syscalls: Complete redesign v9

At early days we've been using only a few syscalls
which together with debug compiler options always
produce relative addresses for memory variables used
in parasite and restorer blobs. Thus it came in unnoticed
that there is something worng with syscalls declarations
we use.

Basically all our syscalls are just a wrappers over inline
assembly code in form of

static long syscall2(int nr, long arg0, long arg1)
{
	long ret;
	asm volatile(
		"movl %1, %%eax		\t\n"
		"movq %2, %%rdi		\t\n"
		"movq %3, %%rsi		\t\n"
		"syscall		\t\n"
		"movq %%rax, %0		\t\n"
		: "=r"(ret)
		: "g" ((int)nr), "g" (arg0), "g" (arg1)
		: "rax", "rdi", "rsi", "memory");
	return ret;
}

so every argument treated to be plain long (even if the call
sematics implies it's a memory address passed but not some
integer direct value) and transferred via general purpose
register.

As being mentioned it caused no problems when debug options
specified at compile time, the compiler do not tries to optimize
addressing but generates code which always compute them.

The situation is changed if one is building crtools with
optimization enabled -- the compiler finds that arguments
are rather plain long numbers and might pass direct addresses
of variables, instead of generating relative addresses
(because function declarations have no pointers and 'g' in cope
 with 'mov' is used, which is of course wrong).

To fix all this -- now syscalls declarations are generated from
syscall.def file and function arguments are passed in conform
with x86-64 ABI.

This shrinks amount of source code needed to declare syscalls
and opens a way to use optimization.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Makefile

@@ -48,6 +48,7 @@ OBJS		+= ipc_ns.o
 
 DEPS		:= $(patsubst %.o,%.d,$(OBJS))
 
+include Makefile.syscall
 include Makefile.pie
 
 all: $(PROGRAM)
@@ -64,11 +65,11 @@ all: $(PROGRAM)
 	$(E) "  CC      " $@
 	$(Q) $(CC) -S $(CFLAGS) -fverbose-asm $< -o $@
 
-$(PROGRAM): $(OBJS) | $(PIE-GEN)
+$(PROGRAM): $(OBJS) | $(SYS-OBJ) $(PIE-GEN)
 	$(E) "  LINK    " $@
-	$(Q) $(CC) $(CFLAGS) $(OBJS) $(LIBS) -o $@
+	$(Q) $(CC) $(CFLAGS) $(OBJS) $(LIBS) $(SYS-OBJ) -o $@
 
-%.d: %.c | $(PIE-GEN)
+%.d: %.c | $(SYS-OBJ) $(PIE-GEN)
 	$(Q) $(CC) -M -MT $(patsubst %.d,%.o,$@) $(CFLAGS) $< -o $@
 
 test-legacy: $(PROGRAM)
@@ -90,7 +91,7 @@ rebuild:
 	$(Q) $(MAKE)
 .PHONY: rebuild
 
-clean: cleanpie
+clean: cleanpie cleansyscall
 	$(E) "  CLEAN"
 	$(Q) $(RM) -f ./*.o
 	$(Q) $(RM) -f ./*.d

Makefile.pie

@@ -26,15 +26,15 @@ DEPS		+= $(patsubst %.o,%.d,$(ROBJS))
 PIEFLAGS	:= -fpie
 ASMFLAGS	:= -D__ASSEMBLY__
 
-$(PASM-OBJS): $(PASM-SRC)
+$(PASM-OBJS): $(PASM-SRC) $(SYS-OBJ)
 	$(E) "  CC      " $@
 	$(Q) $(CC) -c $(ASMFLAGS) $(CFLAGS) $(PIEFLAGS) $(patsubst %.o,%.S,$@) -o $@
 
-$(POBJS): $(PSRCS) $(PASM-OBJS)
+$(POBJS): $(PSRCS) $(PASM-OBJS) $(SYS-OBJ)
 	$(E) "  CC      " $@
 	$(Q) $(CC) -c $(CFLAGS) $(PIEFLAGS) $(patsubst %.o,%.c,$@) -o $@
 
-parasite-util-net.o: util-net.c
+parasite-util-net.o: util-net.c $(SYS-OBJ)
 	$(E) "  CC      " $@
 	$(Q) $(CC) -c $(CFLAGS) $(PIEFLAGS) $< -o $@
 
