Hi all. I'm writing some code to make a custom Ubuntu 8.04 Live CD and the code will write the character "x" to all free physical memory. In a shell script, I turn off swapping (swapoff), turn off overcommit (echo 2 > /proc/sys/vm/overcommit_memory), drop_cache, check MemFree in /proc/meminfo and pass the value to a simple C program that malloc's a buffer with the size of that value. But I'm unable to allocate that amount because I think my CommitLimit shown in /proc/meminfo is 700mb out of the 1.4gb MemFree (1.5gb MemTotal). In the C code, the malloc returns a null pointer when I pass it anything greater than the value of CommitLimit, which is consistent with what I think is causing the null pointer.

Is there a way to increase the CommitLimit so I can allocate every bit of the free space? Again the script and C code is being run from a LiveCD environment, so there is no physical hard disk.

In short: no. The design of linux memory allocator is to always be optimistic, and therefore you cannot prevent the OOM killer from running, or from exceeding the commit limit in an allocation. I could be wrong on this, but that is my assumption on how the linux allocation scheme works.

You can turn off the optimistic allocation as described (echo 2 > /proc/sys/vm/overcommit_memory). This will indirectly prevent the OOM killer from running, in theory, by not preventing memory allocation calls from getting more than they're supposed to. However I'm wondering how the kernel arrives at the CommitLimit value. It seems odd that it would only be about half the amount of MemFree.