Re: Smashing the Stack?

From: Dan Kaminsky (dan@doxpara.com)
Date: 07/18/02 

Date: Thu, 18 Jul 2002 10:29:13 -0700
From: Dan Kaminsky <dan@doxpara.com>
To: fila <fila@detack.de>

>
>
>the real question is how did the gcc developers come to this solution ?
>I mean ... it looks pretty stupid. for a 3 chars buffer the compiler
>alocates 24 bytes and for a 4 chars buff only 4 bytes ?!?!?!
>the same problem you encounter if you go further with your exploration:
>for a 7 chars buff -> 24 bytes
>for a 8 chars buff -> 8 bytes :)))))))))))
>
>isn't it dumb ? it's been allready a pretty long time since I keep
>looking for somebody to explain why does the gcc alocate space in this
>manner.
>
>
>
4 chars = 4 bytes = 32 bits
8 chars = 8 bytes = 64 bits

Unsurprisingly, it looks like GCC is trying to be more efficient when
operating on variables or structures that align across 2^n (hardware
boundries / programmer fetish). My guess, completely pulled out of my
ass, is that non-2^n stuff is thrown on the stack inline with all
metadata, while 2^n stuff has all the metadata piled on in the beginning
with pointers to actual data. So to read a bunch of vars, you just do a
mass copy, instead of repeatedly skipping around.

I don't do much with stacksmashing, but that's what I assume is going on.

--Dan

Relevant Pages

  • Re: Using dlls
    ... >:> I am probably doing something very stupid indeed, ... >:> gently tell me what it is I'd be very grateful! ... that IMPLIB would do this, ... I had to take an educated guess what the utility was called in GCC, ...
    (comp.os.os2.programmer.misc)
  • Re: Vexing GCC warnings: "assignment discards qualifiers from pointer target type"
    ... snip ... ... > GCC probably realizes that and tries to prevent you from doing ... > something stupid by warning about returning a string literal as ...
    (comp.lang.c)
  • Re: __TIME_MIN/__TIME_MAX
    ... On Fri, 14 Nov 2003, Jacques Vidrine wrote: ... >> I prefer the cast. ... > stupid) warnings emitted by GCC. ... If time_t is actually signed then we may get a GCC ...
    (freebsd-arch)
  • Re: [git patches 1/2] warnings: attack valid cases spotted by warnings
    ... (rofl, look at that mess: it was utterly impractical, unrealistic and ... stupid for gcc to go and UTFify the compiler output. ... I absolutely *detest* how gcc does idiotic quoting just because you ...
    (Linux-Kernel)
  • Re: Learning C with Older books ?.
    ... you're also at odds with the gcc developers themselves. ... warnings not included in -Wall are the controversial ones: ... unless you happen to like that particular programming style. ...
    (comp.lang.c)