[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
#define FIO_HAVE_E4_ENG

/*
 * 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 FIO_HAVE_LINUX_FALLOCATE
#define FIO_HAVE_FALLOC_ENG
#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

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	#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	#define FIO_HAVE_E4_ENG
	49
	50	/*
	51	* Can only enable this for newer glibcs, or the header and defines are
	52	* missing
	53	*/
	54	#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 6
	55	#define FIO_HAVE_FALLOCATE
	56	#endif
	57	#if __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 8
	58	#define FIO_HAVE_LINUX_FALLOCATE
	59	#endif
	60
	61	#ifdef FIO_HAVE_LINUX_FALLOCATE
	62	#define FIO_HAVE_FALLOC_ENG
	63	#endif
	64
	65	#ifdef SYNC_FILE_RANGE_WAIT_BEFORE
	66	#define FIO_HAVE_SYNC_FILE_RANGE
	67	#endif
	68
	69	#define OS_MAP_ANON MAP_ANONYMOUS
	70
	71	#ifndef CLOCK_MONOTONIC
	72	#define CLOCK_MONOTONIC 1
	73	#endif
	74
	75	typedef cpu_set_t os_cpu_mask_t;
	76
	77	typedef struct drand48_data os_random_state_t;
	78
	79	/*
	80	* we want fadvise64 really, but it's so tangled... later
	81	*/
	82	#ifdef FIO_HAVE_FADVISE
	83	#define fadvise(fd, off, len, advice) \
	84	posix_fadvise((fd), (off_t)(off), (len), (advice))
	85	#endif
	86
	87	/*
	88	* If you are on an ancient glibc (2.3.2), then define GLIBC_2_3_2 if you want
	89	* the affinity helpers to work.
	90	*/
	91	#ifndef GLIBC_2_3_2
	92	#define fio_setaffinity(pid, cpumask) \
	93	sched_setaffinity((pid), sizeof(cpumask), &(cpumask))
	94	#define fio_getaffinity(pid, ptr) \
	95	sched_getaffinity((pid), sizeof(cpu_set_t), (ptr))
	96	#else
	97	#define fio_setaffinity(pid, cpumask) \
	98	sched_setaffinity((pid), &(cpumask))
	99	#define fio_getaffinity(pid, ptr) \
	100	sched_getaffinity((pid), (ptr))
	101	#endif
	102
	103	#define fio_cpu_clear(mask, cpu) (void) CPU_CLR((cpu), (mask))
	104	#define fio_cpu_set(mask, cpu) (void) CPU_SET((cpu), (mask))
	105
	106	static inline int fio_cpuset_init(os_cpu_mask_t *mask)
	107	{
	108	CPU_ZERO(mask);
	109	return 0;
	110	}
	111
	112	static inline int fio_cpuset_exit(os_cpu_mask_t *mask)
	113	{
	114	return 0;
	115	}
	116
	117	#define FIO_MAX_CPUS CPU_SETSIZE
	118
	119	static inline int ioprio_set(int which, int who, int ioprio)
	120	{
	121	return syscall(__NR_ioprio_set, which, who, ioprio);
	122	}
	123
	124	static inline int gettid(void)
	125	{
	126	return syscall(__NR_gettid);
	127	}
	128
	129	/*
	130	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	131	* aren't either.
	132	*/
	133	#ifndef SPLICE_F_MOVE
	134	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	135	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	136	/* we may still block on the fd we splice */
	137	/* from/to, of course */
	138	#define SPLICE_F_MORE (0x04) /* expect more data */
	139	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
	140
	141	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
	142	size_t len, unsigned int flags)
	143	{
	144	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
	145	}
	146
	147	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	148	{
	149	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
	150	}
	151
	152	static inline int vmsplice(int fd, const struct iovec *iov,
	153	unsigned long nr_segs, unsigned int flags)
	154	{
	155	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
	156	}
	157	#endif
	158
	159	#define SPLICE_DEF_SIZE (64*1024)
	160
	161	#ifdef FIO_HAVE_SYSLET
	162
	163	struct syslet_uatom;
	164	struct async_head_user;
	165
	166	/*
	167	* syslet stuff
	168	*/
	169	static inline struct syslet_uatom *
	170	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
	171	{
	172	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
	173	}
	174
	175	static inline long
	176	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	177	struct async_head_user *ahu)
	178	{
	179	return syscall(__NR_async_wait, min_wait_events,
