On 3/1/2011 12:27 PM, ABRAHAM, KISHON VIJAY wrote:
On Tue, Mar 1, 2011 at 4:47 PM, Cousson, Benoit<b-cousson@xxxxxx> wrote:
Hi Kishon,
Sorry for this late reply.
I'm fine with this omap_device API which is well aligned with the discussion
we had. I just have few minor comments about the split omap_hwmod /
omap_device.
To summarize, you should not hack directly hwmod internal attributes from
the upper layer.
You should simply expose the same number of API at hwmod level instead of
relying on a single omap_hwmod_set_slave_idlemode with miscellaneous
parameters.
Should it then be like one-one mapping between omap_device APIs and
omap_hwmod APIs, or have a single omap_device API, that takes a additional
parameter and based on the parameter, call the appropriate omap_hwmod API?
A one-to-one is the best thing to do for my point of view.
Benoit
On 1/31/2011 1:34 PM, ABRAHAM, KISHON VIJAY wrote:
Provide APIs to be used by the driver in order to modify AUTOIDLE
and SIDLE bits. These APIs in turn call hwmod APIs to modify the
SYSCONFIG register.
Signed-off-by: Kishon Vijay Abraham I<kishon@xxxxxx>
Signed-off-by: Benoit Cousson<b-cousson@xxxxxx>
Cc: Paul Walmsley<paul@xxxxxxxxx>
---
arch/arm/plat-omap/include/plat/omap_device.h | 6 +
arch/arm/plat-omap/omap_device.c | 176
+++++++++++++++++++++++++
2 files changed, 182 insertions(+), 0 deletions(-)
diff --git a/arch/arm/plat-omap/include/plat/omap_device.h
b/arch/arm/plat-omap/include/plat/omap_device.h
index e4c349f..47ad0ab 100644
--- a/arch/arm/plat-omap/include/plat/omap_device.h
+++ b/arch/arm/plat-omap/include/plat/omap_device.h
@@ -110,6 +110,12 @@ struct powerdomain *omap_device_get_pwrdm(struct
omap_device *od);
u32 omap_device_get_context_loss_count(struct platform_device *pdev);
/* Other */
+int omap_device_noidle(struct omap_device *od);
+int omap_device_smartidle(struct omap_device *od);
+int omap_device_forceidle(struct omap_device *od);
+int omap_device_default_idle(struct omap_device *od);
+int omap_device_enable_autoidle(struct omap_device *od);
+int omap_device_disable_autoidle(struct omap_device *od);
int omap_device_idle_hwmods(struct omap_device *od);
int omap_device_enable_hwmods(struct omap_device *od);
diff --git a/arch/arm/plat-omap/omap_device.c
b/arch/arm/plat-omap/omap_device.c
index 57adb27..da8609a 100644
--- a/arch/arm/plat-omap/omap_device.c
+++ b/arch/arm/plat-omap/omap_device.c
@@ -726,6 +726,182 @@ void __iomem *omap_device_get_rt_va(struct
omap_device *od)
return omap_hwmod_get_mpu_rt_va(od->hwmods[0]);
}
+/**
+ * omap_device_noidle - set the device's slave idlemode to no idle
+ * @od: struct omap_device *
+ *
+ * Sets the IP block's OCP slave idlemode in hardware to no idle.
+ * Intended to be used by drivers that have some erratum that requires
direct
+ * manipulation of the SIDLEMODE bits. Returns -EINVAL if @od is in
invalid
+ * state, or passes along the return value from
+ * omap_hwmod_set_slave_idlemode().
+ */
+int omap_device_noidle(struct omap_device *od)
+{
+ int retval = 0, i;
+
+ if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
+ WARN(1, "omap_device: %s.%d: %s() called from invalid
state "
+ "%d\n", od->pdev.name, od->pdev.id, __func__,
+ od->_state);
+ return -EINVAL;
+ }
+
+ for (i = 0; i< od->hwmods_cnt; i++)
+ retval |= omap_hwmod_set_slave_idlemode(od->hwmods[i],
+ HWMOD_IDLEMODE_NO);
+
+ return retval;
+}
+
+
+/**
+ * omap_device_smartidle - set the device's slave idlemode to smart idle
+ * @od: struct omap_device *
+ *
+ * Sets the IP block's OCP slave idlemode in hardware to smart idle.
