Message-ID: <1375075224.25230.1.camel@linux-fkkt.site>
Subject: Re: [PATCH v2 1/3] net/usb/r815x: replace USB buffer from stack to
 DMA-able
From: Oliver Neukum <oneukum@suse.de>
To: David Miller <davem@davemloft.net>
Cc: hayeswang@realtek.com, gregkh@linuxfoundation.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-usb@vger.kernel.org
Date: Mon, 29 Jul 2013 07:20:24 +0200
In-Reply-To: <20130727.202107.555068116238400560.davem@davemloft.net>
References: <20130724.174947.452272360732442554.davem@davemloft.net>
	 <1374739144-732-1-git-send-email-hayeswang@realtek.com>
	 <20130727.202107.555068116238400560.davem@davemloft.net>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1128
Lines: 35

On Sat, 2013-07-27 at 20:21 -0700, David Miller wrote:
> From: Hayes Wang <hayeswang@realtek.com>
> Date: Thu, 25 Jul 2013 15:59:02 +0800
> 
> > Some USB buffers use stack which may not be DMA-able.
> > Use the buffers from kmalloc to replace those one.
> > 
> > Signed-off-by: Hayes Wang <hayeswang@realtek.com>
> 
> I don't think it's reasonable to kmalloc() a small integer every time
> you want to use a USB message transfer to read or write chip
> registers.
> 
> Instead, add a scratch buffer to struct r8152 which is allocated once
> at driver attach time and which you can use for the transfers.
> 
> I think you only need an array of two u32's so something like:
> 
> 	u32		transfer_buf[2];
> 
> ought to be sufficient.

We cannot do that. It would violate the rules about DMA coherency.
We must not touch the same cacheline while DMA is in operation.

	Regards
		Oliver


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/