i386 Linux 2.2+ Syscalls -1-

|                                                                                                        |

Copyright © h-peter recktenwald, berlin, 2000 - free for any related work and, non-commercial distribution.
text mode (Lynx, Printer) tables arranged to 110 chars/line.

[intro] [a:index] [s:index] [#:index] [sys_fcntl]
[1..64] [65..128] [129..192] [193..256] [257..] [ref] [struc] [fcntl] [ioctl] [pguide] [man2]
[next] [back] [linux] [main] [bot] [top]
Syscalls pages archive (110K->1M), updated (2.3.0, ver. 1.66) man 2 pages (170K->1,2M)

Fight Patents abuse by EC authorities - Gegen Patentmißbrauch durch die EU-Behörden.

syscall-no. in eax; args left to right in ebx,ecx,edx,esi,edi(,ebp), all preserved.
returned 0xfffff000 < eax <= 0xffffffff may be an error code (-4096 < eax < 0).
Common reference:
- <unistd.h>, <linux/types.h>, <asm/types.h>, <asm/posix_types.h>, and <linux/kernel.h> for data structures.
- The syscall interface is defined in arch/i386/kernel/entry.S, symbols for loadable modules kernel/ksyms.c.
- Error numbers and messages in include/asm/errno.h.
- Basic data types
  short: word (16bit), int: dword (32bit), long: dword (32bit), pointer: dword (32bit).

P	exit	int status		1
	terminate current process
	arg	eax	1
		ebx	exit code
	return	eax	-/-
	errors	eax	-/-

	source		kernel/exit.c

P	fork	struct pt_regs		2
	create a child process resource; resource utilizations:=0, file locks & pending signals not inherited. simplified "clone" w/ flags set to SIGCHLD and caller's ESP
	arg	eax	2
		ebx...	all register values passed to duplicate job
	return	eax	clone : zero caller: pid_t pid (dword) or, -ve error code
	errors	EAGAIN, ENOMEM
	ref	linux/sched.h, asm/ptrace.h.
	source	arch/i386/kernel/process.c, kernel/fork.c

P	read	int fd,	void *buf,	size_t count	3
	read up to count bytes from file fd into buffer *buf
	arg	eax	3
		ebx	file descriptor
		ecx	ptr to input buffer
		edx	buffer size, max. count of bytes to receive
	return	eax	no. of bytes received, file pointer advanced accordingly
	errors	eax	EAGAIN, EBADF, EFAULT, EINTR, EINVAL, EIO, EISDIR
		note that "buffer overflow" is not immanent to the system call but, was introduced by ignorant "programming" rsp (lib)"C" useage/design...
	source		fs/read_write.c

P	write	int fd,	void *buf,	size_t count,	4
	write (up to) count bytes of data from *buf to file fd
	arg	eax	4
		ebx	file descriptor
		ecx	ptr to output buffer
		edx	count of bytes to send
	return	eax	no. of sent bytes (if POSIX conforming f.s.)
	errors	eax	EAGAIN, EBADF, EFAULT, EINTR, EINVAL, EIO, ENOSPC, EPIPE
	source	fs/read_write.c

P	open		const char *pathname,	int flags,	mode_t mode	5
		open a file or device, create/truncate if corresponding mode flag(s) set
	arg		eax	5
			ebx	ptr to asciz abs or rel pathname
			ecx	file access bits
			edx	file permissions, mode
	return		eax	int fd (16bit filedescriptor)
	errors		eax	ACCESS, EXIST, FAULT, ISDIR, LOOP, MFILE, NAMETOOLONG, NFILE, NOENT, NODEV, NODIR, NOMEM, NOSPC, NXIO, ROFS, TXTBSY
	ref		linux/types.h, linux/posix_types.h, linux/stat.h -re- struct stat, asm/fcntl.h, asm/types.h, asm/posix_types.h, fcntl.h, sys/stat.h. sys/types.h
	source		fs/open.c

