we are now using only one memory block for both sg rings. This is
required so that the SG interface can benefit from a read cache
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
polling HW is slow (>1us). Polling RAM is faster. This is a first implementation which only polls the first BD that is active. This is why this commit also removes the second read in nodes/fpga. This is not really useful anyways.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
remove explicit mentioning of PCIe in the use of Device as a preparation
for integrating platform devices. auto formatted some files.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
This edits the headers in every file so the copyright notice mentions RWTH Aachen University. We also update some copyright years and fix various comments so the header is the same across all of VILLASframework.
* Harmonize comment and code-style
Signed-off-by: Steffen Vogel <steffen.vogel@opal-rt.com>
* Harmonize comment and code-style
Signed-off-by: Steffen Vogel <steffen.vogel@opal-rt.com>
---------
Signed-off-by: Steffen Vogel <steffen.vogel@opal-rt.com>
introduce new struct Completion that is returned by Dma::readCompletion
and Dma::writeCompletion
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
Modify villas-fpga-ctrl to fit the new behavior of Dma.
Makes reading from DMA work even when we are too slow and
only receive partial batches of BDs.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
this fixes that the wrong order of allocating and PciCard destruction
causes an undefined state.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
The project is now also REUSE compliant: https://reuse.software/
Previous copyright holders have provided their
acknowledgement to transition to the new license in the
following GitHub PR: https://github.com/VILLASframework/fpga/pull/66
Signed-off-by: Steffen Vogel <post@steffenvogel.de>
unmapping of the scatter gather attribute memory was done
after the DMA destructor was called, leading to Card trying to
unmap memory that was already freed.
This lead to crashing during cleaning up.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
The changes lead to crashing because intc will not be configured
properly. Also revert removing configDone in ips/dma
This reverts commit 3b8949afe9.
Signed-off-by: Niklas Eiling <niklas.eiling@eonerc.rwth-aachen.de>
