[PATCH v3] arm64: fault: Don't leak data in ESR context for user fault on kernel VA

[peter.maydell@linaro.org (Peter Maydell)]

If userspace faults on a kernel address, handing them the raw ESR
value on the sigframe as part of the delivered signal can leak data
useful to attackers who are using information about the underlying hardware
fault type (e.g. translation vs permission) as a mechanism to defeat KASLR.

However there are also legitimate uses for the information provided
in the ESR -- notably the GCC and LLVM sanitizers use this to report
whether wild pointer accesses by the application are reads or writes
(since a wild write is a more serious bug than a wild read), so we
don't want to drop the ESR information entirely.

For faulting addresses in the kernel, sanitize the ESR. We choose
to present userspace with the illusion that there is nothing mapped
in the kernel's part of the address space at all, by reporting all
faults as level 0 translation faults taken to EL1.

These fields are safe to pass through to userspace as they depend
only on the instruction that userspace used to provoke the fault:
 EC IL (always)
 ISV CM WNR (for all data aborts)
All the other fields in ESR except DFSC are architecturally RES0
for an L0 translation fault taken to EL1, so can be zeroed out
without confusing userspace.

The illusion is not entirely perfect, as there is a tiny wrinkle
where we will report an alignment fault that was not due to the memory
type (for instance a LDREX to an unaligned address) as a translation
fault, whereas if you do this on real unmapped memory the alignment
fault takes precedence. This is not likely to trip anybody up in
practice, as the only users we know of for the ESR information who
care about the behaviour for kernel addresses only really want to
know about the WnR bit.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
This patch is an alternative proposal to Will's patch
https://patchwork.kernel.org/patch/10258781/
which simply removed the ESR record entirely for kernel addresses.

Changes v1->v2:
 * rebased on master
 * commit message tweak
 * DABT_CUR and IABT_CUR moved to "can't happen" default case
 * explicitly clear the bits which are RES0 if ISV == 0
 * comment text tweaks
Changes v2->v3:
 * remove the support for reporting ESRs with ISV == 1 (this
   can't happen, and we probably don't want to tell userspace
   that the exception was taken to EL2 if in some hypothetical
   future that becomes possible)
 * rebased on 4.17-rc6
---
 arch/arm64/mm/fault.c | 51 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 51 insertions(+)

diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
index 4165485e8b6e..2af3dd89bcdb 100644
--- a/arch/arm64/mm/fault.c
+++ b/arch/arm64/mm/fault.c
@@ -293,6 +293,57 @@ static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 static void __do_user_fault(struct siginfo *info, unsigned int esr)
 {
 	current->thread.fault_address = (unsigned long)info->si_addr;
+
+	/*
+	 * If the faulting address is in the kernel, we must sanitize the ESR.
+	 * From userspace's point of view, kernel-only mappings don't exist
+	 * at all, so we report them as level 0 translation faults.
+	 * (This is not quite the way that "no mapping there at all" behaves:
+	 * an alignment fault not caused by the memory type would take
+	 * precedence over translation fault for a real access to empty
+	 * space. Unfortunately we can't easily distinguish "alignment fault
+	 * not caused by memory type" from "alignment fault caused by memory
+	 * type", so we ignore this wrinkle and just return the translation
+	 * fault.)
+	 */
+	if (current->thread.fault_address >= TASK_SIZE) {
+		switch (ESR_ELx_EC(esr)) {
+		case ESR_ELx_EC_DABT_LOW:
+			/*
+			 * These bits provide only information about the
+			 * faulting instruction, which userspace knows already.
+			 * We explicitly clear bits which are architecturally
+			 * RES0 in case they are given meanings in future.
+			 * We always report the ESR as if the fault was taken
+			 * to EL1 and so ISV and the bits in ISS[23:14] are
+			 * clear. (In fact it always will be a fault to EL1.)
+			 */
+			esr &= ESR_ELx_EC_MASK | ESR_ELx_IL |
+				ESR_ELx_CM | ESR_ELx_WNR;
+			esr |= ESR_ELx_FSC_FAULT;
+			break;
+		case ESR_ELx_EC_IABT_LOW:
+			/*
+			 * Claim a level 0 translation fault.
+			 * All other bits are architecturally RES0 for faults
+			 * reported with that DFSC value, so we clear them.
+			 */
+			esr &= ESR_ELx_EC_MASK | ESR_ELx_IL;
+			esr |= ESR_ELx_FSC_FAULT;
+			break;
+		default:
+			/*
+			 * This should never happen (entry.S only brings us
+			 * into this code for insn and data aborts from a lower
+			 * exception level). Fail safe by not providing an ESR
+			 * context record at all.
+			 */
+			WARN(1, "ESR 0x%x is not DABT or IABT from EL0\n", esr);
+			esr = 0;
+			break;
+		}
+	}
+
 	current->thread.fault_code = esr;
 	arm64_force_sig_info(info, esr_to_fault_info(esr)->name, current);
 }
-- 
2.17.0