Message-ID: <99d5f95e0085ff738e4877f9e2aa7960d3972565.camel@sipsolutions.net>
Subject: Re: [PATCH v2] rhashtable: make walk safe from softirq context
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Herbert Xu <herbert@gondor.apana.org.au>
Cc:     linux-wireless@vger.kernel.org, netdev@vger.kernel.org,
        Jouni Malinen <j@w1.fi>, Thomas Graf <tgraf@suug.ch>
Date:   Thu, 07 Feb 2019 22:56:06 +0100
In-Reply-To: <20190207214834.awrpkxunnvfdorbe@gondor.apana.org.au>
References: <20190206090721.8001-1-johannes@sipsolutions.net>
         <20190207134006.gmuooqmyc5womcaf@gondor.apana.org.au>
         <fee1bdca8bb21a472a75c1a43fa61aad7fe4bff5.camel@sipsolutions.net>
         <20190207214834.awrpkxunnvfdorbe@gondor.apana.org.au>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk

On Fri, 2019-02-08 at 05:48 +0800, Herbert Xu wrote:
> On Thu, Feb 07, 2019 at 02:50:54PM +0100, Johannes Berg wrote:
> > 
> > > This interface wasn't designed for use in softirq contexts.
> > 
> > Well, it clearly was used there. You even gave it a gfp_t argument in
> > rhashtable_walk_init(), so you can't really claim it wasn't designed for
> > this. I see now that it's ignored, but still?
> 
> I see.  This was added behind my back so I wasn't aware of it.

It's not used and actually I was wrong anyway: this would have also
allowed doing the walk while holding a spinlock or with softirqs
disabled, rather than from IRQ/softirq context.

In any case, it's clearly working to iterate from this context, and
doing a spinlock_bh vs. a spinlock in the rhashtable code isn't really
that big a deal? Not sure I really understand your objection.

johannes