The function meson_clk_pll_enable() can be invoked under the enable_lock
spinlock from the clk core logic, which risks a kernel panic during the
usleep_range() call:
BUG: scheduling while atomic: kworker/u4:2/36/0x00000002
Modules linked in: g_ffs usb_f_fs libcomposite
CPU: 1 PID: 36 Comm: kworker/u4:2 Not tainted 6.4.0-rc5 #273
Workqueue: events_unbound async_run_entry_fn
Call trace:
dump_backtrace+0x9c/0x128
show_stack+0x20/0x38
dump_stack_lvl+0x48/0x60
dump_stack+0x18/0x28
__schedule_bug+0x58/0x78
__schedule+0x828/0xa88
schedule+0x64/0xd8
schedule_hrtimeout_range_clock+0xd0/0x208
schedule_hrtimeout_range+0x1c/0x30
usleep_range_state+0x6c/0xa8
meson_clk_pll_enable+0x1f4/0x310
clk_core_enable+0x78/0x200
clk_core_enable+0x58/0x200
clk_core_enable+0x58/0x200
clk_core_enable+0x58/0x200
clk_enable+0x34/0x60
So it is required to use the udelay() function instead of usleep_range()
for the atomic context safety.
Fixes: b6ec400aa153 ("clk: meson: introduce new pll power-on sequence for A1 SoC family")
Reported-by: Jan Dakinevich <[email protected]>
Signed-off-by: Dmitry Rokosov <[email protected]>
---
Changes v2 since v1 at [1]:
- remove redundant SoB from commit msg
- drop unuseful reference to clk_core_enable() from commit msg
Links:
[1] https://lore.kernel.org/all/[email protected]/
---
drivers/clk/meson/clk-pll.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/clk/meson/clk-pll.c b/drivers/clk/meson/clk-pll.c
index 56ec2210f1ad..eef6f37c8d8d 100644
--- a/drivers/clk/meson/clk-pll.c
+++ b/drivers/clk/meson/clk-pll.c
@@ -367,9 +367,9 @@ static int meson_clk_pll_enable(struct clk_hw *hw)
* 3. enable the lock detect module
*/
if (MESON_PARM_APPLICABLE(&pll->current_en)) {
- usleep_range(10, 20);
+ udelay(10);
meson_parm_write(clk->map, &pll->current_en, 1);
- usleep_range(40, 50);
+ udelay(40);
};
if (MESON_PARM_APPLICABLE(&pll->l_detect)) {
--
2.36.0
On Wed 05 Jul 2023 at 00:54, Dmitry Rokosov <[email protected]> wrote:
> The function meson_clk_pll_enable() can be invoked under the enable_lock
> spinlock from the clk core logic, which risks a kernel panic during the
> usleep_range() call:
>
> BUG: scheduling while atomic: kworker/u4:2/36/0x00000002
> Modules linked in: g_ffs usb_f_fs libcomposite
> CPU: 1 PID: 36 Comm: kworker/u4:2 Not tainted 6.4.0-rc5 #273
> Workqueue: events_unbound async_run_entry_fn
> Call trace:
> dump_backtrace+0x9c/0x128
> show_stack+0x20/0x38
> dump_stack_lvl+0x48/0x60
> dump_stack+0x18/0x28
> __schedule_bug+0x58/0x78
> __schedule+0x828/0xa88
> schedule+0x64/0xd8
> schedule_hrtimeout_range_clock+0xd0/0x208
> schedule_hrtimeout_range+0x1c/0x30
> usleep_range_state+0x6c/0xa8
> meson_clk_pll_enable+0x1f4/0x310
> clk_core_enable+0x78/0x200
> clk_core_enable+0x58/0x200
> clk_core_enable+0x58/0x200
> clk_core_enable+0x58/0x200
> clk_enable+0x34/0x60
>
> So it is required to use the udelay() function instead of usleep_range()
> for the atomic context safety.
>
> Fixes: b6ec400aa153 ("clk: meson: introduce new pll power-on sequence for A1 SoC family")
> Reported-by: Jan Dakinevich <[email protected]>
> Signed-off-by: Dmitry Rokosov <[email protected]>
Applied, Thx
> ---
> Changes v2 since v1 at [1]:
> - remove redundant SoB from commit msg
> - drop unuseful reference to clk_core_enable() from commit msg
>
> Links:
> [1] https://lore.kernel.org/all/[email protected]/
> ---
> drivers/clk/meson/clk-pll.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/clk/meson/clk-pll.c b/drivers/clk/meson/clk-pll.c
> index 56ec2210f1ad..eef6f37c8d8d 100644
> --- a/drivers/clk/meson/clk-pll.c
> +++ b/drivers/clk/meson/clk-pll.c
> @@ -367,9 +367,9 @@ static int meson_clk_pll_enable(struct clk_hw *hw)
> * 3. enable the lock detect module
> */
> if (MESON_PARM_APPLICABLE(&pll->current_en)) {
> - usleep_range(10, 20);
> + udelay(10);
> meson_parm_write(clk->map, &pll->current_en, 1);
> - usleep_range(40, 50);
> + udelay(40);
> };
>
> if (MESON_PARM_APPLICABLE(&pll->l_detect)) {