	file access bits
	O_ACCMODE 0003 O_RDONLY 00 O_WRONLY 01 O_RDWR 02 O_CREAT 0100 O_EXCL 0200 O_NOCTTY 0400 O_TRUNC 01000 O_APPEND 02000 O_NONBLOCK 04000 O_NDELAY O_NONBLOCK O_SYNC 010000 specific to ext2 fs and block devices FASYNC 020000 fcntl, for BSD compatibility O_DIRECT 040000 direct disk access hint - currently ignored O_LARGEFILE 0100000 O_DIRECTORY 0200000 must be a directory O_NOFOLLOW 0400000 don't follow links

	file permissions flags
	S_ISUID 04000 set user ID on execution S_ISGID 02000 set group ID on execution S_ISVTX 01000 sticky bit S_IRUSR 00400 read by owner (S_IREAD) S_IWUSR 00200 write by owner (S_IWRITE) S_IXUSR 00100 execute/search by owner (S_IEXEC) S_IRGRP 00040 read by group S_IWGRP 00020 write by group S_IXGRP 00010 execute/search by group S_IROTH 00004 read by others (R_OK) S_IWOTH 00002 write by others (W_OK) S_IXOTH 00001 execute/search by others (X_OK) S_IRWXUGO (S_IRWXU\|S_IRWXG\|S_IRWXO) -re- umask S_IALLUGO (S_ISUID\|S_ISGID\|S_ISVTX\|S_IRWXUGO) S_IRUGO (S_IRUSR\|S_IRGRP\|S_IROTH) S_IWUGO (S_IWUSR\|S_IWGRP\|S_IWOTH) S_IXUGO (S_IXUSR\|S_IXGRP\|S_IXOTH)

P	close		int fd		6
		close a file by fd reference
	arg		eax	6
			ebx	file descriptor
	return		eax	-/-
	errors		eax	EBADF (-1)
	source		fs/open.c

waitpid		pid_t pid,	int *status,	int options	7
	wait for a specified child process termination; simplified call to wait4
arg		eax	7
		ebx	2nd-ary process id or specification
		ecx	NULL or, ptr to buf for exitted 2nd-ary job status
		edx	option flags, zero or WNOHANG, WUNTRACED
return		eax	pid_t, pid of 2ndary job which exitted
		ecx	if buffer supplied, exit state of 2nd-ary job
errors		eax	CHILD, INVAL, RESTARTSYS
ref		libc: sys/wait.h, wait.c &c, [types]
source		kernel/exit.c

creat		const char *pathname	mode_t mode	8
	create a file
arg		eax	8
		ebx	ptr to asciz abs or rel pathname
		ecx	file permissions, mode
return		eax	int fd filedescriptor
errors		eax	access, exist, fault, isdir, loop, mfile, nametoolong nfile, noent, nodev, nodir, nomem, nospc, nxio rofs, txtbsy
ref		[types], sys/stat.h, fcntl.h
source		fs/open.c
note		sys_creat is identical to sys_open w. O_CREAT\|O_WRONLY\|O_TRUNC flags.

P	link		const char *oldpath	const char *newpath	9
		create hard link to an existing file
	arg		eax	9
			ebx	ptr to asciz abs or rel name of existing file
			ecx	ptr to asciz abs or rel name of new file-name
	return		eax	-/-
	errors		eax	acces, io, perm, exist, fault, loop mlink, nametoolong, noent, nomem, nospc, notdir, perm, rofs, xdev
	ref		-/-
	source		fs/namei.c

P	unlink		const char *pathname		10
		delete a name and remove file when not busy
	arg		eax	10
			ebx	ptr to asciz abs or rel file-name
	return		eax	-/-
	errors		eax	acces, fault, io, isdir, loop, nametoolong, noent, nomem, notdir, perm, rofs
	ref		-/-
	source		fs/namei.c

