[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"
#include "binject.h"
#include "../file.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_LINUX_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
#define FIO_HAVE_BINJECT
#define FIO_HAVE_CLOCK_MONOTONIC

#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(struct fio_file *f)
{
	return ioctl(f->fd, BLKFLSBUF);
}

static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)
{
	if (!ioctl(f->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	16	#include "indirect.h"
79a43187	17	#include "binject.h"
e2e58886	18	#include "../file.h"
f3de88a7	19
ebac4655 JA	20	#define FIO_HAVE_LIBAIO
	21	#define FIO_HAVE_POSIXAIO
	22	#define FIO_HAVE_FADVISE
	23	#define FIO_HAVE_CPU_AFFINITY
	24	#define FIO_HAVE_DISK_UTIL
	25	#define FIO_HAVE_SGIO
ba4f8923	26	#define FIO_HAVE_IOPRIO
8756e4d4	27	#define FIO_HAVE_SPLICE
22f78b32	28	#define FIO_HAVE_IOSCHED_SWITCH
2c0ecd28	29	#define FIO_HAVE_ODIRECT
74b025b0	30	#define FIO_HAVE_HUGETLB
07e5b264	31	#define FIO_HAVE_RAWBIND
5e62c22a	32	#define FIO_HAVE_BLKTRACE
5921e80c	33	#define FIO_HAVE_STRSEP
fffca02d	34	#define FIO_HAVE_FALLOCATE
a596f047	35	#define FIO_HAVE_LINUX_FALLOCATE
207cb0f0	36	#define FIO_HAVE_POSIXAIO_FSYNC
f356d01d	37	#define FIO_HAVE_PSHARED_MUTEX
eb7ccf38	38	#define FIO_HAVE_CL_SIZE
a696fa2a	39	#define FIO_HAVE_CGROUPS
c36d16f5	40	#define FIO_HAVE_FDATASYNC
2e3bd4c2	41	#define FIO_HAVE_FS_STAT
a5f3027c	42	#define FIO_HAVE_TRIM
79a43187	43	#define FIO_HAVE_BINJECT
ecc314ba	44	#define FIO_HAVE_CLOCK_MONOTONIC
b6959b55	45
093729f7 JA	46	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	47	#define FIO_HAVE_SYNC_FILE_RANGE
	48	#endif
	49
dc873b6f	50	#define OS_MAP_ANON MAP_ANONYMOUS
ebac4655	51
4d8947de SL	52	#ifndef CLOCK_MONOTONIC
	53	#define CLOCK_MONOTONIC 1
	54	#endif
	55
ebac4655	56	typedef cpu_set_t os_cpu_mask_t;
4e78e405	57
6dfd46b9	58	typedef struct drand48_data os_random_state_t;
ebac4655 JA	59
	60	/*
	61	* we want fadvise64 really, but it's so tangled... later
	62	*/
4d8947de	63	#ifdef FIO_HAVE_FADVISE
ebac4655 JA	64	#define fadvise(fd, off, len, advice) \
ebac4655 JA	65	posix_fadvise((fd), (off_t)(off), (len), (advice))
4d8947de	66	#endif
ebac4655	67
c8f025f8 JA	68	/*
	69	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	70	* the affinity helpers to work.
	71	*/
	72	#ifndef GLIBC_2_3_2
e8462bd8 JA	73	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	74	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
ebac4655 JA	75	#define fio_getaffinity(pid, ptr) \
ebac4655 JA	76	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
c8f025f8	77	#else
e8462bd8 JA	78	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	79	sched_setaffinity((pid), &(cpumask))
c8f025f8 JA	80	#define fio_getaffinity(pid, ptr) \
	81	sched_getaffinity((pid), (ptr))
	82	#endif
be4ecfdf JA	83
be4ecfdf JA	84	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
6d459ee7	85	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
d2ce18b5 JA	86
	87	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	88	{
	89	CPU_ZERO(mask);
	90	return 0;
	91	}
	92
	93	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	94	{
	95	return 0;
	96	}
6d459ee7 JA	97
6d459ee7 JA	98	#define FIO_MAX_CPUS CPU_SETSIZE
4d8947de	99
ebac4655 JA	100	static inline int ioprio_set(int which, int who, int ioprio)
	101	{
	102	return syscall(__NR_ioprio_set, which, who, ioprio);
	103	}
	104
97fbdfa1 JA	105	/*
	106	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	107	* aren't either.
