Clock unregistration support in the common clock framework
Subsystems (4)
cover Clock unregistration support in the common clock framework
Hello,
This short patch set aims to fix issues in the common clock framework WRT support of the clock suppliers as loadable modules. The thread [1] might be a good summary and pre-requisiste reading on what this patch series is trying to achieve.
The first patch adds common implementation of the __clk_get(), __clk_put() helpers for the common clock framework. Currently these functions are empty and the modules that registered clocks are prone to removal and leaving invalid clock references.
The second patch adds implementation of the clk_unregister() function and is based on patch [2].
I have some doubts whether we need to be taking reference on a module in clk_get() _and_ use kref to keep track of references to a clock. Taking reference on a module only seems insufficient, since a clock supplier driver can be unbound from its device through sysfs, even if the module stays in place.
We could be doing only reference counting on the clock, but then there are issues as Russell clearly explained in [1]. It is not obvious what to do with a clock when it has consumers and its supplier driver is being deinitialized/unloaded [3]. IMHO clock suppliers should be normally prevented from being removed when resources they provide are in use, otherwise it all may get a bit hairy.
Thanks,
Sylwester
[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/183302.html
[2] http://www.spinics.net/lists/arm-kernel/msg247548.html
[3] http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/183292.html
Sylwester Nawrocki (2):
clk: add common __clk_get(), __clk_put() implementations
clk: implement clk_unregister
arch/arm/include/asm/clkdev.h | 2 + arch/blackfin/include/asm/clkdev.h | 2 + arch/mips/include/asm/clkdev.h | 2 + arch/sh/include/asm/clkdev.h | 2 + drivers/clk/clk.c | 147 +++++++++++++++++++++++++++++++++++- include/linux/clk-private.h | 5 ++ include/linux/clkdev.h | 5 ++ 7 files changed, 162 insertions(+), 3 deletions(-)
--
1.7.9.5
RFC PATCH 1/2 clk: add common __clk_get(), __clk_put() implementations
This patch adds common __clk_get(), __clk_put() clkdev helpers which replace their platform specific counterparts when the common clock API enabled.
An owner module pointer is added to struct clk so a reference to the clock supplier module is taken when the clock has active consumers.
--- arch/arm/include/asm/clkdev.h | 2 ++ arch/blackfin/include/asm/clkdev.h | 2 ++ arch/mips/include/asm/clkdev.h | 2 ++ arch/sh/include/asm/clkdev.h | 2 ++ drivers/clk/clk.c | 24 ++++++++++++++++++++++++ include/linux/clk-private.h | 3 +++ include/linux/clkdev.h | 5 +++++ 7 files changed, 40 insertions(+)
arch/arm/include/asm/clkdev.h +2 -0
@@ -14,12 +14,14 @@#include <linux/slab.h> +#ifndef CONFIG_COMMON_CLK #ifdef CONFIG_HAVE_MACH_CLKDEV #include <mach/clkdev.h> #else #define __clk_get (clk) ({ 1; })#define __clk_put (clk) do { } while (0)#endif +#endif static inline struct clk_lookup_alloc *__clkdev_alloc (size_t size) {
arch/blackfin/include/asm/clkdev.h +2 -0
@@ -8,7 +8,9 @@ static inline struct clk_lookup_alloc *__clkdev_alloc(size_t size) return kzalloc(size, GFP_KERNEL); } +#ifndef CONFIG_COMMON_CLK #define __clk_put (clk)#define __clk_get (clk) ({ 1; }) +#endif #endif
arch/mips/include/asm/clkdev.h +2 -0