+ * Intended to be used by drivers that have some erratum that requires
direct
+ * manipulation of the SIDLEMODE bits. Returns -EINVAL if @od is in
invalid
+ * state, or passes along the return value from
+ * omap_hwmod_set_slave_idlemode().
+ */
+int omap_device_smartidle(struct omap_device *od)
+{
+ int retval = 0, i;
+
+ if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
+ WARN(1, "omap_device: %s.%d: %s() called from invalid
state "
+ "%d\n", od->pdev.name, od->pdev.id, __func__,
+ od->_state);
+ return -EINVAL;
+ }
+
+ for (i = 0; i< od->hwmods_cnt; i++)
+ retval |= omap_hwmod_set_slave_idlemode(od->hwmods[i],
+ HWMOD_IDLEMODE_SMART);
+
+ return retval;
+}
+
+
+/**
+ * omap_device_forceidle - set the device's slave idlemode to force idle
+ * @od: struct omap_device *
+ *
+ * Sets the IP block's OCP slave idlemode in hardware to force idle.
+ * Intended to be used by drivers that have some erratum that requires
direct
+ * manipulation of the SIDLEMODE bits. Returns -EINVAL if @od is in
invalid
+ * state, or passes along the return value from
+ * omap_hwmod_set_slave_idlemode().
+ */
+int omap_device_forceidle(struct omap_device *od)
+{
+ int retval = 0, i;
+
+ if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
+ WARN(1, "omap_device: %s.%d: %s() called from invalid
state "
+ "%d\n", od->pdev.name, od->pdev.id, __func__,
+ od->_state);
+ return -EINVAL;
+ }
+
+ for (i = 0; i< od->hwmods_cnt; i++)
+ retval |= omap_hwmod_set_slave_idlemode(od->hwmods[i],
+ HWMOD_IDLEMODE_FORCE);
+
+ return retval;
+}
+
+/**
+ * omap_device_default_idle - set the device's slave idlemode to no idle
or
+ * smart idle based on the hwmod flag
+ * @od: struct omap_device *
+ *
+ * Sets the IP block's OCP slave idlemode in hardware to no idle or smart
idle
+ * depending on status of the flag HWMOD_SWSUP_SIDLE.
+ * Intended to be used by drivers that have some erratum that requires
direct
+ * manipulation of the SIDLEMODE bits. Returns -EINVAL if @od is in
invalid
+ * state, or passes along the return value from
+ * omap_hwmod_set_slave_idlemode().
+ */
+int omap_device_default_idle(struct omap_device *od)
+{
+ int retval = 0, i;
+ u8 idlemode;
+
+ if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
+ WARN(1, "omap_device: %s.%d: %s() called from invalid
state "
+ "%d\n", od->pdev.name, od->pdev.id, __func__,
+ od->_state);
+ return -EINVAL;
+ }
+
+ for (i = 0; i< od->hwmods_cnt; i++) {
+ idlemode = (od->hwmods[i]->flags& HWMOD_SWSUP_SIDLE) ?
+ HWMOD_IDLEMODE_NO : HWMOD_IDLEMODE_SMART;
+ retval |= omap_hwmod_set_slave_idlemode(od->hwmods[i],
+ idlemode);
That one is the worst :-) You should definitively not hack directly hwmod
internal attributes from the upper layer.
+ }
+
+ return retval;
+}
+
+/*
+ * omap_device_enable_autoidle - enable the device's OCP slave autoidle
+ * @od: struct omap_device *
+ *
+ * Sets the IP block's OCP slave autoidle in hardware.
+ * Intended to be used by drivers that have some erratum that requires
direct
+ * manipulation of the AUTOIDLE bits. Returns -EINVAL if @od is in
invalid
+ * state, or passes along the return value from
+ * omap_hwmod_set_slave_autoidle().
+ */
+int omap_device_enable_autoidle(struct omap_device *od)
+{
+ int retval = 0, i;
+
+ if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
+ WARN(1, "omap_device: %s.%d: %s() called from invalid
state "
+ "%d\n", od->pdev.name, od->pdev.id, __func__,
+ od->_state);
+ return -EINVAL;
+ }
+
+ for (i = 0; i< od->hwmods_cnt; i++)
+ retval |= omap_hwmod_set_slave_autoidle(od->hwmods[i],
+ true);
You are using the same API to send either HWMOD_XXX flags or boolean, and
that's quite confusing.
As I said, you'd better duplicate these APIs at hwmod level.
OK.
Benoit
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
