[net-next,v8,00/15] ACPI support for dpaa2 driver

Message ID 20210610163917.4138412-1-ciorneiioana@gmail.com
Headers show
Series
  • ACPI support for dpaa2 driver
Related show

Message

Ioana Ciornei June 10, 2021, 4:39 p.m.
From: Ioana Ciornei <ioana.ciornei@nxp.com>

This patch set provides ACPI support to DPAA2 network drivers.

It also introduces new fwnode based APIs to support phylink and phy
layers
    Following functions are defined:
      phylink_fwnode_phy_connect()
      fwnode_mdiobus_register_phy()
      fwnode_get_phy_id()
      fwnode_phy_find_device()
      device_phy_find_device()
      fwnode_get_phy_node()
      fwnode_mdio_find_device()
      acpi_get_local_address()

    First one helps in connecting phy to phylink instance.
    Next three helps in getting phy_id and registering phy to mdiobus
    Next two help in finding a phy on a mdiobus.
    Next one helps in getting phy_node from a fwnode.
    Last one is used to get local address from _ADR object.

    Corresponding OF functions are refactored.

Tested-on: LX2160ARDB

Changes in v8:
 - fixed some checkpatch warnings/checks
 - included linux/fwnode_mdio.h in fwnode_mdio.c (fixed the build warnings)
 - added fwnode_find_mii_timestamper() and
   fwnode_mdiobus_phy_device_register() in order to get rid of the cycle
   dependency.
 - change to 'depends on (ACPI || OF) || COMPILE_TEST (for FWNODE_MDIO)
 - remove the fwnode_mdiobus_register from fwnode_mdio.c since it
   introduces a cycle of dependencies.

Changes in v7:
- correct fwnode_mdio_find_device() description
- check NULL in unregister_mii_timestamper()
- Call unregister_mii_timestamper() without NULL check
- Create fwnode_mdio.c and move fwnode_mdiobus_register_phy()
- include fwnode_mdio.h
- Include headers directly used in acpi_mdio.c
- Move fwnode_mdiobus_register() to fwnode_mdio.c
- Include fwnode_mdio.h
- Alphabetically sort header inclusions
- remove unnecassary checks

Changes in v6:
- Minor cleanup
- fix warning for function parameter of fwnode_mdio_find_device()
- Initialize mii_ts to NULL
- use GENMASK() and ACPI_COMPANION_SET()
- some cleanup
- remove unwanted header inclusion
- remove OF check for fixed-link
- use dev_fwnode()
- remove useless else
- replace of_device_is_available() to fwnode_device_is_available()

Changes in v5:
- More cleanup
- Replace fwnode_get_id() with acpi_get_local_address()
- add missing MODULE_LICENSE()
- replace fwnode_get_id() with OF and ACPI function calls
- replace fwnode_get_id() with OF and ACPI function calls

Changes in v4:
- More cleanup
- Improve code structure to handle all cases
- Remove redundant else from fwnode_mdiobus_register()
- Cleanup xgmac_mdio_probe()
- call phy_device_free() before returning

Changes in v3:
- Add more info on legacy DT properties "phy" and "phy-device"
- Redefine fwnode_phy_find_device() to follow of_phy_find_device()
- Use traditional comparison pattern
- Use GENMASK
- Modified to retrieve reg property value for ACPI as well
- Resolved compilation issue with CONFIG_ACPI = n
- Added more info into documentation
- Use acpi_mdiobus_register()
- Avoid unnecessary line removal
- Remove unused inclusion of acpi.h

Changes in v2:
- Updated with more description in document
- use reverse christmas tree ordering for local variables
- Refactor OF functions to use fwnode functions