P	execve		const char *filename, char const argv[], char const envp[]	11
	execute a program
	arg	eax:	11
		ebx:	ptr to <nul> terminated string of program path&name
		ecx:	ptr to zero terminated list of ptrs to <nul> terminated program argument stgs
		edx:	ptr to zero terminated list of ptrs to <nul> terminated environment strings
	return		no return, executed prog inherits resources and overwrites caller
	errors	eax:	2big, acces, inval, io, isdir, libbad loop, nfile, noexec, noent, nomem, notdir, fault, nametoolong, perm, txtbusy
	ref	arch/i386/kernel/process.c
	note		waiting caller of a "fork"ed or "clone"d job being released immediately after entry to execve.

P	chdir	const char *path		12
	change working directory
	arg	eax	12
		ebx	ptr to asciz abs or rel file-name
	return	eax	-/-
	errors	eax	acces, badf, fault, io, loop, nametoolong, noent, nomem, notdir
	ref	-/-
	source	fs/open.c

time	time_t *t		13
get time in seconds since 1-jan-1970
arg	eax	13
	ebx	NULL or, ptr to buffer which receives a copy of the return value
return	eax	((time_t)-1) seconds
errors	eax	fault
ref	linux/time.h
source	kernel/time.c

P	mknod	const char *pathname	mode_t mode	dev_t dev	14
	create a filesystem node (file, device, special file, named pipe)
	return	-/-
	errors	acces, exist, fault, inval, loop, nametoolong, noent, nomem, nospc, notdir, perm, rofs
	ref	sys/stat.h (mode flags), fcntl.h, [types]
	source	fs/namei.c; include/linux/major.h, Documentation/devices.txt

P	chmod	const char *path	mode_t mode	15
	change file permissions (attributes)
	return	-/-
	errors	acces, badf, fault, io, loop, nametoolong, noent, nomem, notdir, perm, rofs
	ref	[types], sys/stat.h, asm/stat.h
	source	fs/open.c

	-re- access mode flags

lchown	const char *path	uid_t owner	gid_t group	16
change owner of a file, don't follow symbolic links
return	-/-
errors	(many...)
source	fs/open.c

	break								17
			n.i.

	oldstat								18
	(abandoned)

P	lseek		int fd	off_t offset	int whence	19
		change file pointer of file ref-d by fd
	arg		eax	19
			ebx	file descriptor
			ecx	disp to the rsp. file position -re - "whence"
			edx	"whence" flag, either one of SEEK_SET 0 add disp to beginning of file, SEEK_CUR 1 add disp to current position, SEEK_END 2 add disp to end of file.
	return		eax	off_t ptr posn. re beginning of file
	errors		eax	badf, inval, ispipe
	ref		unistd.h, linux/unistd.h, [types]
	source		fs/read_write.c

getpid	void	20
get pid of current process
return	pid_t pid
errors	-/-
source	kernel/sched.c

mount

c ch *specialfile

c ch *dir

c ch *fstype

ul rwflag

c v *data

mount a filesystem

arg

eax

ebx

ptr to asciz name of device/specialfile

ecx

ptr to asciz name of mount-point

edx

ptr to asciz name of file system type

flags

esi

MS_MGC_MSK	0xC0ED	in m.s. 16-bit; 'magic', required till 2.4.0-t9
MS_RDONLY	1	Mount read-only
MS_NOSUID	2	Ignore suid and sgid bits
MS_NODEV	4	Disallow access to device special files
MS_NOEXEC	8	Disallow program execution
MS_SYNCHRONOUS	16	Writes are synced at once
MS_REMOUNT	32	Alter flags of a mounted FS
MS_MANDLOCK	64	Allow mandatory locks on an FS
MS_NOATIME	1024	Do not update access times
MS_NODIRATIME	2048	Do not update directory access times
MS_BIND	4096
MS_REC	16384
MS_VERBOSE	32768
MS_ACTIVE	(1<<30)
MS_NOUSER	(1<<31)

edi

ptr to device independent, additional data (max. page-size)

return

eax

-/-

errors

eax

any which can occur in the particular f.s. or, in the kernel

K4+

multiple identical mounts apparently not considered an error, any more, which,
could result to quite dangerous situations, e.g. after single user emergency
startup, where the entire file system might become mounted on the root device!

