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

3 files changed, 31 insertions, 110 deletions

diff --git a/sys/sparc64/include/proc.h b/sys/sparc64/include/proc.h
index 3c23663..dceea17 100644
--- a/sys/sparc64/include/proc.h
+++ b/sys/sparc64/include/proc.h
@@ -53,4 +53,16 @@ struct mdproc {
 
 #define	KINFO_PROC_SIZE 1088
 
+#ifdef _KERNEL
+
+struct syscall_args {
+	u_int code;
+	struct sysent *callp;
+	register_t args[8];
+	int narg;
+};
+#define	HAVE_SYSCALL_ARGS_DEF 1
+
+#endif
+
 #endif /* !_MACHINE_PROC_H_ */
diff --git a/sys/sparc64/sparc64/elf_machdep.c b/sys/sparc64/sparc64/elf_machdep.c
index 2f9ee39..8ddf476 100644
--- a/sys/sparc64/sparc64/elf_machdep.c
+++ b/sys/sparc64/sparc64/elf_machdep.c
@@ -47,6 +47,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/exec.h>
 #include <sys/imgact.h>
 #include <sys/linker.h>
+#include <sys/proc.h>
 #include <sys/sysent.h>
 #include <sys/imgact_elf.h>
 #include <sys/syscall.h>
@@ -88,7 +89,10 @@ static struct sysentvec elf64_freebsd_sysvec = {
 	.sv_setregs	= exec_setregs,
 	.sv_fixlimit	= NULL,
 	.sv_maxssiz	= NULL,
-	.sv_flags	= SV_ABI_FREEBSD | SV_LP64
+	.sv_flags	= SV_ABI_FREEBSD | SV_LP64,
+	.sv_set_syscall_retval = cpu_set_syscall_retval,
+	.sv_fetch_syscall_args = cpu_fetch_syscall_args,
+	.sv_syscallnames = syscallnames,
 };
 
 static Elf64_Brandinfo freebsd_brand_info = {
diff --git a/sys/sparc64/sparc64/trap.c b/sys/sparc64/sparc64/trap.c
index c0772a1..8ce6970 100644
--- a/sys/sparc64/sparc64/trap.c
+++ b/sys/sparc64/sparc64/trap.c
@@ -94,18 +94,9 @@ __FBSDID("$FreeBSD$");
 #include <machine/tsb.h>
 #include <machine/watch.h>
 
-struct syscall_args {
-	u_long code;
-	struct sysent *callp;
-	register_t args[8];
-	register_t *argp;
-	int narg;
-};
-
 void trap(struct trapframe *tf);
 void syscall(struct trapframe *tf);
 
-static int fetch_syscall_args(struct thread *td, struct syscall_args *sa);
 static int trap_cecc(void);
 static int trap_pfault(struct thread *td, struct trapframe *tf);
 
@@ -123,8 +114,6 @@ 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",
@@ -579,11 +568,12 @@ trap_pfault(struct thread *td, struct trapframe *tf)
 /* Maximum number of arguments that can be passed via the out registers. */
 #define	REG_MAXARGS	6
 
-static int
-fetch_syscall_args(struct thread *td, struct syscall_args *sa)
+int
+cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
 {
 	struct trapframe *tf;
 	struct proc *p;
+	register_t *argp;
 	int reg;
 	int regcnt;
 	int error;
@@ -595,19 +585,13 @@ fetch_syscall_args(struct thread *td, struct syscall_args *sa)
 
 	sa->code = tf->tf_global[1];
 
-	if (p->p_sysent->sv_prepsyscall) {
-#if 0
-		(*p->p_sysent->sv_prepsyscall)(tf, sa->args, &sa->code,
-		    &params);
-#endif
-	} else if (sa->code == SYS_syscall || sa->code == SYS___syscall) {
+	if (sa->code == SYS_syscall || sa->code == SYS___syscall) {
 		sa->code = tf->tf_out[reg++];
 		regcnt--;
 	}
 
 	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
@@ -617,21 +601,17 @@ fetch_syscall_args(struct thread *td, struct syscall_args *sa)
 	KASSERT(sa->narg <= sizeof(sa->args) / sizeof(sa->args[0]),
 	    ("Too many syscall arguments!"));
 	error = 0;
-	sa->argp = sa->args;
+	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;
+	}
 
