DMA crash ?

[ysapir]

As a sanity check, can you please replace the DMA transfers with memcpy() calls?

[tnt]

[amylaar]

[quote="ysapir"]Just a quick comment - we found out that under some circumstances, the optimizer does not always consider the volatile attribute to the *last_addr pointer, and optimizes out the while() loop.

That volatile doesn't really mean anything, because the compiler sees that that address is set to point inside an object pointed to by tcb->src_addr, so if that address
didn't point to something volatile, casting or assigning the pointer won't make it so (at least as long as the compiler can keep track of the value propagation).

[ysapir]

[tnt]

Did anyone manage or tried to reproduce the issue btw ? Or am I the only one hitting this ?

[ysapir]

[ysapir]

I used your archive as a basis for a modified program, using the latest driver version. Although I don't see anything substantial that may have changed between the driver revisions, I am yet not seeing the strange behavior you encounter. I will let it run for a while to see if this behavior is repeatable.

[tnt]

[ysapir]

[ysapir]