How to use the DMA Engine in FreeBSD?

[Kurt Lidl]

On 12/14/18 11:09 AM, Alan Somers wrote:
> On Fri, Dec 14, 2018 at 8:36 AM Kurt Lidl <lidl at pix.net> wrote:
>>
>> On 12/13/18 10:53 AM, Warner Losh wrote:
>>> On Thu, Dec 13, 2018, 3:04 AM Rajesh Kumar <rajfbsd at gmail.com wrote:
>>>
>>>> Hi,
>>>>
>>>> Is there any good documentation available to understand the existing
>>>> support, API's and how-to use the DMA Engine in FreeBSD?
>>>>
>>>
>>>
>>> Usually you just use pci busmastering and it just works.
>>>
>>> I am trying to write a test driver which will use DMA Engine to do the data
>>>> transfer (rather than plain memcpy which involves cpu).  Can anyone point
>>>> to any driver implementation which has similar functions implemented?  I
>>>> see references to SYS_RES_DRQ to allocate DMA channels and play around. But
>>>> that seems to be specific to ISA. Can it be used for PCI drivers as well?
>>>>
>>>
>>> No. ISA DMA is only for really old hardware without it's own DMA engine.
>>>
>>> Look at the busdma api/man page.
>>
>> For some Intel based server hardware, there is the "ioat" driver, which
>> allows for user code to schedule DMA operations.  See ioat(4) for
>> details, including a pointer to the test program.
>>
>> -Kurt
> 
> ioat(4) looks cool.  But the man page is vague on a few points.  Do
> you know the answers to these questions?
> * What happened to ioatcontrol(8)?  It's reference by the man page,
> but doesn't exist anywhere.

root at busybox: locate ioatcontrol
/usr/src/tools/tools/ioat/ioatcontrol.8
/usr/src/tools/tools/ioat/ioatcontrol.c

> * In what context are callbacks called?  Are they called from a signal
> handler, or in a separate thread, or something else?

I don't know.

> * Why isn't ioat.h installed?

I don't know that either, but it is in the source tree:

root at busybox: locate ioat.h
/usr/src/sys/dev/ioat/ioat.h

> * Are "interrupts" synonymous with callbacks?
> * Do you have a rough idea for about the minimum buffer size that
> makes sense to use with ioat?

I don't know that either -- I was mostly just pointing out that the 
facility and driver existed for (some) Intel server hardware.  Mostly 
I'm aware of this because I was surprised when a machine I using this 
summer started reporting this hardware, and I wasn't familiar with it.

Good luck.

-Kurt