Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <05e14540-7092-5dd2-d503-473b673af716@collabora.com>
Date:   Fri, 7 Apr 2023 16:10:51 +0500
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.9.0
Cc:     Muhammad Usama Anjum <usama.anjum@collabora.com>,
        Mike Rapoport <rppt@kernel.org>, Peter Xu <peterx@redhat.com>,
        David Hildenbrand <david@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Andrei Vagin <avagin@gmail.com>,
        Danylo Mocherniuk <mdanylo@google.com>,
        Paul Gofman <pgofman@codeweavers.com>,
        Cyrill Gorcunov <gorcunov@gmail.com>,
        Nadav Amit <namit@vmware.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Shuah Khan <shuah@kernel.org>,
        Christian Brauner <brauner@kernel.org>,
        Yang Shi <shy828301@gmail.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        "Liam R . Howlett" <Liam.Howlett@oracle.com>,
        Yun Zhou <yun.zhou@windriver.com>,
        Suren Baghdasaryan <surenb@google.com>,
        Alex Sierra <alex.sierra@amd.com>,
        Matthew Wilcox <willy@infradead.org>,
        Pasha Tatashin <pasha.tatashin@soleen.com>,
        Axel Rasmussen <axelrasmussen@google.com>,
        "Gustavo A . R . Silva" <gustavoars@kernel.org>,
        Dan Williams <dan.j.williams@intel.com>,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-mm@kvack.org, linux-kselftest@vger.kernel.org,
        Greg KH <gregkh@linuxfoundation.org>, kernel@collabora.com
Subject: Re: [PATCH v12 2/5] fs/proc/task_mmu: Implement IOCTL to get and
 optionally clear info about PTEs
Content-Language: en-US
To:     =?UTF-8?B?TWljaGHFgiBNaXJvc8WCYXc=?= <emmir@google.com>
References: <20230406074005.1784728-1-usama.anjum@collabora.com>
 <20230406074005.1784728-3-usama.anjum@collabora.com>
 <CABb0KFHZpYVML2e+Xg9+kwjyhqQkikPBhymO=EXoQnO2xjfG4g@mail.gmail.com>
 <0351b563-5193-6431-aa9c-c5bf5741b791@collabora.com>
 <CABb0KFE4ruptVXDpCk5MB6nkh9WeKTcKfROnx0ecoy-k1eCKCw@mail.gmail.com>
 <8a837998-604f-a871-729e-aa274a621481@collabora.com>
 <CABb0KFEBqAMWWpAeBfqzA4JrHo3yLyaT0rqKTUn28O0hE+szBA@mail.gmail.com>
 <c5b9201d-141c-10ae-0475-4b230d36508b@collabora.com>
 <CABb0KFH3mj5qt22qDLHRKjh-wB7Jrn6Pz8h-QARaf9oR65U0Qg@mail.gmail.com>
From:   Muhammad Usama Anjum <usama.anjum@collabora.com>
In-Reply-To: <CABb0KFH3mj5qt22qDLHRKjh-wB7Jrn6Pz8h-QARaf9oR65U0Qg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 4/7/23 3:14 PM, Michał Mirosław wrote:
> On Fri, 7 Apr 2023 at 12:04, Muhammad Usama Anjum
> <usama.anjum@collabora.com> wrote:
>> On 4/7/23 12:34 PM, Michał Mirosław wrote:
>>> On Thu, 6 Apr 2023 at 23:04, Muhammad Usama Anjum
>>> <usama.anjum@collabora.com> wrote:
>>>> On 4/7/23 1:00 AM, Michał Mirosław wrote:
>>>>> On Thu, 6 Apr 2023 at 19:58, Muhammad Usama Anjum
>>>>> <usama.anjum@collabora.com> wrote:
> [...]
>>>>>>>> +       /*
>>>>>>>> +        * Allocate smaller buffer to get output from inside the page walk
>>>>>>>> +        * functions and walk page range in PAGEMAP_WALK_SIZE size chunks. As
>>>>>>>> +        * we want to return output to user in compact form where no two
>>>>>>>> +        * consecutive regions should be continuous and have the same flags.
>>>>>>>> +        * So store the latest element in p.cur between different walks and
>>>>>>>> +        * store the p.cur at the end of the walk to the user buffer.
>>>>>>>> +        */
>>>>>>>> +       p.vec = kmalloc_array(p.vec_len, sizeof(struct page_region),
>>>>>>>> +                             GFP_KERNEL);
>>>>>>>> +       if (!p.vec)
>>>>>>>> +               return -ENOMEM;
>>>>>>>> +
>>>>>>>> +       walk_start = walk_end = start;
>>>>>>>> +       while (walk_end < end && !ret) {
>>>>>>>
>>>>>>> The loop will stop if a previous iteration returned ENOSPC (and the
>>>>>>> error will be lost) - is it intended?
>>>>>> It is intentional. -ENOSPC means that the user buffer is full even though
>>>>>> there was more memory to walk over. We don't treat this error. So when
>>>>>> buffer gets full, we stop walking over further as user buffer has gotten
>>>>>> full and return as success.
>>>>>
>>>>> Thanks. What's the difference between -ENOSPC and
>>>>> PM_SCAN_FOUND_MAX_PAGES? They seem to result in the same effect (code
>>>>> flow).
>>>> -ENOSPC --> user buffer has been filled completely
>>>> PM_SCAN_FOUND_MAX_PAGES --> max_pages have been found, user buffer may
>>>>                             still have more space
>>>
>>> What is the difference in code behaviour when those two cases are
>>> compared? (I'd expect none.)
>> There is difference:
>> We add data to user buffer. If it succeeds with return code 0, we engage
>> the WP. If it succeeds with PM_SCAN_FOUND_MAX_PAGES, we still engage the
>> WP. But if we get -ENOSPC, we don't perform engage as the data wasn't added
>> to the user buffer.
> 
> Thanks! I see it now. I see a few more corner cases here:
> 1. If we did engage WP but fail to copy the vector we return -EFAULT
> but the WP is already engaged. I'm not sure this is something worth
> guarding against, but documenting that would be helpful I think.
Sure.

> 2. If uffd_wp_range() fails, but we have already processed pages
> earlier, we should treat the error like ENOSPC and back out the failed
> range (the earier changes would be lost otherwise).
Backing out is easier to do for hugepages. But for normal pages, we'll have
to write some code to find where the current data was added (in cur or in
vec) and back out from that. I'll have to write some more code to avoid the
side-effects as well.

But aren't we going over-engineering here? Error occurred and we are trying
to keep the previously generated correct data and returning successfully
still to the user? I don't think we should do this. An error is error. We
should return the error simply even if the memory flags would get lost. We
don't know what caused the error in uffd_wp_range(). Under normal
situation, we there shouldn't have had error.


> 
> Best Regards
> Michał Mirosław

-- 
BR,
Muhammad Usama Anjum