Re: [PATCH] compatibility syscall layer (lets try again)

[george anzinger <george@mvista.com>]

[-- Attachment #1: Type: text/plain, Size: 1282 bytes --]

Linus Torvalds wrote:
> 
> On Wed, 4 Dec 2002, george anzinger wrote:
> >
> > Once it changes the system call (eax, right), could the new
> > call code then just get the parms from the restart_block.
> 
> Agreed.
> 
> > I think it would be best to keep this as generic as
> > possible, i.e. let the new call code fetch its own
> > paramerers from the restart_block.
> 
> We could even have one _single_ a generic "restart" system call, and have
> the function pointer for that be in the restart block.
> 
> > My question is who sets up these values?  I think you are
> > saying it should be the system call.  Is this right?
> 
> Whatever system call that return -ERESTART_RESTARTBLOCK, yes.
> 
> So it would never get set up at all in the fast path. Only in the error
> case path of a system call that wants to have restarting capabilities.
> 
>                 Linus

Ok, here is a patch to do all this.  This is against
2.5.50-bk4.  (Bk5 did not get posted this evening.)

I think this covers all the bases.  It builds boots and
runs.  I haven't tested nano_sleep to see if it does the
right thing yet...

-- 
George Anzinger   george@mvista.com
High-res-timers: 
http://sourceforge.net/projects/high-res-timers/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml

[-- Attachment #2: regs-fix-2.5.50-bk4.1.0.patch --]
[-- Type: text/plain, Size: 6109 bytes --]

diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/arch/i386/kernel/entry.S linux/arch/i386/kernel/entry.S
--- linux-2.5.50-bk4-kb/arch/i386/kernel/entry.S	Wed Dec  4 23:28:20 2002
+++ linux/arch/i386/kernel/entry.S	Wed Dec  4 23:48:49 2002
@@ -769,6 +769,7 @@
 	.long sys_epoll_wait
  	.long sys_remap_file_pages
  	.long sys_set_tid_address
+	.long sys_restart_syscall
 
 
 	.rept NR_syscalls-(.-sys_call_table)/4
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/arch/i386/kernel/signal.c linux/arch/i386/kernel/signal.c
--- linux-2.5.50-bk4-kb/arch/i386/kernel/signal.c	Thu Oct  3 10:41:57 2002
+++ linux/arch/i386/kernel/signal.c	Thu Dec  5 00:27:21 2002
@@ -507,6 +507,7 @@
 		/* If so, check system call restarting.. */
 		switch (regs->eax) {
 			case -ERESTARTNOHAND:
+		        case -ERESTART_RESTARTBLOCK:
 				regs->eax = -EINTR;
 				break;
 
@@ -589,6 +590,10 @@
 		    regs->eax == -ERESTARTSYS ||
 		    regs->eax == -ERESTARTNOINTR) {
 			regs->eax = regs->orig_eax;
+			regs->eip -= 2;
+		}
+		if (regs->eax == -ERESTART_RESTARTBLOCK){
+			regs->eax = __NR_restart_syscall;
 			regs->eip -= 2;
 		}
 	}
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/include/asm-i386/thread_info.h linux/include/asm-i386/thread_info.h
--- linux-2.5.50-bk4-kb/include/asm-i386/thread_info.h	Mon Sep  9 10:35:03 2002
+++ linux/include/asm-i386/thread_info.h	Thu Dec  5 01:07:23 2002
@@ -20,6 +20,12 @@
  * - if the contents of this structure are changed, the assembly constants must also be changed
  */
 #ifndef __ASSEMBLY__
+struct restart_block {
+	int (*fun)(void *);
+	long arg0;
+	long arg1;
+};
+
 struct thread_info {
 	struct task_struct	*task;		/* main task structure */
 	struct exec_domain	*exec_domain;	/* execution domain */
@@ -31,6 +37,7 @@
 					 	   0-0xBFFFFFFF for user-thead
 						   0-0xFFFFFFFF for kernel-thread
 						*/
+	struct restart_block    restart_block;
 
 	__u8			supervisor_stack[0];
 };
@@ -44,6 +51,7 @@
 #define TI_CPU		0x0000000C
 #define TI_PRE_COUNT	0x00000010
 #define TI_ADDR_LIMIT	0x00000014
+#define TI_RESTART_BLOCK 0x0000018
 
 #endif
 
@@ -63,6 +71,9 @@
 	.cpu		= 0,			\
 	.preempt_count	= 1,			\
 	.addr_limit	= KERNEL_DS,		\
+	.restart_block = {                      \
+		.fun = 0,                       \
+	},		                        \
 }
 
