Re: Re: [PATCH] Revert "serial/8250: Add support for NI-Serial PXI/PXIe+485 devices"
From: Greg KH
Date: Wed Nov 27 2019 - 02:43:25 EST
On Wed, Nov 27, 2019 at 07:23:39AM +0000, Je Yen Tam wrote:
> > Subject: [EXTERNAL] Re: [PATCH] Revert "serial/8250: Add support for NI-Serial
> > PXI/PXIe+485 devices"
> >
> > On Wed, Nov 27, 2019 at 11:17:51AM +0800, Je Yen Tam wrote:
> > > This reverts commit fdc2de87124f5183a98ea7eced1f76dbdba22951 which is
> > > commit 225607f75454 ("serial/8250: Add support for NI-Serial PXI/PXIe+485
> > devices").
> >
> > That sentance does not make sense. fdc2de87124f ("serial/8250: Add support for
> > NI-Serial PXI/PXIe+485 devices") is in the 5.4 kernel tree, where did the commit id
> > "225607f75454" come from? I don't see it anywhere in Linus's tree.
>
> Overlooked the commit 225607f75454 which only visible to my own branch, will
> remove it, sorry for the confusion.
>
> > > The upstream introduced a breakage on NI-Serial PXI(e)-RS485 devices,
> > > RS-232 variants have no issue. The Linux system can enumerate the
> > > Serial
> > > PXI(e)-RS485 devices, but it broke the R/W operation on the ports.
> >
> > I can not parse this paragraph either, what do you mean by "The upstream"?
>
> "The upstream" actually means the commit fdc2de87124f. To be more specific, will
> rephrase it to "The commit fdc2de87124f".
>
> > > However, the implementation is working on the NI internal Linux RT
> > > kernel but it does not work in the Linux main tree kernel. This is
> > > only affecting NI products, specifically the RS-485 variants.
> > > Reverting the upstream until a proper implementation that can apply to
> > > both NI internal Linux kernel and Linux mainline kernel is figured out.
> >
> > So this patch really does not work at all? Can't you just find a simple fix to go on top
> > of it to resolve the issue? You really want it all ripped out for 5.5 and 5.4.y?
>
> It does work in NI Linux kernel but not in generic Linux kernel from the main tree, NI's
> Serial team's decision is to have a cleaner implementation so decided to rip out all of it
> for 5.5 and 5.4.y to prevent unexpected breakage.
Ok, fair enough, that's your call. If you resubmit the patch with the
above changes I'll be glad to queue it up.
thanks,
greg k-h