linux-kernel - Re: [PATCH 1/2] iomap: support tail packing inline read

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]

Message-ID: <YPMkKfegS+9KzEhK@casper.infradead.org>
Date:   Sat, 17 Jul 2021 19:40:41 +0100
From:   Matthew Wilcox <willy@...radead.org>
To:     Andreas Grünbacher 
        <andreas.gruenbacher@...il.com>,
        Christoph Hellwig <hch@...radead.org>,
        linux-erofs@...ts.ozlabs.org,
        Linux FS-devel Mailing List <linux-fsdevel@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        "Darrick J. Wong" <djwong@...nel.org>, Chao Yu <chao@...nel.org>,
        Liu Bo <bo.liu@...ux.alibaba.com>,
        Joseph Qi <joseph.qi@...ux.alibaba.com>,
        Liu Jiang <gerry@...ux.alibaba.com>
Subject: Re: [PATCH 1/2] iomap: support tail packing inline read

On Sat, Jul 17, 2021 at 11:15:58PM +0800, Gao Xiang wrote:
> Hi Matthew,
> 
> On Sat, Jul 17, 2021 at 04:01:38PM +0100, Matthew Wilcox wrote:
> > On Sat, Jul 17, 2021 at 09:38:18PM +0800, Gao Xiang wrote:
> > > Sorry about some late. I've revised a version based on Christoph's
> > > version and Matthew's thought above. I've preliminary checked with
> > > EROFS, if it does make sense, please kindly help check on the gfs2
> > > side as well..
> > 
> > I don't understand how this bit works:
> 
> This part inherited from the Christoph version without change.
> The following thoughts are just my own understanding...
> 
> > 
> > >  	struct page *page = ctx->cur_page;
> > > -	struct iomap_page *iop;
> > > +	struct iomap_page *iop = NULL;
> > >  	bool same_page = false, is_contig = false;
> > >  	loff_t orig_pos = pos;
> > >  	unsigned poff, plen;
> > >  	sector_t sector;
> > >  
> > > -	if (iomap->type == IOMAP_INLINE) {
> > > -		WARN_ON_ONCE(pos);
> > > -		iomap_read_inline_data(inode, page, iomap);
> > > -		return PAGE_SIZE;
> > > -	}
> > > +	if (iomap->type == IOMAP_INLINE && !pos)
> > > +		WARN_ON_ONCE(to_iomap_page(page) != NULL);
> > > +	else
> > > +		iop = iomap_page_create(inode, page);
> > 
> > Imagine you have a file with bytes 0-2047 in an extent which is !INLINE
> > and bytes 2048-2051 in the INLINE extent.  When you read the page, first
> > you create an iop for the !INLINE extent.  Then this function is called
> 
> Yes, it first created an iop for the !INLINE extent.
> 
> > again for the INLINE extent and you'll hit the WARN_ON_ONCE.  No?
> 
> If it is called again with another INLINE extent, pos will be non-0?
> so (!pos) == false. Am I missing something?

Well, either sense of a WARN_ON is wrong.

For a file which is PAGE_SIZE + 3 bytes in size, to_iomap_page() will
be NULL.  For a file which is PAGE_SIZE/2 + 3 bytes in size,
to_iomap_page() will not be NULL.  (assuming the block size is <=
PAGE_SIZE / 2).

I think we need a prep patch that looks something like this:

+++ b/fs/iomap/buffered-io.c
@@ -252,8 +252,12 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
                return PAGE_SIZE;
        }

+       if (offset_in_page(pos) || length < PAGE_SIZE)
+               iop = iomap_page_create(inode, page);
+       else
+               iop = NULL;
+
        /* zero post-eof blocks as the page may be mapped */
-       iop = iomap_page_create(inode, page);
        iomap_adjust_read_range(inode, iop, &pos, length, &poff, &plen);
        if (plen == 0)
                goto done;


ie first get the conditions right under which we should create an iop.