Reorganize syscall entry and leave handling.

Extend struct sysvec with three new elements:
sv_fetch_syscall_args - the method to fetch syscall arguments from
  usermode into struct syscall_args. The structure is machine-depended
  (this might be reconsidered after all architectures are converted).
sv_set_syscall_retval - the method to set a return value for usermode
  from the syscall. It is a generalization of
  cpu_set_syscall_retval(9) to allow ABIs to override the way to set a
  return value.
sv_syscallnames - the table of syscall names.

Use sv_set_syscall_retval in kern_sigsuspend() instead of hardcoding
the call to cpu_set_syscall_retval().

The new functions syscallenter(9) and syscallret(9) are provided that
use sv_*syscall* pointers and contain the common repeated code from
the syscall() implementations for the architecture-specific syscall
trap handlers.

Syscallenter() fetches arguments, calls syscall implementation from
ABI sysent table, and set up return frame. The end of syscall
bookkeeping is done by syscallret().

Take advantage of single place for MI syscall handling code and
implement ptrace_lwpinfo pl_flags PL_FLAG_SCE, PL_FLAG_SCX and
PL_FLAG_EXEC. The SCE and SCX flags notify the debugger that the
thread is stopped at syscall entry or return point respectively.  The
EXEC flag augments SCX and notifies debugger that the process address
space was changed by one of exec(2)-family syscalls.

The i386, amd64, sparc64, sun4v, powerpc and ia64 syscall()s are
changed to use syscallenter()/syscallret(). MIPS and arm are not
converted and use the mostly unchanged syscall() implementation.

Reviewed by:	jhb, marcel, marius, nwhitehorn, stas
Tested by:	marcel (ia64), marius (sparc64), nwhitehorn (powerpc),
	stas (mips)
MFC after:	1 month

1 files changed, 51 insertions, 106 deletions

diff --git a/sys/sun4v/sun4v/trap.c b/sys/sun4v/sun4v/trap.c
index 702fd5ce28fd..d099351e3e62 100644
--- a/sys/sun4v/sun4v/trap.c
+++ b/sys/sun4v/sun4v/trap.c
@@ -117,8 +117,6 @@ extern char fas_fault[];
 extern char fas_nofault_begin[];
 extern char fas_nofault_end[];
 
-extern char *syscallnames[];
-
 const char *const trap_msg[] = {
 	"reserved",
 	"instruction access exception",
@@ -566,132 +564,79 @@ trap_pfault(struct thread *td, struct trapframe *tf, int64_t type, uint64_t data
 /* Maximum number of arguments that can be passed via the out registers. */
 #define	REG_MAXARGS	6
 
-/*
- * Syscall handler. The arguments to the syscall are passed in the o registers
- * by the caller, and are saved in the trap frame. The syscall number is passed
- * in %g1 (and also saved in the trap frame).
- */
-void
-syscall(struct trapframe *tf)
+int
+cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
 {
-	struct sysent *callp;
-	struct thread *td;
-	register_t args[8];
-	register_t *argp;
+	struct trapframe *tf;
 	struct proc *p;
-	u_long code;
+	register_t *argp;
 	int reg;
 	int regcnt;
-	int narg;
 	int error;
 
-	td = curthread;
-	KASSERT(td != NULL, ("trap: curthread NULL"));
-	KASSERT(td->td_proc != NULL, ("trap: curproc NULL"));
-
 	p = td->td_proc;
-
-	PCPU_INC(cnt.v_syscall);
-
-	td->td_pticks = 0;
-	td->td_frame = tf;
-	if (td->td_ucred != p->p_ucred)
-		cred_update_thread(td);
-	code = tf->tf_global[1];
-
-	/*
-	 * For syscalls, we don't want to retry the faulting instruction
-	 * (usually), instead we need to advance one instruction.
-	 */
-	td->td_pcb->pcb_tpc = tf->tf_tpc;
-	TF_DONE(tf);
-
+	tf = td->td_frame;
 	reg = 0;
 	regcnt = REG_MAXARGS;
-	if (p->p_sysent->sv_prepsyscall) {
-		/*
-		 * The prep code is MP aware.
-		 */
-#if 0
-		(*p->p_sysent->sv_prepsyscall)(tf, args, &code, &params);
-#endif	
-	} else 	if (code == SYS_syscall || code == SYS___syscall) {
-		code = tf->tf_out[reg++];
-		regcnt--;
-	}
-
- 	if (p->p_sysent->sv_mask)
- 		code &= p->p_sysent->sv_mask;
 