Calvin Johnson (15):
  Documentation: ACPI: DSD: Document MDIO PHY
  net: phy: Introduce fwnode_mdio_find_device()
  net: phy: Introduce phy related fwnode functions
  of: mdio: Refactor of_phy_find_device()
  net: phy: Introduce fwnode_get_phy_id()
  of: mdio: Refactor of_get_phy_id()
  net: mii_timestamper: check NULL in unregister_mii_timestamper()
  net: mdiobus: Introduce fwnode_mdiobus_register_phy()
  of: mdio: Refactor of_mdiobus_register_phy()
  ACPI: utils: Introduce acpi_get_local_address()
  net: mdio: Add ACPI support code for mdio
  net/fsl: Use [acpi|of]_mdiobus_register
  net: phylink: introduce phylink_fwnode_phy_connect()
  net: phylink: Refactor phylink_of_phy_connect()
  net: dpaa2-mac: Add ACPI support for DPAA2 MAC driver

 Documentation/firmware-guide/acpi/dsd/phy.rst | 133 ++++++++++++++++
 MAINTAINERS                                   |   2 +
 drivers/acpi/utils.c                          |  14 ++
 .../net/ethernet/freescale/dpaa2/dpaa2-mac.c  |  88 ++++++-----
 .../net/ethernet/freescale/dpaa2/dpaa2-mac.h  |   2 +-
 drivers/net/ethernet/freescale/xgmac_mdio.c   |  30 ++--
 drivers/net/mdio/Kconfig                      |  14 ++
 drivers/net/mdio/Makefile                     |   4 +-
 drivers/net/mdio/acpi_mdio.c                  |  56 +++++++
 drivers/net/mdio/fwnode_mdio.c                | 144 ++++++++++++++++++
 drivers/net/mdio/of_mdio.c                    | 138 ++---------------
 drivers/net/phy/mii_timestamper.c             |   3 +
 drivers/net/phy/phy_device.c                  | 109 ++++++++++++-
 drivers/net/phy/phylink.c                     |  41 +++--
 include/linux/acpi.h                          |   7 +
 include/linux/acpi_mdio.h                     |  26 ++++
 include/linux/fwnode_mdio.h                   |  35 +++++
 include/linux/phy.h                           |  32 ++++
 include/linux/phylink.h                       |   3 +
 19 files changed, 691 insertions(+), 190 deletions(-)
 create mode 100644 Documentation/firmware-guide/acpi/dsd/phy.rst
 create mode 100644 drivers/net/mdio/acpi_mdio.c
 create mode 100644 drivers/net/mdio/fwnode_mdio.c
 create mode 100644 include/linux/acpi_mdio.h
 create mode 100644 include/linux/fwnode_mdio.h

Comments

Andrew Lunn June 10, 2021, 4:56 p.m. | #1
On Thu, Jun 10, 2021 at 07:39:02PM +0300, Ioana Ciornei wrote:
> From: Ioana Ciornei <ioana.ciornei@nxp.com>
> 
> This patch set provides ACPI support to DPAA2 network drivers.

Just to be clear and avoid confusion, there is a standing NACK against
this patchset. Please see the discussion here:

https://patchwork.kernel.org/project/linux-acpi/patch/20200715090400.4733-2-calvin.johnson@oss.nxp.com/#23518385

So far, i've not seen any indication the issues raised there have been
resolved. I don't see any Acked-by from an ACPI maintainer. So this
code remains NACKed.

	Andrew
Jon Nettleton June 10, 2021, 5:51 p.m. | #2
On Thu, Jun 10, 2021 at 6:56 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> On Thu, Jun 10, 2021 at 07:39:02PM +0300, Ioana Ciornei wrote:
> > From: Ioana Ciornei <ioana.ciornei@nxp.com>
> >
> > This patch set provides ACPI support to DPAA2 network drivers.
>
> Just to be clear and avoid confusion, there is a standing NACK against
> this patchset. Please see the discussion here:
>
> https://patchwork.kernel.org/project/linux-acpi/patch/20200715090400.4733-2-calvin.johnson@oss.nxp.com/#23518385
>
> So far, i've not seen any indication the issues raised there have been
> resolved. I don't see any Acked-by from an ACPI maintainer. So this
> code remains NACKed.

Andrew,

The ACPI maintainers did bless the use of the ACPI standards followed
in this patchset, and their only abstinence from ACK'ing the patchset
was whether the code was used in production systems.  Well currently,
not only have we, SolidRun, been using this patchset and the associated
ACPI tables in our SystemsReady certified firmware for the HoneyComb,
but we also have customers using this same patchset and firmware on
their systems rolled out to customers.

