NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part1/4425

Internet Message Format | 1996-08-05 | 3.8 KB

Path: keats.ugrad.cs.ubc.ca!not-for-mail
From: [e-mail] (Kazimir Kylheku)
Newsgroups: comp.lang.c
Subject: Re: Allocation of memory
Date: 4 Feb 1996 07:38:45 -0800
Organization: Computer Science, University of B.C., Vancouver, B.C., Canada
Message-ID: <[e-mail]>
References: <4evre9$[e-mail]>
NNTP-Posting-Host: keats.ugrad.cs.ubc.ca
Keywords: realloc

In article <4evre9$[e-mail]>,
Klaus Berndl <[e-mail]> wrote:
>today i have a not-RPC-problem, but perhaps you can solve it:
>
>Please look at the foolwing short piece of code:

Let's not and say we did!

>Now my questions:
>
>1) How is it possible, that my machine (Sparc classic) allocates 100MB
>   Storage? I`m sure i have not so much main memory installed. Does it
>   use swap-file??

It's called "virtual memory". Here is one hypothesis on how it is done: Your
malloc() calls the brk() system call, which tells the operating system that you
have 100MB more worth of pages added to your data segment. These are marked
"not-present". It is when you access these these pages that the OS will
actually try to assign to them a piece of virtual memory. It's sort of an
"allocate on write".

So you won't see any thrashing until you start using the memory.

>2) With ALLOCSIZE set to 100000000 the program works correct. But if i set
>   ALLOCSIZE to 300000000, i reach the following output:
>
>     Klaus
>
>      t-realloc failed
>
>     test1
>     Bus Error (core dumped)

It could be that the OS will refuse to honor a brk() call that would result in
more pages being allocated than the free pages in the virtual memory pool.

>   What is 'realloc' doing exactly? Normally if realloc can`t allocate the
>   requested memory it returns NULL and doesn`t allocate memory. Why the
>   second realloc-call (x = ...) fails? There must be 1000000 Byte memory
>   because of the first try with ALLOCSIZE 100000000 works!

Realloc uses the same routines as malloc() to allocate heap storage. When it
can't allocate, it will use the brk() system call to move the "break" address,
which marks the end of the addresses that exist in the process.

Check how much virtual space your system has. If it's not enough for the
kind of allocation you are doing in your project, you need to add more swap
space and/or buy more RAM.


>3) We are comming to the point: I must succesive realloc storage for a
>   certain variable v. And if realloc fails, i want have the recent correct
>   memory allocation for v containing the right value (look at 'temp' in the
>   code). I need a save solution!

Then write your own realloc() routine or macro which allocates new storage, and
if successful, copies the data into it and gives you the new pointer, and then
frees the original copy. Give it different semantics, something like:

    int my_realloc(void **old2new, size_t size);

If my_realloc() fails, it can be set up so that it will _not_ change the value
of the void *old2new and returns some error code, otherwise it will update the
value of the pointer.


I wish I could tell you that unsuccessful realloc() does no harm to the
original allocated block, but I can't say that this is guaranteed on every
system.

The manual page on the HP-UX system does make such a guarantee:

 RETURN VALUE
      Upon successful completion, malloc(), realloc(), and calloc() return a
      pointer to space suitably aligned (after possible pointer coercion)
      for storage of any type of object.  Otherwise, they return a NULL
      pointer.  If realloc() returns a NULL pointer, the memory pointed to
      by the original pointer is left intact.

It probably works the same way on other UNIX systems. So as long as you
remember what the old pointer was, you should be able to keep using it if
realloc() failed.
--