@@ -14,8 +14,10 @@#include <linux/slab.h> +#ifndef CONFIG_COMMON_CLK #define __clk_get (clk) ({ 1; })#define __clk_put (clk) do { } while (0) +#endif static inline struct clk_lookup_alloc *__clkdev_alloc (size_t size) {
arch/sh/include/asm/clkdev.h +2 -0
@@ -25,7 +25,9 @@ static inline struct clk_lookup_alloc *__clkdev_alloc(size_t size) return kzalloc(size, GFP_KERNEL); } +#ifndef CONFIG_COMMON_CLK #define __clk_put (clk)#define __clk_get (clk) ({ 1; }) +#endif #endif /* __CLKDEV_H__ */
drivers/clk/clk.c +24 -0
@@ -1850,6 +1850,30 @@ void devm_clk_unregister(struct device *dev, struct clk *clk) }EXPORT_SYMBOL_GPL (devm_clk_unregister); +/* +* clkdev helpers +*/ +int __clk_get (struct clk *clk) +{ +if (WARN_ON ((!clk))) +return 0 ; + +if (!try_module_get (clk->owner)) +return 0 ; + +return 1 ; +} +EXPORT_SYMBOL (__clk_get); + +void __clk_put (struct clk *clk) +{ +if (!clk ||IS_ERR (clk)) +return ; + +module_put (clk->owner); +} +EXPORT_SYMBOL (__clk_put); +/*** clk rate change notifiers ***/ /**
include/linux/clk-private.h +3 -0
@@ -25,10 +25,13 @@ #ifdef CONFIG_COMMON_CLK +struct module; + struct clk { const char *name; const struct clk_ops *ops; struct clk_hw *hw; + struct module *owner; struct clk *parent; const char **parent_names; struct clk **parents;
include/linux/clkdev.h +5 -0
@@ -43,4 +43,9 @@ int clk_add_alias(const char *, const char *, char *, struct device *);int clk_register_clkdev (struct clk *,const char *,const char *, ...);int clk_register_clkdevs (struct clk *,struct clk_lookup *,size_t ); +#ifdef CONFIG_COMMON_CLK +int __clk_get (struct clk *clk); +void __clk_put (struct clk *clk); +#endif +#endif
--
1.7.9.5
RFC PATCH 2/2 clk: implement clk_unregister
In reply to Sylwester Nawrockiclk_unregister() is currently not implemented and it is required when a clock provider module needs to be unloaded.
Normally the clock supplier module is prevented to be unloaded by taking reference on the module in clk_get().
For cases when the clock supplier module deinitializes despite the consumers of its clocks holding a reference on the module, e.g. when the driver is unbound through "unbind" sysfs attribute, there are empty clock ops added. These ops are assigned temporarily to struct clk and used until all consumers release the clock, to avoid invoking callbacks from the module which just got its driver remove() callback called on it.
---
Changes since v3 of the original patch [1]:
- reparent all children to the orphan list instead of leaving the clock unregistered when it has child clocks,
- removed unnecessary prerequisite checks in clk_debug_unregister(),
- struct clk is now being freed only when the last clock consumer calls clk_put(),
- empty clock ops are used after clk_unregister() has been called until all references to the clock are released and the clock object is freed.
[1] http://www.spinics.net/lists/arm-kernel/msg247548.html
--- drivers/clk/clk.c | 123 +++++++++++++++++++++++++++++++++++++++++-- include/linux/clk-private.h | 2 + 2 files changed, 122 insertions(+), 3 deletions(-)
drivers/clk/clk.c +120 -3
@@ -342,6 +342,21 @@ out:return ret; } +/** +* clk_debug_unregister - remove a clk node from the debugfs clk tree +* @clk: the clk being removed from the debugfs clk tree +* +* Dynamically removes a clk and all it's children clk nodes from the +* debugfs clk tree if clk->dentry points to debugfs created by +* clk_debug_register in __clk_init. +* +* Caller must hold prepare_lock. +*/ +static void clk_debug_unregister (struct clk *clk) +{ +debugfs_remove_recursive (clk->dentry); +} +/** * clk_debug_reparent - reparent clk node in the debugfs clk tree * @clk: the clk being reparented
@@ -432,6 +447,9 @@ static inline int clk_debug_register(struct clk *clk) { return 0; }static inline void clk_debug_reparent (struct clk *clk,struct clk *new_parent) { } +static inline void clk_debug_unregister (struct clk *clk) +{ +}#endif /* caller must hold prepare_lock */
@@ -1655,6 +1673,7 @@ int __clk_init(struct device *dev, struct clk *clk)clk_debug_register (clk); +kref_init (&clk->ref); out:clk_prepare_unlock ();
@@ -1784,13 +1803,106 @@ fail_out: }EXPORT_SYMBOL_GPL (clk_register); +/* +* Free memory allocated for a clock. +* Caller must hold prepare_lock. +*/ +static void __clk_release (struct kref *ref) +{ + struct clk *clk = container_of(ref, struct clk, ref); +int i = clk->num_parents; + +kfree (clk->parents); +while (--i >=0 ) +kfree (clk->parent_names[i]); + +kfree (clk->parent_names); +kfree (clk->name); +kfree (clk); +} + +/* +* Empty clk_ops for unregistered clocks. These are used temporarily +* after clk_unregister() was called on a clock and until last clock +* consumer calls clk_put() and the struct clk object is freed. +*/ +static int clk_dummy_prepare_enable (struct clk_hw *hw) +{ +return -ENXIO ; +} + +static void clk_dummy_disable_unprepare (struct clk_hw *hw) +{ +WARN_ON (1 ); +} + +static int clk_dummy_set_rate (struct clk_hw *hw,unsigned long rate, +unsigned long parent_rate) +{ +return -ENXIO ; +} + +static int clk_dummy_set_parent (struct clk_hw *hw,u8 index) +{ +return -ENXIO ; +} + +static const struct clk_ops clk_dummy_ops = { + .enable = clk_dummy_prepare_enable, + .disable = clk_dummy_disable_unprepare, + .prepare = clk_dummy_prepare_enable, + .unprepare = clk_dummy_disable_unprepare, + .set_rate = clk_dummy_set_rate, + .set_parent = clk_dummy_set_parent, +}; +/** * clk_unregister - unregister a currently registered clock * @clk: clock to unregister -* -* Currently unimplemented. */ -void clk_unregister (struct clk *clk) {} +void clk_unregister (struct clk *clk) +{ +unsigned long flags; + +clk_prepare_lock (); + +if (!clk ||IS_ERR (clk)) { +pr_err ("%s: invalid clock: %p\n" , __func__, clk); + goto out; + } + +if (clk->ops == &clk_dummy_ops) { +pr_err ("%s: unregistered clock: %s\n" , __func__, clk->name); + goto out; + } +/* +* Assign dummy clock ops for consumers that might still hold +* a reference to this clock. +*/ + flags =clk_enable_lock (); + clk->ops = &clk_dummy_ops; +clk_enable_unlock (flags); + +if (!hlist_empty (&clk->children)) { +struct clk *child; + +/* Reparent all children to the orphan list. */ +hlist_for_each_entry (child, &clk->children, child_node) +clk_set_parent (child,NULL ); + } + +clk_debug_unregister (clk); + +hlist_del_init (&clk->child_node); + +if (clk->prepare_count) +pr_warn ("%s: unregistering prepared clock: %s\n" , + __func__, clk->name); + +kref_put (&clk->ref, __clk_release); +out: +clk_prepare_unlock (); +}EXPORT_SYMBOL_GPL (clk_unregister);static void devm_clk_release (struct device *dev,void *res)
@@ -1861,6 +1973,7 @@ int __clk_get(struct clk *clk)if (!try_module_get (clk->owner))return 0 ; +kref_get (&clk->ref);return 1 ; }EXPORT_SYMBOL (__clk_get);
