linux-kernel - Re: gcc fixed size char array initialization bug

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Fri, 03 Aug 2007 01:27:28 +0200
From:	Stefan Richter <stefanr@...6.in-berlin.de>
To:	Al Viro <viro@....linux.org.uk>
CC:	Guennadi Liakhovetski <g.liakhovetski@....de>,
	Andi Kleen <andi@...stfloor.org>, linux-kernel@...r.kernel.org
Subject: Re: gcc fixed size char array initialization bug - known?

Al Viro wrote:
> On Fri, Aug 03, 2007 at 12:51:16AM +0200, Guennadi Liakhovetski wrote:
>> On Fri, 3 Aug 2007, Stefan Richter wrote:
>>
>>> Guennadi Liakhovetski wrote:
>>>> with
>>>>
>>>> 	char c[4] = "012345";
>>>>
>>>> the compiler warns, but actually allocates a 6-byte long array...
>>> Off-topic here, but:  sizeof c / sizeof *c == 4.
>> Don't think it is OT here - kernel depends on gcc. And, what I meant, is, 
>> that gcc places all 7 (sorry, not 6 as I said above) characters in the 
>> .rodata section of the compiled object file. Of course, it doesn't mean, 
>> that c is 7 characters long.
> 
> So gcc does that kind of recovery, after having warned you.  Makes sense,
> as long as it's for ordinary variables (and not, say it, struct fields) -
> you get less likely runtime breakage on the undefined behaviour (e.g.
> passing c to string functions).  So gcc has generated some padding between
> the global variables, that's all.

No, the fact that the full 012345\0 ends up in the object file is
apparently unrelated to what happens to the variable c...

> It doesn't change the fact that use of c[4] or strlen(c) or strcpy(..., c)
> means nasal demon country for you.
> 
> Now, if gcc does that for similar situation with struct fields, you'd have
> a cause to complain.

...since only 0123 will get into c at runtime, i.e. a 4 bytes long array
without \0 appendix or other extraordinary padding.

#include <stdio.h>
#include <string.h>

int main()
{
	char c[4] = "012345";

	printf("%d %d _%s_\n", sizeof c / sizeof *c, strlen(c), c);
	return 0;
}

$ ./a.out
4 8 _01230®¿_

$ strings a.out |grep 0123
012345

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