Re: [PATCH] FRV: Fix mmap2 error handling

[Matthew Wilcox <matthew@wil.cx>]

On Thu, Nov 27, 2008 at 03:12:38PM +0000, David Howells wrote:
> Fix the error handling in sys_mmap2().  Currently, if the pgoff check fails,
> fput() might have to be called (which it isn't), so do the pgoff check first,
> before fget() is called.

*sigh*.

My reaction was "Why do we have sys_mmap2 in every architecture?"  So I
started looking.  Oh dear, oh dear oh dear.

FRV:
        /* As with sparc32, make sure the shift for mmap2 is constant
           (12), no matter what PAGE_SIZE we have.... */
ia64:
Just uses PAGE_SIZE (currently supported values: 4k, 8k, 16k and 64k).

So what is poor userspace to do?  Check which architecture it's on and
figure out what PAGE_SIZE to use for mmap2 based on that?


How about we introduce a sys_mmap6() in common code which takes 'off'
in multiples of 4k.  Then FRV and other sane architectures can replace
their sys_mmap2 entries in their syscall tables with sys_mmap6.  ia64 has
to keep its insane sys_mmap2 entry, but it can add a sys_mmap6 entry too.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."