Archive for November, 2006

Cisco 7960 buggy SIP firmware rev. 7.5

Sunday, November 5th, 2006

I’ve been getting intermittent complaints recently that our Cisco 7960 phones are refusing to work and displaying an ‘XML Parse Error’. In inability to transfer calls seems also to be part of the equation. Well, I’ve never had the problem myself and didn’t see anything necessarily wrong with the phones or the setup so I’ve been ignoring it for the most part since power cycling the phone seems to resolve the problem for a while. Recently however we have changed our incoming caller pattern to go through a receptionist in all cases, and she’s been having the problem very consistently. Well, it turns out the problem is a deficiency in Cisco’s SIP implementation. Lots more details are available at , but apparently downgrading to revision 7.4 solves the problem. I have no idea if the bug persists in newer (8.x) versions or not.

Fedora Core 5 ATI driver with theater_out tv support

Sunday, November 5th, 2006

I have a radeon 9200 video card in my HTPC, since when I bought it all the HTPC enthusiasts over at seemed to think ATI was better for tv output. I also wanted the possibility of using a component adapter cable (which I never bought). For years I’ve been happily using the ATI proprietary FGLRX drivers under linux to drive my Toshiba 27″ TV with an svideo cable. Alas, all good things come to an end. Xorg 7.0 came along with my Fedora Core 5 and MythTV 0.20 upgrade, and with it came FGLRX breakage. Or to be precise, along with it came an FGLRX upgrade.

After entirely too much digging, I discovered that versions of FGLRX greater than 0.20 have broken XV scaling. No, it doesn’t appear to be documented anywhere at ATI, but what happens is the video only uses part of the screen and appears to be clipped out of the top left corner of the frame. I fought with it for an hour or two thinking I was just drawing offscreen or something, but eventually realized it wasn’t happening.

Finally I gave up and just left XV disabled entirely, which seems to be ok for recorded TV, but DVD playback is noticeably choppy. In both cases CPU usage is well over 50% (most of it being eaten by the X server) on my Athlon XP 2500+, which is ludicrous.

In any case, the poor performance drove me to look for other options. The default ati driver included with x.org unfortunately doesn’t include any tv output support, but the gatos driver folks do support it, so I built up a new version of the latest ATI driver with their patch applied and lo and behold it works! Kudo’s to the fedora x.org packagers for splitting out the drivers so nicely. Here’s a patch against the most recent fc5 ati driver specfile.


--- SPECS/xorg-x11-drv-ati.spec 2006-04-25 21:22:36.000000000 -0400
+++ SPECS/xorg-x11-drv-ati-theater_out.spec     2006-11-05 15:56:11.000000000 -0500
@@ -13,7 +13,7 @@
 Summary:   Xorg X11 ati video driver
 Name:      xorg-x11-drv-ati
 Version:   6.5.8.0
-Release:   1
+Release:   1.theater_out
 URL:       http://www.x.org
 License:   MIT/X11
 Group:     User Interface/X Hardware Support
@@ -24,6 +24,8 @@
 Source2:   r128.xinf
 Source3:   radeon.xinf
 Patch0:    xorg-x11-drv-ati-6.5.7.3-radeon-metamodes-SEGV-fix.patch
+#Patch1:          xorg7-6.5.8.0-tv_output.patch.gz
+Patch1:           http://megahurts.dk/rune/stuff/xorg7-6.5.8.0-tv_output.patch.gz

 ExclusiveArch: %{ix86} x86_64 ia64 ppc alpha sparc sparc64

@@ -42,6 +44,7 @@
 %prep
 %setup -q -n %{tarball}-%{version}
 #%patch0 -p2 -b .radeon-metamodes-SEGV-fix
+%patch1 -p1 -b .theater_out

 %build
 %configure --disable-static
@@ -79,12 +82,16 @@
 %{moduledir}/multimedia/theatre200_drv.so
 %{moduledir}/multimedia/theatre_detect_drv.so
 %{moduledir}/multimedia/theatre_drv.so
+%{moduledir}/multimedia/theater_out_drv.so
 #%dir %{_mandir}/man4x
 %{_mandir}/man4/ati.4*
 %{_mandir}/man4/r128.4*
 %{_mandir}/man4/radeon.4*

 %changelog
+* Sat Nov 04 2006 Erik LaBianca  6.5.8.0-1.theater_out
+- Added gatos theater out patch
+
 * Tue Apr 25 2006 Adam Jackson  6.5.8.0-1
 - Updated to stable branch release from upstream.

