Well, That Was Fun – not

Ah, the life of “the live in tech support” and hacker in residence…

I’d been having a load of “issues” with the WiFi. I won’t go into detail. It was basically having “failed to connect” and “authentication failure” (even on the ‘open’ unauthenticated network) and more. Now the WiFi router was in one end of the house, and the spouse with laptop was in the other, but signal strength was still about 60%…

There was a ‘wall of stuff’ between antenna and laptop, maybe acting as chaff? There was DNS lookup failures, that didn’t seem to care what DNS was used. And more.

I set about taking out potential problems one at a time. Moved the router to the middle of the house (ran wire through wall… nice to own your walls… and a hole punch…). Took down a second WiFi router (maybe signal interference? Ought not to be but… Also simplifies routing tables and more.) Checked router configs. Check this, checked that.

Every change seemed to work sometimes, then fail. Eventually this morning had a tablet working, and the laptop not, in the same room in the same lap. Hmmm… Tablet had fixed IP, laptop DHCP (I’d made it fixed on the encrypting router, but not the “open” one; but that had not helped on the encrypting router). Found one of the old XP boxes had a ‘secondary IP’ on one of the wireless interfaces that would have polluted the router tables to the boundary router and removed it, don’t know if it was an issue, but fixed it anyway. After making that laptop fixed IP on the open network, everything was fine… So is the last “fix” the only problem, or just the last one of several? I’ll never know, as I’m not going to regression test.

As a nice side effect, I also got a wired connect in the living room. I already had a cable laid from the middle room to the living room, so just plugged it into the WiFi router hardwire port. Now I can use my Linux box even when the wireless dongle is having driver issues… which it does on some live-CDs.

That put me to about 4 P.M.

But at last, the wireless networks are stable on all platforms. Tomorrow I can go back and integrate the encrypting router and complicate thing up again (but keeping it stable at each step). Why have two networks? I use the encrypting one most of the time, and power up the open one for guests and such…

The Hanging Debian

I’d installed Debian Wheezy on the EVO, and generally liked it a lot. But it would sporadically hang when using a browser. Didn’t matter if it was Iceweasel, Firefox, Chromium, or even Konqueror. Sometimes it would get past a glitch moment and be fine. Mostly it would just rapidly lock up. Cursor would move, but nothing else could be done at all. I got tired of crashing the box…

I tried different “windows managers” and “desktop environments”. Fluxbox, Openbox, KDE, XFCE, LXDE… The heavier weight ones tended to hang faster. Fluxbox let me stay up longer, sometimes. Using the ‘event monitor’ I found it would ring the bell and have a few useless messages before hanging. (Syslog had nearly nothing useful). Somewhere between browsers, X-windows, and libraries things were not well…

I’d been using the CentOS box and some live-CDs rather than deal with this for the last few weeks. But decided “now was time”.

A bit of pondering “what to install” as a replacement Linux had me bothered. Ever more releases had gone to ‘systemd’ as the initialization program. Classical Unix / Linix used init. I have a few decades of using init, and didn’t like the idea of tossing out everything I know and starting over. Systemd also violates the Unix / Linux ethos of “do one small thing and do it well”. It is from the Microsoft school of “put a load of things in a black box and if something goes wrong, reboot”. I was not liking my choices, as several major (and favorite) releases had gone that way. Fedora, CentOS, even Debian and Ubuntu had gone over. Gnome strongly integrates with systemd, so if you want to run gnome, you get systemd’ed like it or not. (That’s part of the problem with the ‘big ball of wax’ approach. Swallow the whole turkey, or STFU.)

For reasons beyond my caring, the “kitchen sink” folks have been stuffing everything into the “desktop environment”. KDE got junked up trying to be Gnome Lite. Gnome is trying to be everything in the desktop, and forget the old Unix heritage of “one thing and well”. Sigh. I was already stuck with gnome on the ASUS (though being a 6.x level CentOS, not yet systemd) and didn’t want it shoved at me on my ‘play box’ too. What to do, what to do…

Gentoo was attractive, but I didn’t want to do such a radical update / replace if possible. I’d already been saving email and stuff in my home directory on this box. Rebuild from scratch was, er, bothersome.

So while searching for “what release doesn’t do systemd” I found a discussion of the food fight at Debian over it. Several developers up and quit over it. For the same reasons that I was not keen on it.

BUT, the good news is that they started a new fork of Debian that was init based. And, to swap to it, one just changes what ‘repository’ your system points at, and does an “apt-get update” then “apt-get upgrade”. I did that.

The number of things replaced was surprising to me. About 300 MB. Lots of libs and such.

The install went well, nearly flawless, and needed no intervention. (At some point I need to separately upgrade a database component postgresSQL IIRC the log…)


All my user space is as I left it. Wallpaper and all. Launched the Chromium browser and nothing hung. Using it now to do this posting. At the “few minutes into it” time, there was a small ‘video glitch’ with blink to black, but when it returned, things were still working instead of hung. ( I suspect that whatever causes that ‘blink’ is the trigger for the lock up in Debian; but sometimes there isn’t a blink, just a hang.)

This isn’t fully proven yet. I did have Debian work once or twice without fail. It will take a week or two to see real provable stability. But for now, it is sure looking like a fix.

Only “downsides” are that the memory footprint looks larger ( 900+ MB where before it was about 600 MB?) and the Chromium browser has a bit of type-ahead. Probably as this is a more development release and has a lot of diagnostics and such left turned on / compiled in. With luck, the other browsers will not have type-ahead.

For now, I’m liking what they are doing. Devuan is pronounced “Dev-one”, and I’m going to be using it for a while…

Some Links

There’s the generic Debian:


This is the Devuan site:


OK, nevermind the names, but what’s the plan?
Init Freedom logo

Have you tried to opt-out of the systemd change in Debian⁽™⁾ and stay with sysvinit, or whatever other init you prefer? You will quickly notice that is not a matter of choosing packages and in fact Debian offers no choice.

We want freedom of choice, we want Init Freedom!

We are working towards a stable, production ready fork of Debian Jessie, free from the entangling web of dependencies imposed by systemd.

One can just migrate to Devuan by installing devuan-baseconf.deb which will re-configure the system with our source repository:

deb http://packages.devuan.org/merged jessie main

Beware this APT source is exclusive and cannot coexist with Debian’s repositories.

