Add support for dumping board-specific information on assertion. From David Sidrane

1 files changed, 30 insertions, 0 deletions

diff --git a/nuttx/include/nuttx/board.h b/nuttx/include/nuttx/board.h
index 55aa85b60..4c9e1731f 100644
--- a/nuttx/include/nuttx/board.h
+++ b/nuttx/include/nuttx/board.h
@@ -291,4 +291,34 @@ uint8_t board_buttons(void);
 xcpt_t board_button_irq(int id, xcpt_t irqhandler);
 #endif
 
+/****************************************************************************
+ * Name: board_crashdump
+ *
+ * Description:
+ *   If CONFIG_BOARD_CRASHDUMP is selected then up_asseert will call out to
+ *   board_crashdump prior to calling exit in the case of an assertion failure.
+ *   Or in the case of a hardfault looping indefinitely. board_crashdump then
+ *   has a chance to save the state of the machine. The provided
+ *   board_crashdump should save as much information as it can about the cause
+ *   of the fault and then most likely reset the system.
+ *
+ *   N.B. There are limited system resources that can be used by the provided
+ *   board_crashdump function. The tems from the fact that most critical/fatal
+ *   crashes are because of a hard fault or during interrupt processing.
+ *   In these cases, up_assert is running from the context of an interrupt
+ *   handlerand it is impossible to use any device driver in this context.
+ *
+ *   Also consider the following: Who knows what state the system is in? Is
+ *   memory trashed? Is the Heap intact? Therefore all we can expect to do in
+ *   board_crashdump is save the "machine state" in a place where on the next
+ *   reset we can write it to more sophisticated storage in a sane operating
+ *   environment.
+ *
+ ****************************************************************************/
+
+#ifdef CONFIG_BOARD_CRASHDUMP
+void board_crashdump(uint32_t currentsp, void *tcb, uint8_t *filename,
+                     int lineno);
+#endif
+
 #endif /* __INCLUDE_NUTTX_BOARD_H */