x86/entry/64/compat: Preserve r8-r11 in int $0x80

32-bit user code that uses int $80 doesn't care about r8-r11.  There is,
however, some 64-bit user code that intentionally uses int $0x80 to invoke
32-bit system calls.  From what I've seen, basically all such code assumes
that r8-r15 are all preserved, but the kernel clobbers r8-r11.  Since I
doubt that there's any code that depends on int $0x80 zeroing r8-r11,
change the kernel to preserve them.

I suspect that very little user code is broken by the old clobber, since
r8-r11 are only rarely allocated by gcc, and they're clobbered by function
calls, so they only way we'd see a problem is if the same function that
invokes int $0x80 also spills something important to one of these
registers.

The current behavior seems to date back to the historical commit
"[PATCH] x86-64 merge for 2.6.4".  Before that, all regs were
preserved.  I can't find any explanation of why this change was made.

Update the test_syscall_vdso_32 testcase as well to verify the new
behavior, and it strengthens the test to make sure that the kernel doesn't
accidentally permute r8..r15.

Suggested-by: Denys Vlasenko <dvlasenk@redhat.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Link: https://lkml.kernel.org/r/d4c4d9985fbe64f8c9e19291886453914b48caee.1523975710.git.luto@kernel.org

1 files changed, 21 insertions, 14 deletions

diff --git a/tools/testing/selftests/x86/test_syscall_vdso.c b/tools/testing/selftests/x86/test_syscall_vdso.c
index 40370354d4c1..c9c3281077bc 100644
--- a/tools/testing/selftests/x86/test_syscall_vdso.c
+++ b/tools/testing/selftests/x86/test_syscall_vdso.c
@@ -100,12 +100,19 @@ asm (
         "       shl     $32, %r8\n"
         "       orq     $0x7f7f7f7f, %r8\n"
         "       movq    %r8, %r9\n"
-        "       movq    %r8, %r10\n"
-        "       movq    %r8, %r11\n"
-        "       movq    %r8, %r12\n"
-        "       movq    %r8, %r13\n"
-        "       movq    %r8, %r14\n"
-        "       movq    %r8, %r15\n"
+        "       incq    %r9\n"
+        "       movq    %r9, %r10\n"
+        "       incq    %r10\n"
+        "       movq    %r10, %r11\n"
+        "       incq    %r11\n"
+        "       movq    %r11, %r12\n"
+        "       incq    %r12\n"
+        "       movq    %r12, %r13\n"
+        "       incq    %r13\n"
+        "       movq    %r13, %r14\n"
+        "       incq    %r14\n"
+        "       movq    %r14, %r15\n"
+        "       incq    %r15\n"
         "       ret\n"
         "       .code32\n"
         "       .popsection\n"
@@ -128,12 +135,13 @@ int check_regs64(void)
         int err = 0;
         int num = 8;
         uint64_t *r64 = &regs64.r8;
+        uint64_t expected = 0x7f7f7f7f7f7f7f7fULL;
 
         if (!kernel_is_64bit)
                 return 0;
 
         do {
-                if (*r64 == 0x7f7f7f7f7f7f7f7fULL)
+                if (*r64 == expected++)
                         continue; /* register did not change */
                 if (syscall_addr != (long)&int80) {
                         /*
@@ -147,18 +155,17 @@ int check_regs64(void)
                                 continue;
                         }
                 } else {
-                        /* INT80 syscall entrypoint can be used by
+                        /*
+                         * INT80 syscall entrypoint can be used by
                          * 64-bit programs too, unlike SYSCALL/SYSENTER.
                          * Therefore it must preserve R12+
                          * (they are callee-saved registers in 64-bit C ABI).
                          *
-                         * This was probably historically not intended,
-                         * but R8..11 are clobbered (cleared to 0).
-                         * IOW: they are the only registers which aren't
-                         * preserved across INT80 syscall.
+                         * Starting in Linux 4.17 (and any kernel that
+                         * backports the change), R8..11 are preserved.
+                         * Historically (and probably unintentionally), they
+                         * were clobbered or zeroed.
                          */
-                        if (*r64 == 0 && num <= 11)
-                                continue;
                 }
                 printf("[FAIL]\tR%d has changed:%016llx\n", num, *r64);
                 err++;