John's Tech Notes

Bulk copy

2011-05-26T14:33:00.000-07:00

Copying one disk to many disks:


dd if=/dev/sda bs=1M | tee >(dd of=/dev/sdb bs=1M) >(dd of=/dev/sdc bs=1M) > /dev/null

Duplicating the bs argument is important for throughput.

Installing iWork 09 over iWork 09 Trial

2010-12-07T11:32:00.001-08:00

I wanted to play the pdf (yay beamer)->keynote game to get a nice presentation display at the low low price of a student discounted iWork. I had installed and updated the trial earlier, and when I installed the version from the DVD it would ask me for a serial number, which does not exist on any of the packaging. The instructions below the serial number prompt said "Install from DVD and follow the on-screen instructions". This clearly hadn't worked. Nor did just deleting /Applications/iWork 09. You actually need to also delete /Library/Application Support/iWork 09. (I reinstalled at this point, I don't know if ordering is important).

Disable persistent device naming

2010-07-29T16:28:00.000-07:00

Persistent device naming is all fun and games until you start moving a boot drive between machines frequently.

/etc/udev/rules.d/00-no-persistent-net.rules

# Disable net renaming
SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", NAME="$name"

Then go and empty out the 70-persistent-net.rules

adapted from here (Modern udev doesn't like NAME="%k")

multicast dns keeps getting installed on my machines...

2010-05-07T03:19:00.000-07:00

/etc/apt/preferences.d/avahi-daemon-is-annoying:


Package: avahi-daemon
Pin: release a=blacklist-this-thing
Pin-Priority: 1001

Down with flash cookies

2010-03-08T13:11:00.000-08:00

Because clearing your browser history isn't enough.

crontab entry:


0 6 * * * find ~/.macromedia -name '*.sol' -delete

Wiping out host keys for all likely instances.

2010-01-29T14:08:00.000-08:00

For when an ssh host key changes and you don't want to run ssh-keygen -R until it stops complaining. I know it works on zsh, I believe it should be bash compatible.


ssh-wipe ()
{
    addr=`host $1 | awk '{print $4}'`
    names=`host $addr | awk '{print substr($5, 0, -1)}'`

    for x in $1 $addr `host $addr | awk '{print $5}'`
    do
        echo ${x%%.}
        ssh-keygen -R ${x%%.}
    done
}

Xorg DRI with a Radeon HD 4650 on Debian

2010-01-13T13:37:00.000-08:00

X was being slow on this fancy desktop, and I noticed a "Direct Rendering Disabled" message in the log. Dmesg indicated a failure to load the firmware "platform r600_cp.0: firmware: requesting radeon/RV730_cp.bin". A simple


apt-get install linux-firmware

corrected that problem, however, trying to run glxgears results in


drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.

I noticed a few MTRR errors, and after adding enable_mtrr_cleanup to GRUB_CMDLIN_LINUX_DEFAULT in /etc/defaults/grub, rebooting, fixing the fact that a recent grub2 upgrade (boo grub2) had hosed itself (you should run 'grub-setup "(hd0)"' before rebooting if you haven't for a while and don't like getting a rescue cd out).

This fixed the MTRR errors, but the drmRadeonCmdBuffer issue remains, and is related to Mesa and the drm driver being out of sync (https://bugs.launchpad.net/ubuntu/+source/mesa/+bug/419434).

For the moment, X is faster, with no GL.

[[Edit, works now w/drivers from testing (squeeze)]]

One-liner to extract vmlinuz to vmlinux

2009-12-13T16:41:00.000-08:00

The hex string is the gzip offset. I've seen variations of this around, many using od. (The od approach doesn't work when the magic string spans output lines).


x=/boot/vmlinuz-2.6.26-2-amd64 dd if=$x skip=`grep -a -b -o -m 1 -e $'\x1f\x8b\x08\x00 $x | cut -d: -f 1` bs=1 | zcat > /tmp/vmlinux

(edited to fix \xf1 -> \x1f).

pxe booting system rescue cd with a bnx2

2009-10-22T18:28:00.000-07:00

The system rescue cd project has made progress on becoming more pxe bootable in 1.3.1 than it was in 1.1.2 or so. It worked like a charm on a machine of mine with an intel NIC. However, many of our machines have the Broadcom NetXtreme II cards (bnx2) which require firmware to load. Currently, the sysrcd firmware lives in the sysrcd.dat squashfs that should be loaded over tftp/http.

