linux-kernel - Re: [RFC v2 05/14] readahead: align index to mapping_min_order in ondemand_ra and force

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Wed, 14 Feb 2024 16:10:51 +0100
From: "Pankaj Raghav (Samsung)" <kernel@...kajraghav.com>
To: Dave Chinner <david@...morbit.com>
Cc: linux-xfs@...r.kernel.org, linux-fsdevel@...r.kernel.org, 
	mcgrof@...nel.org, gost.dev@...sung.com, akpm@...ux-foundation.org, 
	kbusch@...nel.org, djwong@...nel.org, chandan.babu@...cle.com, p.raghav@...sung.com, 
	linux-kernel@...r.kernel.org, hare@...e.de, willy@...radead.org, linux-mm@...ck.org
Subject: Re: [RFC v2 05/14] readahead: align index to mapping_min_order in
 ondemand_ra and force_ra

> > @@ -324,6 +325,13 @@ void force_page_cache_ra(struct readahead_control *ractl,
> >  	 * be up to the optimal hardware IO size
> >  	 */
> >  	index = readahead_index(ractl);
> > +	if (!IS_ALIGNED(index, min_nrpages)) {
> > +		unsigned long old_index = index;
> > +
> > +		index = round_down(index, min_nrpages);
> > +		nr_to_read += (old_index - index);
> > +	}
> 
> 	new_index = mapping_align_start_index(mapping, index);
> 	if (new_index != index) {
> 		nr_to_read += index - new_index;
> 		index = new_index
Looks good.

> 	}
> 
> > +
> >  	max_pages = max_t(unsigned long, bdi->io_pages, ra->ra_pages);
> >  	nr_to_read = min_t(unsigned long, nr_to_read, max_pages);
> 
> This needs to have a size of at least the minimum folio order size
> so readahead can fill entire folios, not get neutered to the maximum
> IO size the underlying storage supports.

So something like:

> >  	max_pages = max_t(unsigned long, bdi->io_pages, ra->ra_pages);
> >  	nr_to_read = min_t(unsigned long, nr_to_read, max_pages);
nr_to_read = max(nr_to_read, min_order);

> 
> > + * For higher order address space requirements we ensure no initial reads
> > + * are ever less than the min number of pages required.
> > + *
> > + * We *always* cap the max io size allowed by the device.
> >   */
> > -static unsigned long get_init_ra_size(unsigned long size, unsigned long max)
> > +static unsigned long get_init_ra_size(unsigned long size,
> > +				      unsigned int min_nrpages,
> > +				      unsigned long max)
> >  {
> >  	unsigned long newsize = roundup_pow_of_two(size);
> >  
> > +	newsize = max_t(unsigned long, newsize, min_nrpages);
> 
> This really doesn't need to care about min_nrpages. That rounding
> can be done in the caller when the new size is returned.

Sounds good.

> 
> >  	if (newsize <= max / 32)
> >  		newsize = newsize * 4;
> 
> >  
> >  
> > @@ -561,7 +583,11 @@ static void ondemand_readahead(struct readahead_control *ractl,
> >  	unsigned long add_pages;
> >  	pgoff_t index = readahead_index(ractl);
> >  	pgoff_t expected, prev_index;
> > -	unsigned int order = folio ? folio_order(folio) : 0;
> > +	unsigned int min_order = mapping_min_folio_order(ractl->mapping);
> > +	unsigned int min_nrpages = mapping_min_folio_nrpages(ractl->mapping);
> > +	unsigned int order = folio ? folio_order(folio) : min_order;
> 
> Huh? If we have a folio, then the order is whatever that folio is,
> otherwise we use min_order. What if the folio is larger than
> min_order? Doesn't that mean that this:
> 
> > @@ -583,8 +609,8 @@ static void ondemand_readahead(struct readahead_control *ractl,
> >  	expected = round_down(ra->start + ra->size - ra->async_size,
> >  			1UL << order);
> >  	if (index == expected || index == (ra->start + ra->size)) {
> > -		ra->start += ra->size;
> > -		ra->size = get_next_ra_size(ra, max_pages);
> > +		ra->start += round_down(ra->size, min_nrpages);
> > +		ra->size = get_next_ra_size(ra, min_nrpages, max_pages);
> 
> may set up the incorrect readahead range because the folio order is
> larger than min_nrpages?

Hmm... So I think we should just increment ra->start by ra->size, and
make sure to round the new size we get from get_next_ra_size() to
min_nrpages. Then we will not disturb the readahead range and always
increase the range in multiples of min_nrpages:

ra->start += ra->size;
ra->size = round_up(get_next_ra_size(ra, max_pages), min_nrpages);

> 
> >  		ra->async_size = ra->size;
> >  		goto readit;
> >  	}
> > @@ -603,13 +629,18 @@ static void ondemand_readahead(struct readahead_control *ractl,
> >  				max_pages);
> >  		rcu_read_unlock();
> >  
> > +		start = round_down(start, min_nrpages);
> 
> 		start = mapping_align_start_index(mapping, start);
> > +
> > +		VM_BUG_ON(folio->index & (folio_nr_pages(folio) - 1));
> > +
> >  		if (!start || start - index > max_pages)
> >  			return;
> >  
> >  		ra->start = start;
> >  		ra->size = start - index;	/* old async_size */
> > +
> >  		ra->size += req_size;
> > -		ra->size = get_next_ra_size(ra, max_pages);
> > +		ra->size = get_next_ra_size(ra, min_nrpages, max_pages);
> 
> 		ra->size = max(min_nrpages, get_next_ra_size(ra, max_pages));

If this is a round_up of size instead of max operation, we can
always ensure the ra->start from index aligned to min_nrpages. See my
reasoning in the previous comment.

--
Pankaj