mbox series

[0/5] watchdog: eiois200_wdt: Add EIO-IS200 Watchdog Driver

Message ID cover.1696495372.git.advantech.susiteam@gmail.com (mailing list archive)
Headers show
Series watchdog: eiois200_wdt: Add EIO-IS200 Watchdog Driver | expand

Message

Wenkai Oct. 5, 2023, 8:51 a.m. UTC
From: Wenkai <advantech.susiteam@gmail.com>

This patch series aims to add support for the Advantech EIO-IS200
Embedded Controller's watchdog timer to the Linux kernel. The EIO-IS200
is a widely used embedded controller, and this series introduces a
native driver for its watchdog timer functionality within the Linux
ecosystem.

Driver Features:
- Complete support for the Advantech EIO-IS200 Embedded Controller's
  hardware watchdog timer.
- Seamless integration with the Linux Watchdog framework, enabling
  standard watchdog functionality.
- Flexible configuration options for watchdog timeout and event types.
- Module parameters for setting default timeout and event type.
- The EIO-IS200 can select special event output pin such as PWRBTN
  (Power button),SCI (ACPI System Control Interrupt), IRQ, and GPIO

Driver Dependencies:
- The driver relies on the Linux Multi-Function Device (MFD) framework
  and related dependencies.
- Assumption of the presence of the Advantech eiois200_core MFD core
  driver.

Wenkai (5):
  watchdog: eiois200_wdt: Constructing Advantech EIO-IS200 watchdog
			  driver
  watchdog: eiois200_wdt: Add PMC support with eiois200_core.
  watchdog: eiois200_wdt: Implement basic watchdog functionalities
  watchdog: eiois200_wdt: Enhanced watchdog functionality and pretimeout
  watchdog: eiois200_wdt: Enhanced IRQ trigger behavior

 drivers/watchdog/Kconfig        |  14 +
 drivers/watchdog/Makefile       |   1 +
 drivers/watchdog/eiois200_wdt.c | 658 ++++++++++++++++++++++++++++++++
 3 files changed, 673 insertions(+)
 create mode 100644 drivers/watchdog/eiois200_wdt.c


base-commit: 9aab92bc3a8922d4b2e24d10271dfe3034cbf5c2

Comments

Guenter Roeck Oct. 6, 2023, 3:02 a.m. UTC | #1
On Thu, Oct 05, 2023 at 04:51:18PM +0800, advantech.susiteam@gmail.com wrote:
> From: Wenkai <advantech.susiteam@gmail.com>
> 
> This patch series aims to add support for the Advantech EIO-IS200
> Embedded Controller's watchdog timer to the Linux kernel. The EIO-IS200
> is a widely used embedded controller, and this series introduces a
> native driver for its watchdog timer functionality within the Linux
> ecosystem.
> 
I am not going to review this patch series. This is just one watchdog driver.
One patch is sufficient.

Guenter
Wenkai Oct. 6, 2023, 9:27 a.m. UTC | #2
Guenter Roeck 於 10/6/2023 11:02 AM 寫道:
> On Thu, Oct 05, 2023 at 04:51:18PM +0800, advantech.susiteam@gmail.com wrote:
>> From: Wenkai <advantech.susiteam@gmail.com>
>>
>> This patch series aims to add support for the Advantech EIO-IS200
>> Embedded Controller's watchdog timer to the Linux kernel. The EIO-IS200
>> is a widely used embedded controller, and this series introduces a
>> native driver for its watchdog timer functionality within the Linux
>> ecosystem.
>>
> I am not going to review this patch series. This is just ne watchdog driver.
> One patch is sufficient.
>
> Guenter
Hi Guenter,

Advantech's EIO-IS200 watchdog supports 5 output pins: RESET, Power
Button, SCI, IRQ, and GPIO. The most traditional scenario is that the
Pretimeout triggers IRQ, and the timeout triggers RESET.

However, unfortunately, for industrial usages, there are various use
cases, which require certain mechanisms and logic to manage which signal
is output when Pretimeout and timeout expire. I am concerned that
consolidating all these features into a single patch for upstream may
lead to confusion and make the source code less readable and
understandable.

Therefore, I have divided the implementation into 5 separate patches,
aiming to make the code more comprehensible and acceptable. If it's
acceptable to you, I am more than willing to provide a single patch as
per your preference.

