[fio.git] / os / os-linux.h

#ifndef FIO_OS_LINUX_H
#define FIO_OS_LINUX_H

#define	FIO_OS	os_linux

#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_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
#define FIO_USE_GENERIC_INIT_RANDOM_STATE

/*
 * Can only enable this for newer glibcs, or the header and defines are
 * missing
 */
#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 6
#define FIO_HAVE_FALLOCATE
#endif
#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 8
#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)	(void) CPU_CLR((cpu), (mask))
#define fio_cpu_set(mask, cpu)		(void) 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

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

static inline int ioprio_set(int which, int who, int ioprio_class, int ioprio)
{
	/*
	 * If no class is set, assume BE
	 */
	if (!ioprio_class)
		ioprio_class = IOPRIO_CLASS_BE;

	ioprio |= ioprio_class << IOPRIO_CLASS_SHIFT;
	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 */

#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	#define FIO_OS os_linux
	5
	6	#include <sys/ioctl.h>
	7	#include <sys/uio.h>
	8	#include <sys/syscall.h>
	9	#include <sys/vfs.h>
	10	#include <unistd.h>
	11	#include <fcntl.h>
	12	#include <errno.h>
	13	#include <sched.h>
	14	#include <linux/unistd.h>
	15	#include <linux/raw.h>
	16	#include <linux/major.h>
	17	#include <endian.h>
	18
	19	#include "indirect.h"
	20	#include "binject.h"
	21	#include "../file.h"
	22
	23	#define FIO_HAVE_LIBAIO
	24	#define FIO_HAVE_POSIXAIO
	25	#define FIO_HAVE_FADVISE
	26	#define FIO_HAVE_CPU_AFFINITY
	27	#define FIO_HAVE_DISK_UTIL
	28	#define FIO_HAVE_SGIO
	29	#define FIO_HAVE_IOPRIO
	30	#define FIO_HAVE_SPLICE
	31	#define FIO_HAVE_IOSCHED_SWITCH
	32	#define FIO_HAVE_ODIRECT
	33	#define FIO_HAVE_HUGETLB
	34	#define FIO_HAVE_RAWBIND
	35	#define FIO_HAVE_BLKTRACE
	36	#define FIO_HAVE_STRSEP
	37	#define FIO_HAVE_POSIXAIO_FSYNC
	38	#define FIO_HAVE_PSHARED_MUTEX
	39	#define FIO_HAVE_CL_SIZE
	40	#define FIO_HAVE_CGROUPS
	41	#define FIO_HAVE_FDATASYNC
	42	#define FIO_HAVE_FS_STAT
	43	#define FIO_HAVE_TRIM
	44	#define FIO_HAVE_BINJECT
	45	#define FIO_HAVE_CLOCK_MONOTONIC
	46	#define FIO_HAVE_GETTID
	47	#define FIO_USE_GENERIC_INIT_RANDOM_STATE
	48
	49	/*
	50	* Can only enable this for newer glibcs, or the header and defines are
	51	* missing
	52	*/
	53	#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 6
	54	#define FIO_HAVE_FALLOCATE
	55	#endif
	56	#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 8
	57	#define FIO_HAVE_LINUX_FALLOCATE
	58	#endif
	59
	60	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	61	#define FIO_HAVE_SYNC_FILE_RANGE
	62	#endif
	63
	64	#define OS_MAP_ANON MAP_ANONYMOUS
	65
	66	#ifndef CLOCK_MONOTONIC
	67	#define CLOCK_MONOTONIC 1
	68	#endif
	69
	70	typedef cpu_set_t os_cpu_mask_t;
	71
	72	typedef struct drand48_data os_random_state_t;
	73
	74	/*
	75	* we want fadvise64 really, but it's so tangled... later
	76	*/
	77	#ifdef FIO_HAVE_FADVISE
	78	#define fadvise(fd, off, len, advice) \
	79	posix_fadvise((fd), (off_t)(off), (len), (advice))
	80	#endif
	81
	82	/*
	83	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	84	* the affinity helpers to work.
	85	*/
	86	#ifndef GLIBC_2_3_2
	87	#define fio_setaffinity(pid, cpumask) \
