* [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get()
@ 2025-07-31 12:39 Chuan Liu via B4 Relay
2025-07-31 16:45 ` Stephen Boyd
0 siblings, 1 reply; 4+ messages in thread
From: Chuan Liu via B4 Relay @ 2025-07-31 12:39 UTC (permalink / raw)
To: Michael Turquette, Stephen Boyd; +Cc: linux-clk, linux-kernel, Chuan Liu
From: Chuan Liu <chuan.liu@amlogic.com>
In specific cases, even a clk_provider managing only a single clock may
reference of_clk_hw_onecell_get() to access its member clocks, as seen
in implementations like clk-scmi.
For a clk_provider with only one clock, when calling
of_parse_phandle_with_args(), the phandle_args->args[] members are not
assigned. In this case, the reference to phandle_args->args[0] in
of_clk_hw_onecell_get() becomes invalid. If phandle_args->args[0]
initially contains a non-zero value, this will trigger an error.
Signed-off-by: Chuan Liu <chuan.liu@amlogic.com>
---
Error conditions observed:
scmi_clk: protocol@14 {
reg = <0x14>;
#clock-cells = <0>;
};
phandle1: clock-controller@1 {
#clock-cells = <1>;
}
clock-consumer@2 {
assigned-clocks = <&phandle1 1>,
<&scmi_clk>;
assigned-clock-rates = <xxx>,
<xxx>;
}
Under these conditions, executing of_clk_set_defaults() triggers the
error: 'of_clk_hw_onecell_get: invalid index 1'.
While the root cause lies in invalid input data to
of_clk_hw_onecell_get(), checking phandle->args_count for data validity
improves driver fault tolerance.
---
drivers/clk/clk.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
index 0565c87656cf..4994c551befe 100644
--- a/drivers/clk/clk.c
+++ b/drivers/clk/clk.c
@@ -4966,7 +4966,7 @@ EXPORT_SYMBOL_GPL(of_clk_hw_simple_get);
struct clk *of_clk_src_onecell_get(struct of_phandle_args *clkspec, void *data)
{
struct clk_onecell_data *clk_data = data;
- unsigned int idx = clkspec->args[0];
+ unsigned int idx = clkspec->args_count ? clkspec->args[0] : 0;
if (idx >= clk_data->clk_num) {
pr_err("%s: invalid clock index %u\n", __func__, idx);
@@ -4981,7 +4981,7 @@ struct clk_hw *
of_clk_hw_onecell_get(struct of_phandle_args *clkspec, void *data)
{
struct clk_hw_onecell_data *hw_data = data;
- unsigned int idx = clkspec->args[0];
+ unsigned int idx = clkspec->args_count ? clkspec->args[0] : 0;
if (idx >= hw_data->num) {
pr_err("%s: invalid index %u\n", __func__, idx);
---
base-commit: 58abdca0eb653c1a2e755ba9ba406ee475d87636
change-id: 20250730-add_fault_tolerance_to_of_clk_hw_onecell_get-fdc81f7e35a5
Best regards,
--
Chuan Liu <chuan.liu@amlogic.com>
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get()
2025-07-31 12:39 [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get() Chuan Liu via B4 Relay
@ 2025-07-31 16:45 ` Stephen Boyd
2025-08-01 2:40 ` Chuan Liu
2025-08-05 6:13 ` Chuan Liu
0 siblings, 2 replies; 4+ messages in thread
From: Stephen Boyd @ 2025-07-31 16:45 UTC (permalink / raw)
To: Chuan Liu via B4 Relay, Michael Turquette, chuan.liu
Cc: linux-clk, linux-kernel, Chuan Liu
Quoting Chuan Liu via B4 Relay (2025-07-31 05:39:58)
> From: Chuan Liu <chuan.liu@amlogic.com>
>
> In specific cases, even a clk_provider managing only a single clock may
> reference of_clk_hw_onecell_get() to access its member clocks, as seen
> in implementations like clk-scmi.
>
> For a clk_provider with only one clock, when calling
> of_parse_phandle_with_args(), the phandle_args->args[] members are not
> assigned. In this case, the reference to phandle_args->args[0] in
> of_clk_hw_onecell_get() becomes invalid. If phandle_args->args[0]
> initially contains a non-zero value, this will trigger an error.
>
> Signed-off-by: Chuan Liu <chuan.liu@amlogic.com>
> ---
> Error conditions observed:
>
> scmi_clk: protocol@14 {
> reg = <0x14>;
> #clock-cells = <0>;
> };
>
> phandle1: clock-controller@1 {
> #clock-cells = <1>;
> }
>
> clock-consumer@2 {
> assigned-clocks = <&phandle1 1>,
> <&scmi_clk>;
> assigned-clock-rates = <xxx>,
> <xxx>;
> }
>
> Under these conditions, executing of_clk_set_defaults() triggers the
> error: 'of_clk_hw_onecell_get: invalid index 1'.