The relevant sections of the xorg.conf file needed to make this go are here:


Section "Monitor"
        Option "DPMS"
        HorizSync    30.0 - 40.0
        VertRefresh  60
        Identifier   "Monitor0"
EndSection

Section "Device"
        Driver      "ati"
        Option      "IgnoreEDID" "True"
        Option      "TVOutput" "NTSC"
        Option      "MonitorLayout" "AUTO, NONE"
        Identifier  "ATI Graphics Adapter"
        BusID       "PCI:3:0:0"
EndSection

Section "Screen"
        Identifier "Screen0"
        Device     "ATI Graphics Adapter"
        Monitor    "Monitor0"
        DefaultDepth     24
        SubSection "Display"
                Viewport   0 0
                Depth     24
                Modes    "800x600"
        EndSubSection
EndSection

This all does indeed work, and reasonably well. My picture isn’t perfectly centered, and the UI screens in Myth have a bit of flicker, but it does work, and is even open source, so I’m pretty happy with the change. I’m not sure of the licensing implications of actually distributing an x.org binary with a gpl patch included, so I’m not posting the RPMS here directly.

IVTV Autoloading on Fedora Core 5

Sunday, November 5th, 2006

My MythTV system has been in place for several years now, and has seen many versions of Fedora. Ensuring the IVTV modules were loaded correctly after a system restart has always been a bit of a black art, however, and with Fedora Core 5 this seems to be no less of an exception. However, I think I have got it partially figured out, so here it is.

First things first. Modern 2.6 linux kernels apparently fully support hardware autodetection. I understand that to mean you aren’t support to need to manually (or in a script) modprobe ivtv nor should you have to explicitly put any configuration into modprobe.conf. If your IVTV kernel modules are installed correctly, it’s supposed to be automatically detected, and then udev is supposed to create the relevant devices for you. This is working for me, however changing the permissions on video devices in /etc/udev/rules.d/50-udev.rules isn’t working. I tried to use this line to do it, but it appears to be ignored in general.


KERNEL=="video*",               MODE="0666"

The default is 0660 and I get this:


crw------- 1 mythtv root 81, 0 Nov  4 15:16 /dev/video0

My guess is that ivtv doesn’t play nice with udev, or I just don’t know enough about udev to use it effectively. I did find some good udev documentation so perhaps I’ll figure it out eventually.

Regardless, on my frontend system I also need my hauppauge pvr-250 remove receiver to work, and this is where things got sticky. I had some settings in place for Fedora Core 3 from Jarrod’s guide trying to preload lirc-i2c before loading ivtv which were hanging up on startup, so I had commented them out. That was allowing ivtv to load, but my remote didn’t work. A quick hack to those lines fixed the problem, however. It appears as if nowadays ivtv wants to load first, and then have lirc_i2c stuff in on top, so this seems to work well.


cat > /etc/modprobe.d/ivtv.conf <<-EOF
install ivtv /sbin/modprobe --first-time --ignore-install ivtv; \
    { /sbin/modprobe lirc_dev; /sbin/modprobe lirc_i2c; }
EOF

Oracle Unbreakable Linux?

Friday, November 3rd, 2006

I’d heard in passing that Larry Ellison announced Oracle was going to start selling Red Hat Linux support, and that Red Hat stock and dropped like 43% in a day because of it, but the gravity of the situation never sank in. Then I checked out their FAQ and things became more clear. In my opinion, Red Hat is really going to have to pony up or risk losing a lot of business. Why? Pricing! Oracle is going to give out free binary downloads (why oh why didn’t Red Hat do that?) and their basic ‘just updates’ service is $99/year for unlimited cpu’s. They’re also discounting their Basic Support to $199 per year for up to 2 cpu’s, which appears to entail 24×7 phone support. I’m not sure exactly what premier support is, and what ‘back ports of fixes to earlier releases’ entails, so I’ll refrain from comment on that but suffice it to say that given Red Hat doesn’t return my calls when I try to purchase their product, let alone actually get support, I’m pretty interested in an alternative.

In any case, $100 bucks a year and the base platform for free is pretty compelling pricing. Giving out the binaries is a huge step. Regardless of how great you think open source is, nobody can use it without binaries. Refusing to give out ‘real’ binaries without a support contract is assinine business in my opinion. How do you expect to get people hooked on your product if they can’t even try it out for free? And we wonder why Ubuntu is taking over the linux hobbyist space?