	88	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
	89	#define fio_getaffinity(pid, ptr) \
	90	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
	91	#else
	92	#define fio_setaffinity(pid, cpumask) \
	93	sched_setaffinity((pid), &(cpumask))
	94	#define fio_getaffinity(pid, ptr) \
	95	sched_getaffinity((pid), (ptr))
	96	#endif
	97
	98	#define fio_cpu_clear(mask, cpu) (void) CPU_CLR((cpu), (mask))
	99	#define fio_cpu_set(mask, cpu) (void) CPU_SET((cpu), (mask))
	100
	101	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	102	{
	103	CPU_ZERO(mask);
	104	return 0;
	105	}
	106
	107	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	108	{
	109	return 0;
	110	}
	111
	112	#define FIO_MAX_CPUS CPU_SETSIZE
	113
	114	enum {
	115	IOPRIO_CLASS_NONE,
	116	IOPRIO_CLASS_RT,
	117	IOPRIO_CLASS_BE,
	118	IOPRIO_CLASS_IDLE,
	119	};
	120
	121	enum {
	122	IOPRIO_WHO_PROCESS = 1,
	123	IOPRIO_WHO_PGRP,
	124	IOPRIO_WHO_USER,
	125	};
	126
	127	#define IOPRIO_BITS 16
	128	#define IOPRIO_CLASS_SHIFT 13
	129
	130	static inline int ioprio_set(int which, int who, int ioprio_class, int ioprio)
	131	{
	132	/*
	133	* If no class is set, assume BE
	134	*/
	135	if (!ioprio_class)
	136	ioprio_class = IOPRIO_CLASS_BE;
	137
	138	ioprio \|= ioprio_class << IOPRIO_CLASS_SHIFT;
	139	return syscall(__NR_ioprio_set, which, who, ioprio);
	140	}
	141
	142	static inline int gettid(void)
	143	{
	144	return syscall(__NR_gettid);
	145	}
	146
	147	/*
	148	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	149	* aren't either.
	150	*/
	151	#ifndef SPLICE_F_MOVE
	152	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	153	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	154	/* we may still block on the fd we splice */
	155	/* from/to, of course */
	156	#define SPLICE_F_MORE (0x04) /* expect more data */
	157	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
	158
	159	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
	160	size_t len, unsigned int flags)
	161	{
	162	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
	163	}
	164
	165	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	166	{
	167	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
	168	}
	169
	170	static inline int vmsplice(int fd, const struct iovec *iov,
	171	unsigned long nr_segs, unsigned int flags)
	172	{
	173	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
	174	}
	175	#endif
	176
	177	#define SPLICE_DEF_SIZE (64*1024)
	178
	179	#ifdef FIO_HAVE_SYSLET
	180
	181	struct syslet_uatom;
	182	struct async_head_user;
	183
	184	/*
	185	* syslet stuff
	186	*/
	187	static inline struct syslet_uatom *
	188	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
	189	{
	190	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
	191	}
	192
	193	static inline long
	194	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	195	struct async_head_user *ahu)
	196	{
	197	return syscall(__NR_async_wait, min_wait_events,
	198	user_ring_idx, ahu);
	199	}
	200
	201	static inline long async_thread(void event, struct async_head_user ahu)
	202	{
	203	return syscall(__NR_async_thread, event, ahu);
	204	}
	205
	206	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	207	{
	208	return syscall(__NR_umem_add, uptr, inc);
	209	}
	210	#endif /* FIO_HAVE_SYSLET */
	211
