| Message ID | 20130115224502.23734.7414@quantum |
|---|---|
| State | New |
| Headers | show |
On 16 January 2013 04:15, Mike Turquette <mturquette@linaro.org> wrote: > Quoting Rajagopal Venkat (2013-01-08 22:29:48) >> while reparenting a clock, NULL check is done for clock in >> consideration and its new parent. So re-check is not required. >> If done, else part becomes unreachable. >> >> Signed-off-by: Rajagopal Venkat <rajagopal.venkat@linaro.org> >> --- >> drivers/clk/clk.c | 13 ++----------- >> 1 file changed, 2 insertions(+), 11 deletions(-) >> >> diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c >> index 251e45d..1c4097c 100644 >> --- a/drivers/clk/clk.c >> +++ b/drivers/clk/clk.c >> @@ -1040,7 +1040,6 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) >> { >> #ifdef CONFIG_COMMON_CLK_DEBUG >> struct dentry *d; >> - struct dentry *new_parent_d; >> #endif >> >> if (!clk || !new_parent) >> @@ -1048,22 +1047,14 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) >> >> hlist_del(&clk->child_node); >> >> - if (new_parent) >> - hlist_add_head(&clk->child_node, &new_parent->children); >> - else >> - hlist_add_head(&clk->child_node, &clk_orphan_list); >> + hlist_add_head(&clk->child_node, &new_parent->children); > > Rajagopal, > > You found a bug, but not the right one :) > > This change would result in never moving a clock into the orphan list if > the parent is missing. The right thing to do is to allow the operation > to succeed and migrate this clock into the orphan list. In keeping > with the clk.h api we need to treat struct clk new_parent as an opaque > cookie and not care whether or not it is NULL. > Mike, The only path where __clk_reparent() can get new parent as NULL is from clk_set_parent() api. Tracing this path, here are the reasons why I submitted this patch. 1. The clk_set_parent() function takes new parent for granted and tries to access new_parent->rate when notifiers are registered, 2. For clocks with multiple parents, the __clk_set_parent() function also takes new parent for granted and tries to access parent name while finding the index of new parent from cached parent pointers. But if the design decision is to allow clock to be re-parented to orphan list when new parent is missing, I can provide the new working patch. > Untested patch below fixes the real bug: > > > > From a4d56e3ee51452366365749873710e16631e9de7 Mon Sep 17 00:00:00 2001 > From: Mike Turquette <mturquette@linaro.org> > Date: Tue, 15 Jan 2013 14:39:06 -0800 > Subject: [PATCH] clk: allow re-parenting to NULL clks > > __clk_reparent presently bails early if the new parent of a clk is NULL. > This is wrong and prevents dynamically migrating clocks into the orphan > list. The fix is to remove the NULL pointer check for new_parent in > __clk_parent. > > Signed-off-by: Mike Turquette <mturquette@linaro.org> > --- > drivers/clk/clk.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > index 593a2e4..f056230 100644 > --- a/drivers/clk/clk.c > +++ b/drivers/clk/clk.c > @@ -1186,7 +1186,7 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) > struct dentry *new_parent_d; > #endif > > - if (!clk || !new_parent) > + if (!clk) > return; > > hlist_del(&clk->child_node); > -- > 1.7.10.4 >
Quoting Rajagopal Venkat (2013-01-16 04:45:13) > On 16 January 2013 04:15, Mike Turquette <mturquette@linaro.org> wrote: > > Quoting Rajagopal Venkat (2013-01-08 22:29:48) > >> while reparenting a clock, NULL check is done for clock in > >> consideration and its new parent. So re-check is not required. > >> If done, else part becomes unreachable. > >> > >> Signed-off-by: Rajagopal Venkat <rajagopal.venkat@linaro.org> > >> --- > >> drivers/clk/clk.c | 13 ++----------- > >> 1 file changed, 2 insertions(+), 11 deletions(-) > >> > >> diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > >> index 251e45d..1c4097c 100644 > >> --- a/drivers/clk/clk.c > >> +++ b/drivers/clk/clk.c > >> @@ -1040,7 +1040,6 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) > >> { > >> #ifdef CONFIG_COMMON_CLK_DEBUG > >> struct dentry *d; > >> - struct dentry *new_parent_d; > >> #endif > >> > >> if (!clk || !new_parent) > >> @@ -1048,22 +1047,14 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) > >> > >> hlist_del(&clk->child_node); > >> > >> - if (new_parent) > >> - hlist_add_head(&clk->child_node, &new_parent->children); > >> - else > >> - hlist_add_head(&clk->child_node, &clk_orphan_list); > >> + hlist_add_head(&clk->child_node, &new_parent->children); > > > > Rajagopal, > > > > You found a bug, but not the right one :) > > > > This change would result in never moving a clock into the orphan list if > > the parent is missing. The right thing to do is to allow the operation > > to succeed and migrate this clock into the orphan list. In keeping > > with the clk.h api we need to treat struct clk new_parent as an opaque > > cookie and not care whether or not it is NULL. > > > Mike, > > The only path where __clk_reparent() can get new parent as NULL is from > clk_set_parent() api. Tracing this path, here are the reasons why I submitted > this patch. > > 1. The clk_set_parent() function takes new parent for granted and tries to > access new_parent->rate when notifiers are registered, > 2. For clocks with multiple parents, the __clk_set_parent() function also > takes new parent for granted and tries to access parent name while finding > the index of new parent from cached parent pointers. > > But if the design decision is to allow clock to be re-parented to orphan list > when new parent is missing, I can provide the new working patch. > Rajagopal, Yes, reparenting a clock to the orphan list should be supported. Let me know if you are going to spin a new patch or not. Thanks, Mike > > Untested patch below fixes the real bug: > > > > > > > > From a4d56e3ee51452366365749873710e16631e9de7 Mon Sep 17 00:00:00 2001 > > From: Mike Turquette <mturquette@linaro.org> > > Date: Tue, 15 Jan 2013 14:39:06 -0800 > > Subject: [PATCH] clk: allow re-parenting to NULL clks > > > > __clk_reparent presently bails early if the new parent of a clk is NULL. > > This is wrong and prevents dynamically migrating clocks into the orphan > > list. The fix is to remove the NULL pointer check for new_parent in > > __clk_parent. > > > > Signed-off-by: Mike Turquette <mturquette@linaro.org> > > --- > > drivers/clk/clk.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > > index 593a2e4..f056230 100644 > > --- a/drivers/clk/clk.c > > +++ b/drivers/clk/clk.c > > @@ -1186,7 +1186,7 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) > > struct dentry *new_parent_d; > > #endif > > > > - if (!clk || !new_parent) > > + if (!clk) > > return; > > > > hlist_del(&clk->child_node); > > -- > > 1.7.10.4 > > > > > > -- > Regards, > Rajagopal
diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index 593a2e4..f056230 100644 --- a/drivers/clk/clk.c +++ b/drivers/clk/clk.c @@ -1186,7 +1186,7 @@ void __clk_reparent(struct clk *clk, struct clk *new_parent) struct dentry *new_parent_d; #endif - if (!clk || !new_parent) + if (!clk) return; hlist_del(&clk->child_node);