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