Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 6 Apr 2018 12:35:45 +0100
From:   Alan Cox <gnomes@lxorguk.ukuu.org.uk>
To:     "Michael S. Tsirkin" <mst@redhat.com>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        stable <stable@vger.kernel.org>,
        syzbot+6304bf97ef436580fede@syzkaller.appspotmail.com,
        linux-mm <linux-mm@kvack.org>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Huang Ying <ying.huang@intel.com>,
        Jonathan Corbet <corbet@lwn.net>,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Thorsten Leemhuis <regressions@leemhuis.info>
Subject: Re: [PATCH] gup: return -EFAULT on access_ok failure
Message-ID: <20180406123545.24953eb4@alans-desktop>
In-Reply-To: <20180405211945-mutt-send-email-mst@kernel.org>
References: <1522431382-4232-1-git-send-email-mst@redhat.com>
        <20180405045231-mutt-send-email-mst@kernel.org>
        <CA+55aFwpe92MzEX2qRHO-MQsa1CP-iz6AmanFqXCV6_EaNKyMg@mail.gmail.com>
        <20180405171009-mutt-send-email-mst@kernel.org>
        <CA+55aFz_mCZQPV6ownt+pYnLFf9O+LUK_J6y4t1GUyWL1NJ2Lg@mail.gmail.com>
        <20180405211945-mutt-send-email-mst@kernel.org>
Organization: Intel Corporation
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

> so an error on the 1st page gets propagated to the caller,
> and that get_user_pages_unlocked eventually calls __get_user_pages
> so it does return an error sometimes.
> 
> Would it be correct to apply the second part of the patch then
> (pasted below for reference) or should get_user_pages_fast
> and all its callers be changed to return 0 on error instead?

0 isn't an error. As SuS sees it (ie from the userspace end of the pile)

returning the number you asked for means it worked

returning a smaller number means it worked partially and that much was
consumed (or in some cases more and the rest if so was lost - depends
what you are reading/writing)

returning 0 means you read nothing as you were at the end of file

returning an error code means it broke, or you should try again
(EAGAIN/EWOULDBLOCK)

The ugly bit there is the try-again semantics needs to exactly match the
attached poll() behaviour or you get busy loops.

Alan