@@ -42,21 +42,21 @@ POBJS		+= parasite-util-net.o
 
 $(PBLOB-BIN): $(PBLOB-LDS) $(POBJS) $(PASM-OBJS)
 	$(E) "  GEN     " $@
-	$(Q) $(LD) --oformat=binary -T $(PBLOB-LDS) -o $(PBLOB-BIN) $(POBJS) $(PASM-OBJS)
-	$(Q) $(LD) --oformat=elf64-x86-64 -T $(PBLOB-LDS) -o $(PBLOB-BIN).o $(POBJS) $(PASM-OBJS)
+	$(Q) $(LD) --oformat=binary -T $(PBLOB-LDS) -o $(PBLOB-BIN) $(POBJS) $(PASM-OBJS) $(SYS-OBJ)
+	$(Q) $(LD) --oformat=elf64-x86-64 -T $(PBLOB-LDS) -o $(PBLOB-BIN).o $(POBJS) $(PASM-OBJS) $(SYS-OBJ)
 
 $(PBLOB-HDR): $(PBLOB-BIN) $(GEN-OFFSETS)
 	$(E) "  GEN     " $@
 	$(Q) $(SH) $(GEN-OFFSETS) $(PBLOB-NAME) > $@ || rm -f $@
 
-$(ROBJS): $(RSRCS)
+$(ROBJS): $(RSRCS) $(SYS-OBJ)
 	$(E) "  CC      " $@
 	$(Q) $(CC) -c $(CFLAGS) $(PIEFLAGS) $(patsubst %.o,%.c,$@) -o $@
 
 $(RBLOB-BIN): $(RBLOB-LDS) $(ROBJS)
 	$(E) "  GEN     " $@
-	$(Q) $(LD) --oformat=binary -T $(RBLOB-LDS) -o $(RBLOB-BIN) $(ROBJS)
-	$(Q) $(LD) --oformat=elf64-x86-64 -T $(RBLOB-LDS) -o $(RBLOB-BIN).o $(ROBJS)
+	$(Q) $(LD) --oformat=binary -T $(RBLOB-LDS) -o $(RBLOB-BIN) $(ROBJS) $(SYS-OBJ)
+	$(Q) $(LD) --oformat=elf64-x86-64 -T $(RBLOB-LDS) -o $(RBLOB-BIN).o $(ROBJS) $(SYS-OBJ)
 
 $(RBLOB-HDR): $(RBLOB-BIN) $(GEN-OFFSETS)
 	$(E) "  GEN     " $@

Makefile.syscall

+SYS-DEF		:= include/syscall-x86-64.def
+SYS-ASM-COMMON	:= syscall-common-x86-64.S
+SYS-TYPES	:= include/syscall-types.h
+
+SYS-CODES	:= include/syscall-codes.h
+SYS-PROTO	:= include/syscall.h
+
+SYS-ASM		:= syscall-x86-64.S
+SYS-GEN		:= syscalls-x86-64.sh
+
+SYS-OBJ		:= $(patsubst %.S,%.o,$(SYS-ASM))
+
+SYS-FLAGS	:= -pie -Wstrict-prototypes -D__ASSEMBLY__ -nostdlib -fomit-frame-pointer
+
+$(SYS-ASM): $(SYS-GEN) $(SYS-DEF) $(SYS-ASM-COMMON) $(SYS-TYPES)
+	$(E) "  GEN     " $@
+	$(Q) $(SH)			\
+		$(SYS-GEN)		\
+		$(SYS-DEF)		\
+		$(SYS-CODES)		\
+		$(SYS-PROTO)		\
+		$(SYS-ASM)		\
+		$(SYS-ASM-COMMON)	\
+		$(SYS-TYPES)
+
+$(SYS-OBJ): $(SYS-ASM)
+	$(E) "  CC      " $@
+	$(Q) $(CC) -c $(CFLAGS) $(SYS-FLAGS)  $< -o $@
+
+cleansyscall:
+	$(E) "  CLEAN SYSCALLS"
+	$(Q) $(RM) -f ./$(SYS-ASM)
+	$(Q) $(RM) -f ./$(SYS-CODES)
+	$(Q) $(RM) -f ./$(SYS-PROTO)

