Date:   Fri, 22 Jun 2018 14:30:06 +0000
From:   Eugeniy Paltsev <Eugeniy.Paltsev@...opsys.com>
To:     Vineet Gupta <Vineet.Gupta1@...opsys.com>,
        "linux-snps-arc@...ts.infradead.org" 
        <linux-snps-arc@...ts.infradead.org>
CC:     "hch@....de" <hch@....de>,
        "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>
Subject: Re: [RFC] ARC: allow to use IOC and non-IOC DMA devices
 simultaneously

Hi Vineet, Christoph,
thanks for responding.

On Mon, 2018-06-18 at 15:53 -0700, Vineet Gupta wrote:
> On 06/15/2018 05:58 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.
> 
> This is really nice: having this a per device behaviour has been desirable rather
> than the current blunt system-wide behaviour.
> 
> However the patch doesn't seem to change the routing off non-coherent traffic -
> everything would still go thru it - per the current default setting of
> CREG_AXI_M_*_SLV[0-1] registers. Ideally you would want to disable that as well,
> an addon patch is fine.

Yep, in this patch I only implement infrastructure to be able to choose
coherent/non-coherent cache ops per device. As for today all
*used* (supported and enabled in the mainline kernel) DMA peripherals on AXS103
and HSDK boards are routed via IOC port. So in this patch I'm going to simply enable
io-coherency via device tree for every DMA peripheral on AXS103 and HSDK.

Then I going to modify apertures configuration in separate patch.
It will be something like it is done in u-boot:
https://elixir.bootlin.com/u-boot/v2018.07-rc2/source/board/synopsys/hsdk/hsdk.c#L441

> > 
> > Some recent SoC with ARC HS (like HSDK) allow to select bus
> > port (IOC or non-IOC port) for connecting DMA devices in runtime.
> > 
> > With this patch we can use both HW-coherent and regular DMA
> > peripherals simultaneously.
> > 
> > For example we can connect USB and SDIO controllers through IOC port
> > (so we don't need to need to maintain cache coherency for these
> > devices manualy. All cache sync ops will be nop)
> > And we can connect Ethernet directly to RAM port (so we had to
> > maintain cache coherency manualy. Cache sync ops will be real
> > flush/invalidate operations)
> > 
> > Cache ops are set per-device and depends on "dma-coherent" device
> > tree property:
> > "dma_noncoherent_ops" are used if no "dma-coherent" property is
> > present (or IOC is disabled)
> > "dma_direct_ops" are used if "dma-coherent" property is present.
> 
> I agree with Christoph that creating a new file for this seems excessive.
Ok, I'll move it to arch/arc/mm/dma.c

> > NOTE 1:
> > It works perfectly fine only if we don't have ZONE_HIGHMEM used
> > as IOC doesn't cover all physical memory. As for today It configured
> > to cover 1GiB starting from 0x8z (which is ZONE_NORMAL memory for
> > us). Transactions outside this region are sent on the non-coherent
> > I/O bus interface.
> > We can't configure IOC to cover all physical memory as it has several
> > limitations relating to aperture size and start address.
> > 
> > And if we get DMA buffer from ZONE_HIGHMEM memory we need to
> > do real flush/invalidate operations on that buffer, which is obviously
> > not done by "dma_direct_ops".
> > 
> > So I am not sure about "dma_direct_ops" using - probably we need to
> > create our special cache ops like "arc_ioc_ops" which will handle
> > ZONE_HIGHMEM case.

Any ideas about ZONE_HIGHMEM? Should I implement separate cache ops
like "arc_ioc_ops" where I'll check and process HIGHMEM pages special way?

> > (BTW: current ARC dma_noncoherent_ops implementation also has same
> > problem if IOC and HIGHMEM are enabled.)
> 
> Can we highlight this fact, add error prints somewhere ?
The easiest solution is to panic if both IOC and ZONE_HIGHMEM enabled.