This wiki discusses the issues with systemd:


While clearly intending to be promoting the good things it does, this description just looks like a huge ball of ‘fingers in every pie” and not so much like “one thing small and well”:

Poettering describes systemd development as “never finished, never complete, but tracking progress of technology”. In May 2014, Poettering further defined systemd as aiming to unify “pointless differences between distributions”, by providing the following three general functions:

* A system and service manager (manages both the system, as by applying various configurations, and its services)

* A software platform (serves as a basis for developing other software)
* The glue between applications and the kernel (provides various interfaces that expose functionalities provided by the kernel)

systemd is not just the name of the init daemon but also refers to the entire software bundle around it, which, in addition to the systemd init daemon, includes the daemons journald, logind and networkd, and many other low-level components. In January 2013, Poettering described systemd not as one program, but rather a large software suite that includes 69 individual binaries. As an integrated software suite, systemd replaces the startup sequences and runlevels controlled by the traditional init daemon, along with the shell scripts executed under its control. systemd also integrates many other services that are common on Linux systems by handling user logins, the system console, device hotplugging (see udev), scheduled execution (replacing cron) logging, hostnames and locales.

It is just entirely wrong from a Unix ethos point of view, and very Microsoft like.

In a September 2014 ZDNet interview, Theodore Ts’o expressed his opinion that the dispute over systemd’s centralized design philosophy, more than technical concerns, indicates a dangerous general trend toward uniformizing the Linux ecosystem, alienating and marginalizing parts of the open-source community, and leaving little room for alternative projects. In this he found similarities with the attitude he found in the GNOME project toward non-standard configurations. On social media, Ts’o also later compared the attitudes of two key developers to that of GNOME’s developers.

GNOME is now dependent on systemd, and they are pushing to have logind hooked into GNOME as well. As someone who likes other windows managers and desktops better, being forced into a Gnome Centric World against my will is not attractive.


By Jim Lee Feb 16, 2015, 17:28

I wouldn’t have a problem with systemd, except for 3 issues I feel need to be addressed: (1) if systemd is going to be some monocultural thing to all distros (like MS became to all PC’s) then I hope the code is bulletproof – one single exploit will be all you need to screw every single PC, laptop, tablet, server, etc running Linux. I’m sure the NSA would love that. (2) One thing which pushed me over to Linux more than a decade ago is, if something goes wrong with my OS, I can roll up my sleeves and get to work fixing it; the OS will tell you what’s wrong, it just means paying attention to the various logs and the boot sequence. Systemd, OTOH, seems to obfuscate these clues. Will I wake up one day and find Linux just as cryptic to diagnose as Windows? (3) I really, REALLY have a problem with how this seemed to be just shoved down everyone’s throats. Could the adoption of systemd not have been handled a little better than this? No wonder Poettering seems to have left so many people with a bad taste in their mouths… Don’t get me wrong: I’m not some sort of techo-Luddite, and systemd seems (at first glance) to be an attempt to rectify some of the flaws with the way Linux presently handles some things. I simply wonder why did its scope have to go beyond handling init functions. The registry was Windows’ Achilles’ Heel for a long time; is systemd eventually going to be that for Linux?

Reply to this comment Reply to this comment
By Rainer Weikusat Feb 17, 2015, 15:31
systemd is supposed to ‘rectify’ the flaw that some code is written in programming languages its authors really don’t like and maybe additionally, to ensure that every part of the system is exclusively accessible to other parts of the system using the freedesktop/ GNOME invented RPC protocol. Anything beyond that is lipstick supposed to make you confuse the pig with a damsell in distress. In particular, it doesn’t provide anything new in the areas of startup organization or service management unless you’re someone who never ventured beyond “whatever RedHat happens to ship today”. Recoding a massive part of the system in a low-level compiled language designed in the 1970s is not ‘a modernization’ and replacing some 20,000 (overestimate) lines of code written in a mix of higher-level and lower level languages with 600,000 lines of code written in a lower level languages is as close to a techno luddite’s dreams as it could possibly get.

Oddly, one of the log entries that came just before the hang had “org.freedesktop” in it…

Looking around on systemd “issues” and problems finds a lot more fire and noise like this.

On to other things…

At some point I’m going to install a Gentoo with ulibC (used in routers and embedded systems a lot, smaller and more secure libraries) and may end up there for my Daily Driver someday. ( I was headed down the path of buying a new WiFi router as the one with DHCP issues is from about 2002 and only goes 11 MB anyway… but now that I have time, I’ll likely ‘build my own’ so I have more control of it. OpenWRT first, but maybe with Gentoo for the experience later…)

Gentoo site:


The Gentoo Linux style is small, efficient, and secure. They compile from source code instead of downloading mystery meat binaries. They support both the larger fatter libraries and the small ulibC libraries aimed at embedded systems. Definitely more of a hacker oriented style than a “want Windows look but never touched Linux before” approach.

In Conclusion

At some point I may be pushed back into the BSD world. There are OpenBSD and FreeBSD kernel version using Linux tools. For now, Devuan still has the easy package and easy install of Debian, but without the style pollution of Gnome / systemd. I’m likely to ride this horse for a while and only move if forced out by some problem. The very light live-CD distributions (like CrunchBang that has now stopped development, and the Puppy series) still are systemd-free) are also an option. I’m running a CentOS 6.x release, and Red Hat went to systemd in the 7.x series, so eventually I’ll have to abandon CentOS (or stay back rev). Probably not much of an issue as I tend to stay on a release for a long time…

But what is clear to me is that Red Hat is taking their flavor of Linux in a direction very much against the ethos of Unix and the original Linux. The other thing that is clear is that in true Open Source style, those of us who value that original approach can just fork our own path and carry on.

But enough for one day. Now I have a working 64 bit development box with CentOS for GIStemp and running some of the climate models. I have a stable WiFi (with stable and secure coming tomorrow). And from the looks of it, a stable Linux Daily Driver – now that systemd has been rooted out of it. And the old XP box working too, for any legacy stuff I might need. With all the infrastructure pretty much working now, looks like all that is left is to get a project plan laid out and start on the “Dig Here!” list ;-)

Subscribe to feed

About E.M.Smith

A technical managerial sort interested in things from Stonehenge to computer science. My present "hot buttons' are the mythology of Climate Change and ancient metrology; but things change...
This entry was posted in Tech Bits and tagged , , , , . Bookmark the permalink.

25 Responses to Well, That Was Fun – not

  1. p.g.sharrow says:

    @EMSmith; Glad to see you are having fun ;-)
    We have router problems from time to time as well. Organized confusion with satellite modem to main router, Ethernet to 2 routers with WiFi, each about 100feet from the main as well as an WiFi extender/router. All to cover our spread out compound.