include/syscall-codes.h

-#ifndef CR_SYSCALL_CODES_H_
-#define CR_SYSCALL_CODES_H_
-
-#ifdef CONFIG_X86_64
-
-#define __NR_read		0
-#define __NR_write		1
-#define __NR_open		2
-#define __NR_close		3
-#define __NR_lseek		8
-#define __NR_mmap		9
-#define __NR_mprotect		10
-#define __NR_munmap		11
-#define __NR_brk		12
-#define __NR_rt_sigaction	13
-#define __NR_rt_sigprocmask	14
-#define __NR_rt_sigreturn	15
-#define __NR_mincore		27
-#define __NR_shmat		30
-#define __NR_dup		32
-#define __NR_dup2		33
-#define __NR_pause		34
-#define __NR_nanosleep		35
-#define __NR_getitimer		36
-#define __NR_setitimer		38
-#define __NR_getpid		39
-#define __NR_socket		41
-#define __NR_sendmsg		46
-#define __NR_recvmsg		47
-#define __NR_bind		49
-#define __NR_setsockopt		54
-#define __NR_getsockopt		55
-#define __NR_clone		56
-#define __NR_exit		60
-#define __NR_wait4		61
-#define __NR_kill		62
-#define __NR_fcntl		72
-#define __NR_flock		73
-#define __NR_unlink		87
-#define __NR_setresuid		117
-#define __NR_setresgid		119
-#define __NR_setfsuid		122
-#define __NR_setfsgid		123
-#define __NR_capset		126
-#define __NR_tgkill		131
-#define __NR__sysctl		156
-#define __NR_prctl		157
-#define __NR_arch_prctl		158
-#define __NR_gettid		186
-#define __NR_futex		202
-#define __NR_set_thread_area	205
-#define __NR_get_thread_area	211
-#define __NR_set_tid_address	218
-#define __NR_restart_syscall	219
-#define __NR_msync		227
-#define __NR_setns		308
-#define __NR_kcmp		312
-
-#else /* CONFIG_X86_64 */
-# error x86-32 bit mode not yet implemented
-#endif /* CONFIG_X86_64 */
-
-#endif /* CR_SYSCALL_CODES_H_ */

include/syscall-x86-64.def

+#
+# System calls table, please make sure the table consist only the syscalls
+# really used somewhere in project.
+#
+# The template is (name and srguments are optinal if you need only __NR_x
+# defined, but no realy entry point in syscalls lib).
+#
+# name			code		name			arguments
+# -----------------------------------------------------------------------
+#
+__NR_read		0		sys_read		(int fd, void *buf, unsigned long count)
+__NR_write		1		sys_write		(int fd, const void *buf, unsigned long count)
+__NR_open		2		sys_open		(const char *filename, unsigned long flags, unsigned long mode)
+__NR_close		3		sys_close		(int fd)
+__NR_lseek		8		sys_lseek		(int fd, unsigned long offset, unsigned long origin)
+__NR_mmap		9		sys_mmap		(void *addr, unsigned long len, unsigned long prot, unsigned long flags, unsigned long fd, unsigned long offset)
+__NR_mprotect		10		sys_mprotect		(const void *addr, unsigned long len, unsigned long prot)
+__NR_munmap		11		sys_munmap		(void *addr, unsigned long len)
+__NR_brk		12		sys_brk			(void *addr)
+__NR_rt_sigaction	13		sys_sigaction		(int signum, const rt_sigaction_t *act, rt_sigaction_t *oldact, size_t sigsetsize)
+__NR_rt_sigprocmask	14		sys_sigprocmask		(int how, k_rtsigset_t *set, k_rtsigset_t *old, size_t sigsetsize)
+__NR_rt_sigreturn	15		sys_rt_sigreturn	(void)
+__NR_mincore		27		sys_mincore		(void *addr, unsigned long size, unsigned char *vec)
+__NR_shmat		30		sys_shmat		(int shmid, void *shmaddr, int shmflag)
+__NR_pause		34		sys_pause		(void)
+__NR_nanosleep		35		sys_nanosleep		(struct timespec *req, struct timespec *rem)
+__NR_getitimer		36		sys_getitimer		(int which, const struct itimerval *val)
+__NR_setitimer		38		sys_setitimer		(int which, const struct itimerval *val, struct itimerval *old)
+__NR_getpid		39		sys_getpid		(void)
+__NR_socket		41		sys_socket		(int domain, int type, int protocol)
+__NR_sendmsg		46		sys_sendmsg		(int sockfd, const struct msghdr *msg, int flags)
+__NR_recvmsg		47		sys_recvmsg		(int sockfd, struct msghdr *msg, int flags)
+__NR_bind		49		sys_bind		(int sockfd, const struct sockaddr *addr, int addrlen)
+__NR_setsockopt		54		sys_setsockopt		(int sockfd, int level, int optname, const void *optval, socklen_t optlen)
+__NR_getsockopt		55		sys_getsockopt		(int sockfd, int level, int optname, const void *optval, socklen_t *optlen)
+__NR_clone		56		sys_clone		(unsigned long flags, void *child_stack, void *parent_tid, void *child_tid)
+__NR_exit		60		sys_exit		(unsigned long error_code)
+__NR_wait4		61		sys_waitpid		(int pid, int *status, int options, struct rusage *ru)
+__NR_kill		62		sys_kill		(long pid, int sig)
+__NR_fcntl		72		sys_fcntl		(int fd, int type, long arg)
+__NR_flock		73		sys_flock		(int fd, unsigned long cmd)
+__NR_unlink		87		sys_unlink		(char *pathname)
+__NR_setresuid		117		sys_setresuid		(int uid, int euid, int suid)
+__NR_setresgid		119		sys_setresgid		(int gid, int egid, int sgid)
+__NR_setfsuid		122		sys_setfsuid		(int fsuid)
+__NR_setfsgid		123		sys_setfsgid		(int fsgid)
+__NR_capset		126		sys_capset		(struct cap_header *h, struct cap_data *d)
+__NR_personality	135		sys_personality		(unsigned int personality)
+__NR_prctl		157		sys_prctl		(int option, unsigned long arg2, unsigned long arg3, unsigned long arg4, unsigned long arg5)
+__NR_arch_prctl		158		sys_arch_prctl		(int option, unsigned long addr)
+__NR_gettid		186		sys_gettid		(void)
+__NR_futex		202		sys_futex		(u32 *uaddr, int op, u32 val, struct timespec *utime, u32 *uaddr2, u32 val3)
+__NR_set_thread_area	205		sys_set_thread_area	(user_desc_t *info)
+__NR_get_thread_area	211		sys_get_thread_area	(user_desc_t *info)
+__NR_set_tid_address	218		sys_set_tid_address	(int *tid_addr)
+__NR_restart_syscall	219		sys_restart_syscall	(void)
+__NR_setns		308		sys_setns		(int fd, int nstype)
+__NR_kcmp		312		sys_kcmp		(pid_t pid1, pid_t pid2, int type, unsigned long idx1, unsigned long idx2)