To fix this, I wedged the firmware in question onto the initram.igz filesystem and tweaked the /sbin/init script to fire up mdev (the udev replacement for busybox that is not told to load until after module probing and hotplug isn't initialized until much later in the boot process) so that the firmware can be loaded.

To unpack the initrd


mkdir initrd
cd initrd
gzip -dc path/to/initram.igz | cpio -id

Move the firmware over


mkdir lib/firmware
cp path/to/mounted/sysrcd.dat/from/cd/or/otherwise/lib/firmware/bnx2 lib/firmware
cd ..

in sbin/init, I added


mount_sysfs
setup_hotplug
start_dev_mgr

just above the line


good_msg "Loading the disk and network drivers required to boot..."

Repack:


find ./ | cpio -H newc -o > /tmp/new-initrd.cpio
gzip /tmp/new-initrd.cpio
mv new-initrd.cpio.gz initram-bnx2.igz

My pxelinux.cfg entry looks like this:


label netboot64
  kernel sysrcd/isolinux/rescue64
  append initrd=sysrcd/isolinux/initram-bnx2.igz netboot=tftp://192.168.1.10/pxelinux/sysrcd/sysrcd.dat console=tty0 console=ttyS0,9600n1 udev

Fetching sysrcd.dat over tftp isn't very fast, and it may be worth looking into the http method.

vimoutliner on debian

2009-05-20T15:21:00.000-07:00

After looking for a clever way to keep notes in a marginally platform independent way, I found vimoutliner which adds some interesting features for folding and making todo lists and so on. I noticed that there was a debian package and I thought "cool I can easily check it out". It turned out to be a bit more complicated than I expected and I replicate my steps here for posterity:


apt-get install vim-vimoutliner vim-addon-manager
vim-addons install vimoutliner

If it doesn't seem to work correctly when you open a *.otl file, ensure than your vimrc has


filetype plugin on

The other note, which is in the /usr/share/doc/vim-vimoutliner/README.Debian is that the ,, commands are in fact \ commands.

See :help vo for instructions.

Happy outlining.

Debian on 730i Nvidia Board w/HDMI

2009-03-17T02:00:00.000-07:00

Put together a machine to run my TV at 1080p. The board is a evga 113-yw-e115-tr. Installed Debian 5.0 (amd64) from a usb key. The biggest hiccup was getting sound output via the HDMI port.
The major steps were:

Install alsa-sources from unstable (to version 1.0.19) (apt-get ...; m-a a-i alsa)

Install libasound2 from unstable

Install nvidia-kernel-source from unstable (m-a a-i nvidia)

Edit: When trying to repeat this process, I discovered that hdmi output would not load without an up-to-date nvidia module that had been not only loaded, but initialized by the X driver.

This gets the HDMI device into Alsa's view:


proost:~$ aplay -l
**** List of PLAYBACK Hardware Devices ****
card 0: NVidia [HDA NVidia], device 0: ALC888 Analog [ALC888 Analog]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 0: NVidia [HDA NVidia], device 1: ALC888 Digital [ALC888 Digital]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 0: NVidia [HDA NVidia], device 3: NVIDIA HDMI [NVIDIA HDMI]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

It took a while to discover that, and then I found that I could playback wav files via aplay -Dplughw:0,3 .... After setting up a basic asound.conf I discovered that not all players output 48kHz. I found some hints that alsa can ensure a specific output rate and that worked well. Then I discovered that dmix (multiple streams) wasn't working. The net result (/etc/asound.conf) is:


pcm.!default {
    type plug
    slave.pcm "dmixer"
}
pcm.dsp0 {
    type plug
    slave.pcm "dmixer"
}
pcm.dmixer
{
    type dmix
    ipc_key 1234
    slave {
        pcm "digital-hw"
        rate 48000
    }
}
pcm.digital-hw {
        type hw
        card 0
        device 3
}
ctl.digital-hw {
        type hw
        card 0
}
ctl.mixer0 {
    type hw
    card 0
}

With the slave line alsa ensures that output is at 48kHz.

