Re: inode->i_count security hole

Alan Cox (alan@lxorguk.ukuu.org.uk)
Tue, 13 Jan 1998 00:34:17 +0000 (GMT)


> PS I assume making inode->i_count a long works, because
>
> pages in address space * max processes << 2^32 (for ix86)

The actual maths is something like

2^32 bytes per process (actually less) 2^12 bytes/page = 2^20 pages per
process * a max of about 1024 processes. Its good to about 6000 processes.

Really we should also keep a limit on the vma's and pages mapped per process
(the latter btw is an rlimit)

Alan