netdev - Re: [chrony-dev] Support for Multiple PPS Inputs on single PHC

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:   Thu, 23 Feb 2023 17:18:06 -0800
From:   Matt Corallo <ntp-lists@...tcorallo.com>
To:     Richard Cochran <richardcochran@...il.com>
Cc:     Miroslav Lichvar <mlichvar@...hat.com>,
        chrony-dev@...ony.tuxfamily.org,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [chrony-dev] Support for Multiple PPS Inputs on single PHC

On 2/23/23 4:19 PM, Richard Cochran wrote:
> On Thu, Feb 23, 2023 at 12:56:34PM -0800, Matt Corallo wrote:
> 
>> There's two separate questions here - multiple readers receiving the same
>> data, and multiple readers receiving data exclusively about one channel.
>>
>> I'd imagine the second is (much?) easier to implement, whereas the first is a bunch of complexity.
> 
> This second idea would require a new API, so that user could select a
> particular channel.
> 
> First idea would only change kernel behavior without changing the API.

Fair point. I figured a new IOCTL to filter was a lighter lift, even if a bunch of boilerplate to 
define it.

I'm happy to take a crack at something to get the ball rolling, though not this week. I'm sure I 
could copy+paste to make a new IOCTL work, but adding relevant queue limiting means I have to go 
read much more kernel code to figure out which datastructures already exist there :).

It sounds like I should go replace the extts queue with a circular buffer, have every reader socket 
store an index in the buffer, and new sockets read only futures pulses? I assume a new pulse already 
wakes all select()ers on the sockets so nothing would need to change there. Is there some existing 
code somewhere I should crib off of or just run and see where I get?

Thanks,
Matt