arm64: handle sys and undef traps consistently

If an EL0 instruction in the SYS class triggers an exception, do_sysintr
looks for a sys64_hook matching the instruction, and if none is found,
injects a SIGILL. This mirrors what we do for undefined instruction
encodings in do_undefinstr, where we look for an undef_hook matching the
instruction, and if none is found, inject a SIGILL.

Over time, new SYS instruction encodings may be allocated. Prior to
allocation, exceptions resulting from these would be handled by
do_undefinstr, whereas after allocation these may be handled by
do_sysintr.

To ensure that we have consistent behaviour if and when this happens, it
would be beneficial to have do_sysinstr fall back to do_undefinstr.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Suzuki Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

arch/arm64/kernel/traps.c

@@ -531,7 +531,12 @@ asmlinkage void __exception do_sysinstr(unsigned int esr, struct pt_regs *regs)
 			return;
 		}
 
-	force_signal_inject(SIGILL, ILL_ILLOPC, regs, 0);
+	/*
+	 * New SYS instructions may previously have been undefined at EL0. Fall
+	 * back to our usual undefined instruction handler so that we handle
+	 * these consistently.
+	 */
+	do_undefinstr(regs);
 }
 
 long compat_arm_syscall(struct pt_regs *regs);