netdev - Re: Device mem changes vs pinning/zerocopy changes

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

Message-ID: <1099957.1749052763@warthog.procyon.org.uk>
Date: Wed, 04 Jun 2025 16:59:23 +0100
From: David Howells <dhowells@...hat.com>
To: Mina Almasry <almasrymina@...gle.com>
Cc: dhowells@...hat.com, willy@...radead.org, hch@...radead.org,
    Jakub Kicinski <kuba@...nel.org>, Eric Dumazet <edumazet@...gle.com>,
    netdev@...r.kernel.org, linux-mm@...ck.org,
    linux-kernel@...r.kernel.org
Subject: Re: Device mem changes vs pinning/zerocopy changes

(Apologies, I accidentally sent the incomplete email)

> I think you need to modify the existing sk_buff. I think adding
> a new struct and migrating the entire net stack to use that is a bit
> too ambitious. But up to you. Just my 2 cents here.

It may come down to that, and if it does, we'll need to handle frags
differently.  Basically, for zerocopy, the following will all apply or come to
apply sometime in the future:

 (1) We're going to be getting arrays of {physaddr,len} from the higher
     layers.  I think Christoph's idea is that this makes DMA mapping easier.
     We will need to retain this.

 (2) There will be no page refcount.  We will obtain a pin (user zc) or there
     will be a destructor (kernel zc).  If the latter, it may have to be
     shared amongst multiple skbuffs.

 (3) We can't assume that a frag refers to a single page - or that there's
     even a necessarily a page struct.  And if there is a page struct, the
     metadata isn't going to be stored in it.  The page struct will contain a
     typed pointer to some other struct (e.g. folio).

 (4) We can't assume anything about the memory type of the frag (could be
     slab, for instance).

Obviously, if we copy the data into netmem buffers, we have full control of
that memory type.

David