On 11/2/2023 2:19 AM, Sean Anderson wrote:
On 8/17/23 13:04, Yang Xiwen via B4 Relay wrote:
From: Yang Xiwen forbidden405@outlook.com
Calling into CCF framework will cause a clock being enabled twice instead of once (clk->enable_count becomes 2 rather than 1), thus making it hard to disable (needs to call clk_disable() twice). Fix that by calling clock provided ops directly.
Can you describe this scenario more? From what I can tell, clk_enable doesn't increment enable_count for CCF clocks.
Well, it's hard to describe clearly. But I can only tell this patch fixed the issue when i was trying to write an Ethernet driver[1] which calls clk_disable() and expects the clock to be disabled after that. Also I found that CCF driver does not have a corresponding test file. I will try to write a test for that in next release.
--Sean
Signed-off-by: Yang Xiwen forbidden405@outlook.com
drivers/clk/clk.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index a38daaac0c..00d082c46f 100644 --- a/drivers/clk/clk.c +++ b/drivers/clk/clk.c @@ -14,6 +14,7 @@ #include <dm/uclass.h> #include <dm/lists.h> #include <dm/device-internal.h> +#include <linux/clk-provider.h> int clk_register(struct clk *clk, const char *drv_name, const char *name, const char *parent_name) @@ -115,11 +116,20 @@ int ccf_clk_set_parent(struct clk *clk, struct clk *parent) static int ccf_clk_endisable(struct clk *clk, bool enable) { struct clk *c; + const struct clk_ops *ops; int err = clk_get_by_id(clk->id, &c); if (err) return err; - return enable ? clk_enable(c) : clk_disable(c); + else + ops = clk_dev_ops(c->dev);
+ if (enable && ops->enable) + return ops->enable(c); + else if (!enable && ops->disable) + return ops->disable(c);
+ return -ENOSYS; } int ccf_clk_enable(struct clk *clk)
[1] https://lore.kernel.org/all/20230814-wip-hisi_femac-trunk-v2-0-1e29f400585d@...