Additionally we have an entire new product line based on Marvell's
Armada CN913x series, which also needs this patchset to be fully
functional.

I am quite certain this is more than enough production systems using
this ACPI description method for networking to progress this patchset
forward.

-Jon
Grant Likely June 10, 2021, 6:19 p.m. | #3
On 10/06/2021 17:39, Ioana Ciornei wrote:
> From: Ioana Ciornei <ioana.ciornei@nxp.com>
> 
> This patch set provides ACPI support to DPAA2 network drivers.
> 
> It also introduces new fwnode based APIs to support phylink and phy
> layers
>      Following functions are defined:
>        phylink_fwnode_phy_connect()
>        fwnode_mdiobus_register_phy()
>        fwnode_get_phy_id()
>        fwnode_phy_find_device()
>        device_phy_find_device()
>        fwnode_get_phy_node()
>        fwnode_mdio_find_device()
>        acpi_get_local_address()
> 
>      First one helps in connecting phy to phylink instance.
>      Next three helps in getting phy_id and registering phy to mdiobus
>      Next two help in finding a phy on a mdiobus.
>      Next one helps in getting phy_node from a fwnode.
>      Last one is used to get local address from _ADR object.
> 
>      Corresponding OF functions are refactored.

In terms of design, this whole series looks right to me. The way data is 
encoded in ACPI is entirely appropriate. As long as Rob Herring is okay 
with the DT code changes I support this series being merged.