ref

sys/mount.h linux/mount.h linux/fs.h

source

fs/super.c, fs/namespace.c (2.4)

umount K4:oldumount	const char specialfile \| dir	22
unmount a filesystem
return	-/-
errors	any which can occur in the particular f.s. or, in the kernel
ref	sys/mount.h, /proc/filesystems, umount2
source	fs/super.c, fs/namespace.c (2.4)
NOTE	inconsistently def'd wrt prior sources in kernel 2.4.18! according to 2.4.18 explanation 'oldumount' executes (new) 'mount *device,0' [hp] 2.4.18 variant found to not operating on (linked) device files, any more.

P	setuid	uid_t uid	23
	set real user id
	return	-/-
	errors	eperm
	source	kernel/sys.c

P	getuid	void	24
	get real user id
	return	uid_t uid
	errors	-/-
	ref	unistd.h, linux/unistd.h, [types]
	source	kernel/sched.c

stime	time_t *t	25
set secinds since 1-jan-1970
return	-/-
errors	eperm
ref	time.h
source	kernel/time.c

	ptrace	long request,	pid_t pid,	long addr,	long data	26
	trace a child process
	arg	eax	26
		ebx	requested action
		ecx	process id of the target job	"child"
		edx	address in target job	"address"
		esi	address in tracing job	"data"
	return	eax	as specified by the request
	errors	io, fault, perm, srch
	ref	linux/ptrace.h. asm/ptrace.h, linux/sched.h. asm/user.h,
	source	arch/i386/kernel/ptrace.c

requests		PTRACE_.. ret
		PEEKTEXT pid,addr,data read word at location addr. PEEKDATA pid,addr,data read the word at location addr in the USER area. PEEKUSR pid,addr,data " POKETEXT pid,addr,data write the word at location addr. POKEDATA pid,addr,data " POKEUSR pid,addr,data write the word at location addr in the USER area TRACEME -/- set the ptrace bit in the process flags. CONT pid,,signo restart after signal. KILL pid make the child exit, send it a sigkill. SINGLESTEP pid sigtrap set the trap flag. n.n. n.n. GETREGS pid,,data get all gp regs from the child. re register. SETREGS pid,,data set all gp regs in the child. GETFPREGS pid,,data get the child FPU state. SETFPREGS pid,,data set the child FPU state. ATTACH pid attach to already running process DETACH pid detach a process that was attached. K4 GETFPXREGS pid,,data get the child extended FPU state. K4 SETFPXREGS pid,,data set the child extended FPU state. n.n. K4 SETOPTIONS pid,,data -no comment- (option 'PTRACE_O_TRACESYSGOOD' or none) n.n. n.n. SYSCALL pid,,signo continue and stop at next (return from) syscall
	registers	EBX ECX EDX ESI EDI EBP EAX DS ES FS GS ORIG_EAX EIP CS EFL UESP SS FRAME_SIZE

P	alarm	u int seconds	27
	send SIGALARM at a specified time
	return	0 or unsigned int seconds until prev. alarms are due to execute
	errors	-/-
	source	kernel/sched.c sys_alarm is a simplified call to sys_setitimer.

	oldfstat								28
	(abandoned)

pause	void	29
sleep until signal
return	EINTR if signal caught
errors	restartnohand
source	kernel/sched.c

P	utime	char *filename	struct utimbuf *buf	30
	change access and/or modification time of an inode
	return	-/-
	errors	acces, noent
	ref	sys/time.h, utime.h, [types]
	source	fs/open.c

	stty								31
			n.i.

	gtty								32
			n.i.

