USB在PIC18F13K50:SIE永不释放BD

我正面临一个问题，我正在为这个PIC写一个堆栈。测试代码当前定义了三个端点；EP0在EP0 OUT和EP1中。除了EP0以外，我用乒乓球在所有的EPS上。堆栈枚举完美（至少在Linux上）。USB处理是在低优先级ISR上完成的。端点0除了标准请求之外，还处理一些自定义请求来控制HW。端点1被设置为EP中的一个块，具有64字节缓冲器（2，如乒乓），捕获一些引脚状态并将其写入EP缓冲器。基本代码。测试代码进入一个循环，试图写入EP缓冲区并将其安排到SIE中。同样，非常基本的东西。我的问题发生在前两个缓冲区（偶数和奇数）被发送，一个在DATA0包中，另一个在DATA1中。一旦两者被发送，SIE在DATA0包上发送第三个缓冲器，从主机接收ACK，但显然忽略它，因为它不释放BD，并且只保留重新发送T。同样的数据，一遍又一遍。我把UUND、DTSN和DTS的状态都放在循环中的缓冲器上，我可以看到，这两个缓冲器最终都被安排到SIE（UWITE位集、DTSN位集和DTS位置奇数和偶数重置）上，并保持这种方式，同时SIE重新发送相同的数据。被捕获的硬件逻辑分析仪连接到UC附近，所以我不认为UC可能误读ACK包。我期待着对SIE做点什么，让它困惑，但是什么？在更新BDS之前，代码专门检查UWORE＝0，但它可以在任何时间读取它们，因为它轮询它们的状态。但这应该是安全的。该部分不是A3版本，所以它不是“PP上所有的，但EP0”的问题。任何想法，什么可能导致SIE继续重新发送相同的缓冲区，而不释放它，尽管ACKs从主机？
                    
                    
               
                                                                                      以下为原文

                                                                            I'm faced with an issue on a stack I'm writing for this PIC. The test code currently defines three endpoins; EP0 IN, EP0 OUT and EP1 IN. I'm using ping-pong on all EPs except EP0. The stack enumerates perfectly (well, at least on Linux). The USB handling is done on a low priority ISR.
Endpoint 0, apart from the standard requests, handles some custom requests to control the HW. Those work perfectly.
Endpoint 1 is setup as a Bulk IN EP with 64 byte buffers (2, as per ping-pong), that captures some pins states and writes it on the EP buffers. Basic stuff.
The test code goes into a loop trying to write into an EP buffer and scheduling it into the SIE as soon as its full. Again, pretty basic stuff. My problem happens after the first two buffers (Even and Odd) are sent, one in a DATA0 packet and the other in a DATA1.
Once both are sent the SIE sends a third buffer on a DATA0 packet, receives an ACK from the host but apparently ignores it as it does not releases the BD and just keeps resending the same data, over and over. I'm dumping the state of the UOWN, DTSEN and DTS on both IN buffers on the loop and I can see that eventually both buffers are scheduled into the SIE (UOWN bits set, DTSEN bits set and DTS bits set on Odd and reset on Even) and remain this way while the SIE resends the same data.
The USB traffic is being captured with an hardware logic analyzer hooked up close to the uC so I don't think its possible for the uC to be misreading the ACK packet. I expect to be doing something to the SIE that is confusing it, but what ?
The code specifically checks UOWN=0 before updating the BDs but it can read them at any time, since its polling their state. But this is supposed to be safe. The part is not an A3 revision so it's not the 'PP on all but EP0' issue. Any thoughs on what could cause the SIE to keep resending the same buffer, without releasing it, despite getting ACKs from the host ?