Merge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
arch/tile: handle rt_sigreturn() more cleanly
arch/tile: handle CLONE_SETTLS in copy_thread(), not user space

arch/tile/include/asm/signal.h

···
 
 #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
 struct pt_regs;
-int restore_sigcontext(struct pt_regs *, struct sigcontext __user *, long *);
+int restore_sigcontext(struct pt_regs *, struct sigcontext __user *);
 int setup_sigcontext(struct sigcontext __user *, struct pt_regs *);
 void do_signal(struct pt_regs *regs);
 #endif

arch/tile/kernel/compat_signal.c

···
 	return ret;
 }
 
+/* The assembly shim for this function arranges to ignore the return value. */
 long compat_sys_rt_sigreturn(struct pt_regs *regs)
 {
 	struct compat_rt_sigframe __user *frame =
 		(struct compat_rt_sigframe __user *) compat_ptr(regs->sp);
 	sigset_t set;
-	long r0;
 
 	if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
 		goto badframe;
···
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
 
-	if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &r0))
+	if (restore_sigcontext(regs, &frame->uc.uc_mcontext))
 		goto badframe;
 
 	if (compat_sys_sigaltstack(&frame->uc.uc_stack, NULL, regs) != 0)
 		goto badframe;
 
-	return r0;
+	return 0;
 
 badframe:
 	force_sig(SIGSEGV, current);

arch/tile/kernel/intvec_32.S

···
 	lw      r20, r20
 
 	/* Jump to syscall handler. */
-	jalr    r20; .Lhandle_syscall_link:
-	FEEDBACK_REENTER(handle_syscall)
+	jalr    r20
+.Lhandle_syscall_link: /* value of "lr" after "jalr r20" above */
 
 	/*
 	 * Write our r0 onto the stack so it gets restored instead
···
 	 */
 	PTREGS_PTR(r29, PTREGS_OFFSET_REG(0))
 	sw      r29, r0
+
+.Lsyscall_sigreturn_skip:
+	FEEDBACK_REENTER(handle_syscall)
 
 	/* Do syscall trace again, if requested. */
 	lw	r30, r31
···
 	};                                              \
 	STD_ENDPROC(_##x)
 
+/*
+ * Special-case sigreturn to not write r0 to the stack on return.
+ * This is technically more efficient, but it also avoids difficulties
+ * in the 64-bit OS when handling 32-bit compat code, since we must not
+ * sign-extend r0 for the sigreturn return-value case.
+ */
+#define PTREGS_SYSCALL_SIGRETURN(x, reg)                \
+	STD_ENTRY(_##x);                                \
+	addli   lr, lr, .Lsyscall_sigreturn_skip - .Lhandle_syscall_link; \
+	{                                               \
+	 PTREGS_PTR(reg, PTREGS_OFFSET_BASE);           \
+	 j      x                                       \
+	};                                              \
+	STD_ENDPROC(_##x)
+
 PTREGS_SYSCALL(sys_execve, r3)
 PTREGS_SYSCALL(sys_sigaltstack, r2)
-PTREGS_SYSCALL(sys_rt_sigreturn, r0)
+PTREGS_SYSCALL_SIGRETURN(sys_rt_sigreturn, r0)
 PTREGS_SYSCALL(sys_cmpxchg_badaddr, r1)
 
 /* Save additional callee-saves to pt_regs, put address in r4 and jump. */

arch/tile/kernel/process.c

···
 	childregs->sp = sp;  /* override with new user stack pointer */
 
 	/*
+	 * If CLONE_SETTLS is set, set "tp" in the new task to "r4",
+	 * which is passed in as arg #5 to sys_clone().
+	 */
+	if (clone_flags & CLONE_SETTLS)
+		childregs->tp = regs->regs[4];
+
+	/*
 	 * Copy the callee-saved registers from the passed pt_regs struct
 	 * into the context-switch callee-saved registers area.
 	 * This way when we start the interrupt-return sequence, the
···
 	return __switch_to(prev, next, next_current_ksp0(next));
 }
 
+/* Note there is an implicit fifth argument if (clone_flags & CLONE_SETTLS). */
 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
 		void __user *, parent_tidptr, void __user *, child_tidptr,
 		struct pt_regs *, regs)

arch/tile/kernel/signal.c

···
  */
 
 int restore_sigcontext(struct pt_regs *regs,
-		       struct sigcontext __user *sc, long *pr0)
+		       struct sigcontext __user *sc)
 {
 	int err = 0;
 	int i;
···
 
 	regs->faultnum = INT_SWINT_1_SIGRETURN;
 
-	err |= __get_user(*pr0, &sc->gregs[0]);
 	return err;
 }
 
-/* sigreturn() returns long since it restores r0 in the interrupted code. */
+/* The assembly shim for this function arranges to ignore the return value. */
 SYSCALL_DEFINE1(rt_sigreturn, struct pt_regs *, regs)
 {
 	struct rt_sigframe __user *frame =
 		(struct rt_sigframe __user *)(regs->sp);
 	sigset_t set;
-	long r0;
 
 	if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
 		goto badframe;
···
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
 
-	if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &r0))
+	if (restore_sigcontext(regs, &frame->uc.uc_mcontext))
 		goto badframe;
 
 	if (do_sigaltstack(&frame->uc.uc_stack, NULL, regs->sp) == -EFAULT)
 		goto badframe;
 
-	return r0;
+	return 0;
 
 badframe:
 	force_sig(SIGSEGV, current);