Re: [Intel-gfx] [PATCH v2 2/8] drm/i915: Add more control to wait_for routines

From: Chris Wilson
Date: Fri Dec 01 2017 - 12:44:27 EST


Quoting Sean Paul (2017-12-01 17:20:24)
> /**
> - * _wait_for - magic (register) wait macro
> + * __wait_for - magic wait macro
> *
> - * Does the right thing for modeset paths when run under kdgb or similar atomic
> - * contexts. Note that it's important that we check the condition again after
> + * Macro to help avoid open coding check/wait/timeout patterns, will do the
> + * right think wrt to choosing msleep vs usleep_range based on how long the wait
> + * interval is. Note that it's important that we check the condition again after
> * having timed out, since the timeout could be due to preemption or similar and
> * we've never had a chance to check the condition before the timeout.
> */
> -#define _wait_for(COND, US, W) ({ \
> +#define __wait_for(OP, COND, US, W) ({ \
> unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \
> int ret__; \
> might_sleep(); \
> for (;;) { \
> bool expired__ = time_after(jiffies, timeout__); \
> + OP; \
> if (COND) { \
> ret__ = 0; \
> break; \
> @@ -62,11 +64,16 @@
> ret__ = -ETIMEDOUT; \
> break; \
> } \
> - usleep_range((W), (W) * 2); \
> + if (W > (20 * 1000)) \
> + msleep(W / 1000); \
> + else \
> + usleep_range((W), (W) * 2); \

The current wait_for() is a little more complicated nowadays (variable
W).

Are ms intervals going to be that common? Using a state-machine springs
to mind, but you could argue that msleep() is just a yield. Using msleep
though is going to leave D processes visible and a bump in load :|

> } \
> ret__; \
> })
>
> +#define _wait_for(COND, US, W) __wait_for(;,(COND), US, W)
> +
> #define wait_for(COND, MS) _wait_for((COND), (MS) * 1000, 1000)
>
> /* If CONFIG_PREEMPT_COUNT is disabled, in_atomic() always reports false. */
> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
> index b4621271e7a2..c851b0c0602d 100644
> --- a/drivers/gpu/drm/i915/intel_uncore.c
> +++ b/drivers/gpu/drm/i915/intel_uncore.c
> @@ -1770,12 +1770,14 @@ int __intel_wait_for_register_fw(struct drm_i915_private *dev_priv,
> }
>
> /**
> - * intel_wait_for_register - wait until register matches expected state
> + * __intel_wait_for_register - wait until register matches expected state
> * @dev_priv: the i915 device
> * @reg: the register to read
> * @mask: mask to apply to register value
> * @value: expected value
> - * @timeout_ms: timeout in millisecond
> + * @fast_timeout_us: fast timeout in microsecond for atomic/tight wait
> + * @slow_timeout_ms: slow timeout in millisecond
> + * @out_value: optional placeholder to hold registry value
> *
> * This routine waits until the target register @reg contains the expected
> * @value after applying the @mask, i.e. it waits until ::
> @@ -1786,15 +1788,18 @@ int __intel_wait_for_register_fw(struct drm_i915_private *dev_priv,
> *
> * Returns 0 if the register matches the desired condition, or -ETIMEOUT.
> */
> -int intel_wait_for_register(struct drm_i915_private *dev_priv,
> +int __intel_wait_for_register(struct drm_i915_private *dev_priv,
> i915_reg_t reg,
> u32 mask,
> u32 value,
> - unsigned int timeout_ms)
> + unsigned int fast_timeout_us,
> + unsigned int slow_timeout_ms,
> + u32 *out_value)
> {
> unsigned fw =
> intel_uncore_forcewake_for_reg(dev_priv, reg, FW_REG_READ);
> int ret;
> + u32 reg_value;
>
> might_sleep();
>
> @@ -1803,14 +1808,18 @@ int intel_wait_for_register(struct drm_i915_private *dev_priv,
>
> ret = __intel_wait_for_register_fw(dev_priv,
> reg, mask, value,
> - 2, 0, NULL);
> + fast_timeout_us, 0, &reg_value);
>
> intel_uncore_forcewake_put__locked(dev_priv, fw);
> spin_unlock_irq(&dev_priv->uncore.lock);
>
> if (ret)
> - ret = wait_for((I915_READ_NOTRACE(reg) & mask) == value,
> - timeout_ms);
> + ret = __wait_for(reg_value = I915_READ_NOTRACE(reg),
> + (reg_value & mask) == value,
> + slow_timeout_ms * 1000, 1000);
> +
> + if (out_value)
> + *out_value = reg_value;

Looks good.
-Chris