	108	*/
	109	#ifndef SPLICE_F_MOVE
	110	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	111	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	112	/* we may still block on the fd we splice */
	113	/* from/to, of course */
	114	#define SPLICE_F_MORE (0x04) /* expect more data */
	115	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
8756e4d4 JA	116
8756e4d4 JA	117	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
495ee9b6	118	size_t len, unsigned int flags)
8756e4d4	119	{
97fbdfa1	120	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
8756e4d4 JA	121	}
	122
	123	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	124	{
97fbdfa1	125	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
8756e4d4 JA	126	}
	127
	128	static inline int vmsplice(int fd, const struct iovec *iov,
	129	unsigned long nr_segs, unsigned int flags)
	130	{
97fbdfa1	131	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
8756e4d4	132	}
97fbdfa1	133	#endif
8756e4d4	134
3feedc60 JA	135	#define SPLICE_DEF_SIZE (64*1024)
3feedc60 JA	136
b6959b55	137	#ifdef FIO_HAVE_SYSLET
bf0dc8fa IM	138
	139	struct syslet_uatom;
	140	struct async_head_user;
	141
a4f4fdd7 JA	142	/*
	143	* syslet stuff
	144	*/
bf0dc8fa IM	145	static inline struct syslet_uatom *
bf0dc8fa IM	146	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
a4f4fdd7	147	{
2b52511f	148	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
a4f4fdd7 JA	149	}
a4f4fdd7 JA	150
bf0dc8fa IM	151	static inline long
	152	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	153	struct async_head_user *ahu)
a4f4fdd7	154	{
bf0dc8fa IM	155	return syscall(__NR_async_wait, min_wait_events,
bf0dc8fa IM	156	user_ring_idx, ahu);
a4f4fdd7 JA	157	}
a4f4fdd7 JA	158
7756b0d0	159	static inline long async_thread(void event, struct async_head_user ahu)
a4f4fdd7	160	{
7756b0d0	161	return syscall(__NR_async_thread, event, ahu);
a4f4fdd7 JA	162	}
	163
	164	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	165	{
	166	return syscall(__NR_umem_add, uptr, inc);
	167	}
b6959b55	168	#endif /* FIO_HAVE_SYSLET */
a4f4fdd7	169
6cefbe33 JA	170	enum {
	171	IOPRIO_CLASS_NONE,
	172	IOPRIO_CLASS_RT,
	173	IOPRIO_CLASS_BE,
	174	IOPRIO_CLASS_IDLE,
	175	};
	176
ebac4655 JA	177	enum {
	178	IOPRIO_WHO_PROCESS = 1,
	179	IOPRIO_WHO_PGRP,
	180	IOPRIO_WHO_USER,
	181	};
	182
6cefbe33	183	#define IOPRIO_BITS 16
ebac4655 JA	184	#define IOPRIO_CLASS_SHIFT 13
	185
	186	#ifndef BLKGETSIZE64
	187	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	188	#endif
	189
e5b401d4 JA	190	#ifndef BLKFLSBUF
	191	#define BLKFLSBUF _IO(0x12,97)
	192	#endif
	193
a5f3027c JA	194	#ifndef BLKDISCARD
	195	#define BLKDISCARD _IO(0x12,119)
	196	#endif
	197
e2e58886	198	static inline int blockdev_invalidate_cache(struct fio_file *f)
e5b401d4	199	{
ecc314ba	200	return ioctl(f->fd, BLKFLSBUF);
e5b401d4 JA	201	}
e5b401d4 JA	202
