Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sat, 17 Aug 2019 14:15:06 -0700 (PDT)
From:   Paul Walmsley <paul.walmsley@sifive.com>
To:     Christoph Hellwig <hch@lst.de>
cc:     Arnd Bergmann <arnd@arndb.de>, Guo Ren <guoren@kernel.org>,
        Michal Simek <monstr@monstr.eu>,
        Greentime Hu <green.hu@gmail.com>,
        Vincent Chen <deanbo422@gmail.com>,
        Guan Xuetao <gxt@pku.edu.cn>, x86@kernel.org,
        linux-arch@vger.kernel.org, linux-s390@vger.kernel.org,
        linux-ia64@vger.kernel.org, linux-parisc@vger.kernel.org,
        linux-sh@vger.kernel.org, linux-hexagon@vger.kernel.org,
        linux-xtensa@linux-xtensa.org, linux-mips@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-m68k@lists.linux-m68k.org,
        openrisc@lists.librecores.org, linux-mtd@lists.infradead.org,
        linux-alpha@vger.kernel.org, sparclinux@vger.kernel.org,
        nios2-dev@lists.rocketboards.org, linux-riscv@lists.infradead.org,
        linux-snps-arc@lists.infradead.org,
        linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH 16/26] asm-generic: don't provide ioremap for
 CONFIG_MMU
In-Reply-To: <20190817073253.27819-17-hch@lst.de>
Message-ID: <alpine.DEB.2.21.9999.1908171414260.4130@viisi.sifive.com>
References: <20190817073253.27819-1-hch@lst.de> <20190817073253.27819-17-hch@lst.de>
User-Agent: Alpine 2.21.9999 (DEB 301 2018-08-15)
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, 17 Aug 2019, Christoph Hellwig wrote:

> All MMU-enabled ports have a non-trivial ioremap and should thus provide
> the prototype for their implementation instead of providing a generic
> one unless a different symbol is not defined.  Note that this only
> affects sparc32 nds32 as all others do provide their own version.
> 
> Also update the kerneldoc comments in asm-generic/io.h to explain the
> situation around the default ioremap* implementations correctly.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/nds32/include/asm/io.h    |  2 ++
>  arch/sparc/include/asm/io_32.h |  1 +
>  include/asm-generic/io.h       | 29 ++++++++---------------------
>  3 files changed, 11 insertions(+), 21 deletions(-)
> 

[ ... ]

> diff --git a/include/asm-generic/io.h b/include/asm-generic/io.h
> index a98ed6325727..6a5edc23afe2 100644
> --- a/include/asm-generic/io.h
> +++ b/include/asm-generic/io.h
> @@ -922,28 +922,16 @@ static inline void *phys_to_virt(unsigned long address)
>  /**
>   * DOC: ioremap() and ioremap_*() variants
>   *
> - * If you have an IOMMU your architecture is expected to have both ioremap()
> - * and iounmap() implemented otherwise the asm-generic helpers will provide a
> - * direct mapping.
> + * Architectures with an MMU are expected to provide ioremap() and iounmap()
> + * themselves.  For NOMMU architectures we provide a default nop-op
> + * implementation that expect that the physical address used for MMIO are
> + * already marked as uncached, and can be used as kernel virtual addresses.
>   *
> - * There are ioremap_*() call variants, if you have no IOMMU we naturally will
> - * default to direct mapping for all of them, you can override these defaults.
> - * If you have an IOMMU you are highly encouraged to provide your own
> - * ioremap variant implementation as there currently is no safe architecture
> - * agnostic default. To avoid possible improper behaviour default asm-generic
> - * ioremap_*() variants all return NULL when an IOMMU is available. If you've
> - * defined your own ioremap_*() variant you must then declare your own
> - * ioremap_*() variant as defined to itself to avoid the default NULL return.
> + * ioremap_wc() and ioremap_wt() can provide more relaxed caching attributes
> + * for specific drivers if the architecture choses to implement them.  If they
                                               ^^^ chooses


- Paul