Re: [PATCH] Introduce a method to catch mmap_region (was: Recent kernel "mount" slow)

[Al Viro <viro@ZenIV.linux.org.uk>]

On Wed, Nov 28, 2012 at 10:16:21PM -0800, Linus Torvalds wrote:
> On Wed, Nov 28, 2012 at 6:58 PM, Linus Torvalds
> <torvalds@linux-foundation.org> wrote:
> >
> > But the fact that the code wants to do things like
> >
> >         block = (sector_t)page->index << (PAGE_CACHE_SHIFT - bbits);
> >
> > seriously seems to be the main thing that keeps us using
> > 'inode->i_blkbits'. Calculating bbits from bh->b_size is just costly
> > enough to hurt (not everywhere, but on some machines).
> >
> > Very annoying.
> 
> Hmm. Here's a patch that does that anyway. I'm not 100% happy with the
> whole ilog2 thing, but at the same time, in other cases it actually
> seems to improve code generation (ie gets rid of the whole unnecessary
> two dereferences through page->mapping->host just to get the block
> size, when we have it in the buffer-head that we have to touch
> *anyway*).
> 
> Comments? Again, untested.
> 
> And I notice that Al Viro hasn't been cc'd, which is sad, since he's
> been involved in much of fs/block_dev.c.
> 
> Al - this is an independent patch to fs/buffer.c to make
> fs/block_dev.c able to change the block size of a block device while
> there is IO in progress that may still use the old block size. The
> discussion has been on fsdevel and lkml, but you may have missed it...

	Umm...  set_blocksize() is calling kill_bdev(), which does
truncate_inode_pages(mapping, 0).  What's going to happen to data in
the dirty pages?  IO in progress is not the only thing to worry about...