[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_SYNC_FILE_RANGE
#define FIO_HAVE_FS_STAT

#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
44f29692	38	#define FIO_HAVE_SYNC_FILE_RANGE
2e3bd4c2	39	#define FIO_HAVE_FS_STAT
b6959b55	40
dc873b6f	41	#define OS_MAP_ANON MAP_ANONYMOUS
ebac4655	42
4d8947de SL	43	#ifndef CLOCK_MONOTONIC
	44	#define CLOCK_MONOTONIC 1
	45	#endif
	46
ebac4655	47	typedef cpu_set_t os_cpu_mask_t;
4e78e405	48
6dfd46b9	49	typedef struct drand48_data os_random_state_t;
ebac4655 JA	50
	51	/*
	52	* we want fadvise64 really, but it's so tangled... later
	53	*/
4d8947de	54	#ifdef FIO_HAVE_FADVISE
ebac4655 JA	55	#define fadvise(fd, off, len, advice) \
ebac4655 JA	56	posix_fadvise((fd), (off_t)(off), (len), (advice))
4d8947de	57	#endif
ebac4655	58
c8f025f8 JA	59	/*
	60	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	61	* the affinity helpers to work.
	62	*/
	63	#ifndef GLIBC_2_3_2
e8462bd8 JA	64	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	65	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
ebac4655 JA	66	#define fio_getaffinity(pid, ptr) \
ebac4655 JA	67	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
c8f025f8	68	#else
e8462bd8 JA	69	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	70	sched_setaffinity((pid), &(cpumask))
c8f025f8 JA	71	#define fio_getaffinity(pid, ptr) \
	72	sched_getaffinity((pid), (ptr))
	73	#endif
be4ecfdf JA	74
be4ecfdf JA	75	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
6d459ee7	76	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
d2ce18b5 JA	77
	78	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	79	{
	80	CPU_ZERO(mask);
	81	return 0;
	82	}
	83
	84	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	85	{
	86	return 0;
	87	}
6d459ee7 JA	88
6d459ee7 JA	89	#define FIO_MAX_CPUS CPU_SETSIZE
4d8947de	90
ebac4655 JA	91	static inline int ioprio_set(int which, int who, int ioprio)
	92	{
	93	return syscall(__NR_ioprio_set, which, who, ioprio);
	94	}
	95
97fbdfa1 JA	96	/*
	97	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	98	* aren't either.
	99	*/
	100	#ifndef SPLICE_F_MOVE
	101	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	102	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	103	/* we may still block on the fd we splice */
	104	/* from/to, of course */
	105	#define SPLICE_F_MORE (0x04) /* expect more data */
	106	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
8756e4d4 JA	107
8756e4d4 JA	108	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
495ee9b6	109	size_t len, unsigned int flags)
8756e4d4	110	{
97fbdfa1	111	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
8756e4d4 JA	112	}
	113
	114	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	115	{
97fbdfa1	116	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
8756e4d4 JA	117	}
	118
	119	static inline int vmsplice(int fd, const struct iovec *iov,
	120	unsigned long nr_segs, unsigned int flags)
	121	{
97fbdfa1	122	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
8756e4d4	123	}
97fbdfa1	124	#endif
8756e4d4	125
3feedc60 JA	126	#define SPLICE_DEF_SIZE (64*1024)
3feedc60 JA	127
b6959b55	128	#ifdef FIO_HAVE_SYSLET
bf0dc8fa IM	129
	130	struct syslet_uatom;
	131	struct async_head_user;
	132
a4f4fdd7 JA	133	/*
	134	* syslet stuff
	135	*/
bf0dc8fa IM	136	static inline struct syslet_uatom *
bf0dc8fa IM	137	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
a4f4fdd7	138	{
2b52511f	139	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
a4f4fdd7 JA	140	}
a4f4fdd7 JA	141
bf0dc8fa IM	142	static inline long
	143	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	144	struct async_head_user *ahu)
a4f4fdd7	145	{
bf0dc8fa IM	146	return syscall(__NR_async_wait, min_wait_events,
