linux-kernel - Re: [PATCH v2 05/12] mm: Add and use find_lock

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:   Tue, 29 Sep 2020 13:48:06 +0100
From:   Matthew Wilcox <willy@...radead.org>
To:     Jan Kara <jack@...e.cz>
Cc:     linux-mm@...ck.org, Andrew Morton <akpm@...ux-foundation.org>,
        Hugh Dickins <hughd@...gle.com>,
        William Kucharski <william.kucharski@...cle.com>,
        Johannes Weiner <hannes@...xchg.org>,
        Yang Shi <yang.shi@...ux.alibaba.com>,
        Dave Chinner <dchinner@...hat.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 05/12] mm: Add and use find_lock_entries

On Tue, Sep 29, 2020 at 10:58:55AM +0200, Jan Kara wrote:
> On Mon 14-09-20 14:00:35, Matthew Wilcox (Oracle) wrote:
> > We have three functions (shmem_undo_range(), truncate_inode_pages_range()
> > and invalidate_mapping_pages()) which want exactly this function, so
> > add it to filemap.c.
> > 
> > Signed-off-by: Matthew Wilcox (Oracle) <willy@...radead.org>
> > diff --git a/mm/shmem.c b/mm/shmem.c
> ...
> > index b65263d9bb67..a73ce8ce28e3 100644
> > --- a/mm/shmem.c
> > +++ b/mm/shmem.c
> > @@ -905,12 +905,8 @@ static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
> >  
> >  	pagevec_init(&pvec);
> >  	index = start;
> > -	while (index < end) {
> > -		pvec.nr = find_get_entries(mapping, index,
> > -			min(end - index, (pgoff_t)PAGEVEC_SIZE),
> > -			pvec.pages, indices);
> > -		if (!pvec.nr)
> > -			break;
> > +	while (index < end && find_lock_entries(mapping, index, end - 1,
> > +			&pvec, indices)) {
> >  		for (i = 0; i < pagevec_count(&pvec); i++) {
> >  			struct page *page = pvec.pages[i];
> >  
> > @@ -925,18 +921,10 @@ static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
> >  								index, page);
> >  				continue;
> >  			}
> > +			index += thp_nr_pages(page) - 1;
> >  
> > -			VM_BUG_ON_PAGE(page_to_pgoff(page) != index, page);
> > -
> > -			if (!trylock_page(page))
> > -				continue;
> > -
> > -			if ((!unfalloc || !PageUptodate(page)) &&
> > -			    page_mapping(page) == mapping) {
> > -				VM_BUG_ON_PAGE(PageWriteback(page), page);
> > -				if (shmem_punch_compound(page, start, end))
> > -					truncate_inode_page(mapping, page);
> > -			}
> > +			if (!unfalloc || !PageUptodate(page))
> > +				truncate_inode_page(mapping, page);
> 
> Is dropping shmem_punch_compound() really safe? AFAICS it can also call
> split_huge_page() which will try to split THP to be able to truncate it.
> That being said there's another loop in shmem_undo_range() which will try
> again so what you did might make a difference with performance but not much
> else. But still it would be good to at least comment about this in the
> changelog...

OK, I need to provide better argumentation in the changelog.

shmem_punch_compound() handles partial THPs.  By the end of this series,
we handle the partial pages in the next part of the function ... the
part where we're handling partial PAGE_SIZE pages.  At this point in
the series, it's safe to remove the shmem_punch_compound() call because
the new find_lock_entries() loop will only return THPs that lie entirely
within the range.