The other hiccup was getting the picture to fill the screen. Standard TV signals crop the edges a fair bit. This didn't show up on the vga cable, but it did for HDMI. After searching the Internet far and wide, it turned out that I can set the picture size to "Just Scan" and it works beautifully. The only other issue was setting the default font size to be less huge (Option "DPI" "120 x 120") in xorg.conf.

Gnuplot Umlauts when input isn't in iso_8859_1

2008-11-17T14:01:00.000-08:00


# Blah
set output 'thingie'
set encoding iso_8859_1
plot 'datafile' title 'Na\357ve' # gets i umlaut, a is 345 or so.

Port bouncing with iptables.

2008-11-13T16:56:00.000-08:00

This just turned out to be useful to someone other than me, and I forget how I did it periodically, so:

The situation is that I am on a network where there are machines with globally routable addresses and some with private addresses. Suppose I want to get traffic from outside on the internet, but I don't want to add port-forwards to the NAT machine (Don't want to make the sysadmin manage it and I don't want to do mysterious things to those machines). Since I have a machine with a globally routable address, I can bounce the packets to the internal machine via:


DPORT=XXX
PRIVADDR=XXX.YYY....
# To
iptables -t nat -A PREROUTING -i eth0 -p tcp -d $EXTIP --dport $DPORT -j DNAT --to-destination $PRIVADDR
#from
iptables -t nat -A POSTROUTING -p tcp -d $PRIVADDR -s ! $EXTIP --dport $DPORT -j SNAT --to-source $EXTIP

If you need to target different ports, I think you can do something like


iptables -t nat -A PREROUTING -i eth0 -p tcp -d $EXTIP --dport $DPORT -j DNAT --to-destination $PRIVADDR:$OTHERPORT
iptables -t nat -A POSTROUTING -p tcp -d $PRIVADDR -s ! $EXTIP --dport $OTHERPORT -j SNAT --to-source $EXTIP

But I haven't tested it (dig around in the iptables man page under DNAT).

BIOS Flashing a Lenovo X60s from FreeBSD

2008-08-28T08:06:00.000-07:00

I used a 2GB USB Flash Stick in this process. Something smaller could work, something larger would make certain steps take longer.

The general idea is to grab the bootable cdrom and wedge the important pieces on to a flash drive (the laptop doesn't actually have cdrom).

The first challenge is making the Flash Stick bootable. The only Windows Machines I had access to through work are Vista, and I couldn't figure out how to make the little guy work with that. You can follow the general ideas from someplace like this which involves getting windows to make it bootable, or you can do the hack-around way.

This may not be the most efficient way, but it involved the least cognitive load:

Install Qemu.

Put freedos in a qemu image

Insert USB Device, figure out name (/dev/da0 in my case)

Make an empty image that is exactly the same size (dd if=/dev/da0 of=flashthing.hd)

Boot the qemu image with -hdc flashthing.hd

fdisk it (xfdisk, I don't remember, probably reboot, should give you d:)

format \s d:

Get the image onto the flash device (dd if=flashthing.hd of=/dev/da0 ... this took a long time since it wasn't usb2.0ing).

