netdev - Re: [PATCH net-next 0/4] bonding work-queues, try

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]

Message-Id: <20150312.152023.288579610581649568.davem@redhat.com>
Date:	Thu, 12 Mar 2015 15:20:23 -0400 (EDT)
From:	David Miller <davem@...hat.com>
To:	maheshb@...gle.com
Cc:	j.vosburgh@...il.com, andy@...yhouse.net, vfalico@...il.com,
	nikolay@...hat.com, maze@...gle.com, netdev@...r.kernel.org,
	edumazet@...gle.com
Subject: Re: [PATCH net-next 0/4] bonding work-queues, try_rtnl() &
 notifications

From: Mahesh Bandewar <maheshb@...gle.com>
Date: Wed, 11 Mar 2015 22:54:33 -0700

> This patch series tries to address the issue discovered in various work-
> queues and the way these handlers deal with the RTNL. Especially for 
> notification handling. If RTNL can not be acquired, these handlers ignore
> sending notifications and just re-arm the timer. This could be very 
> problematic if the re-arm timer has larger value (e.g. in minutes).

Any timeout selected is completely arbitrary, and if the RTNL semaphore
is really being held for a non-trivial amount of time, you're going to
just thrash and potentially even steal cycles and cache lines from
the thing that's taking so long.

Please do something more sane, perhaps similar to the
netdev_run_todo() function which is invoked at rtnl_unlock() time.

You'll need a different synchronization scheme that that which
netdev_run_todo() uses, but that shouldn't be too hard.

This way, when the RTNL mutex is dropped your notifications will get
generated promptly, and without all of the thrashing side-effects
and arbitrary timeouts.

Thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html