[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 <endian.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
#define FIO_HAVE_GETTID

/*
 * Can only enable this for newer glibcs, or the header and defines are
 * missing
 */
#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 6
#define FIO_HAVE_LINUX_FALLOCATE
#endif

#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);
}

static inline int gettid(void)
{
	return syscall(__NR_gettid);
}

/*
 * 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

#if __BYTE_ORDER == __LITTLE_ENDIAN
#define FIO_LITTLE_ENDIAN
#elif __BYTE_ORDER == __BIG_ENDIAN
#define FIO_BIG_ENDIAN
#else
#error "Unknown endianness"
#endif

#define fio_swap16(x)	__bswap_16(x)
#define fio_swap32(x)	__bswap_32(x)
#define fio_swap64(x)	__bswap_64(x)

#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
	1	#ifndef FIO_OS_LINUX_H
	2	#define FIO_OS_LINUX_H
	3
	4	#include <sys/ioctl.h>
	5	#include <sys/uio.h>
	6	#include <sys/syscall.h>
	7	#include <sys/vfs.h>
	8	#include <unistd.h>
	9	#include <fcntl.h>
	10	#include <errno.h>
	11	#include <sched.h>
	12	#include <linux/unistd.h>
	13	#include <linux/raw.h>
	14	#include <linux/major.h>
	15	#include <endian.h>
	16
	17	#include "indirect.h"
	18	#include "binject.h"
	19	#include "../file.h"
	20
	21	#define FIO_HAVE_LIBAIO
	22	#define FIO_HAVE_POSIXAIO
	23	#define FIO_HAVE_FADVISE
	24	#define FIO_HAVE_CPU_AFFINITY
	25	#define FIO_HAVE_DISK_UTIL
	26	#define FIO_HAVE_SGIO
	27	#define FIO_HAVE_IOPRIO
	28	#define FIO_HAVE_SPLICE
	29	#define FIO_HAVE_IOSCHED_SWITCH
	30	#define FIO_HAVE_ODIRECT
	31	#define FIO_HAVE_HUGETLB
	32	#define FIO_HAVE_RAWBIND
	33	#define FIO_HAVE_BLKTRACE
	34	#define FIO_HAVE_STRSEP
	35	#define FIO_HAVE_FALLOCATE
	36	#define FIO_HAVE_POSIXAIO_FSYNC
	37	#define FIO_HAVE_PSHARED_MUTEX
	38	#define FIO_HAVE_CL_SIZE
	39	#define FIO_HAVE_CGROUPS
	40	#define FIO_HAVE_FDATASYNC
	41	#define FIO_HAVE_FS_STAT
	42	#define FIO_HAVE_TRIM
	43	#define FIO_HAVE_BINJECT
	44	#define FIO_HAVE_CLOCK_MONOTONIC
	45	#define FIO_HAVE_GETTID
	46
	47	/*
	48	* Can only enable this for newer glibcs, or the header and defines are
	49	* missing
	50	*/
	51	#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 6
	52	#define FIO_HAVE_LINUX_FALLOCATE
	53	#endif
	54
	55	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	56	#define FIO_HAVE_SYNC_FILE_RANGE
	57	#endif
	58
	59	#define OS_MAP_ANON MAP_ANONYMOUS
	60
	61	#ifndef CLOCK_MONOTONIC
	62	#define CLOCK_MONOTONIC 1
	63	#endif
	64
	65	typedef cpu_set_t os_cpu_mask_t;
	66
	67	typedef struct drand48_data os_random_state_t;
	68
	69	/*
	70	* we want fadvise64 really, but it's so tangled... later
	71	*/
	72	#ifdef FIO_HAVE_FADVISE
	73	#define fadvise(fd, off, len, advice) \
	74	posix_fadvise((fd), (off_t)(off), (len), (advice))
	75	#endif
	76
	77	/*
	78	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	79	* the affinity helpers to work.
	80	*/
	81	#ifndef GLIBC_2_3_2
	82	#define fio_setaffinity(pid, cpumask) \
	83	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
	84	#define fio_getaffinity(pid, ptr) \
	85	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
	86	#else
	87	#define fio_setaffinity(pid, cpumask) \
	88	sched_setaffinity((pid), &(cpumask))
	89	#define fio_getaffinity(pid, ptr) \
	90	sched_getaffinity((pid), (ptr))
	91	#endif
	92
	93	#define fio_cpu_clear(mask, cpu) CPU_CLR((cpu), (mask))
	94	#define fio_cpu_set(mask, cpu) CPU_SET((cpu), (mask))
	95
	96	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	97	{
	98	CPU_ZERO(mask);
	99	return 0;
	100	}
	101
	102	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	103	{
	104	return 0;
	105	}
	106
	107	#define FIO_MAX_CPUS CPU_SETSIZE
	108
	109	static inline int ioprio_set(int which, int who, int ioprio)
	110	{
	111	return syscall(__NR_ioprio_set, which, who, ioprio);
	112	}
	113
	114	static inline int gettid(void)
	115	{
	116	return syscall(__NR_gettid);
	117	}
	118
	119	/*
	120	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	121	* aren't either.
	122	*/
	123	#ifndef SPLICE_F_MOVE
	124	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	125	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	126	/* we may still block on the fd we splice */
	127	/* from/to, of course */
	128	#define SPLICE_F_MORE (0x04) /* expect more data */
	129	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
	130
	131	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
	132	size_t len, unsigned int flags)
	133	{
	134	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
	135	}
	136
	137	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	138	{
	139	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
	140	}
