[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <56FAAA6D.3070806@candelatech.com>
Date: Tue, 29 Mar 2016 09:16:45 -0700
From: Ben Greear <greearb@...delatech.com>
To: Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Herbert Xu <herbert@...dor.apana.org.au>,
"linux-wireless@...r.kernel.org" <linux-wireless@...r.kernel.org>
Subject: Re: Question on rhashtable in worst-case scenario.
Looks like rhashtable has too much policy in it to properly deal with
cases where there are too many hash collisions, so I am going to work on
reverting it's use in mac80211.
Thanks,
Ben
On 03/28/2016 01:29 PM, Ben Greear wrote:
> Hello!
>
> I have a use case for mac80211 where I create multiple stations to
> the same remote peer MAC address.
>
> I'm seeing cases where the rhashtable logic is returning -16 (EBUSY)
> on insert (see sta_info_hash_add).
> This is with the 4.4.6+ (plus local patches) kernel, and it has the patch mentioned
> here:
>
> https://lkml.org/lkml/2015/12/3/307
>
> If I understand the code properly, my use case is going to be worst-case scenario,
> where all of my items in the hash have the same key (peer mac addr).
>
> I have my own secondary hash to handle most of my hot-path lookups, but I still need
> the main hash to at least function in a linear-search manner.
>
> Any idea what I can do to get rid of the EBUSY return code problem, or how
> to debug it further?
>
> Thanks,
> Ben
>
--
Ben Greear <greearb@...delatech.com>
Candela Technologies Inc http://www.candelatech.com
Powered by blists - more mailing lists