Re: [PATCH] x86/boot: add prototype for __fortify_panic()

From: Jeff Johnson
Date: Fri May 31 2024 - 14:29:27 EST


On 5/31/2024 9:28 AM, Kees Cook wrote:
> On Thu, May 30, 2024 at 09:23:36AM -0700, Jeff Johnson wrote:
>> On 5/30/2024 8:42 AM, Nikolay Borisov wrote:
>>>
>>>
>>> On 29.05.24 г. 21:09 ч., Jeff Johnson wrote:
>>>> As discussed in [1] add a prototype for __fortify_panic() to fix the
>>>> 'make W=1 C=1' warning:
>>>>
>>>> arch/x86/boot/compressed/misc.c:535:6: warning: symbol '__fortify_panic' was not declared. Should it be static?
>>>
>>> Actually doesn't it make sense to have this defined under ../string.h ?
>>> Actually given that we don't have any string fortification under the
>>> boot/ why have the fortify _* functions at all ?
>>
>> I'll let Kees answer these questions since I just took guidance from him :)
>
> Ah-ha, I see what's happening. When not built with
> CONFIG_FORTIFY_SOURCE, fortify-string.h isn't included. But since misc.c
> has the function definition, we get a warning that the function
> declaration was never seen. This is likely the better solution:
>
>
> diff --git a/arch/x86/boot/compressed/misc.c b/arch/x86/boot/compressed/misc.c
> index b70e4a21c15f..3f21a5e218f8 100644
> --- a/arch/x86/boot/compressed/misc.c
> +++ b/arch/x86/boot/compressed/misc.c
> @@ -532,7 +532,9 @@ asmlinkage __visible void *extract_kernel(void *rmode, unsigned char *output)
> return output + entry_offset;
> }
>
> +#ifdef CONFIG_FORTIFY_SOURCE
> void __fortify_panic(const u8 reason, size_t avail, size_t size)
> {
> error("detected buffer overflow");
> }
> +#endif
>
>
> Jeff, can you test this? (I still haven't been able to reproduce the
> warning.)

Adding Dan since this comes during:
CHECK arch/x86/boot/compressed/misc.c

What version of smatch are you using? I'm using v0.5.0-8639-gff1cc4d453ff

In the build where I'm seeing this issue I have:
CONFIG_ARCH_HAS_FORTIFY_SOURCE=y
CONFIG_FORTIFY_SOURCE=y
CONFIG_FORTIFY_KUNIT_TEST=m

So that conditional compilation won't make a difference.

Also note that misc.c doesn't include the standard include/linux/string.h but
instead includes the stripped down arch/x86/boot/string.h, so fortify-string.h
isn't included.

This seems to come back around to the question that Nikolay asked, which part
of the boot code actually needs this?

/jeff