    Working with a bunch of different flavors of OS might be a good learning experience, yuk! glad it’s you and not me, yet. I have been trying to get my grandson to set up one of our better boxes into a Linux work station, while experimenting with our Raspberry PIs. pg

  2. Lars P. says:

    “The Hanging Debian”
    Can it be that the problems you have come from the hardware in use, like the graphic card for instance?

  3. Chuckles says:

    E.M. I’ve found that several Linux flavours are susceptible to the ‘rock solid hang’ in the last year or so, particularly when used in VMs of various types. I suspect, but have also not regression tested, that it’s related to the fashionable ‘no swap’ setup they all usually do these days, as after adding a reasonably sized swap file to the offending instances, it’s never recurred…

  4. beng135 says:

    Both my Puppy and Mint Linux have trouble connecting wirelessly. It will work fine for a day or two, then unable to connect for a day (even tho signal strength is 100%). Searching forums, it seems the Atho9 wireless driver is “buggy”. Some suggestions were to disable any “power management” of the wlan0 network interface, but I surrendered & just use the reliable hard-wired (eth0) interface.

  5. gallopingcamel says:

    My OS is Debian (Mint 17).

    When using Chrome to watch Netflix (connected wirelessly) I get those black screens a few times per hour. Misery loves company…………..at least I now know it is not something I did!

    The bad news is that with your vastly greater knowledge of computer software you can’t fix it. That suggests that someone even higher on the Linux food chain will have to step in. How long that will take?

  6. E.M.Smith says:


    The flavors I’ve been using are not that divergent. Red Hat has been around nearly forever, and I’ve been using some flavor of it the whole time, so it’s not “new” to me at all (other than the systemd conversion that’s hanging out there in 7.x and where I’m unlikely to go).

    I’m not presently using a BSD version, but have spent a good few decades using it, too. Just rarely at home. It is the most divergent from Linux. (Or, more accurately, Linux has wandered further away from BSD as Linux left the Unix roots and mutated…). It is more of a “hands on” and “straight forward for a sysadmin, users not so much” ethos. Harder to initially install. More hidebound on ways of doing things. But rock solid in use.

    Debian is the “Always fresh and user friendly” fork. Lots of “fresh meat” and lots of “eye candy” and lots of “easy install” and… but more prone to “issues” as things are a bit more chaotic and rapidly shifting with little enforced style consistency. Because of that, though, there are many derivatives from it. Ubuntu, for example, is Debian made stable and commercialized.


    has a nice evolutionary list of Linux releases. (Just remember that “In the beginning, Unix Version 7 begat BSD that went to University everywhere and begat all the rest… only later having an incestuous affair with System V that came along later and leading to the systvinit and POSIX).

    This diagram from that article is very enlightening. VERY large, but it shows how releases fork, and then evolve, live, or in some cases die out.


    You can see that there are just 3 mainline stems from which most variations descend. Red Hat, Debian, and Slackware. (Knoppix is a Debian descendant, for example. Then a hoard of Knoppix BBC type releases sprout from it). This matters.

    If you have a lot of Red Hat experience, then the derivatives are generally familiar. CentOS is just a Red Hat with more industrial spin and less eye candy. It is Red Hat Enterprise with the branding stripped out. Similarly timing has the start of Fedora Core. The Red Hat home user oriented fork.

    Same thing for Debian. Tails is a Debian derivative, as is Ubuntu (that sprouted CrunchBang and Chromium among others). So if you have used Ubuntu, moving to straight Debian or booting a CrunchBang live-CD is a small step. I’ve only lately used Debian releases, mostly starting with the bootable CDs in about 2004? or so. So about a decade. (And during most of that time Red Hat was my Daily Driver and on most servers I’d install at clients).

    Now look at that graph. There’s a lot of turf covered by the Debian and Red Hat flavors…

    Slackware is the big one where I have the lightest exposure. Used SuSE for a while (it is derived of Slackware in about 1994, but I tried it from time to time from about 1996 – 2008 or so. Slackware is more of a Unix Traditional POV, while SuSE is more POSIX / authority compliant with a German “Ve Kno Betterrr and You Vill Like It!” overlay. (Fewer choices, more mandates, and sometimes surprising complexity for things that don’t need it. But works.)

    After a while, shifting from one minor variation of one of the major base releases to another minor variation is more like re-skinning than anything else. I’d hardly call them “different operating systems” at all. The word you used, “flavors”, is more accurate.

    Unfortunately for me, since Red Hat seems to have given Poettering free reign to complexify, integrate to excess, and obfuscate what ought to be simple, modular, and clean, CentOS 6.x will be the last Red Hat I run at home.

    Thus my increased exploration of things non-Red. Debian was the primary candidate, but the instability has been ‘not-fun’ and discovering their embrace of systemd a discouragement. We’ll see if Devuan is the fresh horse in this race; or if I’m going off to further explore Slackware and Gentoo. (Both more hacker oriented with ‘build from sources’ more frequent and “roll your own” style of release. Gentoo tilted toward small hardware, old and embedded systems, and always compiled from scratch… you download sources for things and compile locally, so binaries take full advantage of YOUR hardware and don’t try to use things that are not there).

    @Lars P.:

    It would be a reasonable concern, except that the funny video chips are in the ASUS running CentOS (and very stable on it).

    The video glitch / hang is on the Evo, that has vanilla hardware choices, was widely sold, and runs a load of “Live-CDs” fine. So I’m pretty sure it’s not hardware related.


    Interesting. One of the hangs came with a Kernel Panic on a failure to swap a kernel page. I had 4 GB of unused swap mounted then, though. Most of them do not have that message.

    Also, I’ve got 2 GB of memory in the Evo, and it rarely to never actually hits swap.

    I suspect there are two interacting issues here. One with swap. The other with X-Windows.

    So, at about the 3? minute mark after boot up this morning, the screen “glitched” again. The wallpaper went away and became a black background. Usually it is just a flicker of black, and then back to normal (and never returning again after ‘one and done’). Today I was moving the mouse right when it happened. The background stayed black, but the system didn’t hang. Moving a window around would refresh the wallpaper anywhere it had moved. So “whatever” causes the hang in Debian is handled much better here. IMHO it’s an X-Windows bug / glitch that is just a flicker so nobody cares to investigate it, that with systemd causes a hang as everything is integrated in it (and likely doesn’t do an init-respawn as well…)

    It is also likely that the ‘glitch’ only shows up in some older hardware so the folks at big development houses (like Red Hat) never see it and can’t find it. And likely don’t care. They tend to “move on” with new hardware frequently and only care that it works on what large companies are buying now or bought in the last 3 years. At one time I ran the software QA group at a compiler house. We had 72 different systems and crosses in the test set. Just for that kind of reason. I doubt anyone does that anymore… The Microsoft model of “ship it and let the customer complaints roll in” seems the current fad…

    @Beng 135:

    Good to know. But the laptop was a Macintosh, and the other systems (including Windows XP) had issues too. It was the XP running a nice 3rd party wireless management ap that showed me the ‘failed to get IP” on renewal and pointed me toward a DHCP failure. (DHCP seems to work better on the newer Netgear with encryption. I’d only shut it down to simplify the search space and eliminate encryption compatibility issues — the Mac wanted one set of crypto-settings and Linux wanted another…)

    I suspect that the XP was also polluting the router table pool as it had a 192.168.1.x secondary address set on one network interface that connected to the 172.x.x.x wireless and my boundary router is on a 192.168.1.x network… so the router tables would be getting 192.168 in two different directions. Just speculation at this time, but cleaned up anyway.

