mbox series

[v5,00/15] clk: at91: updates for power management and dvfs

Message ID 20211011112719.3951784-1-claudiu.beznea@microchip.com (mailing list archive)
Headers show
Series clk: at91: updates for power management and dvfs | expand

Message

Claudiu Beznea Oct. 11, 2021, 11:27 a.m. UTC
Hi,

This series addresses the clock power management for SAMA7G5 and also
updates the master clock driver and sam9x60-pll driver to accommodate
the requests at [1] and avoid overclocking of CPU and MCK0 domains while
changing the frequency via DVFS.

The power management part is implemented by adding
save_context()/restore_context() on each clock driver (patch 1/15). Since
the PM part is necessary only for backup mode (supported on SAMA5D2 and
SAMA7G5) the pmc.c has been adapted to call the
save_context()/restore_context() only on switches to/from backup mode
(patch 2/15).

Patch 3/15 adds the securam clock on SAMA7G5. This is necessary for
backup mode of SAMA7G5.

Patch 4/15 adapt SAMA7G5 MCK1..4 driver to use the defines at
include/linux/clk/at91_pmc.h introduced in commit ec03f18cc222
("clk: at91: add register definition for sama7g5's master clock").

Patch 5/15 improves a bit readabiblity in some places of master clock
driver.

Patch 6/15 enable the suspend/resume for clocks also for SAMA7G5.

Patches 7-10/15 adds fixes in master clock driver and sam9x60-pll driver.

Patches 11-12/15 address DVFS by adding notifiers for master clock driver
and sam9x60-pll driver to avoid overclocking for CPU domain and MCK0
domain.

Patch 13/15 removes the master clock prescaler from Linux clock tree
as it has been discovered a hardware bug on it and it may not lock
on some scenario although its output clock is stable.

Patch 14/15 decreases the low limit of MCK0 as it can go even to 32KHz.

Patch 15/15 uses clk_core_get_rate_recalc() in clk_rate_get().

Thank you,
Claudiu Beznea

[1] https://lore.kernel.org/lkml/20210105104426.4tmgc2l3vyicwedd@vireshk-i7/
[2] https://lore.kernel.org/lkml/163047507296.42057.10597374695758699868@swboyd.mtv.corp.google.com/ 

Changes in v5:
- dropped patches 16/17, 17/17 from previous version
- in patch 02/15:
	- use read_relaxed() to get the value at address pointed by 
	  at91_pmc_backup_suspend
	- removed of_find_device_by_node() from pmc_register_ops() since its
	  returned value was not used in the function (of_device_is_available()
	  is enough to know before mapping securam). of_find_device_by_node()
	  was the one which needed the changes of postcore_initcall() with
	  subsys_initcall() (as described in commit description from v4) thus
	  kept the postcore_initcall(); also updated the commit description
	  to reflect this.

Changes in v4:
- removed opp dependencies along with get_cpu_device() in patch 11/17 as
  those functionalities are not available at the initialization time for
  clocks instantiated with CLK_OF_DECLARE

Changes in v3:
- minor fixes in patch 1/7 (e.g. use regmap_read() + checks + update +
  regmap_write() sequence instead of regmap_read() + checks +
  regmap_update_bits()
- patch 4/17 has been updated after commit ec03f18cc222
  ("clk: at91: add register definition for sama7g5's master clock")
- patch 6-10/17, 14-17/17 are newly introduced
- notifier for sam9x60 div pll was introduced (patch 11/17) and
  notifier logic for master clock div has been changed (patch 12/17)
  to use safe divider on PRE_RATE_CHANGE events and update clock to highest
  possible rate on POST_RATE_CHANGE events
- master clock prescaler has been removed from Linux clock tree for
  SAMA7G5

Changes in v2:
- addressed code review comments
- collected tags

Claudiu Beznea (15):
  clk: at91: re-factor clocks suspend/resume
  clk: at91: pmc: execute suspend/resume only for backup mode
  clk: at91: sama7g5: add securam's peripheral clock
  clk: at91: clk-master: add register definition for sama7g5's master
    clock
  clk: at91: clk-master: improve readability by using local variables
  clk: at91: pmc: add sama7g5 to the list of available pmcs
  clk: at91: sam9x60-pll: use DIV_ROUND_CLOSEST_ULL
  clk: at91: clk-master: check if div or pres is zero
  clk: at91: clk-master: mask mckr against layout->mask
  clk: at91: clk-master: fix prescaler logic
  clk: at91: clk-sam9x60-pll: add notifier for div part of PLL
  clk: at91: clk-master: add notifier for divider
  clk: at91: sama7g5: remove prescaler part of master clock
  clk: at91: sama7g5: set low limit for mck0 at 32KHz
  clk: use clk_core_get_rate_recalc() in clk_rate_get()

