Re: [RFC PATCH] mm: shm: round up tmpfs size to huge page size when huge=always

[Kirill A. Shutemov]

On Mon, Oct 09, 2017 at 03:51:06AM +0800, Yang Shi wrote:
> 
> 
> On 10/8/17 5:56 AM, Kirill A. Shutemov wrote:
> > On Sat, Oct 07, 2017 at 04:22:10AM +0800, Yang Shi wrote:
> > > When passing "huge=always" option for mounting tmpfs, THP is supposed to
> > > be allocated all the time when it can fit, but when the available space is
> > > smaller than the size of THP (2MB on x86), shmem fault handler still tries
> > > to allocate huge page every time, then fallback to regular 4K page
> > > allocation, i.e.:
> > > 
> > > 	# mount -t tmpfs -o huge,size=3000k tmpfs /tmp
> > > 	# dd if=/dev/zero of=/tmp/test bs=1k count=2048
> > > 	# dd if=/dev/zero of=/tmp/test1 bs=1k count=2048
> > > 
> > > The last dd command will handle 952 times page fault handler, then exit
> > > with -ENOSPC.
> > > 
> > > Rounding up tmpfs size to THP size in order to use THP with "always"
> > > more efficiently. And, it will not wast too much memory (just allocate
> > > 511 extra pages in worst case).
> > 
> > Hm. I don't think it's good idea to silently increase size of fs.
> 
> How about printing a warning to say the filesystem is resized?
> 
> > 
> > Maybe better just refuse to mount with huge=always for too small fs?
> 
> It sounds fine too. When mounting tmpfs with "huge=always", if the size is
> not THP size aligned, it just can refuse to mount, then show warning about
> alignment restriction.

Honestly, I wouldn't bother.

Using filesystem at near-full capacity is not best practice. Performance
penalty for doing is fair enough.

And forcing alignment doesn't really fixes anything: user still allowed
to fill the filesystem with files less than 2M in size.

-- 
 Kirill A. Shutemov