Date: Sun, 7 Oct 2018 12:20:25 -0600 From: Warner Losh <imp@bsdimp.com> To: hruodr@gmail.com Cc: FreeBSD-STABLE Mailing List <freebsd-stable@freebsd.org> Subject: Re: dmesg submission service -- please submit today Message-ID: <CANCZdfriNDP8DD3pkSMeMYFRor=-A2GZ1JuJA8w0xZ0Xm3vZOA@mail.gmail.com> In-Reply-To: <alpine.BSF.2.21.9999.1810071701400.1049@fbsd.local> References: <CANCZdfquWcJDL9J0epARb6=3GWChm6LDVj-wvCGsmmitCPByoQ@mail.gmail.com> <alpine.BSF.2.21.9999.1810071701400.1049@fbsd.local>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Oct 7, 2018 at 11:11 AM Roderick <hruodr@gmail.com> wrote: > > On Sat, 6 Oct 2018, Warner Losh wrote: > > > As you can tell, the project is looking to clear some of the deadwood > from > > its driver lists. One problem is that we have to guess what's in used > based > > on our personal experience. This has proven to be less reliable than > hoped > > in the 10/100 discussions that are going on now. > > Should I dump a dmesg with any device I have connected to every computer > I have? Deadline today? > > Or better I risk not be able to read data on old SCSI discs I am not > using now? > There's no deadline, per se. And there will be at least two drivers that will survive the purge (ahc and mpt). At the present time, they are the only two known to be working for real workloads. So chances are you are already using one of these devices, and if not they are cheap, so there will be a fallback to getting data off of parallel SCSI disks, as well as continued support for parallel scsi tape drives, scsi CD, etc. FreeBSD will support a way to access these devices for years to come. The reason I'm looking at this is because CAM has become rather twisted in places over the years. Many of the reasons for the twistiness can be traced directly to weird workarounds for devices that are almost certainly no longer in use. I'd like to eliminate them. Some can be eliminated because they were for a particular ISA device that's no longer in the tree. But I can't do the others until I know which ones, if any, need to remain. This is why I was asking for dmesgs now so I can start to get a feel for which devices are even still in use on any FreeBSD version, let alone more modern ones. Looking at the NIC data it's clear that we have devices that are wildly popular, and then ones that aren't used with very little in between. Unlike in NIC land, there's a more real and tangible cost that I can point to for continuing old devices in the tree, and I said 'about a month from now' so I can come up with examples of what's a problem, as well as look at data to see what devices still have users, etc, to avoid some of the back-and-forth that we had at the start of the 10/100 discussions as people with devices that turned out to be popular rightly complained. I think having a well articulated reason other than 'it's old' will also help the community understand the cost of keeping things in the tree so we can better judge of the current level of 'benefit' justifies that cost. And any elimination will be safely past 12.0 and won't be back-ported, so anything that works today should work for 12.x releases over the next 3-5 years at the very least. I know that these kinds of discussions can be fraught and I'm doing my best to be honest and open about what's going in, why we want to do thing, and giving plenty of time for the communities that are using FreeBSD to let us know of that use in advance of any decisions being made. I'm hoping the extra data people provide will help us make the initial selection process be more data driven rather than 'best guess' which turned out to be not very good. tl;dr: Not removing anything before 12, devices documented as still working will remain, there's cheap alternatives for removed drivers, don't panic. Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfriNDP8DD3pkSMeMYFRor=-A2GZ1JuJA8w0xZ0Xm3vZOA>