For the whole series:
Acked-by: Grant Likely <grant.likely@arm.com>


> 
> Tested-on: LX2160ARDB
> 
> Changes in v8:
>   - fixed some checkpatch warnings/checks
>   - included linux/fwnode_mdio.h in fwnode_mdio.c (fixed the build warnings)
>   - added fwnode_find_mii_timestamper() and
>     fwnode_mdiobus_phy_device_register() in order to get rid of the cycle
>     dependency.
>   - change to 'depends on (ACPI || OF) || COMPILE_TEST (for FWNODE_MDIO)
>   - remove the fwnode_mdiobus_register from fwnode_mdio.c since it
>     introduces a cycle of dependencies.
> 
> Changes in v7:
> - correct fwnode_mdio_find_device() description
> - check NULL in unregister_mii_timestamper()
> - Call unregister_mii_timestamper() without NULL check
> - Create fwnode_mdio.c and move fwnode_mdiobus_register_phy()
> - include fwnode_mdio.h
> - Include headers directly used in acpi_mdio.c
> - Move fwnode_mdiobus_register() to fwnode_mdio.c
> - Include fwnode_mdio.h
> - Alphabetically sort header inclusions
> - remove unnecassary checks
> 
> Changes in v6:
> - Minor cleanup
> - fix warning for function parameter of fwnode_mdio_find_device()
> - Initialize mii_ts to NULL
> - use GENMASK() and ACPI_COMPANION_SET()
> - some cleanup
> - remove unwanted header inclusion
> - remove OF check for fixed-link
> - use dev_fwnode()
> - remove useless else
> - replace of_device_is_available() to fwnode_device_is_available()
> 
> Changes in v5:
> - More cleanup
> - Replace fwnode_get_id() with acpi_get_local_address()
> - add missing MODULE_LICENSE()
> - replace fwnode_get_id() with OF and ACPI function calls
> - replace fwnode_get_id() with OF and ACPI function calls
> 
> Changes in v4:
> - More cleanup
> - Improve code structure to handle all cases
> - Remove redundant else from fwnode_mdiobus_register()
> - Cleanup xgmac_mdio_probe()
> - call phy_device_free() before returning
> 
> Changes in v3:
> - Add more info on legacy DT properties "phy" and "phy-device"
> - Redefine fwnode_phy_find_device() to follow of_phy_find_device()
> - Use traditional comparison pattern
> - Use GENMASK
> - Modified to retrieve reg property value for ACPI as well
> - Resolved compilation issue with CONFIG_ACPI = n
> - Added more info into documentation
> - Use acpi_mdiobus_register()
> - Avoid unnecessary line removal
> - Remove unused inclusion of acpi.h
> 
> Changes in v2:
> - Updated with more description in document
> - use reverse christmas tree ordering for local variables
> - Refactor OF functions to use fwnode functions
> 
> Calvin Johnson (15):
>    Documentation: ACPI: DSD: Document MDIO PHY
>    net: phy: Introduce fwnode_mdio_find_device()
>    net: phy: Introduce phy related fwnode functions
>    of: mdio: Refactor of_phy_find_device()
>    net: phy: Introduce fwnode_get_phy_id()
>    of: mdio: Refactor of_get_phy_id()
>    net: mii_timestamper: check NULL in unregister_mii_timestamper()
>    net: mdiobus: Introduce fwnode_mdiobus_register_phy()
>    of: mdio: Refactor of_mdiobus_register_phy()
>    ACPI: utils: Introduce acpi_get_local_address()
>    net: mdio: Add ACPI support code for mdio
>    net/fsl: Use [acpi|of]_mdiobus_register
>    net: phylink: introduce phylink_fwnode_phy_connect()
>    net: phylink: Refactor phylink_of_phy_connect()
>    net: dpaa2-mac: Add ACPI support for DPAA2 MAC driver
> 
>   Documentation/firmware-guide/acpi/dsd/phy.rst | 133 ++++++++++++++++
>   MAINTAINERS                                   |   2 +
>   drivers/acpi/utils.c                          |  14 ++
>   .../net/ethernet/freescale/dpaa2/dpaa2-mac.c  |  88 ++++++-----
>   .../net/ethernet/freescale/dpaa2/dpaa2-mac.h  |   2 +-
>   drivers/net/ethernet/freescale/xgmac_mdio.c   |  30 ++--
>   drivers/net/mdio/Kconfig                      |  14 ++
>   drivers/net/mdio/Makefile                     |   4 +-
>   drivers/net/mdio/acpi_mdio.c                  |  56 +++++++
>   drivers/net/mdio/fwnode_mdio.c                | 144 ++++++++++++++++++
>   drivers/net/mdio/of_mdio.c                    | 138 ++---------------
>   drivers/net/phy/mii_timestamper.c             |   3 +
>   drivers/net/phy/phy_device.c                  | 109 ++++++++++++-
>   drivers/net/phy/phylink.c                     |  41 +++--
>   include/linux/acpi.h                          |   7 +
>   include/linux/acpi_mdio.h                     |  26 ++++
>   include/linux/fwnode_mdio.h                   |  35 +++++
>   include/linux/phy.h                           |  32 ++++
>   include/linux/phylink.h                       |   3 +
>   19 files changed, 691 insertions(+), 190 deletions(-)
>   create mode 100644 Documentation/firmware-guide/acpi/dsd/phy.rst
>   create mode 100644 drivers/net/mdio/acpi_mdio.c
>   create mode 100644 drivers/net/mdio/fwnode_mdio.c
>   create mode 100644 include/linux/acpi_mdio.h
>   create mode 100644 include/linux/fwnode_mdio.h
>
Rafael J. Wysocki June 10, 2021, 6:23 p.m. | #4
On Thu, Jun 10, 2021 at 6:40 PM Ioana Ciornei <ciorneiioana@gmail.com> wrote:
>
> From: Calvin Johnson <calvin.johnson@oss.nxp.com>
>
> Modify dpaa2_mac_get_node() to get the dpmac fwnode from either
> DT or ACPI.
>
> Modify dpaa2_mac_get_if_mode() to get interface mode from dpmac_node
> which is a fwnode.
>
> Modify dpaa2_pcs_create() to create pcs from dpmac_node fwnode.
>
> Modify dpaa2_mac_connect() to support ACPI along with DT.
>
> Signed-off-by: Calvin Johnson <calvin.johnson@oss.nxp.com>
> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Rafael J. Wysocki June 10, 2021, 6:25 p.m. | #5
On Thu, Jun 10, 2021 at 7:51 PM Jon Nettleton <jon@solid-run.com> wrote:
>
> On Thu, Jun 10, 2021 at 6:56 PM Andrew Lunn <andrew@lunn.ch> wrote:
> >
> > On Thu, Jun 10, 2021 at 07:39:02PM +0300, Ioana Ciornei wrote:
> > > From: Ioana Ciornei <ioana.ciornei@nxp.com>
> > >
> > > This patch set provides ACPI support to DPAA2 network drivers.
> >
> > Just to be clear and avoid confusion, there is a standing NACK against
> > this patchset. Please see the discussion here:
> >
> > https://patchwork.kernel.org/project/linux-acpi/patch/20200715090400.4733-2-calvin.johnson@oss.nxp.com/#23518385
> >
> > So far, i've not seen any indication the issues raised there have been
> > resolved. I don't see any Acked-by from an ACPI maintainer. So this
> > code remains NACKed.
>
> Andrew,
>
> The ACPI maintainers did bless the use of the ACPI standards followed
> in this patchset, and their only abstinence from ACK'ing the patchset
> was whether the code was used in production systems.  Well currently,
> not only have we, SolidRun, been using this patchset and the associated
> ACPI tables in our SystemsReady certified firmware for the HoneyComb,
> but we also have customers using this same patchset and firmware on
> their systems rolled out to customers.
>
> Additionally we have an entire new product line based on Marvell's
> Armada CN913x series, which also needs this patchset to be fully
> functional.
>
> I am quite certain this is more than enough production systems using
> this ACPI description method for networking to progress this patchset
> forward.