bf0dc8fa IM	147	user_ring_idx, ahu);
a4f4fdd7 JA	148	}
a4f4fdd7 JA	149
7756b0d0	150	static inline long async_thread(void event, struct async_head_user ahu)
a4f4fdd7	151	{
7756b0d0	152	return syscall(__NR_async_thread, event, ahu);
a4f4fdd7 JA	153	}
	154
	155	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	156	{
	157	return syscall(__NR_umem_add, uptr, inc);
	158	}
b6959b55	159	#endif /* FIO_HAVE_SYSLET */
a4f4fdd7	160
6cefbe33 JA	161	enum {
	162	IOPRIO_CLASS_NONE,
	163	IOPRIO_CLASS_RT,
	164	IOPRIO_CLASS_BE,
	165	IOPRIO_CLASS_IDLE,
	166	};
	167
ebac4655 JA	168	enum {
	169	IOPRIO_WHO_PROCESS = 1,
	170	IOPRIO_WHO_PGRP,
	171	IOPRIO_WHO_USER,
	172	};
	173
6cefbe33	174	#define IOPRIO_BITS 16
ebac4655 JA	175	#define IOPRIO_CLASS_SHIFT 13
	176
	177	#ifndef BLKGETSIZE64
	178	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	179	#endif
	180
e5b401d4 JA	181	#ifndef BLKFLSBUF
	182	#define BLKFLSBUF _IO(0x12,97)
	183	#endif
	184
	185	static inline int blockdev_invalidate_cache(int fd)
	186	{
a16211bb	187	return ioctl(fd, BLKFLSBUF);
e5b401d4 JA	188	}
e5b401d4 JA	189
9104f874	190	static inline int blockdev_size(int fd, unsigned long long *bytes)
ebac4655 JA	191	{
	192	if (!ioctl(fd, BLKGETSIZE64, bytes))
	193	return 0;
	194
	195	return errno;
	196	}
	197
32cd46a0 JA	198	static inline unsigned long long os_phys_mem(void)
	199	{
	200	long pagesize, pages;
	201
	202	pagesize = sysconf(_SC_PAGESIZE);
	203	pages = sysconf(_SC_PHYS_PAGES);
	204	if (pages == -1 \|\| pagesize == -1)
	205	return 0;
	206
	207	return (unsigned long long) pages * (unsigned long long) pagesize;
	208	}
	209
6dfd46b9 JA	210	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	211	{
	212	srand48_r(seed, rs);
	213	}
	214
	215	static inline long os_random_long(os_random_state_t *rs)
	216	{
	217	long val;
	218
	219	lrand48_r(rs, &val);
	220	return val;
	221	}
	222
8cc7afa9	223	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
07e5b264 JA	224	{
	225	struct raw_config_request rq;
	226	int fd;
	227
	228	if (major(dev) != RAW_MAJOR)
8cc7afa9	229	return 1;
07e5b264 JA	230
	231	/*
	232	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	233	*/
	234	fd = open("/dev/rawctl", O_RDONLY);
	235	if (fd < 0) {
	236	fd = open("/dev/raw/rawctl", O_RDONLY);
	237	if (fd < 0)
8cc7afa9	238	return 1;
07e5b264 JA	239	}
	240
	241	rq.raw_minor = minor(dev);
	242	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	243	close(fd);
8cc7afa9	244	return 1;
07e5b264 JA	245	}
	246
	247	close(fd);
	248	*majdev = rq.block_major;
	249	*mindev = rq.block_minor;
8cc7afa9	250	return 0;
07e5b264 JA	251	}
07e5b264 JA	252
5b6f5c66	253	#ifdef O_NOATIME
5921e80c	254	#define FIO_O_NOATIME O_NOATIME
5b6f5c66 JA	255	#else
	256	#define FIO_O_NOATIME 0
	257	#endif
5921e80c	258
a1c58075 JA	259	#ifdef MADV_REMOVE
	260	#define FIO_MADV_FREE MADV_REMOVE
	261	#endif
	262
eb7ccf38 JA	263	#define CACHE_LINE_FILE \
	264	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	265
	266	static inline int arch_cache_line_size(void)
	267	{
	268	char size[32];
	269	int fd, ret;
	270
	271	fd = open(CACHE_LINE_FILE, O_RDONLY);
	272	if (fd < 0)
	273	return -1;
	274
	275	ret = read(fd, size, sizeof(size));
ec76f15a JA	276
	277	close(fd);
	278
eb7ccf38 JA	279	if (ret <= 0)
eb7ccf38 JA	280	return -1;
ec76f15a JA	281	else
ec76f15a JA	282	return atoi(size);
eb7ccf38 JA	283	}
eb7ccf38 JA	284
2e3bd4c2 JA	285	static inline unsigned long long get_fs_size(const char *path)
	286	{
	287	unsigned long long ret;
	288	struct statfs s;
	289
	290	if (statfs(path, &s) < 0)
	291	return -1ULL;
	292
	293	ret = s.f_bsize;
	294	ret *= (unsigned long long) s.f_bfree;
	295	return ret;
	296	}
	297
ebac4655	298	#endif