> > NOTE 2:
> > In this RFC only hsdk.dts changes are shown to reduce patch size.
> > AXS103 device tree changes are not shown.
> > 
> > Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@...opsys.com>
> > ---
> >  arch/arc/Kconfig                   |  1 +
> >  arch/arc/boot/dts/hsdk.dts         |  4 ++++
> >  arch/arc/include/asm/dma-mapping.h | 14 ++++++++++++++
> >  arch/arc/mm/Makefile               |  2 +-
> >  arch/arc/mm/cache.c                | 15 +--------------
> >  arch/arc/mm/dma-mapping.c          | 20 ++++++++++++++++++++
> >  arch/arc/mm/dma.c                  | 14 +-------------
> >  7 files changed, 42 insertions(+), 28 deletions(-)
> >  create mode 100644 arch/arc/include/asm/dma-mapping.h
> >  create mode 100644 arch/arc/mm/dma-mapping.c
> > 
> > diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig
> > index e81bcd271be7..0a2fcd2a8c32 100644
> > --- a/arch/arc/Kconfig
> > +++ b/arch/arc/Kconfig
> > @@ -17,6 +17,7 @@ config ARC
> >  	select CLONE_BACKWARDS
> >  	select COMMON_CLK
> >  	select DMA_NONCOHERENT_OPS
> > +	select DMA_DIRECT_OPS
> >  	select DMA_NONCOHERENT_MMAP
> >  	select GENERIC_ATOMIC64 if !ISA_ARCV2 || !(ARC_HAS_LL64 && ARC_HAS_LLSC)
> >  	select GENERIC_CLOCKEVENTS
> > diff --git a/arch/arc/boot/dts/hsdk.dts b/arch/arc/boot/dts/hsdk.dts
> > index 006aa3de5348..ebb686c21393 100644
> > --- a/arch/arc/boot/dts/hsdk.dts
> > +++ b/arch/arc/boot/dts/hsdk.dts
> > @@ -176,6 +176,7 @@
> >  			phy-handle = <&phy0>;
> >  			resets = <&cgu_rst HSDK_ETH_RESET>;
> >  			reset-names = "stmmaceth";
> > +			dma-coherent;
> >  
> >  			mdio {
> >  				#address-cells = <1>;
> > @@ -194,12 +195,14 @@
> >  			compatible = "snps,hsdk-v1.0-ohci", "generic-ohci";
> >  			reg = <0x60000 0x100>;
> >  			interrupts = <15>;
> > +			dma-coherent;
> >  		};
> >  
> >  		ehci@...00 {
> >  			compatible = "snps,hsdk-v1.0-ehci", "generic-ehci";
> >  			reg = <0x40000 0x100>;
> >  			interrupts = <15>;
> > +			dma-coherent;
> >  		};
> >  
> >  		mmc@...0 {
> > @@ -212,6 +215,7 @@
> >  			clock-names = "biu", "ciu";
> >  			interrupts = <12>;
> >  			bus-width = <4>;
> > +			dma-coherent;
> >  		};
> >  	};
> >  
> > diff --git a/arch/arc/include/asm/dma-mapping.h b/arch/arc/include/asm/dma-mapping.h
> > new file mode 100644
> > index 000000000000..640a851bd331
> > --- /dev/null
> > +++ b/arch/arc/include/asm/dma-mapping.h
> > @@ -0,0 +1,14 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +// (C) 2018 Synopsys, Inc. (www.synopsys.com)
> > +
> > +#ifndef ASM_ARC_DMA_MAPPING_H
> > +#define ASM_ARC_DMA_MAPPING_H
> > +
> > +#define arch_setup_dma_ops arch_setup_dma_ops
> > +
> > +#include <asm-generic/dma-mapping.h>
> > +
> > +void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
> > +			const struct iommu_ops *iommu, bool coherent);
> > +
> > +#endif
> > diff --git a/arch/arc/mm/Makefile b/arch/arc/mm/Makefile
> > index 3703a4969349..45683897c27b 100644
> > --- a/arch/arc/mm/Makefile
> > +++ b/arch/arc/mm/Makefile
> > @@ -7,5 +7,5 @@
> >  #
> >  
> >  obj-y	:= extable.o ioremap.o dma.o fault.o init.o
> > -obj-y	+= tlb.o tlbex.o cache.o mmap.o
> > +obj-y	+= tlb.o tlbex.o cache.o mmap.o dma-mapping.o
> >  obj-$(CONFIG_HIGHMEM)	+= highmem.o
> > diff --git a/arch/arc/mm/cache.c b/arch/arc/mm/cache.c
> > index 9dbe645ee127..c5d1f2a2c4da 100644
> > --- a/arch/arc/mm/cache.c
> > +++ b/arch/arc/mm/cache.c
> > @@ -896,15 +896,6 @@ static void __dma_cache_wback_slc(phys_addr_t start, unsigned long sz)
> >  	slc_op(start, sz, OP_FLUSH);
> >  }
> >  
> > -/*
> > - * DMA ops for systems with IOC
> > - * IOC hardware snoops all DMA traffic keeping the caches consistent with
> > - * memory - eliding need for any explicit cache maintenance of DMA buffers
> > - */
> > -static void __dma_cache_wback_inv_ioc(phys_addr_t start, unsigned long sz) {}
> > -static void __dma_cache_inv_ioc(phys_addr_t start, unsigned long sz) {}
> > -static void __dma_cache_wback_ioc(phys_addr_t start, unsigned long sz) {}
> > -
> >  /*
> >   * Exported DMA API
> >   */
> > @@ -1253,11 +1244,7 @@ void __init arc_cache_init_master(void)
> >  	if (is_isa_arcv2() && ioc_enable)
> >  		arc_ioc_setup();
> >  
> > -	if (is_isa_arcv2() && ioc_enable) {
> > -		__dma_cache_wback_inv = __dma_cache_wback_inv_ioc;
> > -		__dma_cache_inv = __dma_cache_inv_ioc;
> > -		__dma_cache_wback = __dma_cache_wback_ioc;
> > -	} else if (is_isa_arcv2() && l2_line_sz && slc_enable) {
> 
> Maybe also tweak the boot printing in setup.c to indicate that we now do per
> peripheral ioc !

