From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michael Schmitz Subject: Re: [PATCH net] macmace: Set platform device coherent_dma_mask Date: Fri, 4 May 2018 20:16:58 +1200 Message-ID: References: <20180503085120.GA14574@lst.de> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Cc: Christoph Hellwig , Finn Thain , "David S. Miller" , linux-m68k , netdev , Linux Kernel Mailing List To: Geert Uytterhoeven Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Hi Geert, Am 04.05.2018 um 19:24 schrieb Geert Uytterhoeven: > Hi Michael, > >>> Yes, that would be useful. The other assumption could be that >>> platform devices always allow an all-0xff dma mask. >> >> That's not always true (Atari NCR5380 SCSI and floppy would use a 24 >> bit DMA mask). We use bounce buffers allocated from a dedicated lowmem >> pool there currently, and for all I know don't use the DMA API yet. >> >> I bet that is a rare exception though. Setting the default DMA mask >> for platform devices to all-0xff and letting the few odd drivers force >> a different setting seems the best way forward. > > I'd say that's usually a property of the platform, not of the device? Right - I was thinking 'm68k' as platform, not a particular machine like Mac or Falcon (the 24 bit mask only applies to that particular model anyway). > So IMHO it belongs in the platform code, not in the device driver code. OK - let's have a default mask of 64 bit, and allow machine specific platform_init() to override using a new helper function. Cheers, Michael > Gr{oetje,eeting}s, > > Geert >