P	access	const char *pathname	int mode	33
	check user's permissions for a file
	arg	eax	33
		ebx	ptr to asciz string w. path name
		ecx	mode flags, R_OK, W_OK, X_OK, F_OK = 0, re 'open'
	return	eax	-/-
	errors	eax	all f.s. and file access related errors can occur
	source	fs/open.c, libc/gcc "unistd.h"

nice	int inc	34
change process priority
return	-/-
errors	perm
source	kernel/sched.c

	ftime								35
			n.i.

P	sync	void	36
	commit buffer cache to disk
	return	always 0
	errors	-/-
	source	fs/buffer.c

P	kill	pid_t pid	int sig	37
	send a signal to a process
	return	-/-
	errors	inval, perm, srch
	ref	kernel/sys.c
	source	kernel/signal.c, [more]



		pid 		signal sent to
		-------- | -------------------------------------------
		>  0       process {pid}
		   0       every process in caller's process group
		  -1       all processes but the very first
		< -1       every process in group |{gid}| ("killpg", re getgid)

		no action if sig = 0, but error checking occurs, i.e.
		this can be used to check whether a process exists.

P	rename	c ch *oldpath	c ch *newpath	38
	move/rename a file
	return	-/-
	errors	busy, exist, isdir, notempty, xdev (and other f.s. errors)
	ref	stdio.h
	source	fs/namei.c

P	mkdir	const char *path	mode_t mode	39
	create a directory
	return	-/-
	errors	(f.s. & perm's related errors)
	ref	sys/stat.h, fcntl.h, unistd.h, linux/unistd.h, [types]
	source	fs/namei.c

P	rmdir	const char *pathname	40
	delete an empty directory
	return	-/-
	errors	access, busy, fault, loop, nametoolong, noent, nomem, notdir, notempty, perm, rofs
	source	fs/namei.c

P	dup	int oldfd	41
	create a file descriptor duplicate
	return	int new fd, the lowest until then un-used number
	errors	badf, mfile (,inval)
	source	fs/fcntl.c
	note	dup as well as dup2 assign new fd-s but don't duplicate the channel's description, any modification, e.g by an ioctl, would affect all instances of a resp. fd!

P	pipe	int fd[2]		42
	create a pipe
	arg	eax	42
		ebx	ptr to writable array of two dwords
	return	[ebx]	fd(0) at [ebx] for input from, fd(1) at [ebx+4] output to pipe
	errors	eax	fault, mfile, nfile
	source		arch/i386/kernel/sys_i386.c
	note:		re mknod for named pipes.

times	struct tms *buf	43
get process times
return	-/-
errors	-/-
ref	sys/times.h
source	kernel/sys.c

	prof								44
			n.i.

P	brk		void *end_data_segment		45
	change data section size, top of .bss
	arg		eax	45
			ebx	(virtual) address of new .bss top in range of data-space bottom and, below any linked library or 16K below .ss bottom, rsp, within the limits of the owner process' data size - re getrlimit.
			ebx := 0	can be used to finding the currently valid top address of .bbs.
	return	eax	currently, after call valid top address
	errors		eax	nomem
	source			mm/mmap.c

P	setgid	gid_t gid	46
	set real group id of current process
	return	-/-
	errors	perm
	source	kernel/sys.c

P	getgid	void	47
	get real group id of current process
	return	gid_t gid
	errors	-/-
	ref	unistd.h, linux/unistd.h, [types]
	source	kernel/sched.c