@@ -1870,6 +1983,10 @@ void __clk_put(struct clk *clk)if (!clk ||IS_ERR (clk))return ; +clk_prepare_lock (); +kref_put (&clk->ref, __clk_release); +clk_prepare_unlock (); +module_put (clk->owner); }EXPORT_SYMBOL (__clk_put);
include/linux/clk-private.h +2 -0
@@ -12,6 +12,7 @@#define __LINUX_CLK_PRIVATE_H#include <linux/clk-provider.h> +#include <linux/kref.h> #include <linux/list.h> /*
@@ -47,6 +48,7 @@ struct clk {#ifdef CONFIG_COMMON_CLK_DEBUG struct dentry *dentry;#endif +struct kref ref; };/*
--
1.7.9.5
Re: [PATCH RFC 2/2] clk: implement clk_unregister
Quoting Sylwester Nawrocki (2013-08-06 08:51:57)
Quoted 24 lines
drivers/clk/clk.c +24
-0+/* +* Empty clk_ops for unregistered clocks. These are used temporarily +* after clk_unregister() was called on a clock and until last clock +* consumer calls clk_put() and the struct clk object is freed. +*/ +static int clk_dummy_prepare_enable (struct clk_hw *hw) +{ +return -ENXIO ; +} + +static void clk_dummy_disable_unprepare (struct clk_hw *hw) +{ +WARN_ON (1 ); +} + +static int clk_dummy_set_rate (struct clk_hw *hw,unsigned long rate, +unsigned long parent_rate) +{ +return -ENXIO ; +} + +static int clk_dummy_set_parent (struct clk_hw *hw,u8 index) +{ +return -ENXIO ;
drivers/clk/clk.c fail_out: · line 1857
Quoted 10 lines
drivers/clk/clk.c +10
-0+} + +static const struct clk_ops clk_dummy_ops = { + .enable = clk_dummy_prepare_enable, + .disable = clk_dummy_disable_unprepare, + .prepare = clk_dummy_prepare_enable, + .unprepare = clk_dummy_disable_unprepare, + .set_rate = clk_dummy_set_rate, + .set_parent = clk_dummy_set_parent, +};
Don't use "clk_dummy_*" here. The use of dummy often implies that operations will return success in the absence of actual hardware but these return an error, and rightly so. So maybe rename the functions and clk_ops instance to something like "clk_nodev_*" or "clk_missing_*"?
Quoted 28 lines
drivers/clk/clk.c +21
-3+/** * clk_unregister - unregister a currently registered clock * @clk: clock to unregister -* -* Currently unimplemented. */ -void clk_unregister (struct clk *clk) {} +void clk_unregister (struct clk *clk) +{ +unsigned long flags; + +clk_prepare_lock (); + +if (!clk ||IS_ERR (clk)) { +pr_err ("%s: invalid clock: %p\n" , __func__, clk); + goto out; + } + +if (clk->ops == &clk_dummy_ops) { +pr_err ("%s: unregistered clock: %s\n" , __func__, clk->name); + goto out; + } +/* +* Assign dummy clock ops for consumers that might still hold +* a reference to this clock. +*/ + flags =clk_enable_lock ();
drivers/clk/clk.c fail_out: · line 1892
Quoted 10 lines
drivers/clk/clk.c +10
-0+ clk->ops = &clk_dummy_ops; +clk_enable_unlock (flags); + +if (!hlist_empty (&clk->children)) { +struct clk *child; + +/* Reparent all children to the orphan list. */ +hlist_for_each_entry (child, &clk->children, child_node) +clk_set_parent (child,NULL ); + }
This looks pretty good. A remaining problem is re-loading the clock provider module will have string name conflicts with the old unregistered clocks (but not yet released) clocks during calls to __clk_lookup.
The best solution would be to refactor clk.c to not use string name lookups but that is probably too big of an issue for the purpose of this series (but it will happen some day).
A short term solution would be to poison the clock's string name here. Reallocate the clk->name string with some poison data so that name conflicts don't occur. What do you think?
