lkml.org 
[lkml]   [2019]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH v6 1/4] can: m_can: Create a m_can platform framework
Date
Wolfgang

On 3/4/19 12:13 PM, Wolfgang Grandegger wrote:
>
>
> Am 04.03.19 um 18:22 schrieb Dan Murphy:
>> Wolfgang
>>
>> On 3/4/19 10:56 AM, Wolfgang Grandegger wrote:
>>> Hello Dan,
>>>
>>> the series already looks quite good. I still realized a few (minor)
>>> issues while browsing the patch/code...
>>>
>>
>> Thanks for the review. It is getting there.
>>
>>> Am 01.03.19 um 19:50 schrieb Dan Murphy:
>>>> Create a m_can platform framework that peripherial
>>>> devices can register to and use common code and register sets.
>>>> The peripherial devices may provide read/write and configuration
>>>> support of the IP.
>>>>
>>>> Signed-off-by: Dan Murphy <dmurphy@ti.com>
>>>> ---
>>>>
>>>> v6 - Squashed platform patch to this patch for bissectablity, fixed coding style
>>>> issues, updated Kconfig help, placed mcan reg offsets back into c file, renamed
>>>> priv->skb to priv->tx_skb and cleared perp interrupts at ISR start -
>>>> Patch 1 comments - https://lore.kernel.org/patchwork/patch/1042446/
>>>> Patch 2 comments - https://lore.kernel.org/patchwork/patch/1042442/
>>>>
>>>> drivers/net/can/m_can/Kconfig | 13 +-
>>>> drivers/net/can/m_can/Makefile | 1 +
>>>> drivers/net/can/m_can/m_can.c | 702 +++++++++++++------------
>>>> drivers/net/can/m_can/m_can.h | 110 ++++
>>>> drivers/net/can/m_can/m_can_platform.c | 198 +++++++
>>>> 5 files changed, 681 insertions(+), 343 deletions(-)
>>>> create mode 100644 drivers/net/can/m_can/m_can.h
>>>> create mode 100644 drivers/net/can/m_can/m_can_platform.c
>>>>
>>>> diff --git a/drivers/net/can/m_can/Kconfig b/drivers/net/can/m_can/Kconfig
>>>> index 04f20dd39007..f7119fd72df4 100644
>>>> --- a/drivers/net/can/m_can/Kconfig
>>>> +++ b/drivers/net/can/m_can/Kconfig
>>>> @@ -1,5 +1,14 @@
>>>> config CAN_M_CAN
>>>> + tristate "Bosch M_CAN support"
>>>> + ---help---
>>>> + Say Y here if you want support for Bosch M_CAN controller framework.
>>>> + This is common support for devices that embed the Bosch M_CAN IP.
>>>> +
>>>> +config CAN_M_CAN_PLATFORM
>>>> + tristate "Bosch M_CAN support for io-mapped devices"
>>>> depends on HAS_IOMEM
>>>> - tristate "Bosch M_CAN devices"
>>>> + depends on CAN_M_CAN
>>>> ---help---
>>>> - Say Y here if you want to support for Bosch M_CAN controller.
>>>> + Say Y here if you want support for IO Mapped Bosch M_CAN controller.
>>>> + This support is for devices that have the Bosch M_CAN controller
>>>> + IP embedded into the device and the IP is IO Mapped to the processor.
>>>> diff --git a/drivers/net/can/m_can/Makefile b/drivers/net/can/m_can/Makefile
>>>> index 8bbd7f24f5be..057bbcdb3c74 100644
>>>> --- a/drivers/net/can/m_can/Makefile
>>>> +++ b/drivers/net/can/m_can/Makefile
>>>> @@ -3,3 +3,4 @@
>>>> #
>>>>
>>>> obj-$(CONFIG_CAN_M_CAN) += m_can.o
>>>> +obj-$(CONFIG_CAN_M_CAN_PLATFORM) += m_can_platform.o
>>>> diff --git a/drivers/net/can/m_can/m_can.c b/drivers/net/can/m_can/m_can.c
>>>> index 9b449400376b..b37d0886f9cb 100644
>>>> --- a/drivers/net/can/m_can/m_can.c
>>>> +++ b/drivers/net/can/m_can/m_can.c
>
> ...snip...
>
>>>> @@ -1451,7 +1459,7 @@ static netdev_tx_t m_can_start_xmit(struct sk_buff *skb,
>>>> netif_stop_queue(dev);
>>>> netdev_warn(dev,
>>>> "TX queue active although FIFO is full.");
>>>> - return NETDEV_TX_BUSY;
>>>> + return;
>>>
>>> m_can_start_xmit() doesn't return NETDEV_TX_BUSY but NETDEV_TX_OK and
>>> the queue is stopped! Also the skb is not freed! The code states
>>> "/* This shouldn't happen */" but then it just prints a warning. Did
>>> you see that message?
>>>
>>
>> No I have not seen this warning but I will re-check to be sure.
>
> If we don't return NETDEV_TX_BUSY but NETDEV_TX_OK, we must handle it
> differently.
>

OK. I see the diff between the perp and io mapped. I will update the code
appropriately.


> ...snip...
>
>>>> +struct m_can_priv;
>>>> +struct m_can_ops {
>>>> + /* Device specific call backs */
>>>> + int (*clr_dev_interrupts)(struct m_can_priv *m_can_class);
>>>
>>> Why not just "clear_interrupt"... to be consistant with the names below.
>>
>> I wanted to be clear in the M_CAN code that these are device interrupts and not M_CAN interrupts.
>>
>> I can change it to clear_interrupt if you think it makes more sense.
>
> Well, like for "read_reg" etc, I think it's clear that it's a
> device-specific function/ops:
>
> cdev->read_reg
> cdev->clear_interrupt
>

OK

>>>> + u32 (*read_reg)(struct m_can_priv *m_can_class, int reg);
>>>> + int (*write_reg)(struct m_can_priv *m_can_class, int reg, int val);
>>>> + u32 (*read_fifo)(struct m_can_priv *m_can_class, int addr_offset);
>>>> + int (*write_fifo)(struct m_can_priv *m_can_class, int addr_offset,
>>>> + int val);
>>>> + int (*device_init)(struct m_can_priv *m_can_class);
>
> And the same here:
>
> cdev->init
>

OK

>>>> + int pm_clock_support;
>>>
>>> A "bool" would be more appropriate, I think.
>>
>>
>> I was abiding by this checkpatch warning I got on the is_peripherial.
>>
>> CHECK: Avoid using bool structure members because of possible alignment issues - see: https://lkml.org/lkml/2017/11/21/384
>> #94: FILE: drivers/net/can/m_can/m_can.h:94:
>> + bool is_peripherial;
>>
>
> Ah, right! I was also surprised to get that warning. The kernel is full
> of bool's, but well, we should make "checkpatch" happy (and Linus).
>
> Wolfgang
>


--
------------------
Dan Murphy

\
 
 \ /
  Last update: 2019-03-04 20:13    [W:0.052 / U:0.028 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site