netdev - Re: [net-next PATCH v8 02/10] net: Add queue and napi association

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]

Date: Wed, 22 Nov 2023 16:56:58 -0800
From: "Nambiar, Amritha" <amritha.nambiar@...el.com>
To: Jakub Kicinski <kuba@...nel.org>
CC: Willem de Bruijn <willemb@...gle.com>, <netdev@...r.kernel.org>,
	<pabeni@...hat.com>, <sridhar.samudrala@...el.com>
Subject: Re: [net-next PATCH v8 02/10] net: Add queue and napi association

On 11/22/2023 2:00 PM, Jakub Kicinski wrote:
> On Wed, 22 Nov 2023 13:28:19 -0800 Nambiar, Amritha wrote:
>> Trying to understand this distinction bit more:
>> So, netdev-genl queue-get shows state of queue objects as reported from
>> the netdev level. Now, unless there's the queue-set command changing the
>> configuration, the state of queue-objects would match the hardware
>> configurations.
>> When the user changes the configuration with a queue-set command:
>> - queue-get would report the new updates (as obtained from the netdev).
>> - The updates would not be reflected in the hardware till a reset is
>> issued. At this point, ethtool or others would report the older
>> configuration (before reset).
>> - After reset, the state of queue objects from queue-get would match the
>> actual hardware configuration.
>>
>> I agree, an explicit "reset" user-command would be great. This way all
>> the set operations for the netdev objects (queue, NAPI, page pool etc.)
>> would stay at the netdev level without needing ndo_op for each, and then
>> the "reset" command can trigger the ndo callback and actuate the
>> hardware changes.
> 
> How the changes are applied is a separate topic. I was only talking
> about the fact that if the settings are controllable both at the device
> level and queue level - the queue state is a result of combining device
> settings with queue settings.

Okay, makes sense. Will fix in v9 and skip listing queues and NAPIs of 
devices which are DOWN.