linux-kernel - Re: [PATCH 08/17] remoteproc/omap: Add support for DRA7xx remote processors

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:   Tue, 12 Nov 2019 10:13:43 -0800
From:   Bjorn Andersson <bjorn.andersson@...aro.org>
To:     Tero Kristo <t-kristo@...com>
Cc:     ohad@...ery.com, linux-remoteproc@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-omap@...r.kernel.org,
        s-anna@...com
Subject: Re: [PATCH 08/17] remoteproc/omap: Add support for DRA7xx remote
 processors

On Tue 12 Nov 00:37 PST 2019, Tero Kristo wrote:

> On 12/11/2019 01:37, Bjorn Andersson wrote:
> > On Mon 28 Oct 05:42 PDT 2019, Tero Kristo wrote:
[..]
> > > +	for (; data && data->device_name; data++) {
> > > +		if (!strcmp(dev_name(&pdev->dev), data->device_name))
> > 
> > I don't fancy the reliance on node names in devicetree, is this well
> > defined in the binding?
> 
> I don't think it is.... So, would it be better to just replace the
> compatible strings for dra7 remoteprocs to be like ti,dra7-dsp1 /
> ti,dra7-dsp2 etc.? I think that would clean up the code also quite a bit.
> 

While it would solve "my" problem I'm not entirely sure about it being
a proper way to flag that they should have different default firmware.

One way would be to simply rely on a "firmware-name" property read from
DeviceTree (this was previously objected to, but we have that for
several bindings now).

Regards,
Bjorn

> > 
> > > +			return data->fw_name;
> > > +	}
> > > +
> > > +	return ERR_PTR(-ENOENT);
> > >   }
> > >   static int omap_rproc_get_boot_data(struct platform_device *pdev,
> > > @@ -334,7 +384,8 @@ static int omap_rproc_get_boot_data(struct platform_device *pdev,
> > >   	int ret;
> > >   	if (!of_device_is_compatible(np, "ti,omap4-dsp") &&
> > > -	    !of_device_is_compatible(np, "ti,omap5-dsp"))
> > > +	    !of_device_is_compatible(np, "ti,omap5-dsp") &&
> > > +	    !of_device_is_compatible(np, "ti,dra7-dsp"))
> > >   		return 0;
> > >   	oproc->boot_data = devm_kzalloc(&pdev->dev, sizeof(*oproc->boot_data),
> > > @@ -360,18 +411,27 @@ static int omap_rproc_get_boot_data(struct platform_device *pdev,
> > >   		return -EINVAL;
> > >   	}
> > > +	if (of_device_is_compatible(np, "ti,dra7-dsp"))
> > > +		oproc->boot_data->boot_reg_shift = 10;
> > 
> > Put this in omap_rproc_dev_data.
> 
> Yeah.
> 
> > 
> > > +
> > >   	return 0;
> > >   }
> > >   static int omap_rproc_of_get_internal_memories(struct platform_device *pdev,
> > >   					       struct rproc *rproc)
> > >   {
> > > -	static const char * const mem_names[] = {"l2ram"};
> > > +	static const char * const ipu_mem_names[] = {"l2ram"};
> > > +	static const char * const dra7_dsp_mem_names[] = {"l2ram", "l1pram",
> > > +								"l1dram"};
> > >   	struct device_node *np = pdev->dev.of_node;
> > >   	struct omap_rproc *oproc = rproc->priv;
> > >   	struct device *dev = &pdev->dev;
> > > +	const char * const *mem_names;
> > >   	struct resource *res;
> > >   	int num_mems;
> > > +	const __be32 *addrp;
> > > +	u32 l4_offset = 0;
> > > +	u64 size;
> > >   	int i;
> > >   	/* OMAP4 and OMAP5 DSPs do not have support for flat SRAM */
> > > @@ -379,7 +439,15 @@ static int omap_rproc_of_get_internal_memories(struct platform_device *pdev,
> > >   	    of_device_is_compatible(np, "ti,omap5-dsp"))
> > >   		return 0;
> > > -	num_mems = ARRAY_SIZE(mem_names);
> > > +	/* DRA7 DSPs have two additional SRAMs at L1 level */
> > > +	if (of_device_is_compatible(np, "ti,dra7-dsp")) {
> > > +		mem_names = dra7_dsp_mem_names;
> > > +		num_mems = ARRAY_SIZE(dra7_dsp_mem_names);
> > > +	} else {
> > > +		mem_names = ipu_mem_names;
> > > +		num_mems = ARRAY_SIZE(ipu_mem_names);
> > > +	}
> > > +
> > >   	oproc->mem = devm_kcalloc(dev, num_mems, sizeof(*oproc->mem),
> > >   				  GFP_KERNEL);
> > >   	if (!oproc->mem)
> > > @@ -395,7 +463,26 @@ static int omap_rproc_of_get_internal_memories(struct platform_device *pdev,
> > >   			return PTR_ERR(oproc->mem[i].cpu_addr);
> > >   		}
> > >   		oproc->mem[i].bus_addr = res->start;
> > > -		oproc->mem[i].dev_addr = OMAP_RPROC_IPU_L2RAM_DEV_ADDR;
> > > +
> > > +		/*
> > > +		 * The DSPs have the internal memories starting at a fixed
> > > +		 * offset of 0x800000 from address 0, and this corresponds to
> > > +		 * L2RAM. The L3 address view has the L2RAM bus address as the
> > > +		 * starting address for the IP, so the L2RAM memory region needs
> > > +		 * to be processed first, and the device addresses for each
> > > +		 * memory region can be computed using the relative offset
> > > +		 * from this base address.
> > > +		 */
> > > +		if (of_device_is_compatible(np, "ti,dra7-dsp") &&
> > 
> > Please don't use a ternary operator repeatedly, it makes the code hard
> > to follow.
> 
> Yeah this parts looks somewhat messy, let me try to fix that.
> 
> -Tero
> 
> > 
> > Regards,
> > Bjorn
> > 
> > > +		    !strcmp(mem_names[i], "l2ram")) {
> > > +			addrp = of_get_address(dev->of_node, i, &size, NULL);
> > > +			l4_offset = of_translate_address(dev->of_node, addrp);
> > > +		}
> > > +		oproc->mem[i].dev_addr =
> > > +			of_device_is_compatible(np, "ti,dra7-dsp") ?
> > > +				res->start - l4_offset +
> > > +				OMAP_RPROC_DSP_LOCAL_MEM_OFFSET :
> > > +				OMAP_RPROC_IPU_L2RAM_DEV_ADDR;
> > >   		oproc->mem[i].size = resource_size(res);
> > >   		dev_dbg(dev, "memory %8s: bus addr %pa size 0x%x va %p da 0x%x\n",
> > > -- 
> > > 2.17.1
> > > 
> > > --
> 
> --
> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki