Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220821055326.GA25950@lst.de> <20220825091752.3535657-1-arnd@kernel.org>
 <YwdQjEJBUwGJB/YK@shell.armlinux.org.uk>
In-Reply-To: <YwdQjEJBUwGJB/YK@shell.armlinux.org.uk>
From:   Arnd Bergmann <arnd@kernel.org>
Date:   Thu, 25 Aug 2022 19:17:22 +0200
Message-ID: <CAK8P3a3xXQ7SbAsFAe637PkpL8_2NYstjOLj35D+QPyy-GzKGg@mail.gmail.com>
Subject: Re: [PATCH 5/4] ARM: footbridge: remove custom DMA address handling
To:     "Russell King (Oracle)" <linux@armlinux.org.uk>
Cc:     Christoph Hellwig <hch@lst.de>, linux-kernel@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org,
        Arnd Bergmann <arnd@arndb.de>,
        Linus Walleij <linus.walleij@linaro.org>,
        Marc Zyngier <maz@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Aug 25, 2022 at 12:35 PM Russell King (Oracle)
<linux@armlinux.org.uk> wrote:
>
> On Thu, Aug 25, 2022 at 11:17:52AM +0200, Arnd Bergmann wrote:
> > -/*
> > - * The footbridge is programmed to expose the system RAM at 0xe0000000.
> > - * The requirement is that the RAM isn't placed at bus address 0, which
> > - * would clash with VGA cards.
> > - */
> > -#define BUS_OFFSET           0xe0000000
> > -#define __virt_to_bus(x)     ((x) + (BUS_OFFSET - PAGE_OFFSET))
> > -#define __bus_to_virt(x)     ((x) - (BUS_OFFSET - PAGE_OFFSET))
>
> ... and this tells you why that is with ISA devices.
>
> So, one hell of a big NAK to this patch, sorry.

Thanks for taking a look. Apparently I misread the interface, and I assumed
that the __bus_to_virt() BUS_OFFSET in set_dma_addr() would
just cancel out the dma_map_single() in isa_enable_dma().

It turns out the logic is actually simpler and we just convert between
virtual address and bus address three times. for most drivers.

I found one driver (cs89x0-isa) that uses isa_virt_to_bus()/virt_to_phys()
instead of dma_map_single(), but that is probably never used here.
Once we move the DMA offset into the device, all other ISA devices will
also get the default zero-offset, but I think that will be fine as long as the
"isa_dma_dev" that is used for the final translation has the correct offset.

I'll try to come up with a new version tomorrow.

       Arnd