Prandr

@geiliwanghaichao > "I want the userspace program to be able to directly specify the DMA target address to avoid data copying from read()/write()." "Does the driver support direct DMA writes...

@dmitrym1 XDMA is fundamentally not RT-capable, because many internal functions (for example memory allocation with `kmalloc`) can sleep. That what [properly implemented] bypass BAR would be good for.

@doonny I mean "Host DMA bypass" AXI-MM interface, that utilises a separate BAR. See PG195 The support for it is [broken in this driver](https://github.com/Xilinx/dma_ip_drivers/issues/335), but restored it in [my reworked...

I am not using them at the moment, just read the code and wanted to confirm that I am not dumb :D

No, it's unclear, how it is supposed to work. However, in my [reworked driver](https://github.com/Prandr/dma_ip_drivers) I also added proper implementation for the bypass BAR that uses direct device I/O and removed...

> I'm not sure having seek is useful in the case of user BAR. As pointed out in the [#96 (comment)](https://github.com/Xilinx/dma_ip_drivers/issues/96#issuecomment-859010927) one should use `mmap` instead of `read/write/seek`. For example,...

Superseded by the reworked driver on https://github.com/Prandr/dma_ip_drivers

@hisilic Yes it is indeed possible. I have specifically reworked the driver to this end. you can find it on my repo under branch reworked_xdma_main: https://github.com/Prandr/dma_ip_drivers

@robert987-dev this is impossible to do with this driver. you can only transfer virtually contiguous memory. You have to either transfer block-by-block. or extend driver functionality (perhaps with additional ioctl...

@dmitrym1 I test on Intel x86. here is example debug output from the driver. Compare it with original that creates an SG entry for every page and a descriptor for...