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