Re: [PATCH v9 3/3] firmware: ti_sci: Partial-IO support
From: Andrew Davis
Date: Mon Nov 03 2025 - 09:18:37 EST
On 11/3/25 4:31 AM, Markus Schneider-Pargmann wrote:
On Thu Oct 30, 2025 at 3:24 PM CET, Andrew Davis wrote:
On 10/30/25 4:26 AM, Markus Schneider-Pargmann (TI.com) wrote:
Add support for Partial-IO poweroff. In Partial-IO pins of a few
hardware units can generate system wakeups while DDR memory is not
powered resulting in a fresh boot of the system. These hardware units in
the SoC are always powered so that some logic can detect pin activity.
If the system supports Partial-IO as described in the fw capabilities, a
sys_off handler is added. This sys_off handler decides if the poweroff
is executed by entering normal poweroff or Partial-IO instead. The
decision is made by checking if wakeup is enabled on all devices that
may wake up the SoC from Partial-IO.
The possible wakeup devices are found by checking which devices
reference a "Partial-IO" system state in the list of wakeup-source
system states. Only devices that are actually enabled by the user will
be considered as an active wakeup source. If none of the wakeup sources
is enabled the system will do a normal poweroff. If at least one wakeup
source is enabled it will instead send a TI_SCI_MSG_PREPARE_SLEEP
message from the sys_off handler. Sending this message will result in an
immediate shutdown of the system. No execution is expected after this
point. The code will wait for 5s and do an emergency_restart afterwards
if Partial-IO wasn't entered at that point.
A short documentation about Partial-IO can be found in section 6.2.4.5
of the TRM at
https://www.ti.com/lit/pdf/spruiv7
Signed-off-by: Markus Schneider-Pargmann (TI.com) <msp@xxxxxxxxxxxx>
---
drivers/firmware/ti_sci.c | 132 +++++++++++++++++++++++++++++++++++++++++++++-
drivers/firmware/ti_sci.h | 5 ++
2 files changed, 136 insertions(+), 1 deletion(-)
diff --git a/drivers/firmware/ti_sci.c b/drivers/firmware/ti_sci.c
index 4db84a92a517b0aa7bb8d47e809d9848a16e2cc4..f2922fccfbe748a436cb9aa0a8c8e5f48db02ef9 100644
--- a/drivers/firmware/ti_sci.c
+++ b/drivers/firmware/ti_sci.c
@@ -6,6 +6,7 @@
* Nishanth Menon
*/
+#include "linux/dev_printk.h"
#define pr_fmt(fmt) "%s: " fmt, __func__
#include <linux/bitmap.h>
@@ -3663,6 +3664,116 @@ devm_ti_sci_get_resource(const struct ti_sci_handle *handle, struct device *dev,
}
EXPORT_SYMBOL_GPL(devm_ti_sci_get_resource);
+/*
+ * Enter Partial-IO, which disables everything including DDR with only a small
+ * logic being active for wakeup.
+ */
+static int ti_sci_enter_partial_io(struct ti_sci_info *info)
+{
+ struct ti_sci_msg_req_prepare_sleep *req;
+ struct ti_sci_xfer *xfer;
+ struct device *dev = info->dev;
+ int ret = 0;
+
+ xfer = ti_sci_get_one_xfer(info, TI_SCI_MSG_PREPARE_SLEEP,
+ TI_SCI_FLAG_REQ_GENERIC_NORESPONSE,
+ sizeof(*req), sizeof(struct ti_sci_msg_hdr));
+ if (IS_ERR(xfer)) {
+ ret = PTR_ERR(xfer);
+ dev_err(dev, "Message alloc failed(%d)\n", ret);
+ return ret;
+ }
+
+ req = (struct ti_sci_msg_req_prepare_sleep *)xfer->xfer_buf;
+ req->mode = TISCI_MSG_VALUE_SLEEP_MODE_PARTIAL_IO;
This whole function is almost identical to ti_sci_cmd_prepare_sleep() other
than you use a different mode here, which this different mode can be passed
into ti_sci_cmd_prepare_sleep() just the same. Only other difference would
be the NORESPONSE flag which you could just check "mode" passed in and
use the right flag for the mode.
I thought this to be nicer as it avoided ifs, but as Kendall and you
both requested this, I changed it for the next version.
+ req->ctx_lo = 0;
+ req->ctx_hi = 0;
+ req->debug_flags = 0;
+
+ ret = ti_sci_do_xfer(info, xfer);
+ if (ret) {
+ dev_err(dev, "Mbox send fail %d\n", ret);
+ goto fail;
+ }
+
+fail:
+ ti_sci_put_one_xfer(&info->minfo, xfer);
+
+ return ret;
+}
+
+/*
+ * Iterate all device nodes that have a wakeup-source property and check if one
+ * of the possible phandles points to a Partial-IO system state. If it
+ * does resolve the device node to an actual device and check if wakeup is
+ * enabled.
+ */
+static bool ti_sci_partial_io_wakeup_enabled(struct ti_sci_info *info)
+{
+ struct device_node *wakeup_node = NULL;
+
+ for_each_node_with_property(wakeup_node, "wakeup-source") {
+ struct of_phandle_iterator it;
+ int err;
+
+ of_for_each_phandle(&it, err, wakeup_node, "wakeup-source", NULL, 0) {
+ struct platform_device *pdev;
+ bool may_wakeup;
+
+ /*
+ * Continue if idle-state-name is not off-wake. Return
+ * value is the index of the string which should be 0 if
+ * off-wake is present.
+ */
+ if (of_property_match_string(it.node, "idle-state-name", "off-wake"))
+ continue;
+
+ pdev = of_find_device_by_node(wakeup_node);
+ if (!pdev)
+ continue;
+
+ may_wakeup = device_may_wakeup(&pdev->dev);
+ put_device(&pdev->dev);
+
+ if (may_wakeup) {
+ dev_dbg(info->dev, "%pOF identified as wakeup source for Partial-IO\n",
+ wakeup_node);
+ of_node_put(it.node);
+ of_node_put(wakeup_node);
+ return true;
+ }
+ }
+ }
+
+ return false;
+}
+
+static int ti_sci_sys_off_handler(struct sys_off_data *data)
+{
+ struct ti_sci_info *info = data->cb_data;
+ bool enter_partial_io = ti_sci_partial_io_wakeup_enabled(info);
+ int ret;
+
+ if (!enter_partial_io)
+ return NOTIFY_DONE;
+
+ dev_info(info->dev, "Entering Partial-IO because a powered wakeup-enabled device was found.\n");
+
+ ret = ti_sci_enter_partial_io(info);
+
No need for newline here.
+ if (ret) {
+ dev_err(info->dev,
+ "Failed to enter Partial-IO %pe, trying to do an emergency restart\n",
+ ERR_PTR(ret));
Why cast this int to a pointer before printing it out?
I am casting this to an error pointer to get the resolution of the error
value to a symbolic name with %pe. This will print 'EBUSY' etc.
Ah I see, but why not use errname() directly? I see dev_err_probe() also goes though
a ERR_PTR() cast, so maybe there is some reason I'm not seeing, so I'm fine either way.
Thanks,
Andrew
Thanks!
Best
Markus