	212	#ifndef BLKGETSIZE64
	213	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	214	#endif
	215
	216	#ifndef BLKFLSBUF
	217	#define BLKFLSBUF _IO(0x12,97)
	218	#endif
	219
	220	#ifndef BLKDISCARD
	221	#define BLKDISCARD _IO(0x12,119)
	222	#endif
	223
	224	static inline int blockdev_invalidate_cache(struct fio_file *f)
	225	{
	226	return ioctl(f->fd, BLKFLSBUF);
	227	}
	228
	229	static inline int blockdev_size(struct fio_file f, unsigned long long bytes)
	230	{
	231	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
	232	return 0;
	233
	234	return errno;
	235	}
	236
	237	static inline unsigned long long os_phys_mem(void)
	238	{
	239	long pagesize, pages;
	240
	241	pagesize = sysconf(_SC_PAGESIZE);
	242	pages = sysconf(_SC_PHYS_PAGES);
	243	if (pages == -1 \|\| pagesize == -1)
	244	return 0;
	245
	246	return (unsigned long long) pages * (unsigned long long) pagesize;
	247	}
	248
	249	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	250	{
	251	srand48_r(seed, rs);
	252	}
	253
	254	static inline long os_random_long(os_random_state_t *rs)
	255	{
	256	long val;
	257
	258	lrand48_r(rs, &val);
	259	return val;
	260	}
	261
	262	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
	263	{
	264	struct raw_config_request rq;
	265	int fd;
	266
	267	if (major(dev) != RAW_MAJOR)
	268	return 1;
	269
	270	/*
	271	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	272	*/
	273	fd = open("/dev/rawctl", O_RDONLY);
	274	if (fd < 0) {
	275	fd = open("/dev/raw/rawctl", O_RDONLY);
	276	if (fd < 0)
	277	return 1;
	278	}
	279
	280	rq.raw_minor = minor(dev);
	281	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	282	close(fd);
	283	return 1;
	284	}
	285
	286	close(fd);
	287	*majdev = rq.block_major;
	288	*mindev = rq.block_minor;
	289	return 0;
	290	}
	291
	292	#ifdef O_NOATIME
	293	#define FIO_O_NOATIME O_NOATIME
	294	#else
	295	#define FIO_O_NOATIME 0
	296	#endif
	297
	298	#ifdef MADV_REMOVE
	299	#define FIO_MADV_FREE MADV_REMOVE
	300	#endif
	301
	302	#if __BYTE_ORDER == __LITTLE_ENDIAN
	303	#define FIO_LITTLE_ENDIAN
	304	#elif __BYTE_ORDER == __BIG_ENDIAN
	305	#define FIO_BIG_ENDIAN
	306	#else
	307	#error "Unknown endianness"
	308	#endif
	309
	310	#define fio_swap16(x) __bswap_16(x)
	311	#define fio_swap32(x) __bswap_32(x)
	312	#define fio_swap64(x) __bswap_64(x)
	313
	314	#define CACHE_LINE_FILE \
	315	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	316
	317	static inline int arch_cache_line_size(void)
	318	{
	319	char size[32];
	320	int fd, ret;
	321
	322	fd = open(CACHE_LINE_FILE, O_RDONLY);
	323	if (fd < 0)
	324	return -1;
	325
	326	ret = read(fd, size, sizeof(size));
	327
	328	close(fd);
	329
	330	if (ret <= 0)
	331	return -1;
	332	else
	333	return atoi(size);
	334	}
	335
	336	static inline unsigned long long get_fs_size(const char *path)
	337	{
	338	unsigned long long ret;
	339	struct statfs s;
	340
	341	if (statfs(path, &s) < 0)
	342	return -1ULL;
	343
	344	ret = s.f_bsize;
	345	ret *= (unsigned long long) s.f_bfree;
	346	return ret;
	347	}
	348
	349	static inline int os_trim(int fd, unsigned long long start,
	350	unsigned long long len)
	351	{
	352	uint64_t range[2];
	353
	354	range[0] = start;
	355	range[1] = len;
	356
	357	if (!ioctl(fd, BLKDISCARD, range))
	358	return 0;
	359
	360	return errno;
	361	}
	362
	363	#endif