Re: [PATCH 1/3] mm/mincore: make mincore() more conservative
From: Jiri Kosina
Date: Thu Mar 07 2019 - 00:46:09 EST
On Thu, 7 Mar 2019, Dominique Martinet wrote:
> > >
> > > addr = 4095
> > > vma->vm_end = 4096
> > > pages = 1000
> > >
> > > then `end' is 4096 and `(end - addr) << PAGE_SHIFT' is zero, but it
> > > should have been 1.
> > Good catch! It should rather be something like
> > unsigned long pages = (end >> PAGE_SHIFT) - (addr >> PAGE_SHIFT);
> That would be 0 for addr = 0 and vma->vm_end = 1; I assume we would
> still want to count that as one page.
Yeah, that was bogus as well, ETOOTIRED yesterday, sorry for the noise.
Both the variants are off.
> I'm not too familiar with this area of the code, but I think there's a
> handy macro we can use for this, perhaps
> DIV_ROUND_UP(end - addr, PAGE_SIZE) ?
> kernel/kexec_core.c has defined PAGE_COUNT() which seems more
> appropriate but I do not see a global equivalent
> #define PAGE_COUNT(x) (((x) + PAGE_SIZE - 1) >> PAGE_SHIFT)
I'll fix that up when doing the other changes requested by Andrew.