linux-kernel - Re: [PATCH v3] mm, netfs, fscache: Stop read optimisation when folio removed from pagecache

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] [day] [month] [year] [list]

Date:   Wed, 23 Nov 2022 12:25:56 -0800
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     David Howells <dhowells@...hat.com>
Cc:     willy@...radead.org, dwysocha@...hat.com,
        Rohith Surabattula <rohiths.msft@...il.com>,
        Steve French <sfrench@...ba.org>,
        Shyam Prasad N <nspmangalore@...il.com>,
        Dominique Martinet <asmadeus@...ewreck.org>,
        Ilya Dryomov <idryomov@...il.com>, linux-cachefs@...hat.com,
        linux-cifs@...r.kernel.org, linux-afs@...ts.infradead.org,
        v9fs-developer@...ts.sourceforge.net, ceph-devel@...r.kernel.org,
        linux-nfs@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3] mm, netfs, fscache: Stop read optimisation when folio
 removed from pagecache

On Wed, Nov 23, 2022 at 12:03 PM David Howells <dhowells@...hat.com> wrote:
>
> Linus Torvalds <torvalds@...ux-foundation.org> wrote:
>
> > But I also think it's strange in another way, with that odd placement of
> >
> >         mapping_clear_release_always(inode->i_mapping);
> >
> > at inode eviction time. That just feels very random.
>
> I was under the impression that a warning got splashed if unexpected
> address_space flags were set when ->evict_inode() returned.  I may be thinking
> of page flags.  If it doesn't, fine, this isn't required.

I don't know if the warning happens or not, but the thing I reacted to
was just how *random* this was. There was no logic to it, nor any
explanation.

I *suspect* that if we add this kind of new generic address space
flag, then that flag should just be cleared by generic code when the
address space is released.

But I'm not saying it has to be done that way - I'm just saying that
however it is done, please don't make it this random mess with no
explanation.

The *setting* of the flag was at least fairly obvious. I didn't find
code like this odd:

+       if (v9inode->netfs.cache)
+               mapping_set_release_always(inode->i_mapping);

and it makes all kinds of sense (ie I can read it as a "if I use netfs
caching for this inode, then I want to be informed when a folio is
released from this mapping").

It's just the clearing that looked very random to me.

Maybe just a comment would have helped, but I get the feeling that it
migth as well just be cleared in "clear_inode()" or something like
that.

> > That code makes no sense what-so-ever. Why isn't it using
> > "folio_has_private()"?
>
> It should be, yes.
>
> > Why is this done as an open-coded - and *badly* so - version of
> > !folio_needs_release() that you for some reason made private to mm/vmscan.c?
>
> Yeah, in retrospect, I should have put that in mm/internal.h.

So if folio_needs_release() is in mm/internal.h, and then mm/filemap.c
uses it in filemap_release_folio() instead of the odd open-coding, I
think that would clear up my worries about both mm/filemap.c and
mm/vmscan.c.

                Linus