Re: [PATCHv5 2/5] arm: socfpga: Enable OCRAM ECC on startup.

[Thor Thayer <tthayer-yzvPICuk2ABMcg4IHK0kFoH6Mc4MB0Vx@public.gmane.org>]

On 12/02/2014 09:11 AM, Mark Rutland wrote:
> On Wed, Nov 12, 2014 at 12:14:20AM +0000, tthayer-yzvPICuk2ABMcg4IHK0kFoH6Mc4MB0Vx@public.gmane.org wrote:
>> From: Thor Thayer <tthayer-yzvPICuk2ABMcg4IHK0kFoH6Mc4MB0Vx@public.gmane.org>
>>
>> This patch enables the ECC for On-Chip RAM on machine
>> startup.  The ECC has to be enabled before data is
>> is stored in memory otherwise the ECC will fail on
>> reads.
>>
>> Signed-off-by: Thor Thayer <tthayer-yzvPICuk2ABMcg4IHK0kFoH6Mc4MB0Vx@public.gmane.org>
>> ---
>> v2: Split OCRAM ECC portion separately. Addition of iounmap()
>> and reorganization of gen_pool_free. Remove defconfig from patch.
>>
>> v3/4: No change
>>
>> v5: Remove ocram.h, use io.h instead of clk-provider.h
>>      Check prop in correct place. Add ECC EN defines.
>> ---
<snip>

>> +
>> +void socfpga_init_ocram_ecc(void)
>> +{
>> +	struct device_node *np;
>> +	const __be32 *prop;
>
> Please don't use accessors which return raw __be32s in drivers,
> typically it's the wrong thing to do and leaves horrible bugs and/or
> quirks that are painful to fix up.

OK. Thanks.

>> +	u32 ocr_edac_addr, iram_addr, len;
>> +	void __iomem  *mapped_ocr_edac_addr;
>> +	size_t size;
>> +	struct gen_pool *gp;
>> +
>> +	np = of_find_compatible_node(NULL, NULL, "altr,ocram-edac");
>> +	if (!np) {
>> +		pr_err("SOCFPGA: Unable to find altr,ocram-edac in dtb\n");
>> +		return;
>> +	}
>> +
>> +	prop = of_get_property(np, "reg", &size);
>> +	if (!prop || size < sizeof(*prop)) {
>> +		pr_err("SOCFPGA: Unable to find OCRAM ECC mapping in dtb\n");
>> +		return;
>> +	}
>> +	ocr_edac_addr = be32_to_cpup(prop++);
>> +	len = be32_to_cpup(prop);
>
> Use of_iomap(np, 0). You don't seem to pass the address around, so that
> should be sufficient.
>
>> +
>> +	gp = of_get_named_gen_pool(np, "iram", 0);
>> +	if (!gp) {
>> +		pr_err("SOCFPGA: OCRAM cannot find gen pool\n");
>> +		return;
>> +	}
>> +
>> +	np = of_find_compatible_node(NULL, NULL, "mmio-sram");
>> +	if (!np) {
>> +		pr_err("SOCFPGA: Unable to find mmio-sram in dtb\n");
>> +		return;
>> +	}
>> +
>> +	/* Determine the OCRAM address and size */
>> +	prop = of_get_property(np, "reg", &size);
>> +	if (!prop || size < sizeof(*prop)) {
>> +		pr_err("SOCFPGA: Unable to find OCRAM mapping in dtb\n");
>> +		return;
>> +	}
>> +	iram_addr = be32_to_cpup(prop++);
>> +	len = be32_to_cpup(prop);
>
> This address is overwritten below. What's going on?
>

Thanks, leftovers from debugging. I will look for a different way of 
getting the information without the be32 stuff.

>> +
>> +	iram_addr = gen_pool_alloc(gp, len);
>> +	if (iram_addr == 0) {
>> +		pr_err("SOCFPGA: cannot alloc from gen pool\n");
>> +		return;
>> +	}
>> +
>> +	memset((void *)iram_addr, 0, len);
>
> How is the iram mapped? Is memset to it safe (e.g. it unaligned accesses
> are made)?
>

I will need to look at this further - good points. The entire iram 
should be used as a pool so I didn't expect anything unusual even though 
I just realized allocations are not contiguous.

>> +
>> +	gen_pool_free(gp, iram_addr, len);
>> +
>> +	mapped_ocr_edac_addr = ioremap(ocr_edac_addr, 4);
>> +	if (!mapped_ocr_edac_addr) {
>> +		pr_err("SOCFPGA: Unable to map OCRAM ecc regs.\n");
>> +		return;
>> +	}
>> +
>> +	/* Clear any pending OCRAM ECC interrupts, then enable ECC */
>> +	writel(ALTR_OCRAM_CLEAR_ECC, mapped_ocr_edac_addr);
>> +	writel(ALTR_OCRAM_ECC_EN, mapped_ocr_edac_addr);
>> +
>> +	iounmap(mapped_ocr_edac_addr);
>> +
>> +	pr_debug("SOCFPGA: Success Initializing OCRAM\n");
>> +}
>> diff --git a/arch/arm/mach-socfpga/socfpga.c b/arch/arm/mach-socfpga/socfpga.c
>> index 0954011..065d80d 100644
>> --- a/arch/arm/mach-socfpga/socfpga.c
>> +++ b/arch/arm/mach-socfpga/socfpga.c
>> @@ -100,6 +100,14 @@ static void socfpga_cyclone5_restart(enum reboot_mode mode, const char *cmd)
>>   	writel(temp, rst_manager_base_addr + SOCFPGA_RSTMGR_CTRL);
>>   }
>>
>> +static void __init socfpga_cyclone5_init(void)
>> +{
>> +	of_platform_populate(NULL, of_default_bus_match_table,
>> +			     NULL, NULL);
>> +	if (IS_ENABLED(CONFIG_EDAC_ALTERA_OCRAM))
>> +		socfpga_init_ocram_ecc();
>
> If it's safe to do this after probing everything else, why can't this be
> a normal device probe?
>
> Thanks,
> Mark.

Good point. Thank you.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html