Anyone running 64bit with SageTV on Linux?

[bcjenkins]

Experiences? Doing a new server build and RAM is so cheap.. I will run in PAE mode but would rather run in 64-bit.

B

[jreichen]

I tried running 64-bit on Gentoo a long time ago but decided to give up until Sage supports it. It doesn't seem like it's a priority for them because that was almost 2 years ago. I've since switched to Ubuntu but stuck with 32-bit.

http://forums.sagetv.com/forums/show...3&postcount=30
http://forums.sagetv.com/forums/show...1&postcount=42

[uggie]

And that's still the case. You can try following what I did a few months ago, but I wasn't completely successful. I think the final failure involved 64bit versions of the ivtv drivers not playing well with 32bit java and sage. I did some searching online and found some tutorials on how to set up a 32bit chroot environment on a 64bit system, but it didn't look like it was worth the hassle.

Anyway, this will at least get sage to start but won't talk to the hardware as is:

http://forums.sagetv.com/forums/showthread.php?t=28396

Regards,
-jason

[drewg]

Originally Posted by uggie

64bit versions of the ivtv drivers not playing well with 32bit

Does anybody know if the DVB drivers do any better? FWIW, 32-bit applications *should* work with a 64-bit kernel. If they don't, it is a kernel or driver bug and should be fixed.

Thanks,

Drew

[uggie]

Unfortunately that is not strictly true. 32bit apps should mostly work, but they can break depending on how addresses and data sizes were coded. The fact that sage isn't compatible with 64bit java (same goes for firefox flash plugins) demonstrates that. Drivers mucking around with hardware are going to be even more finicky and aren't likely to behave well running in a mixed mode.

Having said that, I haven't done a debug trace to see where things are crashing, but the ivtv problem is my best guess as that was the one thing I didn't build in 32bit mode.

-jason

[drewg]

Originally Posted by uggie

Unfortunately that is not strictly true. 32bit apps should mostly work, but they can break depending on how addresses and data sizes were coded. The fact that sage isn't compatible with 64bit java (same goes for firefox flash plugins) demonstrates that. Drivers mucking around with hardware are going to be even more finicky and aren't likely to behave well running in a mixed mode.

Mixing 32- and 64- bit libs and apps is essentially impossible, or at least so difficult that one would not want to try to get it right. I think on amd64, you'd need to change the mode of the CPU between 32- and 64- bit environments. Ick.

Access via ioctls to character drivers is *much* easier to get right. I'm the author of an Linux HPC interconnect driver, as well as the author of the Linux and OSF/1 binary support on the now defunct FreeBSD/alpha port, and I've dealt ioctl binary compatibility several times. For well designed APIs (eg, ones where all the types are explicitly sized, and there are no uses of types like longs or pointers which vary in size between 32- and 64- bit arches, there is no problem. For poorly designed APIs, there can be a shim layer in the kernel or driver which converts the arguments into their proper format. It looks like V4L has such a shim layer, and I suspect that SageTV is triggering a bug in it.

Drew

[ltitran]

I have a Ubuntu 7.10 system running the Placeshifter now, but my new server Phenom is ready to test 64b out.

I thought I read that Sage will run in 32b emulation on a 64b system?

I also know that Ubuntu is releasing a new 64b os at the end of this month...
I might pull the system server drive and test it out...

[bcjenkins]

From an inquiry to SageTV 64-bit is more driver dependent. So as long as there are drivers, you should be good. I did not have the time to want to mess with it, but please keep us posted on success.

B

PS - I thought I had read there were issues with running 32-bit java code in a 64-bit OS, but I didn't spend a ton of time on it either. My argument for native 64-bit is that RAM is cheap and Centrino2 is 64-bit.

[Vox-uLaw]

O/S: OpenSuSE 10.3 (64-Bit)
Tuner: SiliconDust HDHomeRun

• Installed SageTV dependencies, 64-Bit versions; faad2, lame and mad.
• Installed Wine from repository for comskip usage; 32-Bit.
• Downloaded the latest 32-Bit Java from Sun website; jdk-6u5-linux-i586.bin at the time.
• Added the following two lines to the top of the startsage script.
  This allowed me to test many different versions of Java without changing anything on the system.
  

  PATH=/opt/jdk1.6.0_05/bin:$PATH
  export PATH

I was going to install Ubuntu or Gentoo, but their release dates were a few months out so I settled with OpenSuSE, which has very nice RAID detection in the installer. I setup the machine with six available OS partitions so I can install other distributions at later dates.

SageTV detected the HDHomeRun tuner without incident. The primary reason I purchased that tuner was to avoid kernel and driver issues related to on-board tuner cards.

I use an extender rather than hooking up directly to the projector, so I bypassed a lot of video issues there.

I probably did more, but am not sure it relates to getting SageTV up and running on 64-Bit O/S.

[Zervun]

This is really good news, I'm trying the exact same thing as we speak but with centos 5.1 64bit, so I'll chime in how it goes.

[tuxsf]

I'm also thinking of building a Linux box with SageTV, a Silicon HDHomeRun, and a STX-HD100 extender when SageTV gets them into stock again. I'm also considering building a Linux box with MythTV but MythTV doesn't have a nice cheap HD extender with all the GUI's features and menus like SageTV does. I wanted to ask Vox-uLaw and anyone else who has a similar setup - how well do SageTV closed captions and commercial skipping work under Linux? Thanks, John.

[drewg]

Originally Posted by tuxsf

- how well do SageTV closed captions and commercial skipping work under Linux? Thanks, John.

I've never tried closed captions.