    Some helpful history of the WiFi box also can enlighten:

    The WiFi router is a Linksys cheapy from about 2002 and well past ‘use by’ date. It’s gone through some rough times over the years, and it’s sibling Linksys that was newer got fried in a power surge. (The crossbeam on the power pole behind the house ‘let loose’ and dropped high voltage lines onto low voltage distribution lines… the other Linksys was near a window with aluminum trim and has a burn line from the power capacitor on the PC board to the antenna as something shot to ground via the antenna that was very near the aluminum…) I suspect that same power surge may have slightly scrambled the brains in the one that ‘survived’… Lucky for me the computers were not plugged in then as I was in Florida.) At any rate, the boundary router to the Telco ‘sort of worked’ after the surge, but tended to reboot a lot. It eventually died (story posted earlier) too. So there’s lots of reasons to think a power surge took out all three to various degrees. That only DHCP service is flaky is actually the surprise.


    I’ve studiously avoided anything to do with X-Windows as much as possible. It’s just a horrid crate of worms. From having ‘server’ and ‘client’ backwards to over the top complexity and more. Ubuntu is coming out with “Mir” to replace it. GEM was used in Atari and likely would have taken over (IMHO) but for:


    GEM 1.1 was a window manager that supported the desktop metaphor, and used stacking, allowing all windows to overlap. It was released in the early 1980s. GEM is famous for having been included as the main GUI used on the Atari ST, which ran Atari TOS, and was also a popular GUI for MS-DOS prior to the widespread use of Microsoft Windows. As a result of a lawsuit by Apple, GEM was forced to remove the stacking capabilities, making it a tiling window manager.

    As tiling is not my preferred way of doing windows, nor most folks, and that kind of sidelined it.

    There’s a new window system being developed (Wayland) that has hope.


    Wayland consists of a protocol and a reference implementation named Weston. The project is also developing versions of GTK+ and Qt that render to Wayland instead of to X. Most applications are expected to gain support for Wayland through one of these libraries without modification to the application.

    To get an idea of how the “kitchen sink” folks have complexified things into hell, see the chart of “default programs packaged” about 1/2 way down here:


    I count about 20… Now one might wonder why given packages get tied to the X-Window system… IMHO it’s because there isn’t a clean API that is neutral on what window system is behind the interface. Again, a violation of that “simple, one thing, and do it well” ethos of Unix.

    Oh Well. Maybe it will be fixed with a future replacement of X-windows…


    has a nice list of other systems, so you can explore for a long time…:

    List of windowing systems
    For Unix-like operating systems

    8½ and rio for Plan 9
    FramebufferUI (fbui) in-kernel windowing system
    HP Windows
    ManaGeR (MGR)
    NeWS / OpenWindows
    NeXT DPS
    Qt Extended
    Quartz Compositor (Mac OS X)
    Twin (Text WINdows)
    W Window System
    X Window System

    For Windows NT-family operating systems

    Desktop Window Manager (DWM) in Microsoft Windows (Vista and later)
    ReactOS Explorer
    Classic Shell
    Talisman Desktop

    Web windowing systems

    TIBCO General Interface
    Web Window Manager


    Visi On

    Commercial systems such as Microsoft Windows (XP, 9x and earlier), Mac OS (version 9 and earlier), and Palm OS, contain a windowing system which is integrated with the OS.

    So I think you can see why, with that level of complexity and “stuff to sort through”, I’ve avoided “going there”. It’s pretty much a specialty to be the windowing systems guy…

    But, FWIW, testing a Wayland based Linux release of some sort is on my “someday ToDo” list.

  7. E.M.Smith says:

    Well, we have an answer…

    Plugged in the Netgear WiFi router with encryption. In short order found that DNS was screwed up, even on the linux machine that was hard wired to the boundary router.

    Just amazing.

    Even doing an “nslookup yahoo.com” where that’s the boundary router IP would not find any result. I could ping, but DNS lookups failed.

    Unplugged the Netgear. Everything worked again…

    OK, well, nice to know. So pitch that box and buy a new one and move on. Time to go shopping for an OpenWRT capable WiFi router made in this millennium ;-)

    But really…

    Boundary Router <===== Netgear
    ^= Linux box

    and the Linux box can't get DNS service directly from the boundary router if the Netgear is plugged in? …. (And yes, I checked the config on the Netgear a few times, and renumbered things a couple of times to assure no fat-finger-number overlap; even moved it's private side number block from 192.168.0.x to 172.16.17.x to it was not even remotely close. Used DNS via hard numbers and used DNS via DHCP on both clients and Netgear… Oh, for a packet sniffer… I'd love to know what kinds of packets that Netgear is shoving at the boundary router to screw it up…

    Just glad the failure mode showed up in the first 10 minutes instead of working fine for a few hours and popping up after the management station was disconnected and I was doing other things…

    Sometimes it is far more valuable to have a definitive answer even if you don't like the result…

    With that, I'm doing a quick web search on OpenWRT routers and a trip to the store…

  8. Paul Hanlon says:

    Hi Chiefio,
    The Raspberry Pi has a port to Wayland built in. https://www.raspberrypi.org/wayland-preview/. I haven’t tried it myself, but I remember hearing positive reviews at the time. It uses the GPU to do its work, freeing up the CPU for other tasks, instead of putting it all on the CPU with X-Windows.

    My own project of getting a Cubieboard to act as a Personal Home server is progressing nicely, albeit slowly. It’s a great little board that comes with Android 4 installed onto the 4GB NAND flash memory, but I’m using Cubian, which is their equivalent of Raspbian for the Pi. So far it has been running non stop for six months.

    I’m using it as a backup / NFS, sensor and data collector, streaming media and file server, Git / Development platform. I’m trying to do the entire GUI in HTML5, which is the slow bit. I’m also waiting on a Mobile Phone backup battery, which when it is all set up will mean I will have a UPS built in so that it should never die (well, unless we really do have the Zombie Apocalypse in my lifetime). LightInTheBox are doing 10Ah batteries for €20.