ecc314ba	203	static inline int blockdev_size(struct fio_file f, unsigned long long bytes)
ebac4655	204	{
ecc314ba	205	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
ebac4655 JA	206	return 0;
	207
	208	return errno;
	209	}
	210
32cd46a0 JA	211	static inline unsigned long long os_phys_mem(void)
	212	{
	213	long pagesize, pages;
	214
	215	pagesize = sysconf(_SC_PAGESIZE);
	216	pages = sysconf(_SC_PHYS_PAGES);
	217	if (pages == -1 \|\| pagesize == -1)
	218	return 0;
	219
	220	return (unsigned long long) pages * (unsigned long long) pagesize;
	221	}
	222
6dfd46b9 JA	223	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	224	{
	225	srand48_r(seed, rs);
	226	}
	227
	228	static inline long os_random_long(os_random_state_t *rs)
	229	{
	230	long val;
	231
	232	lrand48_r(rs, &val);
	233	return val;
	234	}
	235
8cc7afa9	236	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
07e5b264 JA	237	{
	238	struct raw_config_request rq;
	239	int fd;
	240
	241	if (major(dev) != RAW_MAJOR)
8cc7afa9	242	return 1;
07e5b264 JA	243
	244	/*
	245	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	246	*/
	247	fd = open("/dev/rawctl", O_RDONLY);
	248	if (fd < 0) {
	249	fd = open("/dev/raw/rawctl", O_RDONLY);
	250	if (fd < 0)
8cc7afa9	251	return 1;
07e5b264 JA	252	}
	253
	254	rq.raw_minor = minor(dev);
	255	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	256	close(fd);
8cc7afa9	257	return 1;
07e5b264 JA	258	}
	259
	260	close(fd);
	261	*majdev = rq.block_major;
	262	*mindev = rq.block_minor;
8cc7afa9	263	return 0;
07e5b264 JA	264	}
07e5b264 JA	265
5b6f5c66	266	#ifdef O_NOATIME
5921e80c	267	#define FIO_O_NOATIME O_NOATIME
5b6f5c66 JA	268	#else
	269	#define FIO_O_NOATIME 0
	270	#endif
5921e80c	271
a1c58075 JA	272	#ifdef MADV_REMOVE
	273	#define FIO_MADV_FREE MADV_REMOVE
	274	#endif
	275
eb7ccf38 JA	276	#define CACHE_LINE_FILE \
	277	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	278
	279	static inline int arch_cache_line_size(void)
	280	{
	281	char size[32];
	282	int fd, ret;
	283
	284	fd = open(CACHE_LINE_FILE, O_RDONLY);
	285	if (fd < 0)
	286	return -1;
	287
	288	ret = read(fd, size, sizeof(size));
ec76f15a JA	289
	290	close(fd);
	291
eb7ccf38 JA	292	if (ret <= 0)
eb7ccf38 JA	293	return -1;
ec76f15a JA	294	else
ec76f15a JA	295	return atoi(size);
eb7ccf38 JA	296	}
eb7ccf38 JA	297
2e3bd4c2 JA	298	static inline unsigned long long get_fs_size(const char *path)
	299	{
	300	unsigned long long ret;
	301	struct statfs s;
	302
	303	if (statfs(path, &s) < 0)
	304	return -1ULL;
	305
	306	ret = s.f_bsize;
	307	ret *= (unsigned long long) s.f_bfree;
	308	return ret;
	309	}
	310
a5f3027c JA	311	static inline int os_trim(int fd, unsigned long long start,
	312	unsigned long long len)
	313	{
	314	uint64_t range[2];
	315
	316	range[0] = start;
	317	range[1] = len;
	318
	319	if (!ioctl(fd, BLKDISCARD, range))
	320	return 0;
	321
	322	return errno;
	323	}
	324
ebac4655	325	#endif