Mount the iso and the flash drive then copy:
mdconfig -a -f 7bu???.iso
mount -tcd9660 /dev/md0 /mnt/cdrom
mount -tmsdosfs /dev/da0s1 /mnt/flash
cp -r /mnt/cdrom /mnt/flash

It's important to get the memory setup correct (I got various invalid opcodes, first one, then after I installed HIMEM, a whole lot of them). I copied some pieces from those in the el-torito boot image (you can rip it out with this nice tool). /mnt/flash/FDCONFIG.SYS ended up as:
SET lang=EN
LASTDRIVE=Z
BUFFERS=20
FILES=40
DOS=HIGH,UMB
DOSDATA=UMB
set dircmd=/ogn /4
RED HIMEM is important
DEVICE=C:\FDOS\BIN\HIMEM.EXE
SHELLHIGH=C:\FDOS\bin\command.com C:\FDOS\bin /E:1024 /P=C:\autoexec.bat
REM STACKS is important
STACKS=9,256

After that, reboot onto the flash drive (set it as the first boot device in the bios -- just setting it as the temporary boot device won't work)

cd cdrom and command.com should get you to the point of actually flashing the thing.

Making and viewing pdfs from windows gvim.

2008-08-11T16:59:00.000-07:00

pdfLaTeX is within cygwin, adobe reader is off in windows land, the solution:


map = :!c:\cygwin\bin\bash --login -c "pushd %:p:h:gs?\\?\\\\\\? && pdflatex %" && "C:\Program Files (x86)\Adobe\Reader 8.0\Reader\AcroRd32.exe" %:p:r.pdf

(Windows should stop using the escape character for the directory separator).

Edit
A smarter person would have installed miktex for windows and discovered that "start" launches the program associated with the filetype which reduces this to:


map = :!pdflatex % && start %:p:r.pdf

Vista Caps-Lock to Control Re-Mapping

2008-06-25T10:51:00.000-07:00

Abbreviated from here: Windows Scancode Mapping.

This is the caps/ctrl swap without the ctrl->caps side.

caps2ctl.reg:


Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Keyboard Layout]
"Scancode Map"=hex:00,00,00,00,00,00,00,00,20,00,00,00,1d,00,3a,00,00,00,00,00

IPV6 OpenWrt Kamikaze.

2008-06-15T18:23:00.000-07:00

This is mostly an adaptation of various ipv6 install bits around.
Kamikaze is a little different, and the documentation isn't entirely
complete -- so I thought I would reproduce what I did in here. It isn't
perfect, but it is working for my /64 delegations of the /48 I have
routed to my fat-pipe machine.

Names changed to protect the innocent.


ipkg install kmod-ipv6
ipkg install radvd
ipkg install ip
ipkg install kmod-ip6tables
ipkg install ip6tables

/etc/hotplug.d/iface/10-ipv6:


. /etc/functions.sh
NAME=ipv6
COMMAND=/usr/sbin/ip
IPV6PREFIX="the:/48:subnet:here:2"
REMOTE="fat.pip.es.heh"

[ "$ACTION" = "ifup" -a "$INTERFACE" = "wan" ] && {
    [ -x $COMMAND ] && {
        VLAN=$(nvram get ${INTERFACE}_ifname)
            IFNAME=eth0.${VLAN#vlan}
        IPV4=$(ip -4 addr show $IFNAME | grep inet | cut -f6 -d' ' | cut 
               -f1 -d'/')
            ip tunnel add fatty-ipv6 mode sit ttl 255 remote $REMOTE local $I
            PV4
            ip link set dev fatty-ipv6 up
            ip -6 addr add ${IPV6PREFIX}::2/48 dev fatty-ipv6
            ip -6 route add ::/0 dev fatty-ipv6 metric 1
    }
}
[ "$ACTION" = "ifdown" -a "$INTERFACE" = "wan" ] && {
    [ -x $COMMAND ] && {
        VLAN=$(nvram get ${INTERFACE}_ifname)
            IFNAME=eth0.${VLAN#vlan}
        IPV4=$(ip -4 addr show $IFNAME | grep inet | cut -f6 -d' ' | cut 
               -f1 -d'/')
            ip -6 route flush dev fatty-ipv6
            ip tunnel del fatty-ipv6
    }
}

/etc/config/network:


config interface lan
option type  bridge
option ifname "eth0.0"
option proto static
option ipaddr 192.168.1.1
option ip6addr  the:/48:subnet:here:2::2/64

/etc/config/firewall:


accept:proto=41

/etc/firewall.user:


iptables -t nat -D POSTROUTING -o wan -j MASQUERADE
iptables -t nat -A POSTROUTING --protocol ! 41 -o wan -j MASQUERADE

