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;
From:   Jiaxun Yang <jiaxun.yang@flygoat.com>
Envelope-From: yjx@flygoat.com
To:     Paul Burton <paulburton89@gmail.com>
Cc:     "linux-mips@vger.kernel.org" <linux-mips@vger.kernel.org>,
        "chenhc@lemote.com" <chenhc@lemote.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "hch@lst.de" <hch@lst.de>
In-Reply-To: <20200113185246.zvsahaeh36gdfsv2@pburton-laptop>
References: <20200113140705.74605-1-jiaxun.yang@flygoat.com>
         <20200113140705.74605-2-jiaxun.yang@flygoat.com> <20200113185246.zvsahaeh36gdfsv2@pburton-laptop>
Subject: Re: [PATCH 2/2] MIPS: Loongson64: Add dma iocoherency detection support
MIME-Version: 1.0
Date:   Tue, 14 Jan 2020 11:30:19 +0800
Message-Id: <29582761578972619@iva4-6593cae50902.qloud-c.yandex.net>
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset=utf-8
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


14.01.2020, 02:52, "Paul Burton" <paulburton89@gmail.com>:
> Hi Jiaxun,
>
> On Mon, Jan 13, 2020 at 10:07:05PM +0800, Jiaxun Yang wrote:
>>  diff --git a/arch/mips/include/asm/mach-loongson64/boot_param.h b/arch/mips/include/asm/mach-loongson64/boot_param.h
>>  index 8c286bedff3e..2da2be40ad81 100644
>>  --- a/arch/mips/include/asm/mach-loongson64/boot_param.h
>>  +++ b/arch/mips/include/asm/mach-loongson64/boot_param.h
>>  @@ -115,7 +115,8 @@ struct irq_source_routing_table {
>>           u64 pci_io_start_addr;
>>           u64 pci_io_end_addr;
>>           u64 pci_config_addr;
>>  - u32 dma_mask_bits;
>>  + u16 dma_mask_bits;
>>  + u16 dma_noncoherent;
>>   } __packed;
>
> This struct is generated by the firmware, right? So does this change
> require that firmware be updated along with the kernel? Or was the
> kernel's definition always incorrect/incomplete?
Hi Paul,

This define added to the firmware by 2013 but somehow never upstream.
That's what Loongson always doing.

Some versions of Loongson-3B1500 failed to maintain DMA coherent.
So they add this parameter to the firmware.
As the maximum dma_mask_bits is 64, the upper 16bit of the old u32 will always
be zero. Which means coherent DMA transformed into the new definition,
that's the expected default behavior. Thus it's safe to do so even if firmware
doesn't understand this parameter.

Thanks

>
> Thanks,
>     Paul
--
Jiaxun