MIME-Version: 1.0
In-Reply-To: <20150802002318.GZ17109@ZenIV.linux.org.uk>
References: <alpine.LSU.2.11.1507311022320.10162@eggly.anvils>
	<CA+55aFxGp3Hn-CeFE1NTYDr7vt3xjqe2mV1BbZ2tYHpCxenuAA@mail.gmail.com>
	<alpine.LSU.2.11.1507311207160.11122@eggly.anvils>
	<20150731205036.GA3752@nautica>
	<CA+55aFz9jj_wAjXNoFbKFbV3M1Ki2KhHuoMwQGPi0wiQjMXdsw@mail.gmail.com>
	<20150801072603.GV17109@ZenIV.linux.org.uk>
	<CA+55aFx7OY8X=s=oc5HUeWeA-5vWqkH+gcOV80xC=xXv9Q3AnQ@mail.gmail.com>
	<20150802001402.GY17109@ZenIV.linux.org.uk>
	<20150802002318.GZ17109@ZenIV.linux.org.uk>
Date: Sat, 1 Aug 2015 17:57:44 -0700
Message-ID: <CA+55aFymnDY+kYeNY9Xak9kFGbJr5AMVVPGA0RzENoTVSumLJw@mail.gmail.com>
Subject: Re: [git pull] vfs.git spurious ENOTDIR fix
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dominique Martinet <asmadeus@codewreck.org>,
        Hugh Dickins <hughd@google.com>,
        "J. Bruce Fields" <bfields@fieldses.org>,
        Dominique Martinet <dominique.martinet@cea.fr>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        David Howells <dhowells@redhat.com>
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1339
Lines: 38

On Sat, Aug 1, 2015 at 5:23 PM, Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> Branch head should be at 97242f9, just to make sure you get
> the right one...

Ok, merged in my tree.

However, looking at this, I'm struck by how all the callers of
"link_path_walk()" tend to have very similar patterns wrt error
handling.

And I'm wondering - wouldn't it be nicer to extend that pattern a bit
more, and make the *callers* of link_path_walk() all do

        if (error) {
                if (nd->flags & LOOKUP_RCU) {
                        if (unlazy_walk(nd, NULL, 0))
                                error = -ECHILD;
                }
        }

and maybe even make that part of terminate_walk() that everybody calls
after getting here.

Because it's not just that "!d_can_lookup()" case that triggers it,
you also have that pattern in the RCU error case for may_lookup(), and
get_link().

So why don't we make the rule that *every* single error we get during
an RCU walk should do that unlazy_walk() and turn the error into
ECHILD on failure. Hmm? We're almost there as-is.

             Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/