[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

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	#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	enum {
	120	IOPRIO_CLASS_NONE,
	121	IOPRIO_CLASS_RT,
	122	IOPRIO_CLASS_BE,
	123	IOPRIO_CLASS_IDLE,
	124	};
	125
	126	enum {
	127	IOPRIO_WHO_PROCESS = 1,
	128	IOPRIO_WHO_PGRP,
	129	IOPRIO_WHO_USER,
	130	};
	131
	132	#define IOPRIO_BITS 16
	133	#define IOPRIO_CLASS_SHIFT 13
	134
	135	static inline int ioprio_set(int which, int who, int ioprio_class, int ioprio)
	136	{
	137	/*
	138	* If no class is set, assume BE
	139	*/
	140	if (!ioprio_class)
	141	ioprio_class = IOPRIO_CLASS_BE;
	142
	143	ioprio \|= ioprio_class << IOPRIO_CLASS_SHIFT;
	144	return syscall(__NR_ioprio_set, which, who, ioprio);
	145	}
	146
	147	static inline int gettid(void)
	148	{
	149	return syscall(__NR_gettid);
	150	}
	151
	152	/*
	153	* Just check for SPLICE_F_MOVE, if that isn't there, assume the others
	154	* aren't either.
	155	*/
	156	#ifndef SPLICE_F_MOVE
	157	#define SPLICE_F_MOVE (0x01) /* move pages instead of copying */
	158	#define SPLICE_F_NONBLOCK (0x02) /* don't block on the pipe splicing (but */
	159	/* we may still block on the fd we splice */
	160	/* from/to, of course */
	161	#define SPLICE_F_MORE (0x04) /* expect more data */
	162	#define SPLICE_F_GIFT (0x08) /* pages passed in are a gift */
	163
	164	static inline int splice(int fdin, loff_t off_in, int fdout, loff_t off_out,
	165	size_t len, unsigned int flags)
	166	{
	167	return syscall(__NR_sys_splice, fdin, off_in, fdout, off_out, len, flags);
	168	}
	169
	170	static inline int tee(int fdin, int fdout, size_t len, unsigned int flags)
	171	{
	172	return syscall(__NR_sys_tee, fdin, fdout, len, flags);
	173	}
	174
	175	static inline int vmsplice(int fd, const struct iovec *iov,
	176	unsigned long nr_segs, unsigned int flags)
	177	{
	178	return syscall(__NR_sys_vmsplice, fd, iov, nr_segs, flags);
	179	}
	180	#endif
	181
	182	#define SPLICE_DEF_SIZE (64*1024)
	183
	184	#ifdef FIO_HAVE_SYSLET
	185
	186	struct syslet_uatom;
	187	struct async_head_user;
	188
	189	/*
	190	* syslet stuff
	191	*/
	192	static inline struct syslet_uatom *
	193	async_exec(struct syslet_uatom atom, struct async_head_user ahu)
	194	{
	195	return (struct syslet_uatom *) syscall(__NR_async_exec, atom, ahu);
	196	}
	197
	198	static inline long
	199	async_wait(unsigned long min_wait_events, unsigned long user_ring_idx,
	200	struct async_head_user *ahu)
	201	{
	202	return syscall(__NR_async_wait, min_wait_events,
	203	user_ring_idx, ahu);
	204	}
	205
	206	static inline long async_thread(void event, struct async_head_user ahu)
	207	{
	208	return syscall(__NR_async_thread, event, ahu);
	209	}
	210
	211	static inline long umem_add(unsigned long *uptr, unsigned long inc)
	212	{
	213	return syscall(__NR_umem_add, uptr, inc);
	214	}
	215	#endif /* FIO_HAVE_SYSLET */
	216
	217	#ifndef BLKGETSIZE64
	218	#define BLKGETSIZE64 _IOR(0x12,114,size_t)
	219	#endif
	220
	221	#ifndef BLKFLSBUF
	222	#define BLKFLSBUF _IO(0x12,97)
	223	#endif
	224
	225	#ifndef BLKDISCARD
	226	#define BLKDISCARD _IO(0x12,119)
