[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
#define FIO_HAVE_TRIM

#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

#ifndef BLKDISCARD
#define BLKDISCARD	_IO(0x12,119)
#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;
}

static inline int os_trim(int fd, unsigned long long start,
			  unsigned long long len)
{
	uint64_t range[2];

	range[0] = start;
	range[1] = len;

	if (!ioctl(fd, BLKDISCARD, range))
		return 0;

	return errno;
}

#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
a5f3027c	39	#define FIO_HAVE_TRIM
b6959b55	40
093729f7 JA	41	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	42	#define FIO_HAVE_SYNC_FILE_RANGE
	43	#endif
	44
dc873b6f	45	#define OS_MAP_ANON MAP_ANONYMOUS
ebac4655	46
4d8947de SL	47	#ifndef CLOCK_MONOTONIC
	48	#define CLOCK_MONOTONIC 1
	49	#endif
	50
ebac4655	51	typedef cpu_set_t os_cpu_mask_t;
4e78e405	52
6dfd46b9	53	typedef struct drand48_data os_random_state_t;
ebac4655 JA	54
	55	/*
	56	* we want fadvise64 really, but it's so tangled... later
	57	*/
4d8947de	58	#ifdef FIO_HAVE_FADVISE
ebac4655 JA	59	#define fadvise(fd, off, len, advice) \
ebac4655 JA	60	posix_fadvise((fd), (off_t)(off), (len), (advice))
4d8947de	61	#endif
ebac4655	62
c8f025f8 JA	63	/*
	64	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	65	* the affinity helpers to work.
	66	*/
	67	#ifndef GLIBC_2_3_2
e8462bd8 JA	68	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	69	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
ebac4655 JA	70	#define fio_getaffinity(pid, ptr) \
ebac4655 JA	71	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
c8f025f8	72	#else
e8462bd8 JA	73	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	74	sched_setaffinity((pid), &(cpumask))
c8f025f8 JA	75	#define fio_getaffinity(pid, ptr) \
	76	sched_getaffinity((pid), (ptr))
	77	#endif
be4ecfdf JA	78
be4ecfdf JA	79	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
6d459ee7	80	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
d2ce18b5 JA	81
	82	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	83	{
	84	CPU_ZERO(mask);
	85	return 0;
	86	}
	87
	88	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	89	{
	90	return 0;
	91	}
6d459ee7 JA	92
6d459ee7 JA	93	#define FIO_MAX_CPUS CPU_SETSIZE
4d8947de	94
ebac4655 JA	95	static inline int ioprio_set(int which, int who, int ioprio)
	96	{
	97	return syscall(__NR_ioprio_set, which, who, ioprio);
	98	}
	99
97fbdfa1 JA	100	/*
	101	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	102	* aren't either.
	103	*/
	104	#ifndef SPLICE_F_MOVE
	105	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	106	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	107	/* we may still block on the fd we splice */
	108	/* from/to, of course */
	109	#define SPLICE_F_MORE (0x04) /* expect more data */
	110	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
8756e4d4 JA	111
8756e4d4 JA	112	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
495ee9b6	113	size_t len, unsigned int flags)
8756e4d4	114	{
97fbdfa1	115	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
8756e4d4 JA	116	}
	117
	118	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	119	{
97fbdfa1	120	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
8756e4d4 JA	121	}
	122
	123	static inline int vmsplice(int fd, const struct iovec *iov,
	124	unsigned long nr_segs, unsigned int flags)
	125	{
97fbdfa1	126	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
8756e4d4	127	}
97fbdfa1	128	#endif
8756e4d4	129
3feedc60 JA	130	#define SPLICE_DEF_SIZE (64*1024)
3feedc60 JA	131
b6959b55	132	#ifdef FIO_HAVE_SYSLET
bf0dc8fa IM	133
	134	struct syslet_uatom;
	135	struct async_head_user;
	136
a4f4fdd7 JA	137	/*
	138	* syslet stuff
	139	*/
bf0dc8fa IM	140	static inline struct syslet_uatom *
bf0dc8fa IM	141	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
a4f4fdd7	142	{
2b52511f	143	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
a4f4fdd7 JA	144	}
a4f4fdd7 JA	145
bf0dc8fa IM	146	static inline long
	147	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	148	struct async_head_user *ahu)
a4f4fdd7	149	{
