arm64: mm: allow the kernel to handle alignment faults on user accesses

Although we don't expect to take alignment faults on access to normal
memory, misbehaving (i.e. buggy) user code can pass MMIO pointers into
system calls, leading to things like get_user accessing device memory.

Rather than OOPS the kernel, allow any exception fixups to run and
return something like -EFAULT back to userspace. This makes the
behaviour more consistent with userspace, even though applications with
access to device mappings can easily cause other issues if they try
hard enough.

Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Eun Taik Lee <eun.taik.lee@samsung.com>
[will: dropped __kprobes annotation and rewrote commit mesage]
Signed-off-by: Will Deacon <will.deacon@arm.com>

arch/arm64/mm/fault.c

@@ -371,6 +371,13 @@ static int __kprobes do_translation_fault(unsigned long addr,
 	return 0;
 }
 
+static int do_alignment_fault(unsigned long addr, unsigned int esr,
+			      struct pt_regs *regs)
+{
+	do_bad_area(addr, esr, regs);
+	return 0;
+}
+
 /*
  * This abort handler always returns "fault".
  */
@@ -418,7 +425,7 @@ static struct fault_info {
 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
 	{ do_bad,		SIGBUS,  0,		"unknown 32"			},
-	{ do_bad,		SIGBUS,  BUS_ADRALN,	"alignment fault"		},
+	{ do_alignment_fault,	SIGBUS,  BUS_ADRALN,	"alignment fault"		},
 	{ do_bad,		SIGBUS,  0,		"unknown 34"			},
 	{ do_bad,		SIGBUS,  0,		"unknown 35"			},
 	{ do_bad,		SIGBUS,  0,		"unknown 36"			},