x86/entry/64: Use TASK_SIZE_MAX for canonical RIP test

Using shifts to determine if an address is canonical is difficult for
the compiler to optimize when the virtual address width is variable
(LA57 feature) without using inline assembly.  Instead, compare RIP
against TASK_SIZE_MAX.  The only user executable address outside of that
range is the deprecated vsyscall page, which can fall back to using IRET.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Uros Bizjak <ubizjak@gmail.com>
Link: https://lore.kernel.org/r/20231011224351.130935-3-brgerst@gmail.com

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index 90214652ff06381e868c0aa18e88018fc632bce3..4c7154d097c11eb42f80d8cef48c3bdfa1452779 100644 (file)
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -110,10 +110,10 @@ __visible noinstr bool do_syscall_64(struct pt_regs *regs, int nr)
         * in kernel space.  This essentially lets the user take over
         * the kernel, since userspace controls RSP.
         *
-        * Change top bits to match the most significant bit (47th or 56th bit
-        * depending on paging mode) in the address.
+        * TASK_SIZE_MAX covers all user-accessible addresses other than
+        * the deprecated vsyscall page.
         */
-       if (unlikely(!__is_canonical_address(regs->ip, __VIRTUAL_MASK_SHIFT + 1)))
+       if (unlikely(regs->ip >= TASK_SIZE_MAX))
                return false;
 
        /*