netdev - RE: [PATCH net-next] ethtool: add netlink support for rss set

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]

Date:   Tue, 14 Mar 2023 23:51:00 +0000
From:   "Mogilappagari, Sudheer" <sudheer.mogilappagari@...el.com>
To:     Edward Cree <ecree.xilinx@...il.com>,
        Jakub Kicinski <kuba@...nel.org>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "mkubecek@...e.cz" <mkubecek@...e.cz>,
        "Samudrala, Sridhar" <sridhar.samudrala@...el.com>,
        "Nguyen, Anthony L" <anthony.l.nguyen@...el.com>
Subject: RE: [PATCH net-next] ethtool: add netlink support for rss set



> -----Original Message-----
> From: Edward Cree <ecree.xilinx@...il.com>
> Sent: Tuesday, March 14, 2023 6:35 AM
> To: Jakub Kicinski <kuba@...nel.org>; Mogilappagari, Sudheer
> <sudheer.mogilappagari@...el.com>
> Cc: netdev@...r.kernel.org; mkubecek@...e.cz; Samudrala, Sridhar
> <sridhar.samudrala@...el.com>; Nguyen, Anthony L
> <anthony.l.nguyen@...el.com>
> Subject: Re: [PATCH net-next] ethtool: add netlink support for rss set
> 
> On 13/03/2023 22:53, Jakub Kicinski wrote:
> > Ah, so you do have a feature. Yes, it would be somewhat helpful but my
> > larger concern remains. We skipped the dump implementation when
> > implementing GET. The admin still has no way of knowing what / how
> > many RSS contexts had been created. With the context ID being an
> > unbounded integer just going from 0 until ENOENT is not even an option.
> >
> > So we need to start tracking the contexts.
> 
> Hi Jakub, as the original author of custom RSS contexts I feel like I  owe
> this bit of work; I can take a swing at it, unless Sudheer would  rather do
> it himself.

Hi Edward, would be great if you can add the support. I will modify the RSS_SET
patch accordingly based on your changes.

Got a question wrt dumpit. How to instrument dumpit functionality? ethtool doesn't 
seem to have an explicit option. In ethtool userspace code, NLM_F_DUMP is set 
in below functions 

 - nlsock_prep_get_request: 
	if (devname && !strcmp(devname, WILDCARD_DEVNAME)) {
	       devname = NULL;
	       nlm_flags |= NLM_F_DUMP;
	}
 - stringset_load_request : called with is_dump=false in ethtool functions

How to get devname to be WILDCARD_DEVNAME ?

> 
> > Add a pointer to struct
> > netdevice to hold an "ethtool_settings" struct. In the ethtool
> > settings struct add a list head. Put an object for each created RSS
> > context on that list.

> Would an IDR not be appropriate here, rather than a list?
> AFAICT every driver that supports contexts either treats the context  ID as
> an opaque handle or as an index into a fixed-size array, so as  long as the
> driver reports its max context ID to the core somehow,  the specific ID
> values chosen are arbitrary and the driver doesn't  need to do the choosing,
> it can just take what comes out of the IDR.