netdev - Re: [PATCH net] net: allow small head cache usage with large MAX_SKB

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 for Android: free password hash cracker in your pocket

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

Message-ID: <12199ed2-ca9e-4658-9fc0-44e5b05ca7c3@redhat.com>
Date: Thu, 13 Feb 2025 18:12:58 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: Eric Dumazet <edumazet@...gle.com>
Cc: netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
 Jakub Kicinski <kuba@...nel.org>, Simon Horman <horms@...nel.org>,
 Sabrina Dubroca <sd@...asysnail.net>
Subject: Re: [PATCH net] net: allow small head cache usage with large
 MAX_SKB_FRAGS values

On 2/13/25 2:44 PM, Eric Dumazet wrote:
> On Wed, Feb 12, 2025 at 11:08 PM Paolo Abeni <pabeni@...hat.com> wrote:
>>
>> On 2/12/25 9:47 PM, Eric Dumazet wrote:
>>> This patch still gives a warning if  MAX_TCP_HEADER < GRO_MAX_HEAD +
>>> 64 (in my local build)
>>
>> Oops, I did not consider MAX_TCP_HEADER and GRO_MAX_HEAD could diverge.
>>
>>> Why not simply use SKB_WITH_OVERHEAD(SKB_SMALL_HEAD_CACHE_SIZE) , and
>>> remove the 1024 value ?
>>
>> With CONFIG_MAX_SKB_FRAGS=17, SKB_SMALL_HEAD_CACHE_SIZE is considerably
>> smaller than 1024, I feared decreasing such limit could re-introduce a
>> variation of the issue addressed by commit 3226b158e67c ("net: avoid 32
>> x truesize under-estimation for tiny skbs").
>>
>> Do you feel it would be safe?
> 
> As long as we are using kmalloc() for those, we are good I think.

Due to ENOCOFFEE it took me a while to understand you mean that we just
need to ensure GRO_MAX_HEAD and GOOD_COPY_LEN allocations are backed by
kmalloc to avoid the mentioned issue.

I guess eventual nic drivers shooting themselves in the foot
consistently doing napi_alloc_skb(<max small cache + 1>), if any should
be fixed in their own code.

I concur using SKB_WITH_OVERHEAD(SKB_SMALL_HEAD_CACHE_SIZE) as the limit
would be safe.

Will you send formally the patch or do you prefer otherwise?

Thanks,

Paolo