Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <7ec86c10-3cef-d806-2e27-c99a12704288@bitwise.fi>
Date:   Mon, 30 May 2022 13:55:41 +0300
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.9.0
Subject: Re: [PATCH 09/12] can: kvaser_usb_leaf: Fix silently failing bus
 params setup
Content-Language: en-US
To:     Jimmy Assarsson <extja@kvaser.com>
Cc:     linux-can@vger.kernel.org, Marc Kleine-Budde <mkl@pengutronix.de>,
        linux-kernel@vger.kernel.org
References: <20220516134748.3724796-1-anssi.hannula@bitwise.fi>
 <20220516134748.3724796-10-anssi.hannula@bitwise.fi>
 <39946f68-ac16-9412-fd5e-47e10b84ffbf@kvaser.com>
From:   Anssi Hannula <anssi.hannula@bitwise.fi>
In-Reply-To: <39946f68-ac16-9412-fd5e-47e10b84ffbf@kvaser.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 28.5.2022 10.37, Jimmy Assarsson wrote:
> On 5/16/22 15:47, Anssi Hannula wrote:
>> The device may reject bus params with cmd 45, which is an unhandled
>> error event that does not contain any channel reference.
>>
>> In such cases set_bittiming() callback returns 0 so upper levels will
>> think setting bitrate succeeded and proceed with starting the interface
>> with wrong parameters, causing bus errors (the kernel log will have
>> "Unhandled command (45)", though).
>>
>> Fix that by verifying the bus params took hold by reading them back.
>>
>> Also, add a handler for cmd 45 that simply prints out the error.
>>
>> This condition can be triggered on 0bfd:0124 Kvaser Mini PCI Express
>> 2xHS FW 4.18.778 by trying to set bitrate 1300000 and on 0bfd:0124
>> Kvaser Mini PCI Express 2xHS FW 4.18.778 by trying to set bitrate
>> 1000000.
> The device will send cmd CMD_ERROR_EVENT (45) when invalid data is
> received from the driver and should never occur. CMD_ERROR_EVENT
> indicates a misbehaving driver.
>
> For the Kvaser Mini PCI Express 2xHS case, the problem is settings
> resulting in prescaler equal to 1 are not accepted. This is a bug in the
> firmware and will be fixed in our next release.
>
> And for Kvaser Memorator Professional HS/HS, the can_bittiming_const
> limits are not correct.
> I'll have to look into this a bit more. I'm pretty sure we endup with
> three different can_bittiming_const in kvaser_usb_leaf, depedning on
> firmware/microcontroller.
>
> I created new patches for the CMD_ERROR_EVENT, based on your patch.
> See at end of this mail.
>
> I prefere if we can avoid the paramter readback and
> kvaser_usb_setup_rx_urbs() in kvaser_usb_leaf_set_bittiming().
> With correct can_bittiming_const limits this should not be an issue.
> On the otherhand, since there are existing devices with a bug that will
> reject parameters that are within the correct limits, and the result
> silently failing calls, I cannot see any alternative...
>
> If it ok with you I'll create a new patch based on your readback fix,
> and also implement this in kvaser_usb_hydra?

All OK for me.

-- 
Anssi Hannula / Bitwise Oy
+358 503803997