Date: Wed, 17 Apr 2013 21:05:12 -0700 From: Juli Mallett <jmallett@FreeBSD.org> To: Joe Holden <lists@rewt.org.uk> Cc: "freebsd-mips@FreeBSD.org" <freebsd-mips@freebsd.org> Subject: Re: -HEAD on gxemul/MALTA is broken Message-ID: <CACVs6=9rLMUe23eSootVERkENkoTfk4NVJcULfRWWEQLPjy=LQ@mail.gmail.com> In-Reply-To: <516F46E0.7080204@rewt.org.uk> References: <CAJ-VmonpK2AazaLwc1ZQWve0bTFvtLGeF4K8jWM49rzp1j%2Bk%2BQ@mail.gmail.com> <CAJ-VmomiRo0ZpTd2=qeF9M=nBGmhd3jvqUy-4FtUxx4ok2yRew@mail.gmail.com> <516F46E0.7080204@rewt.org.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Apr 17, 2013 at 6:05 PM, Joe Holden <lists@rewt.org.uk> wrote: > I reported this as a pr for Octeon, I'm seeing the same thing - no respon= se > yet though.... is MIPS ever likely to be a well supported platform given > nobody pays attention? FreeBSD on MIPS is unfortunately not just one platform, and the name is a little misleading. There is substantial difference between the code for Octeon, XLR/XLP, the Atheros SoCs and other board-support packages. It's easy for quirks of one or more to go unnoticed, or for someone to be fixed in one without realizing that the fix is also required in fairly-different code on another. We're talking about -CURRENT, too, which is a moving target. I wouldn't suggest using a -STABLE branch for MIPS. I know several companies using FreeBSD on MIPS without constant issues =E2=80=94 I don't think they'd argue that it's not a well-supported platform. What they've done is more effort than the average casual user is likely to put in, which is to identify a point-in-time at which things are mostly working for them, identify the right people to contact to get their major issues fixed, and then stay with that version for some time, rather than either trying to track top-of-tree or getting stuck with limited support in -STABLE. I run -CURRENT on Octeon on dozens of boards from time to time, but on an ongoing basis on just two. It's mostly stable, although I do notice some crashes on long-term uptime, and as I've mentioned to several people in private, getting the most out of Octeon requires at least doing a lot of your network processing in the kernel, or ideally writing some slightly-specialized software. I have a Broadcom board that I run occasionally, a couple of Atheros ones, and probably a few others I'm forgetting. It's understandable that when just starting to use a port of FreeBSD -CURRENT to MIPS, on finding one or two things broken, to figure that (1) it's never worked (2) it's broken for everyone. Both obviously lead to the conclusion that nobody cares, and that things are incomplete and unmaintained. But the amount of hardware FreeBSD/MIPS supports is incredible. I've personally run FreeBSD/MIPS on perhaps two dozen different hardware configurations, and have helped people run it on probably a dozen more. That's just stuff I take some responsibility for. I know of perhaps a hundred more that FreeBSD runs on. As for not paying attention: I don't update very often, and my focus is really mostly on FreeBSD on Octeon when I can afford to pay any attention at all. Adrian pays a lot of attention, but his focus is mostly on Atheros hardware; he's able to spend the time and have hardware access in part because of his relationship with that company. There are others in similar positions, such as jchandra who puts in huge amounts of work on RMI/NetLogic-based hardware, and with me has done a lot of work on expanding 64-bit support and improving performance. Neel has lots of other responsibilities, but has handled a lot of work on keeping various MIPS hardware platforms working, as well as being responsible for things like a fairly robust SMP abstraction for MIPS, and the first real work on SMP FreeBSD/MIPS, which made it a matter of hours for me to add SMP support for Octeon. What kind of attention would you like to be paid? How much latency in addressing issues do you think is acceptable? How much work do you think it's reasonable to expect users to do to get up and running with what are, almost exclusively, embedded platforms? It may be that the answer here is that we should wait for pfSense MIPS support to become widely-available and stable, with ready-to-use images. FreeBSD/MIPS doesn't even come with a package set, as you know; it's not a one-click operating system, it's more like an operating system base. Is it likely that FreeBSD/MIPS will become more usable? Yes. Due to tremendous amounts of volunteer labour from people who don't have a lot of spare time, and the occasional burst of academic or commercial support which allows for significant improvement to occur in a short period of time. I try to find others to get involved (like Pat Kelsey who has been very active in