Re: [PATCH 2/3] cpuidle/powernv: Change platform init to avoid reparsing dt

From: Gautham R Shenoy
Date: Wed Jun 20 2018 - 00:53:09 EST


Hi Akshay,


On Tue, Jun 19, 2018 at 10:34:27AM +0530, Akshay Adiga wrote:
> The required data is accessible from cpuidle_states structure and
> nr_cpu_idle_states. This patch makes changes to avoid reparsing and use
> data from these structures.
>
> Signed-off-by: Akshay Adiga <akshay.adiga@xxxxxxxxxxxxxxxxxx>
> ---
> arch/powerpc/platforms/powernv/idle.c | 37 ++++++++---------------------------
> 1 file changed, 8 insertions(+), 29 deletions(-)
>
> diff --git a/arch/powerpc/platforms/powernv/idle.c b/arch/powerpc/platforms/powernv/idle.c
> index 07be984..0a62611 100644
> --- a/arch/powerpc/platforms/powernv/idle.c
> +++ b/arch/powerpc/platforms/powernv/idle.c
> @@ -624,8 +624,7 @@ int validate_psscr_val_mask(u64 *psscr_val, u64 *psscr_mask, u32 flags)
> * @dt_idle_states: Number of idle state entries
> * Returns 0 on success
> */
> -static int __init pnv_power9_idle_init(struct device_node *np, u32 *flags,
> - int dt_idle_states)
> +static int __init pnv_power9_idle_init(void)
> {
> u64 max_residency_ns = 0;
> int i;
> @@ -644,7 +643,7 @@ static int __init pnv_power9_idle_init(struct device_node *np, u32 *flags,
> * the shallowest (OPAL_PM_STOP_INST_FAST) loss-less stop state.
> */
> pnv_first_deep_stop_state = MAX_STOP_STATE;
> - for (i = 0; i < dt_idle_states; i++) {
> + for (i = 0; i < nr_pnv_idle_states; i++) {
> int err;
> struct pnv_idle_states_t *state = &pnv_idle_states[i];
> u64 psscr_rl = state->pm_ctrl_reg_val & PSSCR_RL_MASK;

These above two hunks could be folded into the first patch.

> @@ -704,41 +703,21 @@ static int __init pnv_power9_idle_init(struct device_node *np, u32 *flags,
> */
> static void __init pnv_probe_idle_states(void)

Can we move the remaining content of this function into the calling
function pnv_init_idle_states() ? That way...

> {
> - struct device_node *np;
> - int dt_idle_states;
> - u32 *flags = NULL;
> int i;
>
> - np = of_find_node_by_path("/ibm,opal/power-mgt");
> - if (!np) {
> - pr_warn("opal: PowerMgmt Node not found\n");
> - goto out;
> - }
> - dt_idle_states = of_property_count_u32_elems(np,
> - "ibm,cpu-idle-state-flags");
> - if (dt_idle_states < 0) {
> + if (nr_pnv_idle_states < 0) {

... This reduntant check can be removed. Because nr_pnv_idle_states is initialized to
zero) and if there is an error in parsing the device tree we don't set
up this value. If there is no error, then nr_pnv_idle_states is a
positive value.



> pr_warn("cpuidle-powernv: no idle states found in the DT\n");
> - goto out;
> - }
> -
> - flags = kcalloc(dt_idle_states, sizeof(*flags), GFP_KERNEL);
> -
> - if (of_property_read_u32_array(np,
> - "ibm,cpu-idle-state-flags", flags, dt_idle_states)) {
> - pr_warn("cpuidle-powernv: missing ibm,cpu-idle-state-flags in DT\n");
> - goto out;
> + return;
> }
>
> if (cpu_has_feature(CPU_FTR_ARCH_300)) {
> - if (pnv_power9_idle_init(np, flags, dt_idle_states))
> - goto out;
> + if (pnv_power9_idle_init())

There cannot any longer be a failure in pnv_power9_idle_init() after
the first patch, since we always return a 0. So this if condition can be removed.

> + return;
> }
>
> - for (i = 0; i < dt_idle_states; i++)
> - supported_cpuidle_states |= flags[i];
> + for (i = 0; i < nr_pnv_idle_states; i++)
> + supported_cpuidle_states |= pnv_idle_states[i].flags;

Ideally we should be setting the supported_cpuidle_states flags after
all the validatations have been done, including the ones for stop-api
initialization.

>
> -out:
> - kfree(flags);
> }
>
> /*
> --
> 2.5.5
>