- 	if (code >= p->p_sysent->sv_size)
- 		callp = &p->p_sysent->sv_table[0];
-  	else
- 		callp = &p->p_sysent->sv_table[code];
+	sa->code = tf->tf_global[1];
 
-	narg = callp->sy_narg;
-
-	if (narg <= regcnt) {
-		argp = &tf->tf_out[reg];
-		error = 0;
-	} else {
-		KASSERT(narg <= sizeof(args) / sizeof(args[0]),
-		    ("Too many syscall arguments!"));
-		argp = args;
-		bcopy(&tf->tf_out[reg], args, sizeof(args[0]) * regcnt);
-		error = copyin((void *)(tf->tf_out[6] + SPOFF +
-		    offsetof(struct frame, fr_pad[6])),
-		    &args[regcnt], (narg - regcnt) * sizeof(args[0]));
+	if (sa->code == SYS_syscall || sa->code == SYS___syscall) {
+		sa->code = tf->tf_out[reg++];
+		regcnt--;
 	}
 
-	CTR5(KTR_SYSC, "syscall: td=%p %s(%#lx, %#lx, %#lx)", td,
-	    syscallnames[code], argp[0], argp[1], argp[2]);
-
-	/*
-	 * Try to run the syscall without the MP lock if the syscall
-	 * is MP safe.
-	 */
-#ifdef KTRACE
-	if (KTRPOINT(td, KTR_SYSCALL))
-		ktrsyscall(code, narg, argp);
-#endif
+	if (p->p_sysent->sv_mask)
+		sa->code &= p->p_sysent->sv_mask;
+	if (sa->code >= p->p_sysent->sv_size)
+		sa->callp = &p->p_sysent->sv_table[0];
+	else
+		sa->callp = &p->p_sysent->sv_table[sa->code];
+
+	sa->narg = sa->callp->sy_narg;
+	KASSERT(sa->narg <= sizeof(sa->args) / sizeof(sa->args[0]),
+	    ("Too many syscall arguments!"));
+	error = 0;
+	argp = sa->args;
+	bcopy(&tf->tf_out[reg], sa->args, sizeof(sa->args[0]) * regcnt);
+	if (sa->narg > regcnt)
+		error = copyin((void *)(tf->tf_out[6] + SPOFF +
+		    offsetof(struct frame, fr_pad[6])), &sa->args[regcnt],
+		    (sa->narg - regcnt) * sizeof(sa->args[0]));
 	if (error == 0) {
 		td->td_retval[0] = 0;
 		td->td_retval[1] = 0;
+	}
 
-		STOPEVENT(p, S_SCE, narg);	/* MP aware */
-
-		PTRACESTOP_SC(p, td, S_PT_SCE);
-
-		AUDIT_SYSCALL_ENTER(code, td);
-		error = (*callp->sy_call)(td, argp);
-		AUDIT_SYSCALL_EXIT(error, td);
+	return (error);
+}
 
-		CTR5(KTR_SYSC, "syscall: p=%p error=%d %s return %#lx %#lx ", p,
-		    error, syscallnames[code], td->td_retval[0],
-		    td->td_retval[1]);
-	}
+/*
+ * Syscall handler. The arguments to the syscall are passed in the o registers
+ * by the caller, and are saved in the trap frame. The syscall number is passed
+ * in %g1 (and also saved in the trap frame).
+ */
+void
+syscall(struct trapframe *tf)
+{
+	struct thread *td;
+	struct syscall_args sa;
+	int error;
 
-	cpu_set_syscall_retval(td, error);
+	td = curthread;
+	td->td_frame = tf;
 
-	/*
-	 * Handle reschedule and other end-of-syscall issues
-	 */
-	userret(td, tf);
+	KASSERT(td != NULL, ("trap: curthread NULL"));
+	KASSERT(td->td_proc != NULL, ("trap: curproc NULL"));
 
-#ifdef KTRACE
-	if (KTRPOINT(td, KTR_SYSRET))
-		ktrsysret(code, error, td->td_retval[0]);
-#endif
 	/*
-	 * This works because errno is findable through the
-	 * register set.  If we ever support an emulation where this
-	 * is not the case, this code will need to be revisited.
+	 * For syscalls, we don't want to retry the faulting instruction
+	 * (usually), instead we need to advance one instruction.
 	 */
-	STOPEVENT(p, S_SCX, code);
-
-	PTRACESTOP_SC(p, td, S_PT_SCX);
+	td->td_pcb->pcb_tpc = tf->tf_tpc;
+	TF_DONE(tf);
 
-	WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
-	    (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
-	mtx_assert(&Giant, MA_NOTOWNED);
+	error = syscallenter(td, &sa);
+	syscallret(td, error, &sa);
 }
+