	180	user_ring_idx, ahu);
	181	}
	182
	183	static inline long async_thread(void event, struct async_head_user ahu)
	184	{
	185	return syscall(__NR_async_thread, event, ahu);
	186	}
	187
	188	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	189	{
	190	return syscall(__NR_umem_add, uptr, inc);
	191	}
	192	#endif /* FIO_HAVE_SYSLET */
	193
	194	enum {
	195	IOPRIO_CLASS_NONE,
	196	IOPRIO_CLASS_RT,
	197	IOPRIO_CLASS_BE,
	198	IOPRIO_CLASS_IDLE,
	199	};
	200
	201	enum {
	202	IOPRIO_WHO_PROCESS = 1,
	203	IOPRIO_WHO_PGRP,
	204	IOPRIO_WHO_USER,
	205	};
	206
	207	#define IOPRIO_BITS 16
	208	#define IOPRIO_CLASS_SHIFT 13
	209
	210	#ifndef BLKGETSIZE64
	211	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	212	#endif
	213
	214	#ifndef BLKFLSBUF
	215	#define BLKFLSBUF _IO(0x12,97)
	216	#endif
	217
	218	#ifndef BLKDISCARD
	219	#define BLKDISCARD _IO(0x12,119)
	220	#endif
	221
	222	static inline int blockdev_invalidate_cache(struct fio_file *f)
	223	{
	224	return ioctl(f->fd, BLKFLSBUF);
	225	}
	226
	227	static inline int blockdev_size(struct fio_file f, unsigned long long bytes)
	228	{
	229	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
	230	return 0;
	231
	232	return errno;
	233	}
	234
	235	static inline unsigned long long os_phys_mem(void)
	236	{
	237	long pagesize, pages;
	238
	239	pagesize = sysconf(_SC_PAGESIZE);
	240	pages = sysconf(_SC_PHYS_PAGES);
	241	if (pages == -1 \|\| pagesize == -1)
	242	return 0;
	243
	244	return (unsigned long long) pages * (unsigned long long) pagesize;
	245	}
	246
	247	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	248	{
	249	srand48_r(seed, rs);
	250	}
	251
	252	static inline long os_random_long(os_random_state_t *rs)
	253	{
	254	long val;
	255
	256	lrand48_r(rs, &val);
	257	return val;
	258	}
	259
	260	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
	261	{
	262	struct raw_config_request rq;
	263	int fd;
	264
	265	if (major(dev) != RAW_MAJOR)
	266	return 1;
	267
	268	/*
	269	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	270	*/
	271	fd = open("/dev/rawctl", O_RDONLY);
	272	if (fd < 0) {
	273	fd = open("/dev/raw/rawctl", O_RDONLY);
	274	if (fd < 0)
	275	return 1;
	276	}
	277
	278	rq.raw_minor = minor(dev);
	279	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	280	close(fd);
	281	return 1;
	282	}
	283
	284	close(fd);
	285	*majdev = rq.block_major;
	286	*mindev = rq.block_minor;
	287	return 0;
	288	}
	289
	290	#ifdef O_NOATIME
	291	#define FIO_O_NOATIME O_NOATIME
	292	#else
	293	#define FIO_O_NOATIME 0
	294	#endif
	295
	296	#ifdef MADV_REMOVE
	297	#define FIO_MADV_FREE MADV_REMOVE
	298	#endif
	299
	300	#if __BYTE_ORDER == __LITTLE_ENDIAN
	301	#define FIO_LITTLE_ENDIAN
	302	#elif __BYTE_ORDER == __BIG_ENDIAN
	303	#define FIO_BIG_ENDIAN
	304	#else
	305	#error "Unknown endianness"
	306	#endif
	307
	308	#define fio_swap16(x) __bswap_16(x)
	309	#define fio_swap32(x) __bswap_32(x)
	310	#define fio_swap64(x) __bswap_64(x)
	311
	312	#define CACHE_LINE_FILE \
	313	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	314
	315	static inline int arch_cache_line_size(void)
	316	{
	317	char size[32];
	318	int fd, ret;
	319
	320	fd = open(CACHE_LINE_FILE, O_RDONLY);
	321	if (fd < 0)
	322	return -1;
	323
	324	ret = read(fd, size, sizeof(size));
	325
	326	close(fd);
	327
	328	if (ret <= 0)
	329	return -1;
	330	else
	331	return atoi(size);
	332	}
	333
	334	static inline unsigned long long get_fs_size(const char *path)
	335	{
	336	unsigned long long ret;
	337	struct statfs s;
	338
	339	if (statfs(path, &s) < 0)
	340	return -1ULL;
	341
	342	ret = s.f_bsize;
	343	ret *= (unsigned long long) s.f_bfree;
	344	return ret;
	345	}
	346
	347	static inline int os_trim(int fd, unsigned long long start,
	348	unsigned long long len)
	349	{
	350	uint64_t range[2];
	351
	352	range[0] = start;
	353	range[1] = len;
	354
	355	if (!ioctl(fd, BLKDISCARD, range))
	356	return 0;
	357
	358	return errno;
	359	}
	360
	361	#endif