[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"

#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
#define FIO_HAVE_BINJECT

#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	16	#include "indirect.h"
79a43187	17	#include "binject.h"
f3de88a7	18
ebac4655 JA	19	#define FIO_HAVE_LIBAIO
	20	#define FIO_HAVE_POSIXAIO
	21	#define FIO_HAVE_FADVISE
	22	#define FIO_HAVE_CPU_AFFINITY
	23	#define FIO_HAVE_DISK_UTIL
	24	#define FIO_HAVE_SGIO
ba4f8923	25	#define FIO_HAVE_IOPRIO
8756e4d4	26	#define FIO_HAVE_SPLICE
22f78b32	27	#define FIO_HAVE_IOSCHED_SWITCH
2c0ecd28	28	#define FIO_HAVE_ODIRECT
74b025b0	29	#define FIO_HAVE_HUGETLB
07e5b264	30	#define FIO_HAVE_RAWBIND
5e62c22a	31	#define FIO_HAVE_BLKTRACE
5921e80c	32	#define FIO_HAVE_STRSEP
fffca02d	33	#define FIO_HAVE_FALLOCATE
207cb0f0	34	#define FIO_HAVE_POSIXAIO_FSYNC
f356d01d	35	#define FIO_HAVE_PSHARED_MUTEX
eb7ccf38	36	#define FIO_HAVE_CL_SIZE
a696fa2a	37	#define FIO_HAVE_CGROUPS
c36d16f5	38	#define FIO_HAVE_FDATASYNC
2e3bd4c2	39	#define FIO_HAVE_FS_STAT
a5f3027c	40	#define FIO_HAVE_TRIM
79a43187	41	#define FIO_HAVE_BINJECT
b6959b55	42
093729f7 JA	43	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	44	#define FIO_HAVE_SYNC_FILE_RANGE
	45	#endif
	46
dc873b6f	47	#define OS_MAP_ANON MAP_ANONYMOUS
ebac4655	48
4d8947de SL	49	#ifndef CLOCK_MONOTONIC
	50	#define CLOCK_MONOTONIC 1
	51	#endif
	52
ebac4655	53	typedef cpu_set_t os_cpu_mask_t;
4e78e405	54
6dfd46b9	55	typedef struct drand48_data os_random_state_t;
ebac4655 JA	56
	57	/*
	58	* we want fadvise64 really, but it's so tangled... later
	59	*/
4d8947de	60	#ifdef FIO_HAVE_FADVISE
ebac4655 JA	61	#define fadvise(fd, off, len, advice) \
ebac4655 JA	62	posix_fadvise((fd), (off_t)(off), (len), (advice))
4d8947de	63	#endif
ebac4655	64
c8f025f8 JA	65	/*
	66	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	67	* the affinity helpers to work.
	68	*/
	69	#ifndef GLIBC_2_3_2
e8462bd8 JA	70	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	71	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
ebac4655 JA	72	#define fio_getaffinity(pid, ptr) \
ebac4655 JA	73	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
c8f025f8	74	#else
e8462bd8 JA	75	#define fio_setaffinity(pid, cpumask) \
e8462bd8 JA	76	sched_setaffinity((pid), &(cpumask))
c8f025f8 JA	77	#define fio_getaffinity(pid, ptr) \
	78	sched_getaffinity((pid), (ptr))
	79	#endif
be4ecfdf JA	80
be4ecfdf JA	81	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
6d459ee7	82	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
d2ce18b5 JA	83
	84	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	85	{
	86	CPU_ZERO(mask);
	87	return 0;
	88	}
	89
	90	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	91	{
	92	return 0;
	93	}
6d459ee7 JA	94
6d459ee7 JA	95	#define FIO_MAX_CPUS CPU_SETSIZE
4d8947de	96
ebac4655 JA	97	static inline int ioprio_set(int which, int who, int ioprio)
	98	{
	99	return syscall(__NR_ioprio_set, which, who, ioprio);
	100	}
	101
97fbdfa1 JA	102	/*
	103	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	104	* aren't either.
