Linux 内核clk框架描述

  The Common Clk Framework  Mike Turquette <[e-mail]>
This document endeavours to explain the common clk framework details,and how to port a platform over to this framework.  It is not yet adetailed explanation of the clock api in include/linux/clk.h, butperhaps someday it will include that information.
 Part 1 - introduction and interface split
The common clk framework is an interface to control the clock nodesavailable on various devices today.  This may come in the form of clockgating, rate adjustment, muxing or other operations.  This framework isenabled with the CONFIG_COMMON_CLK option.
The interface itself is divided into two halves, each shielded from thedetails of its counterpart.  First is the common definition of structclk which unifies the framework-level accounting and infrastructure thathas traditionally been duplicated across a variety of platforms.  Secondis a common implementation of the clk.h api, defined indrivers/clk/clk.c.  Finally there is struct clk_ops, whose operationsare invoked by the clk api implementation.
The second half of the interface is comprised of the hardware-specificcallbacks registered with struct clk_ops and the correspondinghardware-specific structures needed to model a particular clock.  Forthe remainder of this document any reference to a callback in structclk_ops, such as .enable or .set_rate, implies the hardware-specificimplementation of that code.  Likewise, references to struct clk_fooserve as a convenient shorthand for the implementation of thehardware-specific bits for the hypothetical "foo" hardware.
Tying the two halves of this interface together is struct clk_hw, whichis defined in struct clk_foo and pointed to within struct clk_core.  Thisallows for easy navigation between the two discrete halves of the commonclock interface.
 Part 2 - common data structures and api
Below is the common struct clk_core definition fromdrivers/clk/clk.c, modified for brevity:
 struct clk_core {  const char  *name;  const struct clk_ops *ops;  struct clk_hw  *hw;  struct module  *owner;  struct clk_core  *parent;  const char  **parent_names;  struct clk_core  **parents;  u8   num_parents;  u8   new_parent_index;  ... };
The members above make up the core of the clk tree topology.  The clkapi itself defines several driver-facing functions which operate onstruct clk.  That api is documented in include/linux/clk.h.
Platforms and devices utilizing the common struct clk_core use the structclk_ops pointer in struct clk_core to perform the hardware-specific parts ofthe operations defined in clk-provider.h:
 struct clk_ops {  int  (*prepare)(struct clk_hw *hw);  void  (*unprepare)(struct clk_hw *hw);  int  (*is_prepared)(struct clk_hw *hw);  void  (*unprepare_unused)(struct clk_hw *hw);  int  (*enable)(struct clk_hw *hw);  void  (*disable)(struct clk_hw *hw);  int  (*is_enabled)(struct clk_hw *hw);  void  (*disable_unused)(struct clk_hw *hw);  unsigned long (*recalc_rate)(struct clk_hw *hw,      unsigned long parent_rate);  long  (*round_rate)(struct clk_hw *hw,      unsigned long rate,      unsigned long *parent_rate);  int  (*determine_rate)(struct clk_hw *hw,        struct clk_rate_request *req);  int  (*set_parent)(struct clk_hw *hw, u8 index);  u8  (*get_parent)(struct clk_hw *hw);  int  (*set_rate)(struct clk_hw *hw,         unsigned long rate,         unsigned long parent_rate);  int  (*set_rate_and_parent)(struct clk_hw *hw,         unsigned long rate,         unsigned long parent_rate,         u8 index);  unsigned long (*recalc_accuracy)(struct clk_hw *hw,      unsigned long parent_accuracy);  int  (*get_phase)(struct clk_hw *hw);  int  (*set_phase)(struct clk_hw *hw, int degrees);  void  (*init)(struct clk_hw *hw);  int  (*debug_init)(struct clk_hw *hw,           struct dentry *dentry); };
 Part 3 - hardware clk implementations
The strength of the common struct clk_core comes from its .ops and .hw pointerswhich abstract the details of struct clk from the hardware-specific bits, andvice versa.  To illustrate consider the simple gateable clk implementation indrivers/clk/clk-gate.c:
struct clk_gate { struct clk_hw hw; void __iomem    *reg; u8              bit_idx; ...};
struct clk_gate contains struct clk_hw hw as well as hardware-specificknowledge about which register and bit controls this clk's gating.Nothing about clock topology or accounting, such as enable_count ornotifier_count, is needed here.  That is all handled by the commonframework code and struct clk_core.