-	/*
-	 * This may result in two records if debugger modified
-	 * registers or memory during sleep at stop/ptrace point.
-	 */
-#ifdef KTRACE
-	if (KTRPOINT(td, KTR_SYSCALL))
-		ktrsyscall(sa->code, sa->narg, sa->argp);
-#endif
 	return (error);
 }
 
@@ -644,29 +624,16 @@ fetch_syscall_args(struct thread *td, struct syscall_args *sa)
 void
 syscall(struct trapframe *tf)
 {
-	struct syscall_args sa;
 	struct thread *td;
-	struct proc *p;
+	struct syscall_args sa;
 	int error;
 
 	td = curthread;
+	td->td_frame = tf;
+
 	KASSERT(td != NULL, ("trap: curthread NULL"));
 	KASSERT(td->td_proc != NULL, ("trap: curproc NULL"));
 
-	PCPU_INC(cnt.v_syscall);
-	p = td->td_proc;
-	td->td_syscalls++;
-
-	td->td_pticks = 0;
-	td->td_frame = tf;
-	if (td->td_ucred != p->p_ucred)
-		cred_update_thread(td);
-	if ((p->p_flag & P_TRACED) != 0) {
-		PROC_LOCK(p);
-		td->td_dbgflags &= ~TDB_USERWR;
-		PROC_UNLOCK(p);
-	}
-
 	/*
 	 * For syscalls, we don't want to retry the faulting instruction
 	 * (usually), instead we need to advance one instruction.
@@ -674,68 +641,6 @@ syscall(struct trapframe *tf)
 	td->td_pcb->pcb_tpc = tf->tf_tpc;
 	TF_DONE(tf);
 
-	error = fetch_syscall_args(td, &sa);
-	CTR5(KTR_SYSC, "syscall: td=%p %s(%#lx, %#lx, %#lx)", td,
-	    syscallnames[sa.code], sa.argp[0], sa.argp[1], sa.argp[2]);
-
-	if (error == 0) {
-		td->td_retval[0] = 0;
-		td->td_retval[1] = 0;
-
-		STOPEVENT(p, S_SCE, sa.narg);
-		PTRACESTOP_SC(p, td, S_PT_SCE);
-		if ((td->td_dbgflags & TDB_USERWR) != 0) {
-			/*
-			 * Reread syscall number and arguments if
-			 * debugger modified registers or memory.
-			 */
-			error = fetch_syscall_args(td, &sa);
-			if (error != 0)
-				goto retval;
-			td->td_retval[1] = 0;
-		}
-
-		AUDIT_SYSCALL_ENTER(sa.code, td);
-		error = (*sa.callp->sy_call)(td, sa.argp);
-		AUDIT_SYSCALL_EXIT(error, td);
-
-		CTR5(KTR_SYSC, "syscall: p=%p error=%d %s return %#lx %#lx",
-		    p, error, syscallnames[sa.code], td->td_retval[0],
-		    td->td_retval[1]);
-	}
- retval:
-	cpu_set_syscall_retval(td, error);
-
-	/*
-	 * Check for misbehavior.
-	 */
-	WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
-	    (sa.code >= 0 && sa.code < SYS_MAXSYSCALL) ?
-	    syscallnames[sa.code] : "???");
-	KASSERT(td->td_critnest == 0,
-	    ("System call %s returning in a critical section",
-	    (sa.code >= 0 && sa.code < SYS_MAXSYSCALL) ?
-	    syscallnames[sa.code] : "???"));
-	KASSERT(td->td_locks == 0,
-	    ("System call %s returning with %d locks held",
-	    (sa.code >= 0 && sa.code < SYS_MAXSYSCALL) ?
-	    syscallnames[sa.code] : "???", td->td_locks));
-
-	/*
-	 * Handle reschedule and other end-of-syscall issues.
-	 */
-	userret(td, tf);
-
-#ifdef KTRACE
-	if (KTRPOINT(td, KTR_SYSRET))
-		ktrsysret(sa.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.
-	 */
-	STOPEVENT(p, S_SCX, sa.code);
-
-	PTRACESTOP_SC(p, td, S_PT_SCX);
+	error = syscallenter(td, &sa);
+	syscallret(td, error, &sa);
 }