	227	#endif
	228
	229	static inline int blockdev_invalidate_cache(struct fio_file *f)
	230	{
	231	return ioctl(f->fd, BLKFLSBUF);
	232	}
	233
	234	static inline int blockdev_size(struct fio_file f, unsigned long long bytes)
	235	{
	236	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
	237	return 0;
	238
	239	return errno;
	240	}
	241
	242	static inline unsigned long long os_phys_mem(void)
	243	{
	244	long pagesize, pages;
	245
	246	pagesize = sysconf(_SC_PAGESIZE);
	247	pages = sysconf(_SC_PHYS_PAGES);
	248	if (pages == -1 \|\| pagesize == -1)
	249	return 0;
	250
	251	return (unsigned long long) pages * (unsigned long long) pagesize;
	252	}
	253
	254	static inline void os_random_seed(unsigned long seed, os_random_state_t *rs)
	255	{
	256	srand48_r(seed, rs);
	257	}
	258
	259	static inline long os_random_long(os_random_state_t *rs)
	260	{
	261	long val;
	262
	263	lrand48_r(rs, &val);
	264	return val;
	265	}
	266
	267	static inline int fio_lookup_raw(dev_t dev, int majdev, int mindev)
	268	{
	269	struct raw_config_request rq;
	270	int fd;
	271
	272	if (major(dev) != RAW_MAJOR)
	273	return 1;
	274
	275	/*
	276	* we should be able to find /dev/rawctl or /dev/raw/rawctl
	277	*/
	278	fd = open("/dev/rawctl", O_RDONLY);
	279	if (fd < 0) {
	280	fd = open("/dev/raw/rawctl", O_RDONLY);
	281	if (fd < 0)
	282	return 1;
	283	}
	284
	285	rq.raw_minor = minor(dev);
	286	if (ioctl(fd, RAW_GETBIND, &rq) < 0) {
	287	close(fd);
	288	return 1;
	289	}
	290
	291	close(fd);
	292	*majdev = rq.block_major;
	293	*mindev = rq.block_minor;
	294	return 0;
	295	}
	296
	297	#ifdef O_NOATIME
	298	#define FIO_O_NOATIME O_NOATIME
	299	#else
	300	#define FIO_O_NOATIME 0
	301	#endif
	302
	303	#ifdef MADV_REMOVE
	304	#define FIO_MADV_FREE MADV_REMOVE
	305	#endif
	306
	307	#if __BYTE_ORDER == __LITTLE_ENDIAN
	308	#define FIO_LITTLE_ENDIAN
	309	#elif __BYTE_ORDER == __BIG_ENDIAN
	310	#define FIO_BIG_ENDIAN
	311	#else
	312	#error "Unknown endianness"
	313	#endif
	314
	315	#define fio_swap16(x) __bswap_16(x)
	316	#define fio_swap32(x) __bswap_32(x)
	317	#define fio_swap64(x) __bswap_64(x)
	318
	319	#define CACHE_LINE_FILE \
	320	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
	321
	322	static inline int arch_cache_line_size(void)
	323	{
	324	char size[32];
	325	int fd, ret;
	326
	327	fd = open(CACHE_LINE_FILE, O_RDONLY);
	328	if (fd < 0)
	329	return -1;
	330
	331	ret = read(fd, size, sizeof(size));
	332
	333	close(fd);
	334
	335	if (ret <= 0)
	336	return -1;
	337	else
	338	return atoi(size);
	339	}
	340
	341	static inline unsigned long long get_fs_size(const char *path)
	342	{
	343	unsigned long long ret;
	344	struct statfs s;
	345
	346	if (statfs(path, &s) < 0)
	347	return -1ULL;
	348
	349	ret = s.f_bsize;
	350	ret *= (unsigned long long) s.f_bfree;
	351	return ret;
	352	}
	353
	354	static inline int os_trim(int fd, unsigned long long start,
	355	unsigned long long len)
	356	{
	357	uint64_t range[2];
	358
	359	range[0] = start;
	360	range[1] = len;
	361
	362	if (!ioctl(fd, BLKDISCARD, range))
	363	return 0;
	364
	365	return errno;
	366	}
	367
	368	#endif