I would also like to note that this watchdog driver is part of the
EIO-IS200 MFD (Multi-Function Device) driver family. It serves as one
of the child-drivers of the drivers/mfd/eiois200_core core driver. It's
important to mention that without the presence of
drivers/mfd/eiois200_core, this child-driver eiois200_wdt cannot be
compiled or used.

Should we wait until the core driver (drivers/mfd/eiois200_core) is
successfully incorporated before upstreaming the watchdog child-driver,
or would you prefer to proceed with these patches independently?

Best regards,
Wenkai
Guenter Roeck Oct. 6, 2023, 2:16 p.m. UTC | #3
On Fri, Oct 06, 2023 at 05:27:48PM +0800, Wenkai wrote:
> 
> 
> Guenter Roeck 於 10/6/2023 11:02 AM 寫道:
> > On Thu, Oct 05, 2023 at 04:51:18PM +0800, advantech.susiteam@gmail.com wrote:
> > > From: Wenkai <advantech.susiteam@gmail.com>
> > > 
> > > This patch series aims to add support for the Advantech EIO-IS200
> > > Embedded Controller's watchdog timer to the Linux kernel. The EIO-IS200
> > > is a widely used embedded controller, and this series introduces a
> > > native driver for its watchdog timer functionality within the Linux
> > > ecosystem.
> > > 
> > I am not going to review this patch series. This is just ne watchdog driver.
> > One patch is sufficient.
> > 
> > Guenter
> Hi Guenter,
> 
> Advantech's EIO-IS200 watchdog supports 5 output pins: RESET, Power
> Button, SCI, IRQ, and GPIO. The most traditional scenario is that the
> Pretimeout triggers IRQ, and the timeout triggers RESET.
> 
> However, unfortunately, for industrial usages, there are various use
> cases, which require certain mechanisms and logic to manage which signal
> is output when Pretimeout and timeout expire. I am concerned that
> consolidating all these features into a single patch for upstream may
> lead to confusion and make the source code less readable and
> understandable.
> 

The 1st patch in your series doesn't even compile. I don't call that
understandable.

Oh, it fails to compile because you include a non-existing file from
../mfd directly and because you select a non-existing configuration option
instead of depending on it.

None of those is even remotely acceptable. Are you seriously sending me
a series of patches that don't even build to review ?

> Therefore, I have divided the implementation into 5 separate patches,
> aiming to make the code more comprehensible and acceptable. If it's
> acceptable to you, I am more than willing to provide a single patch as
> per your preference.
> 
Frankly, your series is one more nail in the coffin. I am now seriously
considering to resign as co-maintainer of the watchdog subsystem.

Guenter
Wenkai Oct. 11, 2023, 4:08 a.m. UTC | #4
Guenter Roeck 於 10/6/2023 10:16 PM 寫道:
> On Fri, Oct 06, 2023 at 05:27:48PM +0800, Wenkai wrote:
>>
>> Guenter Roeck 於 10/6/2023 11:02 AM 寫道:
>>> On Thu, Oct 05, 2023 at 04:51:18PM +0800, advantech.susiteam@gmail.com wrote:
>>>> From: Wenkai <advantech.susiteam@gmail.com>
>>>>
>>>> This patch series aims to add support for the Advantech EIO-IS200
>>>> Embedded Controller's watchdog timer to the Linux kernel. The EIO-IS200
>>>> is a widely used embedded controller, and this series introduces a
>>>> native driver for its watchdog timer functionality within the Linux
>>>> ecosystem.
>>>>
>>> I am not going to review this patch series. This is just ne watchdog driver.
>>> One patch is sufficient.
>>>
>>> Guenter
>> Hi Guenter,
>>
>> Advantech's EIO-IS200 watchdog supports 5 output pins: RESET, Power
>> Button, SCI, IRQ, and GPIO. The most traditional scenario is that the
>> Pretimeout triggers IRQ, and the timeout triggers RESET.
>>
>> However, unfortunately, for industrial usages, there are various use
>> cases, which require certain mechanisms and logic to manage which signal
>> is output when Pretimeout and timeout expire. I am concerned that
>> consolidating all these features into a single patch for upstream may
>> lead to confusion and make the source code less readable and
>> understandable.
>>
> The 1st patch in your series doesn't even compile. I don't call that
> understandable.
>
> Oh, it fails to compile because you include a non-existing file from
> ../mfd directly and because you select a non-existing configuration option
> instead of depending on it.
>
> None of those is even remotely acceptable. Are you seriously sending me
> a series of patches that don't even build to review ?