P	signal	signal(int signum, void (handler)(int))	48
	setup a signal handler
	return	void previous value of signal hnd
	errors	sig_err
	source	kernel/signal.c

geteuid	void	49
get effective user id of current process
return	uid_t uid
errors	-/-
ref	unistd.h, linux/unistd.h, [types]
source	kernel/sched.c

getegid	void	50
get effective group id of current process
return	gid_t gid
errors	-/-
ref	unistd.h, linux/unistd.h, [types]
source	kernel/sched.c

acct	int acct(const char *filename)	51
switch process accounting on/off
return	0 or -ve ernum
errors	nosys, nomem, perm, access, io, users
source	kernel/sys.c

umount2 K4:umount	const char specialfile \| dir	int flag	52
arg	*specialfile ptr to asciz mountpoint(?) or device/specialfile name flag 1 to forcedly un-mount, else 0.
return	zero after success, else ernum
errors	any f.s. error
source	fs/namespace.c
ref	oldumount (2.4, formerly named "umount")
<manpages-1.50> include ref to "umount2" by 'man 2 umount'

NOTE: renamed to "umount" w. kernel 2.4.18(+?); from the source file:
"Now umount can handle mount points as well as block devices. This is important for filesystems which use unnamed block devices. We now support a flag for forced unmount like the other 'big iron' unixes. Our API is identical to OSF/1 to avoid making a mess of AMD.

	lock								53
			n.i.

P	ioctl	int fd	int request	char *argp	54
	manipulate a character device
	arg	eax	54
		ebx	file descriptor
		ecx	command code
		edx	ptr to writable data area or other arg structure
	return	eax	zero on success
	errors	eax	badf, fault, inval, notty
	ref		asm/ioctl.h - ioctl page.
	source		fs/ioctl.c for instance, console ioctls provide many nice features - most of which aren't overly helpful because, all(!) consoles would globally be affected!

fcntl	int fd	int cmd	[ struct flock *lock ]	55
file/-descriptor control
arg	eax	55
	ebx	file descriptor
	ecx	command code
	edx	file locks:	ptr to writable struct flock
return	eax	according to the rsp command
errors	eax	acces, again, badf, deadlk, fault, intr, inval, mfilem nolock, perm
ref		fcntl page.
source		fs/fcntl.c, unistd.h, linux/unistd.h, fcntl.h.

	mpx								56
			n.i.

P	setpgid	pid_t pid	pid_t pgid	57
	set process group id
	return	-/-
	errors	inval, perm, srch
	source	kernel/sys.c

ulimit	int cmd	(...)	58	libc
get/set resource limits - not a system call
ref	syscalls setrlimit, getrlimit
ref (glibc2)	man (3) ulimit, sysdeps/generic/ulimit.c, sysdeps/unix/sysv/linux/ulimit.c
NOTE:	not a kernel syscall

oldolduname		59
(syscall removed)
source	utsname.h, struc oldoldutsname

umask	mode_t mask		60
set file creation mask
arg	eax	60
	ebx	file creation permissions, masked w. S_IRWXUGO by kernel code
return	eax	mode_t previously valid mask
errors	-/-
ref	sys/stat.h, [types]
source	kernel/sys.c

P	chroot	const char *path	61
	change root directory (tree)
	return	-/-
	errors	any, dependent on f.s. type
	source	fs/open.c

	ustat								62
	n.i. - libc call

P	dup2	int oldfd	int newfd	63
	duplicate file descriptor oldfd, on success replace newfd w/ duplicated.
	arg	eax	63
		ebx	file descriptor to duplicate
		ecx	file descriptor which the duplicate should be assigned to
	return	eax	duplicate fd
	errors	eax	badf, mfile
	source	fs/fcntl.c
	note	dup2 as well as dup assign new fd-s but don't duplicate the channel's description, any modification, e.g by an ioctl, would affect all instances of a resp. fd! 'newfd' will be closed prior to re-assignation, if applicable.

P	getppid	void		64
	get parent process id
	arg	eax	64
	return	eax	pid_t pid
	errors	-/-
	source	kernel/sched.c

[next] [back] [bot] [top]
[intro] [a:index] [#:index] [1..64] [65..128] [129..192] [193..] [ref] [struc] [ioctl] [pguide]

H.-Peter Recktenwald, Berlin, 18.Feb.2000 = .hpr.l0 = : 6913