[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CACKFLi=MGnF3Tp=5_c6jjXzRbnzWL7ec8ecUYNGh-QZ1b4K9_A@mail.gmail.com>
Date: Fri, 15 Dec 2023 12:45:27 -0800
From: Michael Chan <michael.chan@...adcom.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Andy Gospodarek <andrew.gospodarek@...adcom.com>, davem@...emloft.net,
netdev@...r.kernel.org, edumazet@...gle.com, pabeni@...hat.com,
bpf@...r.kernel.org, hawk@...nel.org, ast@...nel.org, daniel@...earbox.net,
john.fastabend@...il.com, Somnath Kotur <somnath.kotur@...adcom.com>
Subject: Re: [PATCH net] bnxt_en: do not map packet buffers twice
On Fri, Dec 15, 2023 at 9:21 AM Jakub Kicinski <kuba@...nel.org> wrote:
>
> Are the XDP_REDIRECT (target) and XDP_TX going to the same rings?
> The locking seems to be missing, and bnxt_tx_int_xdp() does not
> seem to be able to handle the optimization you described if
> a ring contains a mix of XDP_REDIRECT and XDP_TX.
XDP_REDIRECT mixed with XDP_TX won't work well currently. It was
briefly mentioned on the list a few months ago:
https://lore.kernel.org/netdev/CACKFLin+1whPs0qeM5xBb1yXx8FkFS_vGrW6PaGy41_XVH=SGg@mail.gmail.com/
Yes, they share the same set of TX rings in the current code. My plan
is to have a set of dedicated TX rings for XDP_REDIRECT. Adding
locking to properly support XDP_REDIRECT and XDP_TX seems not ideal
for performance.
Download attachment "smime.p7s" of type "application/pkcs7-signature" (4209 bytes)
Powered by blists - more mailing lists