 drivers/clk/at91/at91rm9200.c       |   2 +-
 drivers/clk/at91/at91sam9260.c      |   2 +-
 drivers/clk/at91/at91sam9g45.c      |   2 +-
 drivers/clk/at91/at91sam9n12.c      |   2 +-
 drivers/clk/at91/at91sam9rl.c       |   2 +-
 drivers/clk/at91/at91sam9x5.c       |   2 +-
 drivers/clk/at91/clk-generated.c    |  46 ++-
 drivers/clk/at91/clk-main.c         |  66 ++++
 drivers/clk/at91/clk-master.c       | 463 ++++++++++++++++++++++------
 drivers/clk/at91/clk-peripheral.c   |  40 ++-
 drivers/clk/at91/clk-pll.c          |  39 +++
 drivers/clk/at91/clk-programmable.c |  29 +-
 drivers/clk/at91/clk-sam9x60-pll.c  | 174 +++++++++--
 drivers/clk/at91/clk-system.c       |  20 ++
 drivers/clk/at91/clk-usb.c          |  27 ++
 drivers/clk/at91/clk-utmi.c         |  39 +++
 drivers/clk/at91/dt-compat.c        |   2 +-
 drivers/clk/at91/pmc.c              | 173 +++--------
 drivers/clk/at91/pmc.h              |  29 +-
 drivers/clk/at91/sam9x60.c          |   6 +-
 drivers/clk/at91/sama5d2.c          |   2 +-
 drivers/clk/at91/sama5d3.c          |   2 +-
 drivers/clk/at91/sama5d4.c          |   2 +-
 drivers/clk/at91/sama7g5.c          |  29 +-
 drivers/clk/clk.c                   |   2 +-
 25 files changed, 890 insertions(+), 312 deletions(-)

Comments

Nicolas Ferre Oct. 15, 2021, 8:33 a.m. UTC | #1
Stephen,

On 11/10/2021 at 13:27, Claudiu Beznea wrote:
> Hi,
> 
> This series addresses the clock power management for SAMA7G5 and also
> updates the master clock driver and sam9x60-pll driver to accommodate
> the requests at [1] and avoid overclocking of CPU and MCK0 domains while
> changing the frequency via DVFS.

I've just added my Ack tags to the whole series. I think that it brings 
so much important additions to our new SoC as well as fixing some (non 
urgent) bugs.

In order to streamline the process of including this series to upstream, 
do you want me to send you a pull-request? against latest clk-at91 
branch in your tree?

