mtd: maps: lantiq-flash: Check if the EBU endianness swap is enabled

The purpose of this patch is to enable the software address endianness
swapping only when the in SoC EBU endianness swapping is disabled.
To perform this check, I look at Bit 30 of the EBU_CON_0 register.
Actually, the driver expects that the in SoC swapping is disabled.
This is the case with current bootloaders shuch as U-boot. But,

I have a router which uses a proprietary bootloader which keeps
the in SoC swapping enabled. The SoC in this router is a vrx200 v1.2.
In this SoC version, I can keep the in SoC swapping without any problem.

Signed-off-by: Sebastien Decourriere <sebtx452@gmail.com>
---
 drivers/mtd/maps/lantiq-flash.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

Hi Sebastien,

thanks, comments inline

On 18/01/2017 13:10, Sebastien Decourriere wrote:
> The purpose of this patch is to enable the software address endianness
> swapping only when the in SoC EBU endianness swapping is disabled.
> To perform this check, I look at Bit 30 of the EBU_CON_0 register.
> Actually, the driver expects that the in SoC swapping is disabled.
> This is the case with current bootloaders shuch as U-boot. But,
> 
> I have a router which uses a proprietary bootloader which keeps
> the in SoC swapping enabled. The SoC in this router is a vrx200 v1.2.
> In this SoC version, I can keep the in SoC swapping without any problem.
> 
> Signed-off-by: Sebastien Decourriere <sebtx452@gmail.com>
> ---
>  drivers/mtd/maps/lantiq-flash.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/mtd/maps/lantiq-flash.c b/drivers/mtd/maps/lantiq-flash.c
> index c8febb3..a091efa 100644
> --- a/drivers/mtd/maps/lantiq-flash.c
> +++ b/drivers/mtd/maps/lantiq-flash.c
> @@ -151,6 +151,11 @@ ltq_mtd_probe(struct platform_device *pdev)
>  	ltq_mtd->map->copy_to = ltq_copy_to;
>  
>  	ltq_mtd->map->map_priv_1 = LTQ_NOR_PROBING;

this line should be dropped

> +	/* We swap the addresses only if the EBU endianness swap is disabled */
> +	if (ltq_ebu_r32(LTQ_EBU_BUSCON0) & BIT(30))

add a define for BIT(30) please and we should really check if this a
v1.2 or newer. if my memory is correct this was a silicon bug inside
v1.0 and v1.1

	John

> +		ltq_mtd->map[i].map_priv_1 = LTQ_NOR_NORMAL;
> +	else
> +		ltq_mtd->map[i].map_priv_1 = LTQ_NOR_PROBING;
>  	ltq_mtd->mtd = do_map_probe("cfi_probe", ltq_mtd->map);
>  	ltq_mtd->map->map_priv_1 = LTQ_NOR_NORMAL;
>  
> @@ -163,8 +168,11 @@ ltq_mtd_probe(struct platform_device *pdev)
>  	mtd_set_of_node(ltq_mtd->mtd, pdev->dev.of_node);
>  
>  	cfi = ltq_mtd->map->fldrv_priv;
> -	cfi->addr_unlock1 ^= 1;
> -	cfi->addr_unlock2 ^= 1;
> +	/* We swap the addresses only if the EBU endianness swap is disabled */
> +	if (!(ltq_ebu_r32(LTQ_EBU_BUSCON0) & BIT(30))) {
> +		cfi->addr_unlock1 ^= 1;
> +		cfi->addr_unlock2 ^= 1;
> +	}
>  
>  	err = mtd_device_register(ltq_mtd->mtd, NULL, 0);
>  	if (err) {
>

Hi John,


>>
>>       ltq_mtd->map->map_priv_1 = LTQ_NOR_PROBING;
>
> this line should be dropped

Yes, I noticed this mistake. I fixed it but forgotten to add thix fix
to my commit -_-

>> +     /* We swap the addresses only if the EBU endianness swap is disabled */
>> +     if (ltq_ebu_r32(LTQ_EBU_BUSCON0) & BIT(30))
>
> add a define for BIT(30) please and we should really check if this a
> v1.2 or newer. if my memory is correct this was a silicon bug inside
> v1.0 and v1.1

In my kernel (OpenWrt) boot log I have :

[    0.000000] Linux version 4.4.14 (qa@serveurQA) (gcc version 5.3.0
(OpenWrt GCC 5.3.0 50020) ) #150 SMP Tue Jan 17 09:18:24 UTC 2017
[    0.000000] SoC: xRX200 rev 1.2
[    0.000000] bootconsole [early0] enabled
[    0.000000] CPU0 revision is: 00019556 (MIPS 34Kc)

It would be better to ensure that the SoC version is >= 1.2 (as this
bug was fixed in this version).

You can get more informations in my OpenWrt pull request :

https://github.com/openwrt/openwrt/pull/321


Regards,

Sebastien.

On 18/01/2017 14:48, Seb wrote:
> Hi John,
> 
> 
>>>
>>>       ltq_mtd->map->map_priv_1 = LTQ_NOR_PROBING;
>>
>> this line should be dropped
> 
> Yes, I noticed this mistake. I fixed it but forgotten to add thix fix
> to my commit -_-
> 
>>> +     /* We swap the addresses only if the EBU endianness swap is disabled */
>>> +     if (ltq_ebu_r32(LTQ_EBU_BUSCON0) & BIT(30))
>>
>> add a define for BIT(30) please and we should really check if this a
>> v1.2 or newer. if my memory is correct this was a silicon bug inside
>> v1.0 and v1.1
> 
> In my kernel (OpenWrt) boot log I have :
> 
> [    0.000000] Linux version 4.4.14 (qa@serveurQA) (gcc version 5.3.0
> (OpenWrt GCC 5.3.0 50020) ) #150 SMP Tue Jan 17 09:18:24 UTC 2017
> [    0.000000] SoC: xRX200 rev 1.2
> [    0.000000] bootconsole [early0] enabled
> [    0.000000] CPU0 revision is: 00019556 (MIPS 34Kc)
> 
> It would be better to ensure that the SoC version is >= 1.2 (as this
> bug was fixed in this version).

the bug is also fixed on the 300 and 500 series chips, so we would want
to check for that aswell.

	John




> You can get more informations in my OpenWrt pull request :
> 
> https://github.com/openwrt/openwrt/pull/321
> 
> 
> Regards,
> 
> Sebastien.
>

> the bug is also fixed on the 300 and 500 series chips, so we would want
> to check for that aswell.
>

I added this to the begin of "ltq_mtd_probe" function:

bool mtd_addr_swap;

        if ( ltq_ebu_r32(LTQ_EBU_BUSCON0) & EBU_ENDIAN_SWAP ) {
                switch (ltq_soc_type()) {
                        case SOC_TYPE_VR9_2 :
                        case SOC_TYPE_AR10 :
                                mtd_addr_swap = false;
                                break;
                        default :
                                mtd_addr_swap = true;
                }
        } else mtd_addr_swap = true;


It's easier to add other architectures as needed. I tried it on my
OpenWrt release, and got it working fine. Is it correct in this way ?