Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH] watchdog: bcm7038_wdt: add big endian support
To:     Guenter Roeck <linux@roeck-us.net>,
        =?UTF-8?Q?=c3=81lvaro_Fern=c3=a1ndez_Rojas?= <noltari@gmail.com>
Cc:     Wim Van Sebroeck <wim@linux-watchdog.org>,
        linux-watchdog@vger.kernel.org, linux-kernel@vger.kernel.org
References: <9381ef9e-a569-9bcd-5546-a48922e4961d@roeck-us.net>
 <80DB1B7E-D719-4597-A2B7-7CAD592E1B19@gmail.com>
 <20210222222447.GA177866@roeck-us.net>
From:   Florian Fainelli <f.fainelli@gmail.com>
Message-ID: <339490a4-4ccd-0d83-b6d7-8732bf39608a@gmail.com>
Date:   Mon, 22 Feb 2021 19:41:32 -0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Firefox/78.0 Thunderbird/78.7.1
MIME-Version: 1.0
In-Reply-To: <20210222222447.GA177866@roeck-us.net>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 2/22/2021 2:24 PM, Guenter Roeck wrote:
> On Mon, Feb 22, 2021 at 10:48:09PM +0100, Álvaro Fernández Rojas wrote:
>> Hi Guenter,
>>
>>> El 22 feb 2021, a las 22:24, Guenter Roeck <linux@roeck-us.net> escribió:
>>>
>>> ﻿On 2/22/21 12:03 PM, Álvaro Fernández Rojas wrote:
>>>> bcm7038_wdt can be used on bmips (bcm63xx) devices too.
>>>>
>>> It might make sense to actually enable it for BCM63XX.
>>
>> bcm63xx SoCs are supported in bcm63xx and bmips.
>> bcm63xx doesn’t have device tree support, but bmips does and this watchdog is already enabled for bmips.
>>
> 
> Maybe add a note saying that this will only be supported for devicetree
> based systems.
> 
>>>
>>>> Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
>>>> ---
>>>> drivers/watchdog/bcm7038_wdt.c | 30 ++++++++++++++++++++++++------
>>>> 1 file changed, 24 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/drivers/watchdog/bcm7038_wdt.c b/drivers/watchdog/bcm7038_wdt.c
>>>> index 979caa18d3c8..62494da1ac57 100644
>>>> --- a/drivers/watchdog/bcm7038_wdt.c
>>>> +++ b/drivers/watchdog/bcm7038_wdt.c
>>>> @@ -34,6 +34,24 @@ struct bcm7038_watchdog {                                 
>>>>
>>>> static bool nowayout = WATCHDOG_NOWAYOUT;
>>>>
>>>> +static inline void bcm7038_wdt_write(unsigned long data, void __iomem *reg)
>>>> +{
>>>> +#ifdef CONFIG_CPU_BIG_ENDIAN
>>>> +    __raw_writel(data, reg);
>>>> +#else
>>>> +    writel(data, reg);
>>>> +#endif
>>>> +}
>>>> +
>>>> +static inline unsigned long bcm7038_wdt_read(void __iomem *reg)
>>>> +{
>>>> +#ifdef CONFIG_CPU_BIG_ENDIAN
>>>> +    return __raw_readl(reg);
>>>> +#else
>>>> +    return readl(reg);
>>>> +#endif
>>>> +}
>>>> +
>>>
>>> This needs further explanation. Why not just use __raw_writel() and
>>> __raw_readl() unconditionally ? Also, is it known for sure that,
>>> say, bmips_be_defconfig otherwise uses the wrong endianness
>>> (vs. bmips_stb_defconfig which is a little endian configuration) ?
>>
>> Because __raw_writel() doesn’t have memory barriers and writel() does.
>> Those configs use the correct endiannes, so I don’t know what you mean...
>>
> So are you saying that it already works with bmips_stb_defconfig 
> (because it is little endian), that bmips_stb_defconfig needs memory
> barriers, and that bmips_be_defconfig doesn't need memory barriers ?
> Odd, but I'll take you by your word. And other code does something
> similar, so I guess there must be a reason for it.

It would be so nice to copy people, and the author of that driver who
could give you such an answer. Neither bmips_be_defconfig nor
bmips_stb_defconfig require barrier because the bus interface towards
registers that is used on those SoCs is non-reordering non-buffered.

> 
> Anyway, after looking into that other code, please use something like
> 
>         if (IS_ENABLED(CONFIG_MIPS) && IS_ENABLED(CONFIG_CPU_BIG_ENDIAN))
>                 __raw_writel(value, reg);
>         else
>                 writel(value, reg);

Yes please.
-- 
Florian