Re: [PATCH] x86: Clean up stack access code in irq_32.c

[Jeff Epler]

It looks like the proposed variant still miscompiles in clang 3.4 and 3.5, the
two versions I had handy to test.

I extracted your code to a simple standalone C translation unit and
inspected various compilers' results via objdump.

// cut here for cso.c
struct thread_info { long l[32]; }; // who knows

#define STACK_WARN (1024)
#define PAGE_SIZE (4096)

#define THREAD_SIZE_ORDER      2
#define THREAD_SIZE            (PAGE_SIZE << THREAD_SIZE_ORDER)
#define THREAD_SIZE_MASK       (THREAD_SIZE - 1)
#define CURRENT_MASK           (~THREAD_SIZE_MASK)

/* how to get the current stack pointer from C */
#define current_stack_pointer ({               \
       register unsigned long sp asm("esp");   \
       sp;                                     \
})

int check_stack_overflow0(void)
{
       long sp;

       __asm__ __volatile__("andl %%esp,%0" :
                            "=r" (sp) : "0" (THREAD_SIZE - 1));

       return sp < (sizeof(struct thread_info) + STACK_WARN);
}

int check_stack_overflow1(void)
{
       return (current_stack_pointer & THREAD_SIZE_MASK)
              < sizeof(struct thread_info) + STACK_WARN;
}
// end cso.c

Typical compiler invocation:
        clang-3.5 -m32 -Os -c cso.c

Both clang-3.4 and clang-3.5 as packaged for debian jessie seem to get
check_stack_overflow1 wrong, yielding a function which always returns true:

    00000000 <check_stack_overflow1>:
       0:   b8 01 00 00 00          mov    $0x1,%eax
       5:   c3                      ret    

Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/