/etc/radvd.conf:


interface br-lan
{
    AdvSendAdvert on;

#
# Disable Mobile IPv6 support
#
    AdvHomeAgentFlag off;

#
# example of a standard prefix
#
    prefix the:/48:subnet:here:2::/64
    {
        AdvOnLink on;
        AdvAutonomous on;
        AdvRouterAddr off;
    };

};

/etc/init.d/radvd enable

Reboot, or whatever. {/etc/init.d/(network|firewall|radvd) restart}

FreeBSD 7.0 on Thinkpad x60s

2008-06-08T17:44:00.000-07:00

Installation (Borrowed/adapted from here).

Grab a boot-only nfs image from here.

Copy the cd contents into the filesystem:

mount -o loop /path/to/7.0-RELEASE-i386-bootonly.iso /mnt
mkdir /freebsd
cp -rP /mnt/boot/ /freebsd
umount /mnt

Edit /freebsd/boot/loader.conf to add:
vfs.root.mountfrom="ufs:/dev/md0c"
to use ramdisk root instead of nfs.

Edit /etc/dhcpd.conf to add:

host x60 {
   hardware ethernet xx:xx:xx:xx:xx:xx;
   fixed-address 10.0.0.xx;
   next-server 10.0.0.2;
   filename "pxeboot";
   option root-path "10.0.0.2:/freebsd";
 }

Copy the pxeboot image to your tftp root dir:
cp /freebsd/boot/pxeboot /var/lib/tftpboot

NOTE: I had trouble with the 7.0 pxeboot ("Can't work out which disk we are booting from"). Crawling the web indicated that it was a 7.0 pxeboot problem and pulling the boot/pxeboot binary from a 6.3 iso image worked.
I already had a tftp server installed (I use the tftpd-hpa server).

Install an nfs server (temporarily if necessary (apt-get install nfs-kernel-server)).

/freebsd 10.0.0.xx(ro,subtree_check)

At this point, installation worked for me.

Getting the hardware working

Wireless

Get the drivers loaded

# See man if_wpi if you want to know what you are agreeing to.
legal.intel_wpi.license_ack=1
if_wpi_load="YES"
wpi_load="YES"
wpi_amrr_load="YES"
wpifw_load="YES"
wlan_xauth_load="YES"

WPA:
/etc/wpa_supplicant.conf
```
network={
   ssid="yourssidhere"
   psk="Your clever passphrase with lots of character variation
}
```
/etc/rc.conf
ifconfig_wpi0="WPA DHCP"
To start it, you can run /etc/rc.d/netif start

Sound

snd_hda_load="YES"

X.org

Xorg -configure
cp /root/xorg.conf.new /etc/X11/xorg.conf

Finger Print Reader

To be explored

Miscellany

/etc/make.conf
CPUTYPE?=core

Debian Firefox Flash Fonts Missing/Hidden

2008-05-04T02:48:00.001-07:00

It took me a while to sort this out. You probably want


apt-get install msttcorefonts

Now I can read the score in that word game that I shouldn't be playing instead of doing work.

autotools and cell

2008-03-15T18:54:00.000-07:00

I came up with the following to get cross-compiling for the cell on sdk 2.0 to work within autotools. I make no claims to the robustness. Furthermore, you can't actually get configure to run on an x86, because it tries to execute code it compiles, thus, you have to run configure on the cell machine. This is still useful, because the x86 machines are significantly faster at compilation.

