Serial over USB driver on STM32, cable unplug detecting

[Giovanni]

Usually a backport is allowed if the updated is localized in a single module and does not break compatibility too much. The USB driver is very different now so I have to make some verifications first.

For new projects I recommend using 2.5.x for development, it is stable enough, and wait 2.6.x for deployment. 2.4.x and 2.2.x are legacy.

Giovanni

[Prof. Dr. YoMan]

Ok. Trunk or 2.5.x for development?

[Giovanni]

Trunk is good enough most of time, if I am going to break something usually I post it in the development forum.

Giovanni

[Prof. Dr. YoMan]

I switch the question around: Do I miss a lot, if I am on 2.5.x and not trunk?

And then: Back on topic. Thanks for the break.

[Giovanni]

You do miss a lot, just read the readme in the trunk version, all differences are listed there. The trunk is always the most up to date and includes all fixes for known bugs.

Giovanni

[Prof. Dr. YoMan]

Which git mirror should I use?

Homepage: https://github.com/mabl/ChibiOS
This? https://github.com/ChibiOS-Upstream/ChibiOS-RT
That? https://github.com/ChibiOS/ChibiOS-RT-Community

Is made sure that the git-hashes do not change in the feature, because I want to use one of them as module in my projects.

[Giovanni]

Hi,

The official repository is the subversion one on the SourceForge page. The GIT mirror listed in the homepage is the mirror.

The other two... I simply didn't understand what are they for.

Giovanni

[alexblack]

Hi.
I tested new STM32F429 MCU and found that in usb driver there is no USB_EVENT_SUSPEND detection which I used for usb cable unplug detectiong. Is there any reason fo this?
I pathed the driver and all works now:

Code: Select all

Index: usb_lld.c
===================================================================
--- usb_lld.c   (revision 6883)
+++ usb_lld.c   (working copy)
@@ -616,6 +616,11 @@
     _usb_isr_invoke_event_cb(usbp, USB_EVENT_RESET);
   }
 
+  /* USB bus SUSPEND condition handling.*/
+  if (sts & GINTSTS_USBSUSP) {
+    _usb_isr_invoke_event_cb(usbp, USB_EVENT_SUSPEND);
+  }
+    
   /* Enumeration done.*/
   if (sts & GINTSTS_ENUMDNE) {
     (void)otgp->DSTS;
@@ -915,10 +920,10 @@
     otgp->DOEPMSK  = 0;
     otgp->DAINTMSK = 0;
     if (usbp->config->sof_cb == NULL)
-      otgp->GINTMSK  = GINTMSK_ENUMDNEM | GINTMSK_USBRSTM /*| GINTMSK_USBSUSPM |
+      otgp->GINTMSK  = GINTMSK_ENUMDNEM | GINTMSK_USBRSTM | GINTMSK_USBSUSPM /*|
                        GINTMSK_ESUSPM  |*/;
     else
-      otgp->GINTMSK  = GINTMSK_ENUMDNEM | GINTMSK_USBRSTM /*| GINTMSK_USBSUSPM |
+      otgp->GINTMSK  = GINTMSK_ENUMDNEM | GINTMSK_USBRSTM | GINTMSK_USBSUSPM /*|
                        GINTMSK_ESUSPM */ | GINTMSK_SOFM;
     otgp->GINTSTS  = 0xFFFFFFFF;         /* Clears all pending IRQs, if any. */

[Giovanni]

Hi,

The problem is the other way, 2.4.x is unable to detect that #SV10 error, 2.6.x has a state checker.

The lock is required because an higher priority interrupt could preempt chVTSetI() in the middle corrupting the timers list.

The lock-unlock protocol ensures that there cannot be race conditions related to kernel calls.

Giovanni

[GSE]

In current version of 2.6.x, STM32 usb_lld.c usb event USB_EVENT_SUSPEND no handling more. How to realize USB cable unplug event in this case? 5V monitoring not so clean way. Why this event handling removed?