And I believe that you have all of the requisite ACKs from the ACPI
side now, so it is up to the networking guys to decide what to do
next.

Thanks,
Rafael
Rafael J. Wysocki June 10, 2021, 6:31 p.m. | #6
On Thu, Jun 10, 2021 at 8:23 PM Grant Likely <grant.likely@arm.com> wrote:
>
> On 10/06/2021 19:05, Rafael J. Wysocki wrote:
> > On Thu, Jun 10, 2021 at 6:40 PM Ioana Ciornei <ciorneiioana@gmail.com> wrote:
> >>
> >> From: Calvin Johnson <calvin.johnson@oss.nxp.com>
> >>
> >> Introduce ACPI mechanism to get PHYs registered on a MDIO bus and
> >> provide them to be connected to MAC.
> >
> > This is not an "ACPI mechanism", because it is not part of the ACPI
> > specification or support documentation thereof.
> >
> > I would call it "a mechanism based on generic ACPI _DSD device
> > properties definition []1]".  And provide a reference to the _DSD
> > properties definition document.
> >
> > With that changed, you can add
> >
> > Acked-by: Rafael J. Wysocki <rafael@kernel.org>
> >
> > to this patch.
> >
> > Note, however, that within the traditional ACPI framework, the _DSD
> > properties are consumed by the driver that binds to the device
> > represented by the ACPI device object containing the _DSD in question
> > in its scope, while in this case IIUC the properties are expected to
> > be consumed by the general networking code in the kernel.  That is not
> > wrong in principle, but it means that operating systems other than
> > Linux are not likely to be using them.
> >
>
> Doesn't this land at the level of device drivers though? None of this
> data needs to be consumed by the OS generic ACPI parsing code, but the
> network device driver can use it to parse the MDIO and MAC configuraiton
> and set itself up appropriately.

That's right in general, which is why I said that doing it this way
wasn't wrong.
Andy Shevchenko June 11, 2021, 8:57 a.m. | #7
On Thu, Jun 10, 2021 at 7:40 PM Ioana Ciornei <ciorneiioana@gmail.com> wrote:
>

> From: Calvin Johnson <calvin.johnson@oss.nxp.com>

>

> Depending on the device node type, call the specific OF or ACPI

> mdiobus_register function.

>

> Note: For both ACPI and DT cases, endianness of MDIO controller


controllers

> need to be specified using "little-endian" property.


using the

...

> Changes in v8:

> - Directly call the OF or ACPI variants of registering the MDIO bus.