	141
	142	static inline int vmsplice(int fd, const struct iovec *iov,
	143	unsigned long nr_segs, unsigned int flags)
	144	{
	145	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
	146	}
	147	#endif
	148
	149	#define SPLICE_DEF_SIZE (64*1024)
	150
	151	#ifdef FIO_HAVE_SYSLET
	152
	153	struct syslet_uatom;
	154	struct async_head_user;
	155
	156	/*
	157	* syslet stuff
	158	*/
	159	static inline struct syslet_uatom *
	160	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
	161	{
	162	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
	163	}
	164
	165	static inline long
	166	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	167	struct async_head_user *ahu)
	168	{
	169	return syscall(__NR_async_wait, min_wait_events,
	170	user_ring_idx, ahu);
	171	}
	172
	173	static inline long async_thread(void event, struct async_head_user ahu)
	174	{
	175	return syscall(__NR_async_thread, event, ahu);
	176	}
	177
	178	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	179	{
	180	return syscall(__NR_umem_add, uptr, inc);
	181	}
	182	#endif /* FIO_HAVE_SYSLET */
	183
	184	enum {
	185	IOPRIO_CLASS_NONE,
	186	IOPRIO_CLASS_RT,
	187	IOPRIO_CLASS_BE,
	188	IOPRIO_CLASS_IDLE,
	189	};
	190
	191	enum {
	192	IOPRIO_WHO_PROCESS = 1,
	193	IOPRIO_WHO_PGRP,
	194	IOPRIO_WHO_USER,
	195	};
	196
	197	#define IOPRIO_BITS 16
	198	#define IOPRIO_CLASS_SHIFT 13
	199
	200	#ifndef BLKGETSIZE64
	201	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	202	#endif
	203
	204	#ifndef BLKFLSBUF
	205	#define BLKFLSBUF _IO(0x12,97)
	206	#endif
	207
	208	#ifndef BLKDISCARD
	209	#define BLKDISCARD _IO(0x12,119)
	210	#endif
	211
	212	static inline int blockdev_invalidate_cache(struct fio_file *f)
	213	{
	214	return ioctl(f->fd, BLKFLSBUF);
	215	}
	216
	217	static inline int blockdev_size(struct fio_file f, unsigned long long bytes)
	218	{
	219	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
	220	return 0;
	221
	222	return errno;
	223	}
	224
	225	static inline unsigned long long os_phys_mem(void)
	226	{
	227	long pagesize, pages;
	228
	229	pagesize = sysconf(_SC_PAGESIZE);
	230	pages = sysconf(_SC_PHYS_PAGES);
	231	if (pages == -1 \|\| pagesize == -1)
	232	return 0;
	233
	234	return (unsigned long long) pages * (unsigned long long) pagesize;
	235	}
	236
	237	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	238	{
	239	srand48_r(seed, rs);
	240	}
	241
	242	static inline long os_random_long(os_random_state_t *rs)
	243	{
	244	long val;
	245
	246	lrand48_r(rs, &val);
	247	return val;
	248	}
	249
	250	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
	251	{
	252	struct raw_config_request rq;
	253	int fd;
	254
	255	if (major(dev) != RAW_MAJOR)
	256	return 1;
	257
	258	/*
	259	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	260	*/
	261	fd = open("/dev/rawctl", O_RDONLY);
	262	if (fd < 0) {
	263	fd = open("/dev/raw/rawctl", O_RDONLY);
	264	if (fd < 0)
	265	return 1;
	266	}
	267
	268	rq.raw_minor = minor(dev);
	269	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	270	close(fd);
	271	return 1;
	272	}
	273
	274	close(fd);
	275	*majdev = rq.block_major;
	276	*mindev = rq.block_minor;
	277	return 0;
	278	}
	279
	280	#ifdef O_NOATIME
	281	#define FIO_O_NOATIME O_NOATIME
	282	#else
	283	#define FIO_O_NOATIME 0
	284	#endif
	285
	286	#ifdef MADV_REMOVE
	287	#define FIO_MADV_FREE MADV_REMOVE
	288	#endif
	289
	290	#if __BYTE_ORDER == __LITTLE_ENDIAN
	291	#define FIO_LITTLE_ENDIAN
	292	#elif __BYTE_ORDER == __BIG_ENDIAN
	293	#define FIO_BIG_ENDIAN
	294	#else
	295	#error "Unknown endianness"
	296	#endif
	297
	298	#define fio_swap16(x) __bswap_16(x)
	299	#define fio_swap32(x) __bswap_32(x)
	300	#define fio_swap64(x) __bswap_64(x)
	301
	302	#define CACHE_LINE_FILE \
	303	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	304
	305	static inline int arch_cache_line_size(void)
	306	{
	307	char size[32];
	308	int fd, ret;
	309
	310	fd = open(CACHE_LINE_FILE, O_RDONLY);
	311	if (fd < 0)
	312	return -1;
	313
	314	ret = read(fd, size, sizeof(size));
	315
	316	close(fd);
	317
	318	if (ret <= 0)
	319	return -1;
	320	else
	321	return atoi(size);
	322	}
	323
	324	static inline unsigned long long get_fs_size(const char *path)
	325	{
	326	unsigned long long ret;
	327	struct statfs s;
	328
	329	if (statfs(path, &s) < 0)
	330	return -1ULL;
	331
	332	ret = s.f_bsize;
	333	ret *= (unsigned long long) s.f_bfree;
	334	return ret;
	335	}
	336
	337	static inline int os_trim(int fd, unsigned long long start,
	338	unsigned long long len)
	339	{
	340	uint64_t range[2];
	341
	342	range[0] = start;
	343	range[1] = len;
	344
	345	if (!ioctl(fd, BLKDISCARD, range))
	346	return 0;
	347
	348	return errno;
	349	}
	350
	351	#endif