From: Wolfgang Grandegger <wg@grandegger.com>
To: Dan Murphy <dmurphy@ti.com>, mkl@pengutronix.de, davem@davemloft.net
Cc: linux-can@vger.kernel.org, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v7 1/4] can: m_can: Create a m_can platform framework
Date: Fri, 8 Mar 2019 15:41:39 +0100 [thread overview]
Message-ID: <cea76d80-d85c-2963-2745-4509450e2ea9@grandegger.com> (raw)
In-Reply-To: <bf2613ae-67f6-661d-0faa-786308c249a0@grandegger.com>
Hello Dan,
thinking more about it...
Am 08.03.19 um 14:29 schrieb Wolfgang Grandegger:
> Hello Dan,
>
> Am 08.03.19 um 13:44 schrieb Dan Murphy:
>> Wolfgang
>>
>> On 3/8/19 4:10 AM, Wolfgang Grandegger wrote:
>>> Hallo Dan,
>>>
>>> Am 05.03.19 um 16:52 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>
>>>> ---
>>>>
>>>>
>>>> v7 - Fixed remaining new checkpatch issues, removed CSR setting, fixed tx hard
>>>> start function to return tx_busy, and renamed device callbacks - https://lore.kernel.org/patchwork/patch/1047220/
>>>>
>>>> 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 | 700 +++++++++++++------------
>>>> drivers/net/can/m_can/m_can.h | 110 ++++
>>>> drivers/net/can/m_can/m_can_platform.c | 202 +++++++
>>>> 5 files changed, 682 insertions(+), 344 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..a60278d94126 100644
>>>> --- a/drivers/net/can/m_can/m_can.c
>>>> +++ b/drivers/net/can/m_can/m_can.c
>>>
>>> ... snip...
>>>
>>>> +static netdev_tx_t m_can_start_xmit(struct sk_buff *skb,
>>>> + struct net_device *dev)
>>>> +{
>>>> + struct m_can_priv *priv = netdev_priv(dev);
>>>> +
>>>> + if (can_dropped_invalid_skb(dev, skb))
>>>> + return NETDEV_TX_OK;
>>>> +
>>>> + if (priv->is_peripherial) {
>>>> + if (priv->tx_skb) {
>>>> + netdev_err(dev, "hard_xmit called while tx busy\n");
>>>> + return NETDEV_TX_BUSY;
>>>> + }
>>>
>>> The problem with that approach is, that the upper layer will try to
>>> resubmit the current "skb" but not the previous "tx_skb". And the
>>> previous "tx_skb" has not been freed yet. I would just drop and free the
>>> skb and return NETDEV_TX_OK in m_can_tx_handler() for peripheral devices
>>> (like can_dropped_invalid_skb() does).
>>>
>>
>> OK.
>>
>> So would this also be a bug in the hi3110 and mcp251x drivers (line 521) as well because besides checking tx_length
>> this is how these drivers are written.
>
> This is different. When entering the "start_xmit" routine, the previous
> TX is still in progress. It will (hopefully) complete soon. Therefore
> returning NETDEV_TX_BUSY is OK. The "start_xmit" routine will be
> recalled soon with the same "skb". That scenario should/could also not
> happen.
In principle, this also applies to the m_can peripheral devices. If
tx_skb is not NULL, the TX is still in progress and returning
NETDEV_TX_BUSY is just fine.
>
> In contrast, in "m_can_tx_handler()", the skb could not be handled
> because the FIFO is full. The "start_xmit" routine for peripheral
> devices for that skb already returned NETDEV_TX_OK. Therefore the only
> meaningful action is to drop the skb. Also this error should not happen
> and if, something is going really wrong. Therefore I think, a
> WARN_ONCE() would be even more appropriate. But that should be a
> separate patch.
But that's a different issue/error. The tx_skb cannot be processed in
"m_can_tx_handler()". Either we drop it or we re-queue it (retry later).
>>
>> In addition in the peripheral context the work queue does not report up to the upper layer the status.
>> Again the hi3110 and mcp251x drivers are written this way.
>>
>> The only issue I see here is that the dropped and invalid check needs to come after the tx_skb check.
>
> See above.
Wolfgang.
next prev parent reply other threads:[~2019-03-08 14:41 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-03-05 15:52 [PATCH v7 1/4] can: m_can: Create a m_can platform framework Dan Murphy
2019-03-05 15:52 ` [PATCH v7 2/4] can: m_can: Rename m_can_priv to m_can_classdev Dan Murphy
2019-03-08 10:25 ` Wolfgang Grandegger
2019-03-08 12:33 ` Dan Murphy
2019-03-05 15:52 ` [PATCH v7 3/4] dt-bindings: can: tcan4x5x: Add DT bindings for TCAN4x5X driver Dan Murphy
2019-03-05 15:52 ` [PATCH v7 4/4] can: tcan4x5x: Add tcan4x5x driver to the kernel Dan Murphy
2019-03-08 10:10 ` [PATCH v7 1/4] can: m_can: Create a m_can platform framework Wolfgang Grandegger
2019-03-08 12:44 ` Dan Murphy
2019-03-08 13:29 ` Wolfgang Grandegger
2019-03-08 14:41 ` Wolfgang Grandegger [this message]
2019-03-08 15:48 ` Dan Murphy
2019-03-08 17:08 ` Wolfgang Grandegger
2019-03-08 17:25 ` Dan Murphy
2019-03-08 17:40 ` Wolfgang Grandegger
2019-03-08 17:52 ` Dan Murphy
2019-03-08 18:06 ` Wolfgang Grandegger
2019-03-08 20:36 ` Dan Murphy
2019-03-11 11:19 ` Wolfgang Grandegger
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=cea76d80-d85c-2963-2745-4509450e2ea9@grandegger.com \
--to=wg@grandegger.com \
--cc=davem@davemloft.net \
--cc=dmurphy@ti.com \
--cc=linux-can@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mkl@pengutronix.de \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).