bf0dc8fa IM	150	return syscall(__NR_async_wait, min_wait_events,
bf0dc8fa IM	151	user_ring_idx, ahu);
a4f4fdd7 JA	152	}
a4f4fdd7 JA	153
7756b0d0	154	static inline long async_thread(void event, struct async_head_user ahu)
a4f4fdd7	155	{
7756b0d0	156	return syscall(__NR_async_thread, event, ahu);
a4f4fdd7 JA	157	}
	158
	159	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	160	{
	161	return syscall(__NR_umem_add, uptr, inc);
	162	}
b6959b55	163	#endif /* FIO_HAVE_SYSLET */
a4f4fdd7	164
6cefbe33 JA	165	enum {
	166	IOPRIO_CLASS_NONE,
	167	IOPRIO_CLASS_RT,
	168	IOPRIO_CLASS_BE,
	169	IOPRIO_CLASS_IDLE,
	170	};
	171
ebac4655 JA	172	enum {
	173	IOPRIO_WHO_PROCESS = 1,
	174	IOPRIO_WHO_PGRP,
	175	IOPRIO_WHO_USER,
	176	};
	177
6cefbe33	178	#define IOPRIO_BITS 16
ebac4655 JA	179	#define IOPRIO_CLASS_SHIFT 13
	180
	181	#ifndef BLKGETSIZE64
	182	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	183	#endif
	184
e5b401d4 JA	185	#ifndef BLKFLSBUF
	186	#define BLKFLSBUF _IO(0x12,97)
	187	#endif
	188
a5f3027c JA	189	#ifndef BLKDISCARD
	190	#define BLKDISCARD _IO(0x12,119)
	191	#endif
	192
e5b401d4 JA	193	static inline int blockdev_invalidate_cache(int fd)
e5b401d4 JA	194	{
a16211bb	195	return ioctl(fd, BLKFLSBUF);
e5b401d4 JA	196	}
e5b401d4 JA	197
9104f874	198	static inline int blockdev_size(int fd, unsigned long long *bytes)
ebac4655 JA	199	{
	200	if (!ioctl(fd, BLKGETSIZE64, bytes))
	201	return 0;
	202
	203	return errno;
	204	}
	205
32cd46a0 JA	206	static inline unsigned long long os_phys_mem(void)
	207	{
	208	long pagesize, pages;
	209
	210	pagesize = sysconf(_SC_PAGESIZE);
	211	pages = sysconf(_SC_PHYS_PAGES);
	212	if (pages == -1 \|\| pagesize == -1)
	213	return 0;
	214
	215	return (unsigned long long) pages * (unsigned long long) pagesize;
	216	}
	217
6dfd46b9 JA	218	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	219	{
	220	srand48_r(seed, rs);
	221	}
	222
	223	static inline long os_random_long(os_random_state_t *rs)
	224	{
	225	long val;
	226
	227	lrand48_r(rs, &val);
	228	return val;
	229	}
	230
8cc7afa9	231	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
07e5b264 JA	232	{
	233	struct raw_config_request rq;
	234	int fd;
	235
	236	if (major(dev) != RAW_MAJOR)
8cc7afa9	237	return 1;
07e5b264 JA	238
	239	/*
	240	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	241	*/
	242	fd = open("/dev/rawctl", O_RDONLY);
	243	if (fd < 0) {
	244	fd = open("/dev/raw/rawctl", O_RDONLY);
	245	if (fd < 0)
8cc7afa9	246	return 1;
07e5b264 JA	247	}
	248
	249	rq.raw_minor = minor(dev);
	250	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	251	close(fd);
8cc7afa9	252	return 1;
07e5b264 JA	253	}
	254
	255	close(fd);
	256	*majdev = rq.block_major;
	257	*mindev = rq.block_minor;
8cc7afa9	258	return 0;
07e5b264 JA	259	}
07e5b264 JA	260
5b6f5c66	261	#ifdef O_NOATIME
5921e80c	262	#define FIO_O_NOATIME O_NOATIME
5b6f5c66 JA	263	#else
	264	#define FIO_O_NOATIME 0
	265	#endif
5921e80c	266
a1c58075 JA	267	#ifdef MADV_REMOVE
	268	#define FIO_MADV_FREE MADV_REMOVE
	269	#endif
	270
eb7ccf38 JA	271	#define CACHE_LINE_FILE \
	272	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	273
	274	static inline int arch_cache_line_size(void)
	275	{
	276	char size[32];
	277	int fd, ret;
	278
	279	fd = open(CACHE_LINE_FILE, O_RDONLY);
	280	if (fd < 0)
	281	return -1;
	282
	283	ret = read(fd, size, sizeof(size));
ec76f15a JA	284
	285	close(fd);
	286
eb7ccf38 JA	287	if (ret <= 0)
eb7ccf38 JA	288	return -1;
ec76f15a JA	289	else
ec76f15a JA	290	return atoi(size);
eb7ccf38 JA	291	}
eb7ccf38 JA	292
2e3bd4c2 JA	293	static inline unsigned long long get_fs_size(const char *path)
	294	{
	295	unsigned long long ret;
	296	struct statfs s;
	297
	298	if (statfs(path, &s) < 0)
	299	return -1ULL;
	300
	301	ret = s.f_bsize;
	302	ret *= (unsigned long long) s.f_bfree;
	303	return ret;
	304	}
	305
a5f3027c JA	306	static inline int os_trim(int fd, unsigned long long start,
	307	unsigned long long len)
	308	{
	309	uint64_t range[2];
	310
	311	range[0] = start;
	312	range[1] = len;
	313
	314	if (!ioctl(fd, BLKDISCARD, range))
	315	return 0;
	316
	317	return errno;
	318	}
	319
ebac4655	320	#endif