linux-kernel - Re: [RFC net-next 1/1] idpf: Don't hard code napi

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: <b7228426-1f70-4e36-9622-c9b69bfe5be9@intel.com>
Date: Thu, 3 Oct 2024 15:35:54 +0200
From: Alexander Lobakin <aleksander.lobakin@...el.com>
To: Joe Damato <jdamato@...tly.com>, Jakub Kicinski <kuba@...nel.org>
CC: Przemek Kitszel <przemyslaw.kitszel@...el.com>, <netdev@...r.kernel.org>,
	Tony Nguyen <anthony.l.nguyen@...el.com>, "David S. Miller"
	<davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>, Paolo Abeni
	<pabeni@...hat.com>, "moderated list:INTEL ETHERNET DRIVERS"
	<intel-wired-lan@...ts.osuosl.org>, open list <linux-kernel@...r.kernel.org>,
	Simon Horman <horms@...nel.org>
Subject: Re: [RFC net-next 1/1] idpf: Don't hard code napi_struct size

From: Jakub Kicinski <kuba@...nel.org>
Date: Wed, 2 Oct 2024 10:17:27 -0700

> On Tue, 1 Oct 2024 07:44:36 -0700 Joe Damato wrote:
>>> But if you change any core API, let's say rename a field used in several
>>> drivers, you anyway need to adjust the affected drivers.  
>>
>> Sorry, but that's a totally different argument.
>>
>> There are obvious cases where touching certain parts of core would
>> require changes to drivers, yes. I agree on that if I change an API
>> or a struct field name, or remove an enum, then this affects drivers
>> which must be updated.
> 
> +1
> 
> I fully agree with Joe. Drivers asserting the size of core structures
> is both undue burden on core changes and pointless.
> The former is subjective, as for the latter: most core structures 
> will contain cold / slow path data, usually at the end. If you care
> about performance of anything that follows a core struct you need
> to align the next field yourself.
> 
> IDK how you want to fit this into your magic macros but complex
> nested types should be neither ro, rw nor cold. They are separate.

Ok I'm convinced enough. I've just imagined that if every NIC driver had
such assertions, that would've been hell.

napi_struct is the only generic struct whichs size is hardcoded in the
macros (struct dim is already sizeof()ed, as well as cpumask_var_t), so
I'm fine with the change you proposed in your first RFC -- I mean

 libeth_cacheline_set_assert(struct idpf_q_vector, 112,
-			    424 + 2 * sizeof(struct dim),
+			    24 + sizeof(struct napi_struct) +
+			    2 * sizeof(struct dim),
 			    8 + sizeof(cpumask_var_t));

I may revise this later and put generic structs outside CL groups as
Jakub suggested.

Thanks,
Olek