    So Cubie attached to 1TB internal hard disk through SATA (really fast) plus external hard drive through USB (perfect for backup), plus UPS, plus ESP8266 with decent aerial to act as second WIFI access point, possibly MQTT server for sensor data. These are all stacked up on one another so it will fit inside a box about 7″ x 3.5″ x 2.5″. I helped a friend out with some stuff he needed doing, so he’ll print me out a box on his 3D printer to enclose it all. To date it will have cost me $150, or just over a one year subscription to Dropbox or Spotify.

    I’m using no-ip for dynamic DNS, but I’m thinking of moving over to DuckDNS as I believe that is completely free (no threats to switch you off every month if you don’t update). This gives me access from anywhere in the world. No Google, no NSA, except when I invite them in. It’s great that all these services exist online, but even nicer when one can provide it all oneself. It just keeps everything mannerly.

  9. gallopingcamel says:


    My attempts to make the X-server behave have been frustrating.

    I will keep my fingers crossed in the hope that Wayland will be more user friendly.

  10. Chuckles says:

    https://www.wireshark.org/ should help in your researches, E.M.

  11. beng135 says:

    Chief, actually my Win7 also had minor but persistent problems w/the wireless from the start — I would often have to disconnect, then reconnect the wireless before Windows would wirelessly connect (or rarely have to turn off/on the router). But for some reason that issue eventually went away…

    And yes, the Linux connection-problem would always come w/a “DHCP failure” message.

  12. E.M.Smith says:


    Looks like there is a new mis-feature in Firefox. I’d thought it was a bug in Devuan, but it just bit me in MacPup too. Got a good 5 paragraphs typed in a comment, scrolled up with the mouse wheel kind of fast, and it launched one of the links I’d pasted into the comment. Doing “back arrow” got me back here, but the comment text is gone. Dang it. Until I find out how to shut that off, the mouse wheel is a land mine. Lord Save Me from Unwanted Feature Hell…

    Starting over…


    Wireless really needs a ‘do over’ for everyone. It’s just not well thought out yet, IMHO.


    Thanks! Vague memory of it being mentioned years ago, and then I forgot about it… I’ll see about giving it a try after I’ve put the new WiFi in service. ( The old ones will go to the experiments and researches pile…)

    @Paul Hanlon:

    I’ve been planning my own “server farm in a box” too… but with different hardware and a more complicated set up. We’ll see if I ever get around to doing it… ;-)

    Take a look at Banana Pi:

    Likely my next HW buy. Like a R.Pi with a lot more punch for about $10 more…

    That Wayland on Pi looks really interesting…. I’m willing to flash a SD chip for it ;-)


    I’m presently typing this on a MacPup on a USB stick. I’m liking it. Faster than the Debian / Devuan and a lot smaller. No video glitching at all. I’d been running from the Live-CD for a few weeks and finally got around to doing the ‘install to USB’. I’ve not yet tested things like setting up a personal account (by default you are root but can su to non-root accounts “spot”, “fido”, or “webuser”. I’m going to test setting up my own real personal home directory and account and verify file persistence across reboots.

    Frankly, though, this is the experience I’ve been trying to find again. Fast. Clean. Not bloated and glitchy. IF this persists and can be set up like a real home-dir I’m likely to just move onto the USB drive and then Debian / Devuan can just be a technical play land / devo tool. I’d be fine with that.

    At any rate, you ought to burn a Live-CD of it and give it a whirl for the ‘no glitch’ experience ;-)


  13. E.M.Smith says:

    Well, after a bunch of web searches leading to dead ends, I decided to go searching in the config of FireFox.

    I think I’ve fixed the unwanted history scroll on fast mousewheel movement. Using about:config in the address bar, and changing the “2” to a “1” in this line seems to have fixed it. Even though I’m not using any “shift action”…


    This link was kind of helpful, but had the wrong text, still the number value pointed me at “2” as important:


    Modifier key is one of: withaltkey, withcontrolkey, withmetakey, withnokey, or withshiftkey

    Type is then one of:

    integer value that determines the type of action:

    0 – Scroll document by a number of lines (given by the numlines property)
    1 – Scroll document by one page
    2 – Move back/forward in history
    3 – Make text larger/smaller
    4 – Scroll document by a number of pixels (given by the numlines property)

    so I found a “mousewheel” entry with a “2” and made it a one, even if the text didn’t have an exact match.

    In a variety of rapid mouse wheel moves, no bad thing has happened, so I think that fixed it…

  14. Paul Hanlon says:


    Once I’ve done with this, which will take a little more time, I’m happy to share. If you do “roll your own”, you are better off using the Cubieboard. I started off with the BananaPi, and I have to say I liked it. Zippier, SATA, same form factor as RPi. Then I switched harddisks and whatever I had done, the BananaPi stopped working.

    Enter the Cubieboard. By the time you pay for the SATA cable and the power cable (which is free with the Cubieboard), it is slightly cheaper than the BPi, and better packaged. And it comes with the NAND chip which can house your OS. Performance is the same as the BPi, both use the A20, but backing up seemed faster. Switching harddisk (which I don’t do often) posed no problems.

    That said, I did manage to corrupt an SDCard when I accidentally unplugged it. Luckily, it was very early stage then. And now, six months since I last rebooted, and it still keeps plugging away in the background. Just downloaded both Puppy Linux and MacPup. Thanks for the links.

  15. E.M.Smith says:

    Dang it.

    Devuan has hung once, which is a big improvement, but this morning crashed twice just running a browser.

    The machine runs fine on XP and other CD or USB sourced Linux; so HW looks sound.

    I think I am back at “replace OS” …

  16. phodges says:

    Very informative discussion, thanks! From your description I will have to give Gentoo a try. I do not have time anymore to install and configure bombproof distros like Slackware or FreeBSD, so i have been trying the more pre-packaged distros like Ubuntu. But they always seem to have problems for me. Gentoo could be the mix of user-defined vs prepackaged that would work.

    I was going to suggest Enlightenment DM, but I see at the bottom there you are using Macpup. E developers are also working the transition to wayland.

    I started using E several years ago and now cannot use any other DM! I also tend to keep releases on a machine as long as they remain stable….which can be a long time for some Linux flavors or XP!

    One distro you did not mention that has been very useful for me is Slax.

    Thanks Chiefio, your blog is always entertaining!

  17. E.M.Smith says:


    I’ve not tried Slax: http://www.slax.org/

    Looks interesting. Any reviews welcome.

