[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <569f55225581aec50f817818b126986647ba925e.camel@kernel.crashing.org>
Date: Mon, 03 Sep 2018 13:44:53 +1000
From: Benjamin Herrenschmidt <benh@...nel.crashing.org>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Jiri Kosina <jikos@...nel.org>,
Jürgen Groß <jgross@...e.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Michal Hocko <mhocko@...e.com>,
Naoya Horiguchi <n-horiguchi@...jp.nec.com>,
Michael Ellerman <mpe@...erman.id.au>,
Will Deacon <will.deacon@....com>
Subject: Re: Access to non-RAM pages
On Sun, 2018-09-02 at 19:52 -0700, Linus Torvalds wrote:
> On Sun, Sep 2, 2018 at 7:47 PM Linus Torvalds
> <torvalds@...ux-foundation.org> wrote:
> >
> > The comment actually does talk about it, although the comment also
> > claims that the cs read would use load_unaligned_zeropad(), which it
> > no longer does (now it only does the read_word_at_a_time).
>
> IOW, look at commit 12f8ad4b0533 ("vfs: clean up __d_lookup_rcu() and
> dentry_cmp() interfaces") for why the zeropad went away for the cs
> access (but the comment wasn't updated).
>
> And then bfe7aa6c39b1 ("fs/dcache: Use read_word_at_a_time() in
> dentry_string_cmp()") did the "let's make KASAN happy thing.
>
> And yes, the word-at-a-time code actually matters a lot for certain
> loads. The "copy-and-hash" thing for path components ends up being
> pretty critical in all the pathname handling.
Yup, makes sense.
Thanks !
Cheers,
Ben.
Powered by blists - more mailing lists