Re: [KJ] [patch] potential data-corruption fix in md.c

[Jaco Kroon <jaco@kroon.co.za>]

[-- Attachment #1.1: Type: text/plain, Size: 1994 bytes --]

Håkon Løvdal wrote:
> On 12/2/05, Jaco Kroon <jaco@kroon.co.za> wrote:
> 
>>Or just cast to (unsigned int) for the check?
>>Perhaps that happens implicitly?
>>Would that be compiler dependant?
> 
> 
> The conversion does indeed happen implicit, and the C standard
> (ISO/IEC 9899:1990) does very specific specify how this should be done,
> see "6.2.1.5 Usual arithmetic conversions" for details.
> 
> The short version is that a smallest type is promoted to the biggest type
> as well that unsigned "wins" over signed. The last point can result in
> some unexpected results if you are not careful.
> Example: What does the following program print?
> 
> #include <stdio.h>
> int main(int argc, char *argv[])
> {
>         unsigned int i = 12345;
>         if (i < -1)
>                 printf("(%d < -1)\n", i);
>         else
>                 printf("! (%d < -1)\n", i);
>         return 0;
> }

jkroon@pug tmp $ gcc -o signed_unsigned signed_unsigned.c
jkroon@pug tmp $ gcc -o signed_unsigned signed_unsigned.c -Wall -W
signed_unsigned.c: In function `main':
signed_unsigned.c:5: warning: comparison between signed and unsigned
signed_unsigned.c: At top level:
signed_unsigned.c:2: warning: unused parameter 'argc'
signed_unsigned.c:2: warning: unused parameter 'argv'
jkroon@pug tmp $ ./signed_unsigned
(12345 < -1)
jkroon@pug tmp $

Which means that gcc is smart enough to recognise that you are doing
something unusual, and 12345 is indeed < -1 due to the fact that -1 is
in fact 2 ^ 32 - 1.  As such the test in register_md_personality with
pnum >= MAX_PERSONALITY is in fact sufficient since MAX_PERSONALITY is
unsigned, so pnum gets converted to unsigned (negative values becomes
extremely large) and as such the <0 is done "by accident".

Very, very nice, but extremely unclear from the code :).

Jaco
-- 
There are only 10 kinds of people in this world,
  those that understand binary and those that don't.
http://www.kroon.co.za/

[-- Attachment #1.2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3166 bytes --]

[-- Attachment #2: Type: text/plain, Size: 168 bytes --]

_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/kernel-janitors