    Per “Enlightenment”: Yes, very addictive. I’m finding myself using MacPup preferentially just for the Enlightenment windows manager… (Or rather, was… see below about CD failure…)

    At this point, I’m pondering making my own ‘remix’ based loosely on Puppy, with a MacPup like look and feel, but a load more tools in it and likely on a DVD drive (or PXE boot… again, see below).

    The general problem with the ‘bombproof’ ones like BSD, Slackware, Gentoo… are that you must do a lot of the set up work by hand. If you do it a lot, not much of an issue. If you do it once every 5 to 10 years, things change and I forget and, well, it’s a pain. Also BSD tends to be slower on device driver support and there’s a lot more DIY X configuration that I hate.

    But for a really stable really good server, nothing beats BSD, and for routers / embedded systems – Gentoo has the goods, and for a darned good reliable desktop, Slackware works and is efficient.

    Yet I tend to end up in the “binary blob packages” land when I want something “good enough” up without a lot of work. Though this recent experience is starting to make me have second thoughts about that “not a lot of work” aspect…


    OK, where I’m at now:

    Tossed money at the network problem (once I knew one of the boxes was bad, and the other was maybe DHCP addled due to power issues) and for $60 bought a Netgear N600 aka WNDR3400 model that is configure, up, and working fine. All issues resolved. (It also has the ability to be a file server via a USB port, and provides both encrypted and “guest” networks at the same time if desired, so no longer do I need two boxes for that).

    All was good in the world. For a day.

    Then I decided to upgrade the CentOS 6.3 on my USB drive on the Evo to 6.6 (the last non-systemd release). Everything went Just Fine. Until the reboot.

    Then it came up to a black screen with a cursor, and you can’t do anything. Looking around, I found 3 or 4 sites saying “same thing happened to me” and giving various “fixes”, some of which might apply to me, but some clearly didn’t work (as the expected .rpmnew files were not where it said they would be). The general thrust of it all is that the built in ‘upgrade’ program (who’s name escapes me at the moment since I flushed stack on it…) is flaky and sometimes doesn’t complete (so fix by hand with moving .rpmnew to where they belong) or it will happily upgrade to a video driver for X-windows that doesn’t care about old video drivers anymore and screws you (so you can roll back the upgrade, maybe, in various ways, some of which don’t seem to be available on this release…).

    All of which I was about to try yesterday when Yet Another Reboot (to use CTL-ALT-F2 or F3 or some such Magic Cord to get a terminal server) consistently tossed me into GRUB from the hard disk. What happened to my CD-Rom boot? (That uses Plop to send me to the USB disk)…

    Well, about an hour more showed that the machine simply can not see any CDs anymore. Even booting Windows-XP. It sees the drive at boot time, but can’t read a thing in it. I suspect the laser is dead or it needs a good cleaning.

    What to do, what to do…

    So in a strange turn of events, the Antek / ASUS that had been soooo much of a PITA to get running on Linux has now become my “Daily Driver” Linux box in addition to being the GIStemp / Climate Model devo box.

    The Evo is still usable as a Windows XP box, but without any CD capability. I’ll likely raid one of my other “junkers” for a working CD / DVD drive and try a swap. In a few days. Or weeks. Or months. Or…

    But this has given me an excuse to update the general approach of my “operation” to about the 1990s… I’ll be putting a PXE server on the Raspberry Pi along with the USB disk. That little board has been rock solid the whole time, running non-stop as a DNS / Bittorrent server for many months, and being a general play toy for about 2 years. It, being the original B, is just a single core 700 Mhz and a bit “lite’ for use as a desktop; but does great at generic server tasks.

    I’m presently torn between getting one of the “quad core” cards and just using it as a Daily Driver Linux playground, or heading over to Weird Stuff and getting one of their Linux Already Installed boxes for $60. Amazon has a nice R.Pi M2 B kit with case, power, chip, WiFi, and more for $70 and would be fun. OTOH, a $50 to $80 box with DVD / CD / RW in it would be nice, as right now only the ASUS / Antek has a working CD / DVD drive… Decisions decisions…

    With a PXE server, I can get out of the business of doing individual machine installs onto local hard disks and using that machine CD / DVD drive. Any machine can be used to load up the kernel and file systems onto the PXE server drive. They can come from CD or DVD or the net. I can roll forward / back or sideways with the change of one line in the PXE config file. Likely ought to have done it a decade ago… (It’s nearly trivial to set up). And since the R.Pi is just doing “whatever server” duty, putting PXE on it isn’t much more. So that’s where I’m headed.

    Then, with a file server and PXE server, “any box will do” as they are just a thin client with local SWAP on their disk. If anything goes “bump in the night” and hardware dies, I no longer care. Swap to a different thin client. Change out the disk. Find another place to put the NFS disk. (or the backup of it). Or swap the R.Pi card if it dies.

    We’ll see how long it takes me. It just isn’t a priority right now and I’ve sunk far too much time into fooling around with this junk already. While it may have been fun, it isn’t making “progress to goal” at the rate I’d like and there are other more important things I ought to be doing. I have more than enough “infrastructure” working and with enough depth to handle outages, so really this is more “play” than “productive” at this point…

    Which is all a very long way of saying: “Spouse is happy with working network, I have working data archives and 2 working systems, time to move on.”

    @Paul Hanlon:

    After some searching on the Cubieboard and related, it looks like the quad-core R.Pi is my most likely next buy. The A20 dual core is probably “enough” upgrade, but with the quad-core at about $40 I’m not seeing a reason to get the A20. (Well, I’ll likely get one for the SATA as an NFS file / disk server at some point… ). Generally the Cubie does look good and I’ll get at least one some day, but right now I’m too far behind on too many things for a play board.

    My impression of the Banana Pi was that it was more bang/buck but not as, er, ‘well done’. The wiki looked like Chinglish editing, the board is a knockoff with added stuff, but the ‘buzz’ around it isn’t very strong. Somehow I’m just not feeling the love… And the Orange Pi looks nice too, but not as cheap as the R.Pi for yet another strange knock off. (Hummingboard seems expensive).

    So at the end of it all, as I’m only needing simple stuff, I’m thinking R.Pi for now (as I’m already familiar with it and have ‘supporting stuff’ like power and chips and OS saves and…) and a Cubie in the “next buy maybe” pile…

    Or I might just get more frustrated tomorrow when Yet Another PC Box has some parts go FUNT! and place an order for my “all roll your own” kit… In which case it will be a R.Pi M2 w/ case and kit, and a Cubieboard with case and kit. One of each.