>   This is needed because the fwnode_mdio.c module should only implement

>   features which can be achieved without going back to the OF/ACPI

>   variants. Without this restrictions we directly end up in a dependency

>   cycle: of_mdio -> fwnode_mdio -> of_mdio.


Shouldn't be simple fwnode_mdio.h.
The idea of fwnode APIs that they provide a common ground for all
resource providers.

> - Changed the commit title since the fwnode_mdiobus_register() is no

>   longer available

>

> Changes in v7:

> - Include fwnode_mdio.h


> - Alphabetically sort header inclusions


I suppose this should be a separate change.


-- 
With Best Regards,
Andy Shevchenko
Andy Shevchenko June 11, 2021, 9 a.m. | #8
On Thu, Jun 10, 2021 at 7:40 PM Ioana Ciornei <ciorneiioana@gmail.com> wrote:
>

> From: Ioana Ciornei <ioana.ciornei@nxp.com>

>

> This patch set provides ACPI support to DPAA2 network drivers.

>

> It also introduces new fwnode based APIs to support phylink and phy

> layers

>     Following functions are defined:

>       phylink_fwnode_phy_connect()

>       fwnode_mdiobus_register_phy()

>       fwnode_get_phy_id()

>       fwnode_phy_find_device()

>       device_phy_find_device()

>       fwnode_get_phy_node()

>       fwnode_mdio_find_device()

>       acpi_get_local_address()

>

>     First one helps in connecting phy to phylink instance.

>     Next three helps in getting phy_id and registering phy to mdiobus

>     Next two help in finding a phy on a mdiobus.

>     Next one helps in getting phy_node from a fwnode.

>     Last one is used to get local address from _ADR object.

>

>     Corresponding OF functions are refactored.


In general it looks fine to me. What really worries me is the calls like

of_foo -> fwnode_bar -> of_baz.

As I have commented in one patch the idea of fwnode APIs is to have a
common ground for all resource providers. So, at the end it shouldn't
be a chain of calls like above mentioned. Either fix the name (so, the
first one will be in fwnode or device namespace) or fix the API that
it will be fwnode/device API.

-- 
With Best Regards,
Andy Shevchenko
Ioana Ciornei June 11, 2021, 9:17 a.m. | #9
On Fri, Jun 11, 2021 at 12:00:02PM +0300, Andy Shevchenko wrote:
> On Thu, Jun 10, 2021 at 7:40 PM Ioana Ciornei <ciorneiioana@gmail.com> wrote:

> >

> > From: Ioana Ciornei <ioana.ciornei@nxp.com>

> >

> > This patch set provides ACPI support to DPAA2 network drivers.

> >

> > It also introduces new fwnode based APIs to support phylink and phy

> > layers

> >     Following functions are defined:

> >       phylink_fwnode_phy_connect()

> >       fwnode_mdiobus_register_phy()

> >       fwnode_get_phy_id()

> >       fwnode_phy_find_device()

> >       device_phy_find_device()

> >       fwnode_get_phy_node()

> >       fwnode_mdio_find_device()

> >       acpi_get_local_address()

> >

> >     First one helps in connecting phy to phylink instance.

> >     Next three helps in getting phy_id and registering phy to mdiobus

> >     Next two help in finding a phy on a mdiobus.

> >     Next one helps in getting phy_node from a fwnode.

> >     Last one is used to get local address from _ADR object.

> >

> >     Corresponding OF functions are refactored.

> 

> In general it looks fine to me. What really worries me is the calls like

> 

> of_foo -> fwnode_bar -> of_baz.

> 

> As I have commented in one patch the idea of fwnode APIs is to have a

> common ground for all resource providers. So, at the end it shouldn't

> be a chain of calls like above mentioned. Either fix the name (so, the

> first one will be in fwnode or device namespace) or fix the API that

> it will be fwnode/device API.



These types of cyclic calls do not exist anymore.
The fwnode_mdio does not call back into of_mdio but instead it directly
implements any necessary operations using the fwnode_handle.

The only calls happening are 'of_mdio -> fwnode_mdio' so that we
leverage the common fwnode handling and do not duplicate code.

Ioana