	105	*/
	106	#ifndef SPLICE_F_MOVE
	107	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	108	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	109	/* we may still block on the fd we splice */
	110	/* from/to, of course */
	111	#define SPLICE_F_MORE (0x04) /* expect more data */
	112	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
8756e4d4 JA	113
8756e4d4 JA	114	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
495ee9b6	115	size_t len, unsigned int flags)
8756e4d4	116	{
97fbdfa1	117	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
8756e4d4 JA	118	}
	119
	120	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	121	{
97fbdfa1	122	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
8756e4d4 JA	123	}
	124
	125	static inline int vmsplice(int fd, const struct iovec *iov,
	126	unsigned long nr_segs, unsigned int flags)
	127	{
97fbdfa1	128	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
8756e4d4	129	}
97fbdfa1	130	#endif
8756e4d4	131
3feedc60 JA	132	#define SPLICE_DEF_SIZE (64*1024)
3feedc60 JA	133
b6959b55	134	#ifdef FIO_HAVE_SYSLET
bf0dc8fa IM	135
	136	struct syslet_uatom;
	137	struct async_head_user;
	138
a4f4fdd7 JA	139	/*
	140	* syslet stuff
	141	*/
bf0dc8fa IM	142	static inline struct syslet_uatom *
bf0dc8fa IM	143	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
a4f4fdd7	144	{
2b52511f	145	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
a4f4fdd7 JA	146	}
a4f4fdd7 JA	147
bf0dc8fa IM	148	static inline long
	149	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	150	struct async_head_user *ahu)
a4f4fdd7	151	{
bf0dc8fa IM	152	return syscall(__NR_async_wait, min_wait_events,
bf0dc8fa IM	153	user_ring_idx, ahu);
a4f4fdd7 JA	154	}
a4f4fdd7 JA	155
7756b0d0	156	static inline long async_thread(void event, struct async_head_user ahu)
a4f4fdd7	157	{
7756b0d0	158	return syscall(__NR_async_thread, event, ahu);
a4f4fdd7 JA	159	}
	160
	161	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	162	{
	163	return syscall(__NR_umem_add, uptr, inc);
	164	}
b6959b55	165	#endif /* FIO_HAVE_SYSLET */
a4f4fdd7	166
6cefbe33 JA	167	enum {
	168	IOPRIO_CLASS_NONE,
	169	IOPRIO_CLASS_RT,
	170	IOPRIO_CLASS_BE,
	171	IOPRIO_CLASS_IDLE,
	172	};
	173
ebac4655 JA	174	enum {
	175	IOPRIO_WHO_PROCESS = 1,
	176	IOPRIO_WHO_PGRP,
	177	IOPRIO_WHO_USER,
	178	};
	179
6cefbe33	180	#define IOPRIO_BITS 16
ebac4655 JA	181	#define IOPRIO_CLASS_SHIFT 13
	182
	183	#ifndef BLKGETSIZE64
	184	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	185	#endif
	186
e5b401d4 JA	187	#ifndef BLKFLSBUF
	188	#define BLKFLSBUF _IO(0x12,97)
	189	#endif
	190
a5f3027c JA	191	#ifndef BLKDISCARD
	192	#define BLKDISCARD _IO(0x12,119)
	193	#endif
	194
e5b401d4 JA	195	static inline int blockdev_invalidate_cache(int fd)
e5b401d4 JA	196	{
a16211bb	197	return ioctl(fd, BLKFLSBUF);
e5b401d4 JA	198	}
e5b401d4 JA	199
9104f874	200	static inline int blockdev_size(int fd, unsigned long long *bytes)
ebac4655 JA	201	{
	202	if (!ioctl(fd, BLKGETSIZE64, bytes))
	203	return 0;
	204
	205	return errno;
	206	}
	207
32cd46a0 JA	208	static inline unsigned long long os_phys_mem(void)
	209	{
	210	long pagesize, pages;
	211
	212	pagesize = sysconf(_SC_PAGESIZE);
	213	pages = sysconf(_SC_PHYS_PAGES);
	214	if (pages == -1 \|\| pagesize == -1)
	215	return 0;
	216
	217	return (unsigned long long) pages * (unsigned long long) pagesize;
	218	}
	219
6dfd46b9 JA	220	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	221	{
	222	srand48_r(seed, rs);
	223	}
	224
	225	static inline long os_random_long(os_random_state_t *rs)
	226	{
	227	long val;
	228
	229	lrand48_r(rs, &val);
	230	return val;
	231	}
	232
8cc7afa9	233	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
07e5b264 JA	234	{
	235	struct raw_config_request rq;
	236	int fd;
	237
	238	if (major(dev) != RAW_MAJOR)
8cc7afa9	239	return 1;
07e5b264 JA	240
	241	/*
	242	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	243	*/
	244	fd = open("/dev/rawctl", O_RDONLY);
	245	if (fd < 0) {
	246	fd = open("/dev/raw/rawctl", O_RDONLY);
	247	if (fd < 0)
8cc7afa9	248	return 1;
07e5b264 JA	249	}
	250
	251	rq.raw_minor = minor(dev);
	252	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	253	close(fd);
8cc7afa9	254	return 1;
07e5b264 JA	255	}
	256
	257	close(fd);
	258	*majdev = rq.block_major;
	259	*mindev = rq.block_minor;
8cc7afa9	260	return 0;
07e5b264 JA	261	}
07e5b264 JA	262
5b6f5c66	263	#ifdef O_NOATIME
5921e80c	264	#define FIO_O_NOATIME O_NOATIME
5b6f5c66 JA	265	#else
	266	#define FIO_O_NOATIME 0
	267	#endif
5921e80c	268
a1c58075 JA	269	#ifdef MADV_REMOVE
	270	#define FIO_MADV_FREE MADV_REMOVE
	271	#endif
	272
eb7ccf38 JA	273	#define CACHE_LINE_FILE \
	274	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	275
	276	static inline int arch_cache_line_size(void)
	277	{
	278	char size[32];
	279	int fd, ret;
	280
	281	fd = open(CACHE_LINE_FILE, O_RDONLY);
	282	if (fd < 0)
	283	return -1;
	284
	285	ret = read(fd, size, sizeof(size));
ec76f15a JA	286
	287	close(fd);
	288
eb7ccf38 JA	289	if (ret <= 0)
eb7ccf38 JA	290	return -1;
ec76f15a JA	291	else
ec76f15a JA	292	return atoi(size);
eb7ccf38 JA	293	}
eb7ccf38 JA	294
2e3bd4c2 JA	295	static inline unsigned long long get_fs_size(const char *path)
	296	{
	297	unsigned long long ret;
	298	struct statfs s;
	299
	300	if (statfs(path, &s) < 0)
	301	return -1ULL;
	302
	303	ret = s.f_bsize;
	304	ret *= (unsigned long long) s.f_bfree;
	305	return ret;
	306	}
	307
a5f3027c JA	308	static inline int os_trim(int fd, unsigned long long start,
	309	unsigned long long len)
	310	{
	311	uint64_t range[2];
	312
	313	range[0] = start;
	314	range[1] = len;
	315
	316	if (!ioctl(fd, BLKDISCARD, range))
	317	return 0;
	318
	319	return errno;
	320	}
	321
ebac4655	322	#endif