linux-kernel - Re: [PATCH v2 0/4] ARC: allow to use IOC and non-IOC DMA devices simultaneously

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 13 Aug 2018 16:19:21 +0000
From:   Vineet Gupta <Vineet.Gupta1@...opsys.com>
To:     Eugeniy Paltsev <Eugeniy.Paltsev@...opsys.com>,
        "linux-snps-arc@...ts.infradead.org" 
        <linux-snps-arc@...ts.infradead.org>
CC:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-arch@...r.kernel.org" <linux-arch@...r.kernel.org>,
        Alexey Brodkin <Alexey.Brodkin@...opsys.com>,
        "hch@....de" <hch@....de>
Subject: Re: [PATCH v2 0/4] ARC: allow to use IOC and non-IOC DMA devices
 simultaneously

On 07/30/2018 09:26 AM, Eugeniy Paltsev wrote:
> The ARC HS processor provides an IOC port (I/O coherency bus
> interface) that allows external devices such as DMA devices
> to access memory through the cache hierarchy, providing
> coherency between I/O transactions and the complete memory
> hierarchy.

You didn't pay attention to my previous comment on this !
IOC port can be considered a micro-architecture optimization (an important one
though). The main thing is hardware snooping DMA transactions which enabled IOC in
first place.

> Some recent SoC with ARC HS (like HSDK) allow to select bus
> port (IOC or non-IOC port) for connecting DMA devices in runtime.

Again you mention the port but none of your 4 patches actually touch the port
itself in programming it.


> With this patch we can use both HW-coherent and regular DMA
> peripherals simultaneously.
>
> NOTE:
> This patch series was stress tested on HSDK with iperf3 (ethernet)
> and bonie++ (usb and sdio) in three configurations:
>  * IOC enabled globaly
>  * IOC disabled globaly
>  * IOC enabled partially (USB & SDIO are connected via IOC AXI port,
>    ethernet is connected to DDR AXI port (non-IOC port) 
>
> NOTE:
> If you want to test some device without IOC it is not enough
> to remove "dma-coherent" property from dts. You had to remap this
> device to regular DDR AXI port intead of IOC AXI port.

Why are we not doing that for the GPU as part of this series.

> You also need to apply 3 following patches firstly:
> https://www.mail-archive.com/linux-snps-arc@lists.infradead.org/msg03865.html
> https://www.mail-archive.com/linux-snps-arc@lists.infradead.org/msg03889.html
> https://www.mail-archive.com/linux-snps-arc@lists.infradead.org/msg03887.html
>
> NOTE:
> We don't touch any aperture configuration in this patch series. So we
> don't switch any devices between IOC and non-IOC AXI ports on any board.
> It can be done later if it is required.
>
> Changes v1->v2 (Thanks to Christoph):
>  * Don't select DMA_DIRECT_OPS explicitly as it is already selected by
>    DMA_NONCOHERENT_OPS
>  * Remove check for HIGHMEM pages from arch_dma_{alloc, free}
>
> Eugeniy Paltsev (4):
>   ARC: DTS: mark DMA devices connected through IOC port as dma-coherent
>   ARC: allow to use IOC and non-IOC DMA devices simultaneously
>   ARC: IOC: panic if both IOC and ZONE_HIGHMEM enabled
>   ARC: don't check for HIGHMEM pages in arch_dma_alloc
>
>  arch/arc/boot/dts/axc003.dtsi      | 26 ++++++++++++++++
>  arch/arc/boot/dts/axc003_idu.dtsi  | 26 ++++++++++++++++
>  arch/arc/boot/dts/hsdk.dts         |  4 +++
>  arch/arc/include/asm/dma-mapping.h | 13 ++++++++
>  arch/arc/mm/cache.c                | 30 +++++++++---------
>  arch/arc/mm/dma.c                  | 62 +++++++++++++++++++-------------------
>  6 files changed, 115 insertions(+), 46 deletions(-)
>  create mode 100644 arch/arc/include/asm/dma-mapping.h
>