Regards,
Mike
Quoted 57 lines
drivers/clk/clk.c +18
-0+ +clk_debug_unregister (clk); + +hlist_del_init (&clk->child_node); + +if (clk->prepare_count) +pr_warn ("%s: unregistering prepared clock: %s\n" , + __func__, clk->name); + +kref_put (&clk->ref, __clk_release); +out: +clk_prepare_unlock (); +}EXPORT_SYMBOL_GPL (clk_unregister);static void devm_clk_release (struct device *dev,void *res)@@ -1861,6 +1973,7 @@ int __clk_get(struct clk *clk)if (!try_module_get (clk->owner))return 0 ; +kref_get (&clk->ref);return 1 ; }EXPORT_SYMBOL (__clk_get);@@ -1870,6 +1983,10 @@ void __clk_put(struct clk *clk)if (!clk ||IS_ERR (clk))return ; +clk_prepare_lock (); +kref_put (&clk->ref, __clk_release); +clk_prepare_unlock (); +module_put (clk->owner); }EXPORT_SYMBOL (__clk_put);include/linux/clk-private.h +2
-1@@ -12,6 +12,7 @@#define __LINUX_CLK_PRIVATE_H#include <linux/clk-provider.h> +#include <linux/kref.h> #include <linux/list.h> /* @@ -47,6 +48,7 @@ struct clk {#ifdef CONFIG_COMMON_CLK_DEBUG struct dentry *dentry;#endif +struct kref ref; };/* -- 1.7.9.5
Re: [PATCH RFC 2/2] clk: implement clk_unregister
Quoted 32 lines · Mike Turquette
Quoting Sylwester Nawrocki (2013-08-06 08:51:57)
Quoted 31 lines
drivers/clk/clk.c +31
-0+/* +* Empty clk_ops for unregistered clocks. These are used temporarily +* after clk_unregister() was called on a clock and until last clock +* consumer calls clk_put() and the struct clk object is freed. +*/ +static int clk_dummy_prepare_enable (struct clk_hw *hw) +{ +return -ENXIO ; +} + +static void clk_dummy_disable_unprepare (struct clk_hw *hw) +{ +WARN_ON (1 ); +} + +static int clk_dummy_set_rate (struct clk_hw *hw,unsigned long rate, +unsigned long parent_rate) +{ +return -ENXIO ; +} + +static int clk_dummy_set_parent (struct clk_hw *hw,u8 index) +{ +return -ENXIO ; +} + +static const struct clk_ops clk_dummy_ops = { + .enable = clk_dummy_prepare_enable, + .disable = clk_dummy_disable_unprepare, + .prepare = clk_dummy_prepare_enable, + .unprepare = clk_dummy_disable_unprepare,
Quoted 8 lines
drivers/clk/clk.c fail_out: · line 1857
Quoted 3 lines
drivers/clk/clk.c +3
-0+ .set_rate = clk_dummy_set_rate, + .set_parent = clk_dummy_set_parent, +};Don't use "clk_dummy_*" here. The use of dummy often implies that operations will return success in the absence of actual hardware but these return an error, and rightly so. So maybe rename the functions and
clk_opsinstance to something like "clk_nodev_*" or "clk_missing_*"?
Hmm, this is more about a driver being removed rather than the device. Then perhaps we could make it __clk_nodrv_* or clk_nodrv_* ?
Quoted 34 lines
Quoted 34 lines
/**
*clk_unregister- unregister a currently registered clock
* @clk: clock to unregister
- *
- * Currently unimplemented.
*/drivers/clk/clk.c +27
-1-void clk_unregister (struct clk *clk) {} +void clk_unregister (struct clk *clk) +{ +unsigned long flags; + +clk_prepare_lock (); + +if (!clk ||IS_ERR (clk)) { +pr_err ("%s: invalid clock: %p\n" , __func__, clk); + goto out; + } + +if (clk->ops == &clk_dummy_ops) { +pr_err ("%s: unregistered clock: %s\n" , __func__, clk->name); + goto out; + } +/* +* Assign dummy clock ops for consumers that might still hold +* a reference to this clock. +*/ + flags =clk_enable_lock (); + clk->ops = &clk_dummy_ops; +clk_enable_unlock (flags); + +if (!hlist_empty (&clk->children)) { +struct clk *child; + +/* Reparent all children to the orphan list. */
Quoted 8 lines
drivers/clk/clk.c fail_out: · line 1892
Quoted 3 lines
drivers/clk/clk.c +3
-0+hlist_for_each_entry (child, &clk->children, child_node) +clk_set_parent (child,NULL ); + }This looks pretty good. A remaining problem is re-loading the clock provider module will have string name conflicts with the old unregistered clocks (but not yet released) clocks during calls to
__clk_lookup.
But the clock is being dropped from the clock tree immediately in this function. After the hlist_del_init() call below the clock is not present on any clocks list. Upon clock release only the memory allocated for the clock is freed.