Ok, good idea.

> > +	if (is_isa_arcv2() && l2_line_sz && slc_enable) {
> >  		__dma_cache_wback_inv = __dma_cache_wback_inv_slc;
> >  		__dma_cache_inv = __dma_cache_inv_slc;
> >  		__dma_cache_wback = __dma_cache_wback_slc;
> > diff --git a/arch/arc/mm/dma-mapping.c b/arch/arc/mm/dma-mapping.c
> > new file mode 100644
> > index 000000000000..9d0d310bbf5a
> > --- /dev/null
> > +++ b/arch/arc/mm/dma-mapping.c
> > @@ -0,0 +1,20 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +// (C) 2018 Synopsys, Inc. (www.synopsys.com)
> > +
> > +#include <linux/dma-mapping.h>
> > +
> > +void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
> > +			const struct iommu_ops *iommu, bool coherent)
> > +{
> > +	const struct dma_map_ops *dma_ops = &dma_noncoherent_ops;
> > +
> > +	/*
> > +	 * IOC hardware snoops all DMA traffic keeping the caches consistent
> > +	 * with memory - eliding need for any explicit cache maintenance of
> > +	 * DMA buffers - so we can use dma_direct cache ops.
> > +	 */
> > +	if (is_isa_arcv2() && ioc_enable && coherent)
> > +		dma_ops = &dma_direct_ops;
> > +
> > +	set_dma_ops(dev, dma_ops);
> 
> Add a debug printk here maybe ?

Ok.

> > +}
> > diff --git a/arch/arc/mm/dma.c b/arch/arc/mm/dma.c
> > index 8c1071840979..4fd130e786c7 100644
> > --- a/arch/arc/mm/dma.c
> > +++ b/arch/arc/mm/dma.c
> > @@ -33,19 +33,7 @@ void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
> >  	if (!page)
> >  		return NULL;
> >  
> > -	/*
> > -	 * IOC relies on all data (even coherent DMA data) being in cache
> > -	 * Thus allocate normal cached memory
> > -	 *
> > -	 * The gains with IOC are two pronged:
> > -	 *   -For streaming data, elides need for cache maintenance, saving
> > -	 *    cycles in flush code, and bus bandwidth as all the lines of a
> > -	 *    buffer need to be flushed out to memory
> > -	 *   -For coherent data, Read/Write to buffers terminate early in cache
> > -	 *   (vs. always going to memory - thus are faster)
> > -	 */
> > -	if ((is_isa_arcv2() && ioc_enable) ||
> > -	    (attrs & DMA_ATTR_NON_CONSISTENT))
> > +	if (attrs & DMA_ATTR_NON_CONSISTENT)
> >  		need_coh = 0;
> >  
> >  	/*
> 
> 
-- 
 Eugeniy Paltsev

Powered by blists - more mailing lists