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;
Subject: Re: [PATCH 1/2] fuse: Fix race in fuse_writepage_in_flight()
To:     Miklos Szeredi <miklos@szeredi.hu>
Cc:     linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
References: <154322517208.18737.3297786654135648324.stgit@localhost.localdomain>
 <00b1782a-5c5c-5bc8-7ea9-4f8450679fa1@virtuozzo.com>
 <CAJfpeguokUoCPEnK7Q_Rp9cmyBCMfm51=7TmbbS-QiEy2BxxAQ@mail.gmail.com>
From:   Kirill Tkhai <ktkhai@virtuozzo.com>
Message-ID: <a45d54a3-b275-9c95-3c9d-a19566e447dc@virtuozzo.com>
Date:   Thu, 10 Jan 2019 14:03:17 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.3.1
MIME-Version: 1.0
In-Reply-To: <CAJfpeguokUoCPEnK7Q_Rp9cmyBCMfm51=7TmbbS-QiEy2BxxAQ@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 10.01.2019 14:00, Miklos Szeredi wrote:
> On Thu, Jan 10, 2019 at 11:48 AM Kirill Tkhai <ktkhai@virtuozzo.com> wrote:
>>
>> Hi, Miklos,
>>
>> any comments about this?
> 
> Is there a reproducer?  ISTR that fsx-linux with mmaps enabled was
> good for stressing the writeback_cache code.

There is no a reproducer, since I found that by eyes during preparation of another patchset.

>>
>> On 26.11.2018 12:46, Kirill Tkhai wrote:
>>> Checking for FR_PENDING in fuse_writepage_in_flight() is racy.
>>> It does not guarantee the first request in misc.write.next list
>>> is not in userspace, since there we take fc->lock, while
>>> fuse_dev_do_read() takes fiq->waitq.lock:
>>>
>>> fuse_dev_read()                           fuse_writepage_in_flight()
>>>                                             test_bit(FR_PENDING)
>>>   clear_bit(FR_PENDING)
>>> handle old_req->pages[0] in userspace
>>>                                             copy_highpage(old_req->pages[0], page)
>>>                                             ^^^^^
>>>                                             userspace never sees this pages
>>>
>>> The only reliable way to determ, whether we are able to replace
>>> old_req's page, is to completely skip the first request in the list.
>>> This patch makes the function to do that.
>>>
>>> Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
>>> ---
>>>  fs/fuse/file.c |    4 +++-
>>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
>>> index b52f9baaa3e7..c6650c68b31a 100644
>>> --- a/fs/fuse/file.c
>>> +++ b/fs/fuse/file.c
>>> @@ -1740,6 +1740,7 @@ static bool fuse_writepage_in_flight(struct fuse_req *new_req,
>>>  {
>>>       struct fuse_conn *fc = get_fuse_conn(new_req->inode);
>>>       struct fuse_inode *fi = get_fuse_inode(new_req->inode);
>>> +     struct fuse_req *first_req;
>>>       struct fuse_req *tmp;
>>>       struct fuse_req *old_req;
>>>       bool found = false;
>>> @@ -1764,6 +1765,7 @@ static bool fuse_writepage_in_flight(struct fuse_req *new_req,
>>>       }
>>>
>>>       new_req->num_pages = 1;
>>> +     first_req = old_req;
>>>       for (tmp = old_req; tmp != NULL; tmp = tmp->misc.write.next) {
>>>               BUG_ON(tmp->inode != new_req->inode);
>>>               curr_index = tmp->misc.write.in.offset >> PAGE_SHIFT;
>>> @@ -1773,7 +1775,7 @@ static bool fuse_writepage_in_flight(struct fuse_req *new_req,
>>>               }
>>>       }
>>>
>>> -     if (old_req->num_pages == 1 && test_bit(FR_PENDING, &old_req->flags)) {
>>> +     if (old_req->num_pages == 1 && old_req != first_req) {
>>>               struct backing_dev_info *bdi = inode_to_bdi(page->mapping->host);
>>>
>>>               copy_highpage(old_req->pages[0], page);
>>>