Quoted 7 lines
The best solution would be to refactor
clk.cto not use string name lookups but that is probably too big of an issue for the purpose of this series (but it will happen some day).A short term solution would be to poison the clock's string name here. Reallocate the
clk->namestring with some poison data so that name conflicts don't occur. What do you think?
This shouldn't be necessary, for the reason described above. I've tested multiple registrations when a clock was being referenced by a consumer driver and it worked well.
I'm still a bit unsure about the kref reference counting, but I'd would assume it is good to have. It prevents the kernel to crash in some situations. Many other subsystems/drivers crash miserably when a driver gets unbound using the sysfs "unbind" attribute. However, if it is assumed that user space needs to keep track of respective resource references and should know what it does when unbinding drivers, then we could probably do without the kref. I'm seriously sceptical though about letting user space to crash the kernel in fairly simple steps, it just doesn't sound right.
Quoted 31 lines
Regards,
MikeQuoted 28 lines
drivers/clk/clk.c +14
-0+ +clk_debug_unregister (clk); + +hlist_del_init (&clk->child_node); + +if (clk->prepare_count) +pr_warn ("%s: unregistering prepared clock: %s\n" , + __func__, clk->name); + +kref_put (&clk->ref, __clk_release); +out: +clk_prepare_unlock (); +}EXPORT_SYMBOL_GPL (clk_unregister);static void devm_clk_release (struct device *dev,void *res)@@ -1861,6 +1973,7 @@ int __clk_get(struct clk *clk)if (!try_module_get (clk->owner))return 0 ; +kref_get (&clk->ref);return 1 ; }EXPORT_SYMBOL (__clk_get);@@ -1870,6 +1983,10 @@ void __clk_put(struct clk *clk)if (!clk ||IS_ERR (clk))return ;
Quoted 7 lines
drivers/clk/clk.c void __clk_put(struct clk *clk) · line 1992
Quoted 7 lines
drivers/clk/clk.c +4
-0+clk_prepare_lock (); +kref_put (&clk->ref, __clk_release); +clk_prepare_unlock (); +module_put (clk->owner); }EXPORT_SYMBOL (__clk_put);
--
Regards,
Sylwester
RFC PATCH 2/2 clk: implement clk_unregister
Quoting Sylwester Nawrocki (2013-08-19 11:04:33)
Quoted 42 lines · Mike Turquette
Quoted 32 lines
Quoting Sylwester Nawrocki (2013-08-06 08:51:57)
Quoted 31 lines
drivers/clk/clk.c +31
-0+/* +* Empty clk_ops for unregistered clocks. These are used temporarily +* after clk_unregister() was called on a clock and until last clock +* consumer calls clk_put() and the struct clk object is freed. +*/ +static int clk_dummy_prepare_enable (struct clk_hw *hw) +{ +return -ENXIO ; +} + +static void clk_dummy_disable_unprepare (struct clk_hw *hw) +{ +WARN_ON (1 ); +} + +static int clk_dummy_set_rate (struct clk_hw *hw,unsigned long rate, +unsigned long parent_rate) +{ +return -ENXIO ; +} + +static int clk_dummy_set_parent (struct clk_hw *hw,u8 index) +{ +return -ENXIO ; +} + +static const struct clk_ops clk_dummy_ops = { + .enable = clk_dummy_prepare_enable, + .disable = clk_dummy_disable_unprepare, + .prepare = clk_dummy_prepare_enable, + .unprepare = clk_dummy_disable_unprepare,Quoted 8 lines
drivers/clk/clk.c fail_out: · line 1857
Quoted 3 lines
drivers/clk/clk.c +3
-0+ .set_rate = clk_dummy_set_rate, + .set_parent = clk_dummy_set_parent, +};Don't use "clk_dummy_*" here. The use of dummy often implies that operations will return success in the absence of actual hardware but these return an error, and rightly so. So maybe rename the functions and
clk_opsinstance to something like "clk_nodev_*" or "clk_missing_*"?
Quoted 2 lines
Hmm, this is more about a driver being removed rather than the device. Then perhaps we could make it __clk_nodrv_* or clk_nodrv_* ?
clk_nodrv_* sounds good.
