mem(void)
	202	{
	203	long pagesize, pages;
	204
	205	pagesize = sysconf(_SC_PAGESIZE);
	206	pages = sysconf(_SC_PHYS_PAGES);
	207	if (pages == -1 \|\| pagesize == -1)
	208	return 0;
	209
	210	return (unsigned long long) pages * (unsigned long long) pagesize;
	211	}
	212
6dfd46b9 JA	213	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	214	{
	215	srand48_r(seed, rs);
	216	}
	217
	218	static inline long os_random_long(os_random_state_t *rs)
	219	{
	220	long val;
	221
	222	lrand48_r(rs, &val);
	223	return val;
	224	}
	225
8cc7afa9	226	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
07e5b264 JA	227	{
	228	struct raw_config_request rq;
	229	int fd;
	230
	231	if (major(dev) != RAW_MAJOR)
8cc7afa9	232	return 1;
07e5b264 JA	233
	234	/*
	235	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	236	*/
	237	fd = open("/dev/rawctl", O_RDONLY);
	238	if (fd < 0) {
	239	fd = open("/dev/raw/rawctl", O_RDONLY);
	240	if (fd < 0)
8cc7afa9	241	return 1;
07e5b264 JA	242	}
	243
	244	rq.raw_minor = minor(dev);
	245	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	246	close(fd);
8cc7afa9	247	return 1;
07e5b264 JA	248	}
	249
	250	close(fd);
	251	*majdev = rq.block_major;
	252	*mindev = rq.block_minor;
8cc7afa9	253	return 0;
07e5b264 JA	254	}
07e5b264 JA	255
5b6f5c66	256	#ifdef O_NOATIME
5921e80c	257	#define FIO_O_NOATIME O_NOATIME
5b6f5c66 JA	258	#else
	259	#define FIO_O_NOATIME 0
	260	#endif
5921e80c	261
a1c58075 JA	262	#ifdef MADV_REMOVE
	263	#define FIO_MADV_FREE MADV_REMOVE
	264	#endif
	265
eb7ccf38 JA	266	#define CACHE_LINE_FILE \
	267	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	268
	269	static inline int arch_cache_line_size(void)
	270	{
	271	char size[32];
	272	int fd, ret;
	273
	274	fd = open(CACHE_LINE_FILE, O_RDONLY);
	275	if (fd < 0)
	276	return -1;
	277
	278	ret = read(fd, size, sizeof(size));
ec76f15a JA	279
	280	close(fd);
	281
eb7ccf38 JA	282	if (ret <= 0)
eb7ccf38 JA	283	return -1;
ec76f15a JA	284	else
ec76f15a JA	285	return atoi(size);
eb7ccf38 JA	286	}
eb7ccf38 JA	287
2e3bd4c2 JA	288	static inline unsigned long long get_fs_size(const char *path)
	289	{
	290	unsigned long long ret;
	291	struct statfs s;
	292
	293	if (statfs(path, &s) < 0)
	294	return -1ULL;
	295
	296	ret = s.f_bsize;
	297	ret *= (unsigned long long) s.f_bfree;
	298	return ret;
	299	}
	300
ebac4655	301	#endif