From mboxrd@z Thu Jan 1 00:00:00 1970 From: Russell King - ARM Linux Subject: Re: amba pl011: where to submit a fix for x86 builds Date: Fri, 7 Jun 2013 10:50:13 +0100 Message-ID: <20130607095013.GU18614@n2100.arm.linux.org.uk> References: <20130607093854.GA3429@mail.gnudd.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20130607093854.GA3429@mail.gnudd.com> Sender: linux-kernel-owner@vger.kernel.org To: Alessandro Rubini Cc: linux-kernel@vger.kernel.org, gregkh@linuxfoundation.org, jslaby@suse.cz, linux-serial@vger.kernel.org, x86@kernel.org, tglx@linutronix.de, mingo@redhat.com, hpa@zytor.com, Chanho Min , Giancarlo Asnaghi List-Id: linux-serial@vger.kernel.org On Fri, Jun 07, 2013 at 11:38:54AM +0200, Alessandro Rubini wrote: > Hello. > This is cc'd to $(get_maintainer.pl -f drivers/tty/serial/amba-pl011.c) > and x86 maintainers too. > > While working on the pci-to-amba bridge driver (to be resubmitted in > the next days) I have to deal with the problem of pl011 not building, > after phys_to_page was introduced, by Chanho Min (Cc: here) in commit > > cb06ff1 ARM: PL011: Add support for Rx DMA buffer polling. > > because x86 has no phys_to_page. > > The fix I use is trivially this: > > - sg_set_page(&sg->sg, phys_to_page(dma_addr), > + sg_set_page(&sg->sg, virt_to_page(sg->buf), > > (virt_to_page is slower, but other sg_get_page() users do the same > and other solutions are more convoluted). ... and thereby breaks ARM, because sg->buf is not part of the kernel direct mapped RAM region. I don't think there's a portable way to do the above, because the DMA coherent API doesn't really support translating the allocated memory region to a struct page pointer.