[PATCH] MIPS: Modernize READ_IMPLIES_EXEC

[Kees Cook]

I'm doing some thread necromancy of
https://lore.kernel.org/lkml/202007081624.82FA0CC1EA@keescook/

x86, arm64, and arm32 adjusted their READ_IMPLIES_EXEC logic to better
align with the safer defaults and the interactions with other mappings,
which I illustrated with this comment on x86:

/*
 * An executable for which elf_read_implies_exec() returns TRUE will
 * have the READ_IMPLIES_EXEC personality flag set automatically.
 *
 * The decision process for determining the results are:
 *
 *                 CPU: | lacks NX*  | has NX, ia32     | has NX, x86_64 |
 * ELF:                 |            |                  |                |
 * ---------------------|------------|------------------|----------------|
 * missing PT_GNU_STACK | exec-all   | exec-all         | exec-none      |
 * PT_GNU_STACK == RWX  | exec-stack | exec-stack       | exec-stack     |
 * PT_GNU_STACK == RW   | exec-none  | exec-none        | exec-none      |
 *
 *  exec-all  : all PROT_READ user mappings are executable, except when
 *              backed by files on a noexec-filesystem.
 *  exec-none : only PROT_EXEC user mappings are executable.
 *  exec-stack: only the stack and PROT_EXEC user mappings are
 *  executable.
 *
 *  *this column has no architectural effect: NX markings are ignored by
 *   hardware, but may have behavioral effects when "wants X" collides with
 *   "cannot be X" constraints in memory permission flags, as in
 *   https://lkml.kernel.org/r/20190418055759.GA3155@xxxxxxxxxxxx
 *
 */

For MIPS, the "lacks NX" above is the "!cpu_has_rixi" check. On x86,
we decided that the READ_IMPLIES_EXEC flag needed to reflect the
expectations, not the architectural behavior due to bad interactions
as noted above, as always returning "1" on non-NX hardware breaks
some mappings.

The other part of the issue is "what does the MIPS toolchain do for
PT_GNU_STACK?" The answer seems to be "by default, include PT_GNU_STACK,
but mark it executable" (likely due to concerns over non-NX hardware):

$ mipsel-linux-gnu-gcc -o hello_world hello_world.c
$ llvm-readelf -lW hellow_world | grep GNU_STACK
  GNU_STACK      0x000000 0x00000000 0x00000000 0x00000 0x00000 RWE 0x10

Given that older hardware doesn't support non-executable memory, it
seems safe to make the "PT_GNU_STACK is absent" logic mean "assume
non-executable", but this might break very old software running on
modern MIPS. This situation matches the ia32-on-x86_64 logic x86
uses (which assumes needing READ_IMPLIES_EXEC in that situation). But
modern toolchains on modern MIPS hardware should follow a safer default
(assume NX stack).

A follow-up to this change would be to switch the MIPS toolchain to emit
a non-executable PT_GNU_STACK, as this seems to be unneeded.

Cc: Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx>
Cc: Tiezhu Yang <yangtiezhu@xxxxxxxxxxx>
Cc: Xuefeng Li <lixuefeng@xxxxxxxxxxx>
Cc: Juxin Gao <gaojuxin@xxxxxxxxxxx>
Cc: linux-mips@xxxxxxxxxxxxxxx
Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
---
 arch/mips/kernel/elf.c | 16 +++++-----------
 1 file changed, 5 insertions(+), 11 deletions(-)

diff --git a/arch/mips/kernel/elf.c b/arch/mips/kernel/elf.c
index 7b045d2a0b51..5582a4ca1e9e 100644
--- a/arch/mips/kernel/elf.c
+++ b/arch/mips/kernel/elf.c
@@ -328,16 +328,10 @@ void mips_set_personality_nan(struct arch_elf_state *state)
 
 int mips_elf_read_implies_exec(void *elf_ex, int exstack)
 {
-	if (exstack != EXSTACK_DISABLE_X) {
-		/* The binary doesn't request a non-executable stack */
-		return 1;
-	}
-
-	if (!cpu_has_rixi) {
-		/* The CPU doesn't support non-executable memory */
-		return 1;
-	}
-
-	return 0;
+	/*
+	 * Set READ_IMPLIES_EXEC only on non-NX systems that
+	 * do not request a specific state via PT_GNU_STACK.
+	 */
+	return (!cpu_has_rixi && exstack == EXSTACK_DEFAULT);
 }
 EXPORT_SYMBOL(mips_elf_read_implies_exec);
-- 
2.30.2