Le 21/05/2026 à 10:42, Richard GENOUD a écrit :
Le 11/05/2026 à 11:53, Kumar, Udit a écrit :


On 4/30/2026 2:13 PM, Richard Genoud (TI) wrote:
From: "Thomas Richard (TI)" <[email protected]>

Add the capability to detect a resume. This overrides the weak
function j7xx_board_is_resuming(void).
To detect the resume, SPL searches a magic value (0xBA) in a register
of PMICA.
This value is set by DM-Firmware during the suspend sequence.

Signed-off-by: Thomas Richard (TI) <[email protected]>
Co-developed-by: Gregory CLEMENT (TI) <[email protected]>
Signed-off-by: Gregory CLEMENT (TI) <[email protected]>
Co-developed-by: Richard Genoud (TI) <[email protected]>
Signed-off-by: Richard Genoud (TI) <[email protected]>
---
  board/ti/common/k3-lpm.h | 17 +++++++++++++++++
  board/ti/j721e/evm.c     | 37 +++++++++++++++++++++++++++++++++++++
  2 files changed, 54 insertions(+)
  create mode 100644 board/ti/common/k3-lpm.h

diff --git a/board/ti/common/k3-lpm.h b/board/ti/common/k3-lpm.h
new file mode 100644
index 000000000000..a2141a0c25a8
--- /dev/null
+++ b/board/ti/common/k3-lpm.h
@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Copyright (C) 2026, Texas Instruments Incorporated - https:// www.ti.com/
+ * Copyright (C) 2026, Bootlin
+ */
+
+#ifndef __K3_LPM_H
+#define __K3_LPM_H
+
+/* Magic value in PMIC register to indicate the suspend state (SOC_OFF) */
+#define K3_LPM_MAGIC_SUSPEND 0xba
+
+/* PMIC register where the magic value resides */
+#define K3_LPM_SCRATCH_PAD_REG 0xcb
+
+#endif /* __K3_LPM_H */
+
diff --git a/board/ti/j721e/evm.c b/board/ti/j721e/evm.c
index 881392bd963a..67ff13828514 100644
--- a/board/ti/j721e/evm.c
+++ b/board/ti/j721e/evm.c
@@ -17,9 +17,11 @@
  #include <spl.h>
  #include <dm.h>
  #include <asm/arch/k3-ddr.h>
+#include <power/pmic.h>
  #include "../common/board_detect.h"
  #include "../common/fdt_ops.h"
+#include "../common/k3-lpm.h"
  #define board_is_j721e_som()    (board_ti_k3_is("J721EX-PM1-SOM") || \
                   board_ti_k3_is("J721EX-PM2-SOM"))
@@ -469,6 +471,41 @@ err_free_gpio:
      }
  }
+#if (IS_ENABLED(CONFIG_SPL_BUILD) && IS_ENABLED(CONFIG_TARGET_J7200_R5_EVM))
+
+bool j7xx_board_is_resuming(void)
+{
+    struct udevice *pmica;
+    int ret;
+
+    if (gd_k3_resuming() != K3_RESUME_STATE_UNKNOWN)
+        goto end;
+
+    ret = uclass_get_device_by_name(UCLASS_PMIC,
+                    "pmic@48", &pmica);
+    if (ret) {
+        printf("Getting PMICA init failed: %d\n", ret);
+        goto end;
+    }
+    debug("%s: PMICA is detected (%s)\n", __func__, pmica->name);
+
+    if (pmic_reg_read(pmica, K3_LPM_SCRATCH_PAD_REG) == K3_LPM_MAGIC_SUSPEND) {
+        debug("%s: board is resuming\n", __func__);
+        gd_set_k3_resuming(K3_RESUME_STATE_RESUMING);

If you failed to clear magic word then, we are not in correct state.
So doing all work with PMIC, set state to resuming
That's right, I'll check the error.
Sorry, I answered too quickly.
I mean, yes, if we fail to clean the magic value from PMIC, the next boot may be incorrectly detected as resuming.


+
+        /* clean magic suspend */
+        if (pmic_reg_write(pmica, K3_LPM_SCRATCH_PAD_REG, 0))
+            printf("Failed to clean magic value for suspend detection in PMICA\n");
But what could we do in this case?
We could hang() the board, reboot...
Or gd_set_k3_resuming back to K3_RESUME_STATE_UNKNOWN and hope that the next call to j7xx_board_is_resuming() will successfully clean the magic value.

Otherwise, we can reset this magic value in the DM startup?

+    } else {
+        debug("%s: board is booting (no resume detected)\n", __func__);
+        gd_set_k3_resuming(K3_RESUME_STATE_BOOTING);
+    }
+end:
+    return gd_k3_resuming() == K3_RESUME_STATE_RESUMING;
+}
+
+#endif /* CONFIG_SPL_BUILD && CONFIG_TARGET_J7200_R5_EVM */
+
  void spl_board_init(void)
  {
      struct udevice *dev;


Thanks!

Regards,
Richard

Regards,
Richard

Reply via email to