netdev - Re: [PATCH RFC net-next/mm V3 1/2] page_pool: Remove workqueue in new shutdown scheme

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: <87ednxbr3c.fsf@toke.dk>
Date: Wed, 03 May 2023 13:18:47 +0200
From: Toke Høiland-Jørgensen <toke@...hat.com>
To: Jakub Kicinski <kuba@...nel.org>, Jesper Dangaard Brouer
 <brouer@...hat.com>
Cc: Ilias Apalodimas <ilias.apalodimas@...aro.org>, netdev@...r.kernel.org,
 Eric Dumazet <eric.dumazet@...il.com>, linux-mm@...ck.org, Mel Gorman
 <mgorman@...hsingularity.net>, lorenzo@...nel.org, linyunsheng@...wei.com,
 bpf@...r.kernel.org, "David S. Miller" <davem@...emloft.net>, Paolo Abeni
 <pabeni@...hat.com>, Andrew Morton <akpm@...ux-foundation.org>,
 willy@...radead.org
Subject: Re: [PATCH RFC net-next/mm V3 1/2] page_pool: Remove workqueue in
 new shutdown scheme

Jakub Kicinski <kuba@...nel.org> writes:

> On Fri, 28 Apr 2023 18:16:19 +0200 Jesper Dangaard Brouer wrote:
>> This removes the workqueue scheme that periodically tests when
>> inflight reach zero such that page_pool memory can be freed.
>> 
>> This change adds code to fast-path free checking for a shutdown flags
>> bit after returning PP pages.
>
> We can remove the warning without removing the entire delayed freeing
> scheme. I definitely like the SHUTDOWN flag and patch 2 but I'm a bit
> less clear on why the complexity of datapath freeing is justified.
> Can you explain?

You mean just let the workqueue keep rescheduling itself every minute
for the (potentially) hours that skbs will stick around? Seems a bit
wasteful, doesn't it? :)

We did see an issue where creating and tearing down lots of page pools
in a short period of time caused significant slowdowns due to the
workqueue mechanism. Lots being "thousands per second". This is possible
using the live packet mode of bpf_prog_run() for XDP, which will setup
and destroy a page pool for each syscall...

-Toke