syscall-common-x86-64.S

+#include "linkage.h"
+
+#define SYSCALL(name, opcode)		\
+	ENTRY(name);			\
+	movl	$opcode, %eax;		\
+	jmp	__syscall_common;	\
+	END(name)
+
+	.text
+	.align	4
+
+ENTRY(__syscall_common)
+	movq	%rcx, %r10
+	syscall
+	ret
+END(__syscall_common)

syscalls-x86-64.sh

+#!/bin/sh
+
+in=$1
+codesout=$2
+codes=`echo $2 | sed -e 's/include\///g'`
+protosout=$3
+protos=`echo $3 | sed -e 's/include\///g'`
+asmout=$4
+asmcommon=$5
+prototypes=`echo $6 | sed -e 's/include\///g'`
+
+codesdef=`echo $codes | tr "[[:space:]].-" _`
+protosdef=`echo $protos | tr "[[:space:]].-" _`
+
+echo "/* Autogenerated, don't edit */"	>  $codesout
+echo "#ifndef $codesdef"		>> $codesout
+echo "#define $codesdef"		>> $codesout
+
+echo "/* Autogenerated, don't edit */"	>  $protosout
+echo "#ifndef $protosdef"		>> $protosout
+echo "#define $protosdef"		>> $protosout
+echo "#include \"$prototypes\""		>> $protosout
+echo "#include \"$codes\""		>> $protosout
+
+echo "/* Autogenerated, don't edit */"	>  $asmout
+echo "#include \"$codes\""		>> $asmout
+echo "#include \"$asmcommon\""		>> $asmout
+
+cat $in | egrep -v '^#' | sed -e 's/\t\{1,\}/|/g' | awk -F '|' '{print "#define", $1, $2}'		>> $codesout
+cat $in | egrep -v '^#' | sed -e 's/\t\{1,\}/|/g' | awk -F '|' '{print "extern long ", $3, $4, ";"}'	>> $protosout
+cat $in | egrep -v '^#' | sed -e 's/\t\{1,\}/|/g' | awk -F '|' '{print "SYSCALL(", $3, ",", $2, ")"}'	>> $asmout
+
+echo "#endif /* $codesdef */"		>> $codesout
+echo "#endif /* $protosdef */"		>> $protosout