In configure.in:

AC_CHECK_PROGS(SC,spu-gcc, "no")
if test  "$SC" != "no"; then
    SCPPFLAGS="$CPPFLAGS"
    SCFLAGS="$SCFLAGS -I$CELL_TOP/sysroot/usr/spu/include -W -Wall -Winline -Wno-
main -O3 "
    LIBS="$LIBS -lspe2 -lspe"
else
    echo
    echo "ERROR Can't find spu-gcc"
    exit 1
fi

# /usr/include is the wrong architecture for the spu.
SCFLAGS=`echo $SCFLAGS |sed -e 's/-I\/usr\/include //g'`
SCPPFLAGS=`echo $SCPPFLAGS | sed -e 's/-I\/usr\/include //g'`

AC_SUBST(SC)
AC_SUBST(SCFLAGS)
AC_SUBST(SCPPFLAGS)

AC_CHECK_PROGS(PPUEMBED, ppu-embedspu, "no")
if test "$PPUEMBED" != "no"; then
    PPUEMBEDFLAGS="$PPUEMBEFLAGS -m32"
else
    echo "ERRO Can't finde ppu-embedspu"
    exit 1
fi
AC_SUBST(PPUEMBED)
AC_SUBST(PPUEMBEDFLAGS)

In your spu directory's Makefile.am:

noinst_LIBRARIES = libcellblaht.a

SUFFIXES = -embed.c -embed.o

.c-embed.o:
        $(SC) $(SCFLAGS) $(SCPPFLAGS) $(AM_SCFLAGS) $(INCLUDES) -c $<
        $(SC) -o $* $*.o
        $(PPUEMBED) $(PPUEMBEDFLAGS) $* $* $@

libcellblah_a_SOURCES = cell_blah-embed.c

INCLUDES = @INCLUDES@

The actual source file name should be cell_blah.c. This doesn't really support multiple .c files being compiled into a single embedded program. I got tired of futzing with automake.

Favorite new trick.

2008-02-26T23:42:00.001-08:00

Dump all the macros defined by the compiler:


echo "" | gcc -E -dM -

xterm, gcc, and colorization

2008-02-22T11:56:00.000-08:00

I've found gcc giving strange output when colorizing it's errors on remote machines. Usually there is a carat-a followed by some garbage. It makes it very hard to read the output when it isn't there. Turns out it is a symptom of xterm not supporting unicode. The simple solution is:


setenv LANG en_US


export LANG=en_US

depending on the shell.

ns2 configure lame

2008-01-11T11:13:00.000-08:00

ns2 ./configure was failing on tcl in Debian. Having installed tcl8.4-dev and tk8.4-dev:


ln -s /usr/share/tcltk/tcl8.4/init.tcl /usr/lib/tcl8.4
ln -s /usr/share/tcltk/tk8.4/tk.tcl /usr/lib/tk8.4

Avahi broken for evince while building gnucash.

2007-12-06T01:09:00.000-08:00

Still on the path to getting gnucash built, I couldn't get avahi to build as a dependency to evince.

I tweaked /opt/local/var/macports/sources/rsync.macports.org/release/ports/gnome/evince/Portfile to remove the avahi lib depend and add --disable-avahi to the config line. You can also tack add to the Portfile:


variant without_avahi {
    depends_lib-delete port:avahi
    configure.args-delete --disable-avahi
}

and install evince with +without_avahi

There has been so much godawful bloat installing this thing. without_docs seems to take care of some of it.

Other requisite tricks:


pushd /opt/local/bin
ln -s guile16-config guile-config
ln -s guile16 guile
popd
pushd /opt/local/include
ln -s guile16 guile
ln -s libguile16.h libguile.h
ln -s libguile16 libguile
popd
pushd /opt/local/lib
ln -s libguile16.dylib libguile.dylib
popd

guile16 is enabled in the port because "Newer versions of guile don't work with macports". At least at this point, it doesn't compile in a straightforward manner.

My final invocation was:


port install gnucash +without_docs +guile16

osx macports libGL ld cycle.

2007-12-05T22:52:00.000-08:00

I ran into a linker bug trying to get gnucash installed when building the aqbanking package. To fix this, I modified the aqbanking Portfile (/opt/local/var/macports/sources/rsync.macports.org/release/ports/devel/aqbanking/Portfile) to include:

configure.ldflags-append "-Wl,-dylib_file,/System/Library/Frameworks/OpenGL.framework/Versions/A/Libraries/libGL.dylib:/System/Library/Frameworks/OpenGL.framework/Versions/A/Libraries/libGL.dylib"