Non continuous VM address space

Non continuous VM address space

[Phy Prabab]

Hello,

I have a question, I have a little application and sbrks and then
mmaps to get memory from the system.  Originally, this program worked
fine because I had assumed that my applications address space would be
continuous, each subsequent memory request would "mapped" next to each
other, after all this is the address space for this process.  However,
what I am finding is that the space is discontiguous, I am getting
back addresses not next to each other.  Sometimes the next memory
allocation is several megs above where the previous mapping came from.
 Is this possible?  Or am I not accounting for something?  Yes, I have
intentionally defined malloc, calloc, etc. so they can not be used
under the hood within my code.

BTW, seeing this on 2.6.9 and 2.4.(?) - these are both RH supplied kernels

TIA!
Phy

Re: Non continuous VM address space

[Jeremy Fitzhardinge]

Phy Prabab wrote:
> Sometimes the next memory
> allocation is several megs above where the previous mapping came from.
> Is this possible?

The kernel is free to place mmaped memory whereever it likes unless you
use MAP_FIXED and specify an address.

    J