Re: [PATCH for-next] iw_cxgb3: remove iw_cxgb3 module from kernel.

From: Don Dutile <ddutile@redhat.com>
To: Doug Ledford <dledford@redhat.com>, Jason Gunthorpe <jgg@ziepe.ca>
Cc: Potnuri Bharat Teja <bharat@chelsio.com>,
	linux-rdma@vger.kernel.org, nirranjan@chelsio.com
Subject: Re: [PATCH for-next] iw_cxgb3: remove iw_cxgb3 module from kernel.
Date: Fri, 18 Oct 2019 18:50:24 -0400	[thread overview]
Message-ID: <aea0cf04-bcac-0422-f24a-0158aaac2537@redhat.com> (raw)
In-Reply-To: <3b6fbc551c3a1c53e9365f6af5889ca38e141d3d.camel@redhat.com>

On 10/18/2019 04:51 PM, Doug Ledford wrote:
> On Fri, 2019-10-18 at 17:46 -0300, Jason Gunthorpe wrote:
>> On Wed, Oct 16, 2019 at 01:47:29PM -0400, Don Dutile wrote:
>>
>>> Isn't there a better way to mark a driver deprecated?
>>>
>>> This kind of removal makes long-term maintenance of such drivers
>>> painful in downstream distros, as API changes that are rippled from
>>> core through all the drivers, don't update these drivers, and when
>>> backporting such API changes to downstream distros, we have to +1
>>> removed drivers.
>>
>> You still have cxg3 as an enabled & supported driver? In RH8? Why?
>>   
>>> It's much easier if upstream continues to update the drivers for
>>> such across-the-driver-patch-changes.  heck, add a separate patch
>>> that punches out a printk stating DEPRECATED (dropping a patch to
>>> backport is easy! :) ).
>>
>> The whole point of doing this is to avoid this work!
> 
> People don't quit *using* hardware just because a company has quit
> selling it.  When we can quit supporting it is more about whether
> customers ask for support (or our records indicate lots of systems use
> the hardware) than about whether the vendors still care.
> 
> That said, we probably could have dropped cxgb3 from 8.0, but I'm not
> positive about that.  Just a guess.
> 

cxgb3 is dropped from rhel-8;
yeah, we pruned the tree very closely on RHEL8, and many on this list know that all too well --
b/c ... oh yeah, the vendors EOL'd their devices ... but they are in labs' (well-established) OEM servers everywhere, and now they can't run RHEL8. :(

cxgb3 still quite alive in rhel-7 --> 10yr life cycle ! 7.0 shipped June 2014... so in 2024...
RHEL7 has exited Phase I, where we update the RDMA base to upstream en-masse every RHEL-X.Y release,
so the concern is less now, but some bug-fixes require a refactoring that breaks un-sourced drivers.

Hey, this nut case named Doug Ledford wanted me to keep mthca enabled in RHEL8 for his private lab!
hahahahahaha ....

On the flip-side, I wish we could EOL & remove a whole lot of other code in the kernel (/me looking at old x86 compat code as a prime example...).  other old hw on arch's ... would make a lot of mm cleanup possible without the long, protracted effort that Mike Rappaport goes through!

ok, I made my point.  Do as you wish.