Quoted 44 lines
Quoted 35 lines
Quoted 35 lines
/**
*clk_unregister- unregister a currently registered clock
* @clk: clock to unregister
- *
- * Currently unimplemented.
*/drivers/clk/clk.c +27
-1-void clk_unregister (struct clk *clk) {} +void clk_unregister (struct clk *clk) +{ +unsigned long flags; + +clk_prepare_lock (); + +if (!clk ||IS_ERR (clk)) { +pr_err ("%s: invalid clock: %p\n" , __func__, clk); + goto out; + } + +if (clk->ops == &clk_dummy_ops) { +pr_err ("%s: unregistered clock: %s\n" , __func__, clk->name); + goto out; + } +/* +* Assign dummy clock ops for consumers that might still hold +* a reference to this clock. +*/ + flags =clk_enable_lock (); + clk->ops = &clk_dummy_ops; +clk_enable_unlock (flags); + +if (!hlist_empty (&clk->children)) { +struct clk *child; + +/* Reparent all children to the orphan list. */ Quoted 8 lines
drivers/clk/clk.c fail_out: · line 1892
Quoted 3 lines
drivers/clk/clk.c +3
-0+hlist_for_each_entry (child, &clk->children, child_node) +clk_set_parent (child,NULL ); + }This looks pretty good. A remaining problem is re-loading the clock provider module will have string name conflicts with the old unregistered clocks (but not yet released) clocks during calls to
__clk_lookup.
Quoted 4 lines
But the clock is being dropped from the clock tree immediately in this function. After the
hlist_del_init()call below the clock is not present on any clocks list. Upon clock release only the memory allocated for the clock is freed.
You are correct. Not sure why I thought that the clock being unregistered was also getting pushed to the orphan list.
Quoted 13 lines
Quoted 8 lines
The best solution would be to refactor
clk.cto not use string name lookups but that is probably too big of an issue for the purpose of this series (but it will happen some day).A short term solution would be to poison the clock's string name here. Reallocate the
clk->namestring with some poison data so that name conflicts don't occur. What do you think?This shouldn't be necessary, for the reason described above. I've tested multiple registrations when a clock was being referenced by a consumer driver and it worked well.
Quoted 9 lines
I'm still a bit unsure about the kref reference counting, but I'd would assume it is good to have. It prevents the kernel to crash in some situations. Many other subsystems/drivers crash miserably when a driver gets unbound using the sysfs "unbind" attribute. However, if it is assumed that user space needs to keep track of respective resource references and should know what it does when unbinding drivers, then we could probably do without the kref. I'm seriously sceptical though about letting user space to crash the kernel in fairly simple steps, it just doesn't sound right.
Let's leave the kref bits in. If we can prove that they are unnecessary in the future then they can always be removed.
This series looks good, barring the s/dummy/no_drv/ rename. Russell's ACK is needed for patch #1.
Regards,
Mike
Quoted 43 lines
Quoted 32 lines
Regards,
MikeQuoted 28 lines
drivers/clk/clk.c +14
-0+ +clk_debug_unregister (clk); + +hlist_del_init (&clk->child_node); + +if (clk->prepare_count) +pr_warn ("%s: unregistering prepared clock: %s\n" , + __func__, clk->name); + +kref_put (&clk->ref, __clk_release); +out: +clk_prepare_unlock (); +}EXPORT_SYMBOL_GPL (clk_unregister);static void devm_clk_release (struct device *dev,void *res)@@ -1861,6 +1973,7 @@ int __clk_get(struct clk *clk)if (!try_module_get (clk->owner))return 0 ; +kref_get (&clk->ref);return 1 ; }EXPORT_SYMBOL (__clk_get);@@ -1870,6 +1983,10 @@ void __clk_put(struct clk *clk)if (!clk ||IS_ERR (clk))return ;Quoted 7 lines
drivers/clk/clk.c void __clk_put(struct clk *clk) · line 1992
Quoted 7 lines
drivers/clk/clk.c +4
-0+clk_prepare_lock (); +kref_put (&clk->ref, __clk_release); +clk_prepare_unlock (); +module_put (clk->owner); }EXPORT_SYMBOL (__clk_put);--
Regards,
Sylwester