Best regards,
   Nicolas

> The power management part is implemented by adding
> save_context()/restore_context() on each clock driver (patch 1/15). Since
> the PM part is necessary only for backup mode (supported on SAMA5D2 and
> SAMA7G5) the pmc.c has been adapted to call the
> save_context()/restore_context() only on switches to/from backup mode
> (patch 2/15).
> 
> Patch 3/15 adds the securam clock on SAMA7G5. This is necessary for
> backup mode of SAMA7G5.
> 
> Patch 4/15 adapt SAMA7G5 MCK1..4 driver to use the defines at
> include/linux/clk/at91_pmc.h introduced in commit ec03f18cc222
> ("clk: at91: add register definition for sama7g5's master clock").
> 
> Patch 5/15 improves a bit readability in some places of master clock
> driver.
> 
> Patch 6/15 enable the suspend/resume for clocks also for SAMA7G5.
> 
> Patches 7-10/15 adds fixes in master clock driver and sam9x60-pll driver.
> 
> Patches 11-12/15 address DVFS by adding notifiers for master clock driver
> and sam9x60-pll driver to avoid overclocking for CPU domain and MCK0
> domain.
> 
> Patch 13/15 removes the master clock prescaler from Linux clock tree
> as it has been discovered a hardware bug on it and it may not lock
> on some scenario although its output clock is stable.
> 
> Patch 14/15 decreases the low limit of MCK0 as it can go even to 32KHz.
> 
> Patch 15/15 uses clk_core_get_rate_recalc() in clk_rate_get().
> 
> Thank you,
> Claudiu Beznea
> 
> [1] https://lore.kernel.org/lkml/20210105104426.4tmgc2l3vyicwedd@vireshk-i7/
> [2] https://lore.kernel.org/lkml/163047507296.42057.10597374695758699868@swboyd.mtv.corp.google.com/
> 
> Changes in v5:
> - dropped patches 16/17, 17/17 from previous version
> - in patch 02/15:
> 	- use read_relaxed() to get the value at address pointed by
> 	  at91_pmc_backup_suspend
> 	- removed of_find_device_by_node() from pmc_register_ops() since its
> 	  returned value was not used in the function (of_device_is_available()
> 	  is enough to know before mapping securam). of_find_device_by_node()
> 	  was the one which needed the changes of postcore_initcall() with
> 	  subsys_initcall() (as described in commit description from v4) thus
> 	  kept the postcore_initcall(); also updated the commit description
> 	  to reflect this.
> 
> Changes in v4:
> - removed opp dependencies along with get_cpu_device() in patch 11/17 as
>    those functionalities are not available at the initialization time for
>    clocks instantiated with CLK_OF_DECLARE
> 
> Changes in v3:
> - minor fixes in patch 1/7 (e.g. use regmap_read() + checks + update +
>    regmap_write() sequence instead of regmap_read() + checks +
>    regmap_update_bits()
> - patch 4/17 has been updated after commit ec03f18cc222
>    ("clk: at91: add register definition for sama7g5's master clock")
> - patch 6-10/17, 14-17/17 are newly introduced
> - notifier for sam9x60 div pll was introduced (patch 11/17) and
>    notifier logic for master clock div has been changed (patch 12/17)
>    to use safe divider on PRE_RATE_CHANGE events and update clock to highest
>    possible rate on POST_RATE_CHANGE events
> - master clock prescaler has been removed from Linux clock tree for
>    SAMA7G5
> 
> Changes in v2:
> - addressed code review comments
> - collected tags
> 
> Claudiu Beznea (15):
>    clk: at91: re-factor clocks suspend/resume
>    clk: at91: pmc: execute suspend/resume only for backup mode
>    clk: at91: sama7g5: add securam's peripheral clock
>    clk: at91: clk-master: add register definition for sama7g5's master
>      clock
>    clk: at91: clk-master: improve readability by using local variables
>    clk: at91: pmc: add sama7g5 to the list of available pmcs
>    clk: at91: sam9x60-pll: use DIV_ROUND_CLOSEST_ULL
>    clk: at91: clk-master: check if div or pres is zero
>    clk: at91: clk-master: mask mckr against layout->mask
>    clk: at91: clk-master: fix prescaler logic
>    clk: at91: clk-sam9x60-pll: add notifier for div part of PLL
>    clk: at91: clk-master: add notifier for divider
>    clk: at91: sama7g5: remove prescaler part of master clock
>    clk: at91: sama7g5: set low limit for mck0 at 32KHz
>    clk: use clk_core_get_rate_recalc() in clk_rate_get()
> 
>   drivers/clk/at91/at91rm9200.c       |   2 +-
>   drivers/clk/at91/at91sam9260.c      |   2 +-
>   drivers/clk/at91/at91sam9g45.c      |   2 +-
>   drivers/clk/at91/at91sam9n12.c      |   2 +-
>   drivers/clk/at91/at91sam9rl.c       |   2 +-
>   drivers/clk/at91/at91sam9x5.c       |   2 +-
>   drivers/clk/at91/clk-generated.c    |  46 ++-
>   drivers/clk/at91/clk-main.c         |  66 ++++
>   drivers/clk/at91/clk-master.c       | 463 ++++++++++++++++++++++------
>   drivers/clk/at91/clk-peripheral.c   |  40 ++-
>   drivers/clk/at91/clk-pll.c          |  39 +++
>   drivers/clk/at91/clk-programmable.c |  29 +-
>   drivers/clk/at91/clk-sam9x60-pll.c  | 174 +++++++++--
>   drivers/clk/at91/clk-system.c       |  20 ++
>   drivers/clk/at91/clk-usb.c          |  27 ++
>   drivers/clk/at91/clk-utmi.c         |  39 +++
>   drivers/clk/at91/dt-compat.c        |   2 +-
>   drivers/clk/at91/pmc.c              | 173 +++--------
>   drivers/clk/at91/pmc.h              |  29 +-
>   drivers/clk/at91/sam9x60.c          |   6 +-
>   drivers/clk/at91/sama5d2.c          |   2 +-
>   drivers/clk/at91/sama5d3.c          |   2 +-
>   drivers/clk/at91/sama5d4.c          |   2 +-
>   drivers/clk/at91/sama7g5.c          |  29 +-
>   drivers/clk/clk.c                   |   2 +-
>   25 files changed, 890 insertions(+), 312 deletions(-)
>
Stephen Boyd Oct. 27, 2021, 1:35 a.m. UTC | #2
Quoting Nicolas Ferre (2021-10-15 01:33:38)
> Stephen,
> 
> On 11/10/2021 at 13:27, Claudiu Beznea wrote:
> > Hi,
> > 
> > This series addresses the clock power management for SAMA7G5 and also
> > updates the master clock driver and sam9x60-pll driver to accommodate
> > the requests at [1] and avoid overclocking of CPU and MCK0 domains while
> > changing the frequency via DVFS.
> 
> I've just added my Ack tags to the whole series. I think that it brings 
> so much important additions to our new SoC as well as fixing some (non 
> urgent) bugs.
> 
> In order to streamline the process of including this series to upstream, 
> do you want me to send you a pull-request? against latest clk-at91 
> branch in your tree?
> 

That's fine. Just let me know if you want to do that in the future. For
now I picked them all up and pushed it out.