  18. p.g.sharrow says:

    The ability to “Train” a R.Pi to serve in many functions would indicate it might be useful in a node type system. The very large base of creative users developing software for it also is a valuable resource.
    The cost of replacing stand alone dedicated devices with a SD “trained” Pi might simplify this constant battle with enforced obsolescent caused cascade replacements along with back door infested up grades. pg

  19. E.M.Smith says:


    That is basically the historic Unix Way.

    At one time about 20? years ago I ran a small (200 person) company IT shop that way. A rack of 2 or 4 U servers (about one inch high per U of rack) and the mantra “One server One service”.

    Any box or part could die and we were one box swap away from back in service (often just swap disk to new chasis or swap network file mount). Typically only one service ever impacted at any time and no need to annouce down time for other services if a reboot needed: since the one service was already down, just reboot pronto. IFF the spare box was already taken, there was enough capacity in spare on some boxes to double up for a while (though we never had to do that). And generic PCs with Unix was still cheaper.

    Now you can do that on a stack of R.Pis for about $200.

    In a large pro shop, now, folks do almost the same thing via virtual machine dispatches, and that could be done small scale on a DIY VM farm of R.Pis, “white box” PCs, or any size “cloud server farm” you need.

    FWIW, I’m headed toward that structure now. Just way too slowly as “life happens”. One R.Pi is doing DNS, time server, and NFS files already. Another will do PXE boot for the rest. Then just adding the desired boxes as needed. Planned, but delayed about a month by the recent coping with hardware failures, has been to put a VM Farm up on my largest box and lay out the structure.

    The last 24 hours has been moving data archives to clear the disk for the PXE server and VM images. Slow interfaces and big disk… but it is almost done. Once data is separated from hardware and any box can boot a needed OS or service, I’ll see what HW works best and make a buy. With bias to the SBC Single Board Computer class.

    I have a half dozen old crap boxes that I’ll prototype with, but the Pi M2 is faster than them, and they are desktop sized with large fans, so a nice stack of Pis is the end goal (less power and noise, more performance, uniform parts for swapping, and newer so less failure prone.)

    Now you know :-)

  20. p.g.sharrow says:

    @EMSmith; It appears you are heading in the same direction I am herding my grandson toward for our use. ;-)
    The press of “other things” that just, must be done today seems to always get in the way. Your efforts and comments are very helpful in saving me from blind alleys. As they say, “A Win Win” for everyone. pg

  21. Steve C says:

    @EM: Don’t know whether someone of your experience will find this as interesting as someone of my experience did, but the Register have lately done an article on Five lightweight Linux desktop worlds for extreme open-sourcers. You never know, I often find unexpected nuggets in articles which are obviously below my level of expertise! ;-)

    For a bit of light relief over a cup of coffee, though, read their We tried using Windows 10 for real work …. The writer is as kind as he can be.

    @p.g. – I’ve been thinking a Pi would make a perfect “media machine”, and note that there are already several of ’em out there. No fan noise at all, a couple of TB storage on the USB and a DVD burner in the box with it … and at last I could unplug all those big whirring desktop boxes from the audio gear forever. At the size & price of a Pi, it almost seems pointless going for “all-purpose” hardware any more.

    I almost feel I’m cheating here, with my Mint 17.1 running very sweetly on 4 cores @ 3.6G and 8 GB of memory … so far, a perfect transition from my well-used old XP machine for day-to-day use. The shack/workshop machine will remain XP indefinitely though, mainly because there are some seriously slick amateur radio programs out there for Windows; my ol’ 2.4G Celeron has been seen reliably decoding signals 33dB below the noise as it is now, so it’s definitely not broke ‘n’ don’t need fixing! (And, hell, I’ve paid for some of those programs … ;-)

  22. E.M.Smith says:

    @Steve C:

    Never forget the Law Of Mutual Superiority. Anything I do, you can improve…. and vice versa.

    Interesting article. Some I was familiar with, some newer than my last update. Like LxQT.

    I still think that a 128 MB base and 256 MB windowed system ought to be easily available but for developers being on 16 GB 64 core superboxes… I’d really like to know why I could run Linux with browser in 64 MB on 32 bit hw a decade back and now the same hw with Linux and browser needs 640 MB before cacheing the first page…

    I suspect the move to object oriented superfat code with fat libraries all compiled with huge buffer size settings. Linux needs an efficiency review and diet… That it runs in far less memory on routers and SBC boards is telling…

  23. E.M.Smith says:

    Why I love Linux / Unix despite the difficulties of installing some apps some times…

    I was vacating a 111 GB USB disk for use with PXE and NFS. I’d shrunk the FAT32 partition to 20 GB and the rest was EXT3, swap, and EXT4. Then I noticed that only 16 GB was available under the FAT partition…

    Inspection under Windows showed no files nor directories I could dump. Yet “defragment” happily listed the names of 3000+ files as it relocated them and squashed the 3 GB into one end of the partition…

    Once it was done, remounting on Linux showed a $RECYCLE directory that simply could not be seen otherwise. Not by XP. Not as a ‘garbage can’. Nothing. With Linux, I wandered into the directory (likely a ‘left over’ from some other Windoze OS long long ago) and found the files. In a directory named for an account I had at a client site about 2006 and with 2006 dates.

    Likely tossed out on exit of the contract, but not wanting to spend a few ours rooting through it right now and wanting to ‘get on with it’ on reuse of the disk… a few keystrokes later I have that entire directory being zipped, compressed, and moved to an SD card on the Raspberry Pi that’s my NFS server du jour…

    tar cf – directorystuff | gzip > /NFSMOUNT/SDCARD/

    Yeah, that easy. Oh, and a “rm -rf directorystuff’ once shown to have completed correctly…

    So most likely I tossed this in the trash in 2006, and the disk has not been used on that particular MS OS since, and the other versions have a different “trashcan” convention. I’ll check the contents “someday” and likely recycle it again…

    Linux makes easy what Microsoft makes impossible… (and makes hard what ought to be easy… but that’s another topic…)

  24. Pingback: The Working Draft Compute Build | Musings from the Chiefio

  25. E.M.Smith says:

    I may have “got clue” on what’s caused the XWindows hangs.

    I can log in remotely, kill the Xwindow process (that is in runaway and sucking up 100% of available CPU) and then a new window spawns and everything is “fine”.

    I’ve been “up” for about 24 hours now.without a glitch.. Yet video seems slower than expected. Things like dragging a terminal window has “lag” while it catches up with the cursor. As though the CPU were doing all the video work, not the GPU ( and the EVO has a decent one of each).