 #define init_thread_info	(init_thread_union.thread_info)
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/include/asm-i386/unistd.h linux/include/asm-i386/unistd.h
--- linux-2.5.50-bk4-kb/include/asm-i386/unistd.h	Wed Nov 27 15:49:22 2002
+++ linux/include/asm-i386/unistd.h	Wed Dec  4 23:48:46 2002
@@ -263,7 +263,7 @@
 #define __NR_sys_epoll_wait	256
 #define __NR_remap_file_pages	257
 #define __NR_set_tid_address	258
-
+#define __NR_restart_syscall    259
 
 /* user-visible error numbers are in the range -1 - -124: see <asm-i386/errno.h> */
 
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/include/linux/errno.h linux/include/linux/errno.h
--- linux-2.5.50-bk4-kb/include/linux/errno.h	Mon Sep  9 10:35:15 2002
+++ linux/include/linux/errno.h	Wed Dec  4 23:53:21 2002
@@ -10,6 +10,7 @@
 #define ERESTARTNOINTR	513
 #define ERESTARTNOHAND	514	/* restart if no handler.. */
 #define ENOIOCTLCMD	515	/* No ioctl command */
+#define ERESTART_RESTARTBLOCK 516 /* restart by calling sys_restart_syscall */
 
 /* Defined for the NFSv3 protocol */
 #define EBADHANDLE	521	/* Illegal NFS file handle */
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/kernel/signal.c linux/kernel/signal.c
--- linux-2.5.50-bk4-kb/kernel/signal.c	Wed Dec  4 23:27:02 2002
+++ linux/kernel/signal.c	Thu Dec  5 01:18:20 2002
@@ -1351,6 +1351,15 @@
  * System call entry points.
  */
 
+asmlinkage long
+sys_restart_syscall( void *parm)
+{
+	if ( ! current_thread_info()->restart_block.fun){
+		return current_thread_info()->restart_block.fun(&parm);
+	}
+	return -ENOSYS;
+}
+
 /*
  * We don't need to get the kernel lock - this is all local to this
  * particular thread.. (and that's good, because this is _heavily_
diff -urP -I \$Id:.*Exp \$ -X /usr/src/patch.exclude linux-2.5.50-bk4-kb/kernel/timer.c linux/kernel/timer.c
--- linux-2.5.50-bk4-kb/kernel/timer.c	Wed Dec  4 23:27:02 2002
+++ linux/kernel/timer.c	Thu Dec  5 01:16:03 2002
@@ -1020,19 +1020,39 @@
 	return current->pid;
 }
 
+struct nano_sleep_call {
+	struct timespec *rqtp; 
+	struct timespec *rmtp;
+};
+
+asmlinkage long sys_nanosleep_restart( struct nano_sleep_call * parms);
+
 asmlinkage long sys_nanosleep(struct timespec *rqtp, struct timespec *rmtp)
 {
 	struct timespec t;
 	unsigned long expire;
+	struct  restart_block *restart_block;
 
-	if(copy_from_user(&t, rqtp, sizeof(struct timespec)))
-		return -EFAULT;
-
-	if (t.tv_nsec >= 1000000000L || t.tv_nsec < 0 || t.tv_sec < 0)
-		return -EINVAL;
+	if (rqtp) {
+		if(copy_from_user(&t, rqtp, sizeof(struct timespec)))
+			return -EFAULT;
 
-	expire = timespec_to_jiffies(&t) + (t.tv_sec || t.tv_nsec);
+		if (t.tv_nsec >= 1000000000L || t.tv_nsec < 0 || t.tv_sec < 0)
+			return -EINVAL;
+		expire = timespec_to_jiffies(&t) + (t.tv_sec || t.tv_nsec);
 
+	}else{
+		restart_block = &current_thread_info()->restart_block;
+		if( restart_block->fun != 
+		    (int (*)(void *))sys_nanosleep_restart ||
+		    ! restart_block->arg0){
+			return  -EFAULT;
+		}
+		restart_block->fun = NULL;
+		expire = restart_block->arg0 - jiffies;
+		if (expire < 0)
+			return 0;
+	}
 	current->state = TASK_INTERRUPTIBLE;
 	expire = schedule_timeout(expire);
 
@@ -1042,10 +1062,19 @@
 			if (copy_to_user(rmtp, &t, sizeof(struct timespec)))
 				return -EFAULT;
 		}
-		return -EINTR;
+		restart_block = &current_thread_info()->restart_block;
+		restart_block->fun = (int (*)(void *))sys_nanosleep_restart;
+		restart_block->arg0 = jiffies + expire;
+		return -ERESTART_RESTARTBLOCK;
 	}
 	return 0;
 }
+
+asmlinkage long sys_nanosleep_restart( struct nano_sleep_call * parms)
+{
+	return sys_nanosleep(NULL, parms->rmtp);
+}
+
 
 /*
  * sys_sysinfo - fill in sysinfo struct