x86/vsyscall: Document odd SIGSEGV error code for vsyscalls

Even if vsyscall=none, user page faults on the vsyscall page are reported
as though the PROT bit in the error code was set.  Add a comment explaining
why this is probably okay and display the value in the test case.

While at it, explain why the behavior is correct with respect to PKRU.

Modify also the selftest to print the odd error code so that there is a
way to demonstrate the odd behaviour.

If anyone really cares about more accurate emulation, the behaviour could
be changed. But that needs a real good justification.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Florian Weimer <fweimer@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Kernel Hardening <kernel-hardening@lists.openwall.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/r/75c91855fd850649ace162eec5495a1354221aaa.1561610354.git.luto@kernel.org

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 288a5462076f03584c8de9b5d3728ade88796712..58e4f1f00bbc10d601ac08fa1a7982768eb24618 100644 (file)
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -710,6 +710,10 @@ static void set_signal_archinfo(unsigned long address,
         * To avoid leaking information about the kernel page
         * table layout, pretend that user-mode accesses to
         * kernel addresses are always protection faults.
+        *
+        * NB: This means that failed vsyscalls with vsyscall=none
+        * will have the PROT bit.  This doesn't leak any
+        * information and does not appear to cause any problems.
         */
        if (address >= TASK_SIZE_MAX)
                error_code |= X86_PF_PROT;
@@ -1375,6 +1379,9 @@ void do_user_addr_fault(struct pt_regs *regs,
         *
         * The vsyscall page does not have a "real" VMA, so do this
         * emulation before we go searching for VMAs.
+        *
+        * PKRU never rejects instruction fetches, so we don't need
+        * to consider the PF_PK bit.
         */
        if (is_vsyscall_vaddr(address)) {
                if (emulate_vsyscall(hw_error_code, regs, address))

diff --git a/tools/testing/selftests/x86/test_vsyscall.c b/tools/testing/selftests/x86/test_vsyscall.c
index 0b4f1cc2291c660df97a984065b4cdc3384e2c98..4c9a8d76dba0b9d18a82b42d410552e990954e52 100644 (file)
--- a/tools/testing/selftests/x86/test_vsyscall.c
+++ b/tools/testing/selftests/x86/test_vsyscall.c
@@ -183,9 +183,13 @@ static inline long sys_getcpu(unsigned * cpu, unsigned * node,
 }
 
 static jmp_buf jmpbuf;
+static volatile unsigned long segv_err;
 
 static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
 {
+       ucontext_t *ctx = (ucontext_t *)ctx_void;
+
+       segv_err =  ctx->uc_mcontext.gregs[REG_ERR];
        siglongjmp(jmpbuf, 1);
 }
 
@@ -416,8 +420,11 @@ static int test_vsys_r(void)
        } else if (!can_read && should_read_vsyscall) {
                printf("[FAIL]\tWe don't have read access, but we should\n");
                return 1;
+       } else if (can_read) {
+               printf("[OK]\tWe have read access\n");
        } else {
-               printf("[OK]\tgot expected result\n");
+               printf("[OK]\tWe do not have read access: #PF(0x%lx)\n",
+                      segv_err);
        }
 #endif