Fix incorrect usage of plan_set_e_position() in G92 E*

To maintain an accurate step count (which is required for correct
position recovery), any call to plan_set_position&co needs to be done
synchronously and from a halted state.

However, G92 E* is currently special-cased to skip the sync (likely to
avoid the associated performance cost), causing an incorrect E step
count and position to be set. This breaks absolute position recovery,
miscalculation of the LA factor and possibly other weird issues.

We rewrite the handling of G92 to always sync but still special-case the
frequent "G92 E0" for performance by using a free bit in the block flags.

To avoid a sync, we relay the request for reset first to the planner
which clears its internal state and then relays the request to the final
stepper isr.

Firmware/Marlin_main.cpp

@@ -3356,6 +3356,49 @@ static void gcode_PRUSA_BadRAMBoFanTest(){
 #endif
 }
 
+
+// G92 - Set current position to coordinates given
+static void gcode_G92()
+{
+    bool codes[NUM_AXIS];
+    float values[NUM_AXIS];
+
+    // Check which axes need to be set
+    for(uint8_t i = 0; i < NUM_AXIS; ++i)
+    {
+        codes[i] = code_seen(axis_codes[i]);
+        if(codes[i])
+            values[i] = code_value();
+    }
+
+    if((codes[E_AXIS] && values[E_AXIS] == 0) &&
+       (!codes[X_AXIS] && !codes[Y_AXIS] && !codes[Z_AXIS]))
+    {
+        // As a special optimization, when _just_ clearing the E position
+        // we schedule a flag asynchronously along with the next block to
+        // reset the starting E position instead of stopping the planner
+        current_position[E_AXIS] = 0;
+        plan_reset_next_e();
+    }
+    else
+    {
+        // In any other case we're forced to synchronize
+        st_synchronize();
+        for(uint8_t i = 0; i < 3; ++i)
+        {
+            if(codes[i])
+                current_position[i] = values[i] + cs.add_homing[i];
+        }
+        if(codes[E_AXIS])
+            current_position[E_AXIS] = values[E_AXIS];
+
+        // Set all at once
+        plan_set_position(current_position[X_AXIS], current_position[Y_AXIS],
+                          current_position[Z_AXIS], current_position[E_AXIS]);
+    }
+}
+
+
 #ifdef BACKLASH_X
 extern uint8_t st_backlash_x;
 #endif //BACKLASH_X
@@ -5217,20 +5260,8 @@ if(eSoundMode!=e_SOUND_MODE_SILENT)
 
     //! ### G92 - Set position
     // -----------------------------
-    case 92:
-      if(!code_seen(axis_codes[E_AXIS]))
-        st_synchronize();
-      for(int8_t i=0; i < NUM_AXIS; i++) {
-        if(code_seen(axis_codes[i])) {
-           if(i == E_AXIS) {
-             current_position[i] = code_value();
-             plan_set_e_position(current_position[E_AXIS]);
-           }
-           else {
-		current_position[i] = code_value()+cs.add_homing[i];
-            plan_set_position(current_position[X_AXIS], current_position[Y_AXIS], current_position[Z_AXIS], current_position[E_AXIS]);
-           }
-        }
+    case 92: {
+        gcode_G92();
     }
     break;
 

Firmware/planner.cpp

@@ -130,6 +130,10 @@ float extruder_advance_K = LIN_ADVANCE_K;
 float position_float[NUM_AXIS];
 #endif
 
+// Request the next block to start at zero E count
+static bool plan_reset_next_e_queue;
+static bool plan_reset_next_e_sched;
+
 // Returns the index of the next block in the ring buffer
 // NOTE: Removed modulo (%) operator, which uses an expensive divide and multiplication.
 static inline int8_t next_block_index(int8_t block_index) {
@@ -441,6 +445,8 @@ void plan_init() {
   previous_speed[2] = 0.0;
   previous_speed[3] = 0.0;
   previous_nominal_speed = 0.0;
+  plan_reset_next_e_queue = false;
+  plan_reset_next_e_sched = false;
 }
 
 
@@ -658,6 +664,9 @@ void planner_abort_hard()
     previous_speed[2] = 0.0;
     previous_speed[3] = 0.0;
 
+    plan_reset_next_e_queue = false;
+    plan_reset_next_e_sched = false;
+
     // Relay to planner wait routine, that the current line shall be canceled.
     waiting_inside_plan_buffer_line_print_aborted = true;
 }
@@ -721,6 +730,20 @@ void plan_buffer_line(float x, float y, float z, const float &e, float feed_rate
   // Save the global feedrate at scheduling time
   block->gcode_feedrate = feedrate;
 
+  // Reset the starting E position when requested
+  if (plan_reset_next_e_queue)
+  {
+      position[E_AXIS] = 0;
+#ifdef LIN_ADVANCE
+      position_float[E_AXIS] = 0;
+#endif
+
+      // the block might still be discarded later, but we need to ensure the lower-level
+      // count_position is also reset correctly for consistent results!
+      plan_reset_next_e_queue = false;
+      plan_reset_next_e_sched = true;
+  }
+
 #ifdef ENABLE_AUTO_BED_LEVELING
   apply_rotation_xyz(plan_bed_level_matrix, x, y, z);
 #endif // ENABLE_AUTO_BED_LEVELING
@@ -1165,6 +1188,13 @@ Having the real displacement of the head, we can calculate the total movement le
   // Reset the block flag.
   block->flag = 0;
 
+  if (plan_reset_next_e_sched)
+  {
+      // finally propagate a pending reset
+      block->flag |= BLOCK_FLAG_E_RESET;
+      plan_reset_next_e_sched = false;
+  }
+
   // Initial limit on the segment entry velocity.
   float vmax_junction;
 
@@ -1367,6 +1397,11 @@ void plan_set_e_position(const float &e)
   st_set_e_position(position[E_AXIS]);
 }
 
+void plan_reset_next_e()
+{
+    plan_reset_next_e_queue = true;
+}
+
 #ifdef PREVENT_DANGEROUS_EXTRUDE
 void set_extrude_min_temp(float temp)
 {

Firmware/planner.h

@@ -44,6 +44,8 @@ enum BlockFlag {
     // than 32767, therefore the DDA algorithm may run with 16bit resolution only.
     // In addition, the stepper routine will not do any end stop checking for higher performance.
     BLOCK_FLAG_DDA_LOWRES = 8,
+    // Block starts with Zeroed E counter
+    BLOCK_FLAG_E_RESET = 16,
 };
 
 union dda_isteps_t
@@ -168,6 +170,9 @@ void plan_set_position(float x, float y, float z, const float &e);
 void plan_set_z_position(const float &z);
 void plan_set_e_position(const float &e);
 
+// Reset the E position to zero at the start of the next segment
+void plan_reset_next_e();
+
 extern bool e_active();
 
 void check_axes_activity();

Firmware/stepper.cpp

@@ -362,6 +362,10 @@ FORCE_INLINE void stepper_next_block()
     LA_phase = -1;
 #endif
 
+    if (current_block->flag & BLOCK_FLAG_E_RESET) {
+        count_position[E_AXIS] = 0;
+    }
+
     if (current_block->flag & BLOCK_FLAG_DDA_LOWRES) {
       counter_x.lo = -(current_block->step_event_count.lo >> 1);
       counter_y.lo = counter_x.lo;