[dpdk-dev] [PATCH v1 0/4] Generalize PCI specific EAL function/structures

Thomas Monjalon thomas.monjalon at 6wind.com
Thu Oct 6 15:01:01 CEST 2016


2016-10-06 17:13, Shreyansh Jain:
> Hi Thomas,
> 
> On Monday 03 October 2016 07:06 PM, Thomas Monjalon wrote:
> > 2016-10-03 11:07, Shreyansh Jain:
> >> Hi David,
> >>
> >> On Friday 30 September 2016 09:01 PM, David Marchand wrote:
> >>> Those patches move linux specifics (binding pci devices using sysfs)
> >>> to common infrastucture.
> >>> We have no proper hotplug support on bsd, but if we had some common
> >>> code we should at least try to make the apis generic.
> >>>
> >>
> >> I am not sure if I understood your point well. Just to confirm - you are
> >> stating that the movement done in the patches might not suit BSD.
> >> Probably you are talking about (Patch 3/4 and 4/4).
> >> Is my understanding correct?
> >>
> >> So, movement to just Linux area is not enough?
> >> I am not well versed with BSD way of doing something similar so if
> >> someone can point it out, I can integrate that. (I will investigate it
> >> at my end as well).
> >>
> >> This patchset makes the PCI->EAL movement *only* for Linux for sysfs
> >> bind/unbind. (I should add this to cover letter, at the least).
> >
> > The concern is about function declarations in
> > 	lib/librte_eal/common/eal_private.h
> > We cannot be sure it can be applicable to something else than Linux.
> > As it is implemented in Linux only, it should not be in a common header.
> >
> 
> Ok. But, digging a little I found at least one more similar case.
> 'rte_eal_check_module()' which is present in the linuxapp/eal.c and has 
> existence in common, but no parallel implementation for BSD exists. This 
> function is accessing /sys/modules - which might be Linux specific.

Yes the BSD implementation is missing.
But the common function prototype makes sense as it can be called by an
application and could be implemented for BSD.


More information about the dev mailing list