netdev - Re: [PATCHv3 net 1/2] fib: convert fib_nh_is_v6 and nh

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: <e49bcbc4-b7ae-66b1-4964-957983688426@kernel.org>
Date: Thu, 21 Sep 2023 19:51:01 -0600
From: David Ahern <dsahern@...nel.org>
To: Hangbin Liu <liuhangbin@...il.com>
Cc: netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
 Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
 Ido Schimmel <idosch@...dia.com>, Benjamin Poirier <bpoirier@...dia.com>,
 Thomas Haller <thaller@...hat.com>,
 Stephen Hemminger <stephen@...workplumber.org>,
 Eric Dumazet <edumazet@...gle.com>,
 Nicolas Dichtel <nicolas.dichtel@...nd.com>, Ido Schimmel <idosch@...sch.org>
Subject: Re: [PATCHv3 net 1/2] fib: convert fib_nh_is_v6 and nh_updated to use
 a single bit

On 9/21/23 7:29 PM, Hangbin Liu wrote:
> On Thu, Sep 21, 2023 at 07:03:20AM -0600, David Ahern wrote:
>> On 9/20/23 9:14 PM, Hangbin Liu wrote:
>>> The FIB info structure currently looks like this:
>>> struct fib_info {
>>>         struct hlist_node          fib_hash;             /*     0    16 */
>>>         [...]
>>>         u32                        fib_priority;         /*    80     4 */
>>>
>>>         /* XXX 4 bytes hole, try to pack */
>>>
>>>         struct dst_metrics *       fib_metrics;          /*    88     8 */
>>>         int                        fib_nhs;              /*    96     4 */
>>>         bool                       fib_nh_is_v6;         /*   100     1 */
>>>         bool                       nh_updated;           /*   101     1 */
>>>
>>>         /* XXX 2 bytes hole, try to pack */
>>
>> 2B hole here and you want to add a single flag so another bool. I would
>> prefer the delay to a bitfield until all holes are consumed.
>>
> 
> OK, just in case I didn't misunderstand. I should add a `bool pfsrc_removed`
> here and drop the first patch, right?

yes.

The bitfield has higher overhead. Let's reserve that overhead to when
there are no more holes and a new field pushes the struct over 128B.