    So I went looking in the Xorg log files. Here’s what I found (I’ve bolded the bits that I think matter):

    [  2357.417] [mi] EQ overflowing.  Additional events will be discarded until existing events are processed.
    [  2357.417] 
    [  2357.417] Backtrace:
    [  2357.436] 0: /usr/bin/X (xorg_backtrace+0x49) [0xb76fc309]
    [  2357.436] 1: /usr/bin/X (mieqEnqueue+0x22b) [0xb76dab9b]
    [  2357.436] 2: /usr/bin/X (0xb757e000+0x519e5) [0xb75cf9e5]
    [  2357.436] 3: /usr/bin/X (xf86PostMotionEventM+0x24b) [0xb7609bdb]
    [  2357.436] 4: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6ea4000+0x35ad) [0xb6ea75ad]
    [  2357.436] 5: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6ea4000+0x4a2c) [0xb6ea8a2c]
    [  2357.436] 6: /usr/bin/X (0xb757e000+0x7b271) [0xb75f9271]
    [  2357.436] 7: /usr/bin/X (0xb757e000+0xa0fba) [0xb761efba]
    [  2357.436] 8: linux-gate.so.1 (__kernel_sigreturn+0x0) [0xb755c400]
    [  2357.436] 9: linux-gate.so.1 (__kernel_vsyscall+0x10) [0xb755c424]
    [  2357.437] 10: /lib/i386-linux-gnu/i686/cmov/libc.so.6 (ioctl+0x19) [0xb7259219]
    [  2357.437] 11: /usr/lib/i386-linux-gnu/libdrm.so.2 (drmIoctl+0x2e) [0xb6e8ebce]
    [  2357.437] 12: /usr/lib/i386-linux-gnu/libdrm.so.2 (drmCommandNone+0x25) [0xb6e919a5]
    [  2357.437] 13: /usr/lib/xorg/modules/drivers/intel_drv.so (0xb6d93000+0x136b1) [0xb6da66b1]
    [  2357.437] 14: /usr/lib/xorg/modules/drivers/intel_drv.so (0xb6d93000+0xf762) [0xb6da2762]
    [  2357.437] 15: /usr/bin/X (BlockHandler+0x56) [0xb75be936]
    [  2357.437] 16: /usr/bin/X (WaitForSomething+0xfa) [0xb76f94ca]
    [  2357.437] 17: /usr/bin/X (0xb757e000+0x3c6be) [0xb75ba6be]
    [  2357.437] 18: /usr/bin/X (0xb757e000+0x2a205) [0xb75a8205]
    [  2357.437] 19: /lib/i386-linux-gnu/i686/cmov/libc.so.6 (__libc_start_main+0xf3) [0xb7191a63]
    [  2357.437] 20: /usr/bin/X (0xb757e000+0x2a559) [0xb75a8559]
    [  2357.437] 
    [  2357.437] [mi] These backtraces from mieqEnqueue may point to a culprit higher up the stack.
    [  2357.437] [mi] mieq is *NOT* the cause.  It is a victim.
    [  2357.905] [mi] EQ overflow continuing.  100 events have been dropped.
    [  2357.905] 
    [  2357.905] Backtrace:
    [  2357.905] 0: /usr/bin/X (xorg_backtrace+0x49) [0xb76fc309]
    [  2357.905] 1: /usr/bin/X (mieqEnqueue+0xfb) [0xb76daa6b]
    [  2357.905] 2: /usr/bin/X (0xb757e000+0x519e5) [0xb75cf9e5]
    [  2357.905] 3: /usr/bin/X (xf86PostMotionEventM+0x24b) [0xb7609bdb]
    [  2357.905] 4: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6ea4000+0x35ad) [0xb6ea75ad]
    [  2357.905] 5: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6ea4000+0x4a2c) [0xb6ea8a2c]
    [  2357.905] 6: /usr/bin/X (0xb757e000+0x7b271) [0xb75f9271]
    [  2357.906] 7: /usr/bin/X (0xb757e000+0xa0fba) [0xb761efba]
    [  2357.906] 8: linux-gate.so.1 (__kernel_sigreturn+0x0) [0xb755c400]
    [  2357.906] 9: linux-gate.so.1 (__kernel_vsyscall+0x10) [0xb755c424]
    [  2357.906] 10: /lib/i386-linux-gnu/i686/cmov/libc.so.6 (ioctl+0x19) [0xb7259219]
    [  2357.906] 11: /usr/lib/i386-linux-gnu/libdrm.so.2 (drmIoctl+0x2e) [0xb6e8ebce]
    [  2357.906] 12: /usr/lib/i386-linux-gnu/libdrm.so.2 (drmCommandNone+0x25) [0xb6e919a5]
    [  2357.906] 13: /usr/lib/xorg/modules/drivers/intel_drv.so (0xb6d93000+0x136b1) [0xb6da66b1]
    [  2357.906] 14: /usr/lib/xorg/modules/drivers/intel_drv.so (0xb6d93000+0xf762) [0xb6da2762]
    [  2357.906] 15: /usr/bin/X (BlockHandler+0x56) [0xb75be936]
    [  2357.906] 16: /usr/bin/X (WaitForSomething+0xfa) [0xb76f94ca]
    [  2357.906] 17: /usr/bin/X (0xb757e000+0x3c6be) [0xb75ba6be]
    [  2357.906] 18: /usr/bin/X (0xb757e000+0x2a205) [0xb75a8205]
    [  2357.906] 19: /lib/i386-linux-gnu/i686/cmov/libc.so.6 (__libc_start_main+0xf3) [0xb7191a63]
    [  2357.906] 20: /usr/bin/X (0xb757e000+0x2a559) [0xb75a8559]
    [  2357.906] 
    [  2358.360] [mi] Increasing EQ size to 512 to prevent dropped events.
    [  2358.362] [mi] EQ processing has resumed after 167 dropped events.
    [  2358.362] [mi] This may be caused my a misbehaving driver monopolizing the server's resources.
    [  2358.620] (EE) intel(0): Detected a hung GPU, disabling acceleration. 
    [  2358.620] (EE) intel(0): When reporting this, please include i915_error_state from debugfs and the full dmesg.
    [  2427.136] [mi] Increasing EQ size to 1024 to prevent dropped events.

    I think it is a video driver issue causing the GPU to be disabled, and hanging that instance of the Xwindows session, but then new instances work, just slow from no GPU. Especially obvious when driving the box hard (like I am now with 100% CPU on a TB of copy | tar | pipe | process | unpack on other disk).

    Now I just need to find a better driver. Maybe… (Or constantly log in from another machine to kill the first ‘hang’ and live with slow video… kludge / yuck…)

Comments are closed.