Hi, On Thursday 08 February 2018 07:11 AM, Shawn Lin wrote: > Hi Bjorn, > > [+ Kishon] > > On 2018/2/7 22:39, Bjorn Helgaas wrote: > >>> 2. How does device triggers DMA from device to cpu, is it by simply >>> writing to BAR ? >> >> In most cases the device itself performs the DMA. It sounds like >> you have a case where the DMA engine is not part of the device itself. >> I don't know how that would be programmed, except to say that the DMA >> engine presumably operates on DMA addresses, and something would still >> have to DMA-map buffers to obtain those. > > I have a question that why it's always device to performs the DMA for > no matter memory read and write? We have builtin DMA engine inside the The PCI devices are supposed to work with different host controllers and instead of relying on host controller capabilities (in this case DMA), the device can use it's own DMA. > PCIe RC controller, so could the function driver submit the DMA request > to the PCIe RC controller driver and let it fire the DMA to access > device's internal memory? Given that we now support PCIe EP framework, As Bjorne mentioned before, RC accessing the EP memory is mostly for configuring the device and not in the data path. Having said that, there can be a pci_driver which DMAs to a BAR region, if thats your usecase. > and now linux-pci behave like a *device* per your saying, so that is > quite possible for us to consider how we would submit DMA request with > current PCIe EP stuff. Thanks Kishon
