linux-kernel - Re: [PATCH net 1/4] virtchnl: Fix layout of RSS structures

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <a47965c9-c7d6-7349-b717-4eb3bde193c7@intel.com>
Date:   Sun, 28 Mar 2021 20:19:19 -0700
From:   "Samudrala, Sridhar" <sridhar.samudrala@...el.com>
To:     Geert Uytterhoeven <geert@...ux-m68k.org>
Cc:     anthony.l.nguyen@...el.com,
        Norbert Ciosek <norbertx.ciosek@...el.com>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        netdev <netdev@...r.kernel.org>, sassmann@...hat.com,
        Konrad Jankowski <konrad0.jankowski@...el.com>,
        Mateusz Palczewski <mateusz.palczewski@...el.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH net 1/4] virtchnl: Fix layout of RSS structures

On 3/27/2021 2:53 AM, Geert Uytterhoeven wrote:
> Hi Samudrala,
>
> On Fri, Mar 26, 2021 at 11:45 PM Samudrala, Sridhar
> <sridhar.samudrala@...el.com> wrote:
>> On 3/26/2021 1:06 AM, Geert Uytterhoeven wrote:
>>> On Thu, Mar 25, 2021 at 11:29 PM Tony Nguyen <anthony.l.nguyen@...el.com> wrote:
>>> From: Norbert Ciosek <norbertx.ciosek@...el.com>
>>>
>>> Remove padding from RSS structures. Previous layout
>>> could lead to unwanted compiler optimizations
>>> in loops when iterating over key and lut arrays.
>>>
>>>  From an earlier private conversation with Mateusz, I understand the real
>>> explanation is that key[] and lut[] must be at the end of the
>>> structures, because they are used as flexible array members?
>>>
>>> Fixes: 65ece6de0114 ("virtchnl: Add missing explicit padding to structures")
>>> Signed-off-by: Norbert Ciosek <norbertx.ciosek@...el.com>
>>> Tested-by: Konrad Jankowski <konrad0.jankowski@...el.com>
>>> Signed-off-by: Tony Nguyen <anthony.l.nguyen@...el.com>
>>>
>>> --- a/include/linux/avf/virtchnl.h
>>> +++ b/include/linux/avf/virtchnl.h
>>> @@ -476,7 +476,6 @@ struct virtchnl_rss_key {
>>>          u16 vsi_id;
>>>          u16 key_len;
>>>          u8 key[1];         /* RSS hash key, packed bytes */
>>> -       u8 pad[1];
>>>   };
>>>
>>>   VIRTCHNL_CHECK_STRUCT_LEN(6, virtchnl_rss_key);
>>> @@ -485,7 +484,6 @@ struct virtchnl_rss_lut {
>>>          u16 vsi_id;
>>>          u16 lut_entries;
>>>          u8 lut[1];        /* RSS lookup table */
>>> -       u8 pad[1];
>>>   };
>>>
>>> If you use a flexible array member, it should be declared without a size,
>>> i.e.
>>>
>>>      u8 key[];
>>>
>>> Everything else is (trying to) fool the compiler, and leading to undefined
>>> behavior, and people (re)adding explicit padding.
>> This header file is shared across other OSes that use C++ that doesn't support
>> flexible arrays. So the structures in this file use an array of size 1 as a last
>> element to enable variable sized arrays.
> I don't think it is accepted practice to have non-Linux-isms in
> include/*linux*/avf/virtchnl.h header files.  Moreover, using a size
> of 1 is counter-intuitive for people used to Linux kernel development,
> and may lead to off-by-one errors in calculation of sizes.
>
> If you insist on ignoring the above, this definitely deserves a
> comment next to the member's declaration.
Sure. We can add a comment indicating that these fields are used 
variable sized arrays.

Thanks
Sridhar