Commercial skipping works great. Don't get frightened that comskip runs under Wine; it is still 2x to 3x as fast as the MythTV commercial skipping, and it seems to be a bit more accurate. It is much more flexible in terms of tuning.

FWIW, I came from 3 years of doing HD with MythTV. SageTV with an HD100 extender is night and day easier to setup and use.

Drew

[Vox-uLaw]

Originally Posted by tuxsf

... how well do SageTV closed captions and commercial skipping work under Linux?

As drewg stated, comskip works great in Linux.

I installed the CC add-on so I could get subtitles working on SageTV for non-english audio tracks from my DVD collection, and those rare movies which use forced subtitles instead of rendering them on the video. The process to go from a .sub -> .srt -> .smi is anything but automated, but the end product is worth it.

As for recorded viewings, I use ccextractor to extract the captions into a sami file, which the Caption Playback STVI reads. This works on both the placeshifter and the HD Media Extender. This is scripted and on a cron schedule for automation.

Not sure about live TV as I don't watch TV live; I haven't been able to get the embedded captions to display. This is not a problem for me as I like to run comskip before watching TV anyway, the extra overhead for generating SAMI files is negligible.

[bcjenkins]

Just tried to switch this morning to Ubuntu 64-bit 8.04 running a custom 2.6.25 kernel and drivers from ivtv mercurial. Simialr setup in 32-bit is working fine.

My HD tuners seem to be able to tune and record without issue, the analog tuners though are throwing an error in dmesg when trying to tune:

[ 111.079366] ioctl32(java:5893): Unknown cmd fd(35) cmd(c0185647){t:'V';sz:24} arg(ce6acccc) on /dev/video5
[ 38.647273] ioctl32(java:5893): Unknown cmd fd(31) cmd(80685600){t:'V';sz:104} arg(08e9c760) on /dev/video5
[ 38.647661] ioctl32(java:5893): Unknown cmd fd(35) cmd(40085618){t:'V';sz:8} arg(ce6addd8) on /dev/video5
[ 39.038365] ivtv1: Loaded v4l-cx2341x-enc.fw firmware (376836 bytes)
[ 39.110176] ivtv1: Encoder revision: 0x02060039
[ 104.886751] cx25840 2-0044: loaded v4l-cx23885-avcore-01.fw firmware (16382 bytes)
[ 39.471151] ioctl32(java:5893): Unknown cmd fd(38) cmd(c0185647){t:'V';sz:24} arg(ce6acccc) on /dev/video6
[ 39.471246] ioctl32(java:5893): Unknown cmd fd(52) cmd(80685600){t:'V';sz:104} arg(0830a2d8) on /dev/video6
[ 39.471393] ioctl32(java:5893): Unknown cmd fd(38) cmd(40085618){t:'V';sz:8} arg(ce6addd8) on /dev/video6
[ 39.518429] ioctl32(java:5893): Unknown cmd fd(56) cmd(c0185647){t:'V';sz:24} arg(ce6acccc) on /dev/video2
[ 39.518493] ioctl32(java:5893): Unknown cmd fd(57) cmd(80685600){t:'V';sz:104} arg(0830a2d8) on /dev/video2
[ 39.518635] ioctl32(java:5893): Unknown cmd fd(56) cmd(40085618){t:'V';sz:8} arg(ce6addd8) on /dev/video2
[ 146.346095] ioctl32(java:6012): Unknown cmd fd(70) cmd(80685600){t:'V';sz:104} arg(084cf450) on /dev/video2
[ 146.346432] ioctl32(java:6012): Unknown cmd fd(56) cmd(40085618){t:'V';sz:8} arg(ce091c58) on /dev/video2
[ 87.273827] ioctl32(java:6141): Unknown cmd fd(76) cmd(80685600){t:'V';sz:104} arg(081d6510) on /dev/video2
[ 87.273961] ioctl32(java:6141): Unknown cmd fd(56) cmd(40085618){t:'V';sz:8} arg(cd75cad8) on /dev/video2

These are PVR-500 (ivtv) and HVR-1600 (cx18)

Any thoughts?

B

[drewg]

Originally Posted by bcjenkins

Any thoughts?

From a brief read of a slightly older kernel (2.6.23) it looks like ioctl syscalls from 32-bit processes go through a compat handler (fs/compat_ioct). If no .compat_ioctl is registered in the driver's file operations, I think you'll wind up with a default handler which spews the messages you were seeing. It looks like some v4l drivers register compat ioctl handlers and others don't. Ivtv doesn't, but I don't know if that is for a good reason, or if it is just an omission. The first thing to try would be to just add a line:

.compat_ioctl = v4l_compat_ioctl32,

To the ivtv_v4l2_enc_fops and ivtv_v4l2_dec_fops initializations
in ivtv_streams.c. Something will probably explode horribly, but it can't hurt to try...

Drew

[bcjenkins]

wow, thanks for the info! I will post back when I test.

B

[bcjenkins]

So far so good. Will update after I run for a while.

B

[Zervun]

I tried doing 64 bit centos, ran into a snag in that when putting in license number it reported that the encyrption to send back to sage was invalid (32 bit works)

I was using 32 bit java, will poke at it more tonight - I used the centos readme here for install, was perfect

[bastafidli]

Zervun, what tuner are you using? I am currently using my CentOS without tuner but might be interesting in adding one in near future and I would like to get one which works with CentOS.

[bcjenkins]

fwiw - I am running:

1 PVR-500
1 HVR-1600
1 HVR-1800 - Digital only right now
1 Fusion5 GoldRT

If you have your kernel source you can compile the drivers from the V4L Mercurial and it should work on any distro. caveat - I have only tested on Ubuntu, Gentoo, Arch Linux.

B