Let's walk through enabling this clk from driver code:
 struct clk *clk; clk = clk_get(NULL, "my_gateable_clk");
 clk_prepare(clk); clk_enable(clk);
The call graph for clk_enable is very simple:
clk_enable(clk); clk->ops->enable(clk->hw); [resolves to...]  clk_gate_enable(hw);  [resolves struct clk gate with to_clk_gate(hw)]   clk_gate_set_bit(gate);
And the definition of clk_gate_set_bit:
static void clk_gate_set_bit(struct clk_gate *gate){ u32 reg;
 reg = __raw_readl(gate->reg); reg |= BIT(gate->bit_idx); writel(reg, gate->reg);}
Note that to_clk_gate is defined as:
#define to_clk_gate(_hw) container_of(_hw, struct clk_gate, hw)
This pattern of abstraction is used for every clock hardwarerepresentation.
 Part 4 - supporting your own clk hardware
When implementing support for a new type of clock it is only necessary toinclude the following header:
#include <linux/clk-provider.h>
To construct a clk hardware structure for your platform you must definethe following:
struct clk_foo { struct clk_hw hw; ... hardware specific data goes here ...};
To take advantage of your data you'll need to support valid operationsfor your clk:
struct clk_ops clk_foo_ops { .enable  = &clk_foo_enable; .disable = &clk_foo_disable;};
Implement the above functions using container_of:
#define to_clk_foo(_hw) container_of(_hw, struct clk_foo, hw)
int clk_foo_enable(struct clk_hw *hw){ struct clk_foo *foo;
 foo = to_clk_foo(hw);
 ... perform magic on foo ...
 return 0;};
Below is a matrix detailing which clk_ops are mandatory based upon thehardware capabilities of that clock.  A cell marked as "y" meansmandatory, a cell marked as "n" implies that either including thatcallback is invalid or otherwise unnecessary.  Empty cells are eitheroptional or must be evaluated on a case-by-case basis.
                              clock hardware characteristics                -----------------------------------------------------------                | gate | change rate | single parent | multiplexer | root |                |------|-------------|---------------|-------------|------|.prepare        |      |             |               |             |      |.unprepare      |      |             |               |             |      |                |      |             |               |             |      |.enable         | y    |             |               |             |      |.disable        | y    |             |               |             |      |.is_enabled     | y    |             |               |             |      |                |      |             |               |             |      |.recalc_rate    |      | y           |               |             |      |.round_rate     |      | y [1]       |               |             |      |.determine_rate |      | y [1]       |               |             |      |.set_rate       |      | y           |               |             |      |                |      |             |               |             |      |.set_parent     |      |             | n             | y           | n    |.get_parent     |      |             | n             | y           | n    |                |      |             |               |             |      |.recalc_accuracy|      |             |               |             |      |                |      |             |               |             |      |.init           |      |             |               |             |      |                -----------------------------------------------------------[1] either one of round_rate or determine_rate is required.
Finally, register your clock at run-time with a hardware-specificregistration function.  This function simply populates struct clk_foo'sdata and then passes the common struct clk parameters to the frameworkwith a call to:
clk_register(...)
See the basic clock types in drivers/clk/clk-*.c for examples.
 Part 5 - Disabling clock gating of unused clocks
Sometimes during development it can be useful to be able to bypass thedefault disabling of unused clocks. For example, if drivers aren't enablingclocks properly but rely on them being on from the bootloader, bypassingthe disabling means that the driver will remain functional while the issuesare sorted out.
To bypass this disabling, include "clk_ignore_unused" in the bootargs to thekernel.
 Part 6 - Locking
The common clock framework uses two global locks, the prepare lock and theenable lock.
The enable lock is a spinlock and is held across calls to the .enable,.disable and .is_enabled operations. Those operations are thus not allowed tosleep, and calls to the clk_enable(), clk_disable() and clk_is_enabled() APIfunctions are allowed in atomic context.
The prepare lock is a mutex and is held across calls to all other operations.All those operations are allowed to sleep, and calls to the corresponding APIfunctions are not allowed in atomic context.
This effectively divides operations in two groups from a locking perspective.
Drivers don't need to manually protect resources shared between the operationsof one group, regardless of whether those resources are shared by multipleclocks or not. However, access to resources that are shared between operationsof the two groups needs to be protected by the drivers. An example of such aresource would be a register that controls both the clock rate and the clockenable/disable state.
The clock framework is reentrant, in that a driver is allowed to call clockframework functions from within its implementation of clock operations. Thiscan for instance cause a .set_rate operation of one clock being called fromwithin the .set_rate operation of another clock. This case must be consideredin the driver implementations, but the code flow is usually controlled by thedriver in that case.
Note that locking must also be considered when code outside of the commonclock framework needs to access resources used by the clock operations. Thisis considered out of scope of this document.