Re: [PATCH] clk: qcom: clk-rpmh: Wait for completion when enabling clocks

From: Bjorn Andersson
Date: Sat Feb 15 2020 - 02:33:38 EST


On Fri 14 Feb 18:12 PST 2020, Mike Tipton wrote:

> The current implementation always uses rpmh_write_async, which doesn't
> wait for completion. That's fine for disable requests since there's no
> immediate need for the clocks and they can be disabled in the
> background. However, for enable requests we need to ensure the clocks
> are actually enabled before returning to the client. Otherwise, clients
> can end up accessing their HW before the necessary clocks are enabled,
> which can lead to bus errors.
>
> Use the synchronous version of this API (rpmh_write) for enable requests
> in the active set to ensure completion.
>
> Completion isn't required for sleep/wake sets, since they don't take
> effect until after we enter sleep. All rpmh requests are automatically
> flushed prior to entering sleep.
>
> Fixes: 9c7e47025a6b ("clk: qcom: clk-rpmh: Add QCOM RPMh clock driver")
> Signed-off-by: Mike Tipton <mdtipton@xxxxxxxxxxxxxx>

Reviewed-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>

Regards,
Bjorn

> ---
> drivers/clk/qcom/clk-rpmh.c | 18 ++++++++++++++++--
> 1 file changed, 16 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/clk/qcom/clk-rpmh.c b/drivers/clk/qcom/clk-rpmh.c
> index 12bd8715dece..3137595a736b 100644
> --- a/drivers/clk/qcom/clk-rpmh.c
> +++ b/drivers/clk/qcom/clk-rpmh.c
> @@ -143,6 +143,19 @@ static inline bool has_state_changed(struct clk_rpmh *c, u32 state)
> != (c->aggr_state & BIT(state));
> }
>
> +static int clk_rpmh_send(struct clk_rpmh *c, enum rpmh_state state,
> + struct tcs_cmd *cmd, bool wait_for_completion)
> +{
> + int ret;
> +
> + if (wait_for_completion)
> + ret = rpmh_write(c->dev, state, cmd, 1);
> + else
> + ret = rpmh_write_async(c->dev, state, cmd, 1);
> +
> + return ret;
> +}
> +
> static int clk_rpmh_send_aggregate_command(struct clk_rpmh *c)
> {
> struct tcs_cmd cmd = { 0 };
> @@ -159,7 +172,8 @@ static int clk_rpmh_send_aggregate_command(struct clk_rpmh *c)
> if (cmd_state & BIT(state))
> cmd.data = on_val;
>
> - ret = rpmh_write_async(c->dev, state, &cmd, 1);
> + ret = clk_rpmh_send(c, state, &cmd,
> + cmd_state && state == RPMH_ACTIVE_ONLY_STATE);
> if (ret) {
> dev_err(c->dev, "set %s state of %s failed: (%d)\n",
> !state ? "sleep" :
> @@ -267,7 +281,7 @@ static int clk_rpmh_bcm_send_cmd(struct clk_rpmh *c, bool enable)
> cmd.addr = c->res_addr;
> cmd.data = BCM_TCS_CMD(1, enable, 0, cmd_state);
>
> - ret = rpmh_write_async(c->dev, RPMH_ACTIVE_ONLY_STATE, &cmd, 1);
> + ret = clk_rpmh_send(c, RPMH_ACTIVE_ONLY_STATE, &cmd, enable);
> if (ret) {
> dev_err(c->dev, "set active state of %s failed: (%d)\n",
> c->res_name, ret);
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project