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;
Sender: Guenter Roeck <groeck7@gmail.com>
To:     David Miller <davem@davemloft.net>
Cc:     ajk@comnets.uni-bremen.de, kuba@kernel.org,
        linux-hams@vger.kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20210909035743.1247042-1-linux@roeck-us.net>
 <20210909.123442.1648633411296774237.davem@davemloft.net>
 <751f5079-2da1-187e-573c-d7d2d6743bbf@roeck-us.net>
 <20210909.162721.1267526781289116670.davem@davemloft.net>
From:   Guenter Roeck <linux@roeck-us.net>
Subject: Re: [PATCH] net: 6pack: Fix tx timeout and slot time
Message-ID: <aaa2d05e-cdbd-f80c-d85f-1ee92e7a7946@roeck-us.net>
Date:   Thu, 9 Sep 2021 09:44:03 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <20210909.162721.1267526781289116670.davem@davemloft.net>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 9/9/21 8:27 AM, David Miller wrote:
> From: Guenter Roeck <linux@roeck-us.net>
> Date: Thu, 9 Sep 2021 07:53:29 -0700
> 
>> On 9/9/21 4:34 AM, David Miller wrote:
>>> From: Guenter Roeck <linux@roeck-us.net>
>>> Date: Wed,  8 Sep 2021 20:57:43 -0700
>>>
>>>> tx timeout and slot time are currently specified in units of HZ.
>>>> On Alpha, HZ is defined as 1024. When building alpha:allmodconfig,
>>>> this results in the following error message.
>>>>
>>>> drivers/net/hamradio/6pack.c: In function 'sixpack_open':
>>>> drivers/net/hamradio/6pack.c:71:41: error:
>>>> 	unsigned conversion from 'int' to 'unsigned char'
>>>> 	changes value from '256' to '0'
>>>>
>>>> In the 6PACK protocol, tx timeout is specified in units of 10 ms
>>>> and transmitted over the wire. Defining a value dependent on HZ
>>>> doesn't really make sense. Assume that the intent was to set tx
>>>> timeout and slot time based on a HZ value of 100 and use constants
>>>> instead of values depending on HZ for SIXP_TXDELAY and SIXP_SLOTTIME.
>>>>
>>>> Signed-off-by: Guenter Roeck <linux@roeck-us.net>
>>>> ---
>>>> No idea if this is correct or even makes sense. Compile tested only.
>>> These are timer offsets so they have to me HZ based.  Better to make
>>> the
>>> structure members unsigned long, I think.
>>>
>>
>> Hmm, ok. Both tx_delay and slottime are updated in sp_encaps(),
>> though,
>> from data in the transmit buffer. The KISS protocol description states
>> that the values are in units of 10ms; that is where my assumption
>> came from.
> 
> They are ms and must be converted using to HZ in order to use the values as timer offsets.
> 
> The values are perfectly fine, the types used to store them need to be fixed.
> 
>> Anyway, I am inclined to just mark the protocol as dependent on
>> !ALPHA. Would you accept that ?
> 
> No, fix this properly.  Make the unsigfned char members be unsigned long.
> 
> Why do you not want to fix it this way?
> 

All I want is to get alpha:allmodconfig to compile, nothing else,
but at the same time I don't want to introduce new bugs.

If I make tx_delay unsigned long, it is still passed to encode_sixpack()
as parameter, and that parameter is declared unsigned char and put into
a byte sized buffer. If the value is 0x100, it will then be truncated to
0x0, only that is then done silently without generating a compile error.
Sure, that fixes the compile error, but not the underlying issue. Maybe
that truncation is perfectly fine, but I don't understand the code well
enough to make that call.

Never mind, I'll let someone else who has a better understanding of the code
deal with this.

Guenter