responding to SD and MMC threads since I hired him to support MMC over SPI, which was a crucial missing feature on several popular Atheros-based boards.) It's not always easy; sometimes people's priorities or foci change, as with one engineer I was delighted to be working with from a major SoC vendor, who stopped answering E-Mails after committing some of his work had to be delayed due to license issues with some code his employer wanted him to commit. Whether he got reassigned, his bosses soured on FreeBSD's inability to accept GPL'd Linux code into the kernel, or he simply felt there was nothing left to do, I don't know. Being an open source developer is a lot more than waiting around for bug reports to arrive and then fixing them. Even in paid software development that isn't how it works; things get prioritized, and often dealt with opportunistically. There's cases where in my paid work I've known about a bug for years before I've had time to fix it because other, more pressing issues keep coming up. And in open source work, we're not just dealing with prioritizing work, dealing with long-term planning, putting out fires, trying to keep up with others' changes, and deal with bug-fixing. We also help to enable each other to work better, and to become involved, and to stay involved, and to fix things. I'm happiest when I'm able to help others work on things they're excited about, and I think that's when I'm doing my best work as a developer of open source software. Fixing bugs is essential, but it's not always what there's resources for, especially when FreeBSD gets, speaking for myself, such a slim sliver of my time, because I have so much else on my plate. I do the best I can. If you can do better, I'd be thrilled to help you get more involved with FreeBSD. I can respond to E-Mails about things you're passionate about or interested in when I have no time, or energy, or focus, to update to -CURRENT and make sure everything's working great. I would love to see you become a FreeBSD developer and make it better. It doesn't get better magically, but through the commitment and hard work of people like you, or where that's not your skill, through applying whatever yours is. Maybe you can help fund work; maybe you can find bugs and hound people to get them fixed. Making sure bugs gets fixed is important work, that requires discipline and focus; it isn't every software developer's gift, and if it's one you have, I'd encourage you to make use of it. Don't be afraid to harangue people for breaking -CURRENT, but I'd discourage vague insinuations of poor quality because you've hit a couple of roadbumps. (Which is not to say those roadbumps should be there, but to ask whose responsibility you think it is to level them.) I can't do as much as I like, but in addition to supporting other non-profit causes that are important to me, I give money to the FreeBSD Foundation. Their focus isn't FreeBSD/MIPS as such, but I'm pretty confident that they'd help fund FreeBSD/MIPS projects if you can find someone interested in working on them. I benefit a lot from the work I put into FreeBSD; I have had several jobs pretty much entirely because of it. FreeBSD means a lot to me and I don't want it to suck, but there's only so much I can do. So I don't just take all FreeBSD has to give me: an operating system to run, a regular (read: overwhelming) supply of paid work, etc. I find ways to give back. I'm sorry for the length, but I think it's important to be very clear about two things: (1) it's people like you who help make FreeBSD better (2) they work very hard doing it, and the success of that work should not be underestimated, even if it is somewhat undermined when -CURRENT is so routinely in crap shape. I suggest specific and regular technical complaints, though, over insinuations that we simply don't care, or don't pay attention. I want to end there, but I'll say this: in addition to a few messages from you, I've had half a dozen more this week alone from people who need help with some aspect or another of FreeBSD on Octeon (getting hardware, supporting new boards, improving the port, getting started, etc.) I don't even have time to respond to all of them so much as they deserve, but I do the best I can. (Which means that I usually respond to FreeBSD-related E-Mail ahead of other personal correspondance because it's important to me to try to do so.) On top of that, dozens more messages related to FreeBSD in general, including work that I'm funding, work that I'm doing, and the work that I (and most other developers) do reading and reviewing and responding to changes made to FreeBSD. Thanks, Juli.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CACVs6=9rLMUe23eSootVERkENkoTfk4NVJcULfRWWEQLPjy=LQ>