Please write a KUnit test.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get()
2025-07-31 16:45 ` Stephen Boyd
@ 2025-08-01 2:40 ` Chuan Liu
2025-08-05 6:13 ` Chuan Liu
1 sibling, 0 replies; 4+ messages in thread
From: Chuan Liu @ 2025-08-01 2:40 UTC (permalink / raw)
To: Stephen Boyd, Chuan Liu via B4 Relay, Michael Turquette
Cc: linux-clk, linux-kernel
Hi Stephen:
Thanks for the quick feedback. I'll improve the KUnit tests and send v2.
On 8/1/2025 12:45 AM, Stephen Boyd wrote:
> [ EXTERNAL EMAIL ]
>
> Quoting Chuan Liu via B4 Relay (2025-07-31 05:39:58)
>> From: Chuan Liu <chuan.liu@amlogic.com>
>>
>> In specific cases, even a clk_provider managing only a single clock may
>> reference of_clk_hw_onecell_get() to access its member clocks, as seen
>> in implementations like clk-scmi.
>>
>> For a clk_provider with only one clock, when calling
>> of_parse_phandle_with_args(), the phandle_args->args[] members are not
>> assigned. In this case, the reference to phandle_args->args[0] in
>> of_clk_hw_onecell_get() becomes invalid. If phandle_args->args[0]
>> initially contains a non-zero value, this will trigger an error.
>>
>> Signed-off-by: Chuan Liu <chuan.liu@amlogic.com>
>> ---
>> Error conditions observed:
>>
>> scmi_clk: protocol@14 {
>> reg = <0x14>;
>> #clock-cells = <0>;
>> };
>>
>> phandle1: clock-controller@1 {
>> #clock-cells = <1>;
>> }
>>
>> clock-consumer@2 {
>> assigned-clocks = <&phandle1 1>,
>> <&scmi_clk>;
>> assigned-clock-rates = <xxx>,
>> <xxx>;
>> }
>>
>> Under these conditions, executing of_clk_set_defaults() triggers the
>> error: 'of_clk_hw_onecell_get: invalid index 1'.
> Please write a KUnit test.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get()
2025-07-31 16:45 ` Stephen Boyd
2025-08-01 2:40 ` Chuan Liu
@ 2025-08-05 6:13 ` Chuan Liu
1 sibling, 0 replies; 4+ messages in thread
From: Chuan Liu @ 2025-08-05 6:13 UTC (permalink / raw)
To: Stephen Boyd, Chuan Liu via B4 Relay, Michael Turquette
Cc: linux-clk, linux-kernel
Hi Stephen:
After reviewing all CCF drivers referencing of_clk_hw_onecell_get(),
I confirmed their corresponding binding *.yaml files define
'#clock-cells' as 'const: 1'.
Therefore, the case I previously described is currently disallowed.
Please ignore this patch. Thanks.
On 8/1/2025 12:45 AM, Stephen Boyd wrote:
> [ EXTERNAL EMAIL ]
>
> Quoting Chuan Liu via B4 Relay (2025-07-31 05:39:58)
>> From: Chuan Liu <chuan.liu@amlogic.com>
>>
>> In specific cases, even a clk_provider managing only a single clock may
>> reference of_clk_hw_onecell_get() to access its member clocks, as seen
>> in implementations like clk-scmi.
>>
>> For a clk_provider with only one clock, when calling
>> of_parse_phandle_with_args(), the phandle_args->args[] members are not
>> assigned. In this case, the reference to phandle_args->args[0] in
>> of_clk_hw_onecell_get() becomes invalid. If phandle_args->args[0]
>> initially contains a non-zero value, this will trigger an error.
>>
>> Signed-off-by: Chuan Liu <chuan.liu@amlogic.com>
>> ---
>> Error conditions observed:
>>
>> scmi_clk: protocol@14 {
>> reg = <0x14>;
>> #clock-cells = <0>;
>> };
>>
>> phandle1: clock-controller@1 {
>> #clock-cells = <1>;
>> }
>>
>> clock-consumer@2 {
>> assigned-clocks = <&phandle1 1>,
>> <&scmi_clk>;
>> assigned-clock-rates = <xxx>,
>> <xxx>;
>> }
>>
>> Under these conditions, executing of_clk_set_defaults() triggers the
>> error: 'of_clk_hw_onecell_get: invalid index 1'.
> Please write a KUnit test.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2025-08-05 6:13 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-07-31 12:39 [PATCH] clk: Add fault tolerance to of_clk_hw_onecell_get() Chuan Liu via B4 Relay
2025-07-31 16:45 ` Stephen Boyd
2025-08-01 2:40 ` Chuan Liu
2025-08-05 6:13 ` Chuan Liu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).