I understand that the patches don't meet the expected quality standards.
The compile issue is due to my MFD core driver, which is currently under
review and has not been merged yet.

I would also like to seek your advice on how to best proceed with the
sub-drivers like the watchdog driver. Should I wait for my core MFD
driver to be successfully merged before submitting the sub-drivers, or
let Jones Lee review my core MFD driver and all its sub-drivers, or is
there another approach that you recommend?
>> Therefore, I have divided the implementation into 5 separate patches,
>> aiming to make the code more comprehensible and acceptable. If it's
>> acceptable to you, I am more than willing to provide a single patch as
>> per your preference.
>>
> Frankly, your series is one more nail in the coffin. I am now seriously
> considering to resign as co-maintainer of the watchdog subsystem.
>
> Guenter

I also appreciate your patience, dedication, and valuable contributions
to the Linux community. Your longstanding efforts and expertise are
commendable and have been instrumental in advancing the Linux ecosystem.
I understand that upstream review can be a meticulous and vital, albeit
thankless, task. I don't want my actions to cause any inconvenience or
distress, especially to someone as esteemed as you are in the Linux
community. Your insights and guidance are incredibly valuable to all of
us.

Once again, thank you for your understanding, and I am committed to
delivering high-quality code for your review.

Best regards,
Wenkai
Guenter Roeck Oct. 11, 2023, 3:05 p.m. UTC | #5
On Wed, Oct 11, 2023 at 12:08:57PM +0800, Wenkai wrote:
> 
> I understand that the patches don't meet the expected quality standards.
> The compile issue is due to my MFD core driver, which is currently under
> review and has not been merged yet.
> 
> I would also like to seek your advice on how to best proceed with the
> sub-drivers like the watchdog driver. Should I wait for my core MFD
> driver to be successfully merged before submitting the sub-drivers, or
> let Jones Lee review my core MFD driver and all its sub-drivers, or is
> there another approach that you recommend?

If the sub-drivers depend on the mfd driver, at least provide a reference
to the patch or patch series introducing that driver. Either case, a direct
include from "../mfd" is simply unacceptable. include/linux/mfd/ does exist
for a reason, after all.

I don't know the best solution for reviewing all the drivers. I didn't
(and do not plan to) look into the driver-driver API. If the interface
is regmap, reviewing sub-drivers on their own should be straightforward.
If the API is with function calls, things get more complicated because
the API needs the sub-drivers to be tested and everything needs to be
reviewed together.

Guenter
Wenkai Oct. 12, 2023, 7:56 a.m. UTC | #6
Guenter Roeck 於 10/11/2023 11:05 PM 寫道:
> On Wed, Oct 11, 2023 at 12:08:57PM +0800, Wenkai wrote:
>> I understand that the patches don't meet the expected quality standards.
>> The compile issue is due to my MFD core driver, which is currently under
>> review and has not been merged yet.
>>
>> I would also like to seek your advice on how to best proceed with the
>> sub-drivers like the watchdog driver. Should I wait for my core MFD
>> driver to be successfully merged before submitting the sub-drivers, or
>> let Jones Lee review my core MFD driver and all its sub-drivers, or is
>> there another approach that you recommend?
> If the sub-drivers depend on the mfd driver, at least provide a reference
> to the patch or patch series introducing that driver. Either case, a direct
> include from "../mfd" is simply unacceptable. include/linux/mfd/ does exist
> for a reason, after all.

The LKML link is https://lkml.org/lkml/2023/9/6/1245. Is this link to 
the patch
sufficient?

And, I'll move the "eiois200.h" to "include/linux/mfd/".

> I don't know the best solution for reviewing all the drivers. I didn't
> (and do not plan to) look into the driver-driver API. If the interface
> is regmap, reviewing sub-drivers on their own should be straightforward.
> If the API is with function calls, things get more complicated because
> the API needs the sub-drivers to be tested and everything needs to be
> reviewed together.
>
> Guenter

Unfortunately, all sub-drivers mostly communicate with the EC firmware
through the MFD core driver's driver-driver API, only a few uses regmap.
So should the MFD core driver and its sub-drivers be reviewed by the same
maintainers?

Best regards,
Wenkai