Anyone running 64bit with SageTV on Linux?

[stuckless]

My "server" parts arrived today... nothing special... amd quad core with 4 gigs of memory, but it's a step up from my existing single core server. Is there any advantage of running 64bit vs 32bit? I currently use ubuntu 32bit desktop as my "server" and it runs flawlessly, but I'm now considering using ubuntu 32bit server, or 64bit server if there's a distinct advantage.

[bcjenkins]

In 32-bit there is. The kernel is compiled with PAE support. Not sure with 64-bit as I just used the desktop version.

B

[stuckless]

Thanks... I think i'm going with 32bit ubuntu server... btw... on a side note, I think i'm going to go with running the zfs filessytem... I like the simplicity of how you configure it, and want to have a dynamic drive pool with at least raid mirror support. (I was going to use raid+lvm, but I got scared after reading the docs )

[jpwegas]

Originally Posted by stuckless

Thanks... I think i'm going with 32bit ubuntu server... btw... on a side note, I think i'm going to go with running the zfs filessytem... I like the simplicity of how you configure it, and want to have a dynamic drive pool with at least raid mirror support. (I was going to use raid+lvm, but I got scared after reading the docs )

I'm not sure there is a distinct advantage with running 64 bit unless you need the extra memory. My impression is that using 64 bit is more of a hassle due to various driver issues.

Is there a kernel ZFS available or do you still have to use FUSE?

I have basic LVM (ext2) and software RAID5+LVM (xfs) media stores on my Sage host. ZFS looked tempting but at the time I wasn't sure what the performance of ZFS+FUSE was.

The docs for the the LVM and/or RAID setup looks more daunting than they should be, but I'd recommend playing around on some empty drives the first time. :-)

I have notes on what I did last time if anyone's interested.

[videoguy2]

Regarding Unknownhost exception when running 32-bit java on 64-bit Ubuntu 8 -- a similar issue was reported here, for Ubuntu 8.04:

http://developer.amazonwebservices.c...30760&tstart=0

The error does not appear when running on the 32-bit version of Ubuntu 8.04, nor does it when running 64-bit Java on the 64-bit Ubuntu 8.04.

[videoguy2]

https://bugs.launchpad.net/ubuntu/+bug/227111

...editing /etc/nsswitch.conf seems to have worked for me, but according to one of the posts at the above link, YMMV.

[stuckless]

In the end, I went with my original 32bit ubuntu. I just moved my original boot drive from my old server and put it in my new server. Ubuntu and sage both started up fine on the new hardware.

I have setup lvm in the past, although not lvm+raid. In the end, I went with zfs.... it's quite nice to be able to create a 1tb raided volume that is also mounted, using a single command....
zpool create -m /media/media media mirror /dev/sdb /dev/sdc
no formatting required

zfs runs using fuse, but there's a repository for installing it. Also, you can play with zfs using virtual disks (ie, files on the local filesytem).

The performance of zfs is not great, but it's good enough for my needs. I hear they are making improvements al the time, so time will tell.

[kbyrd]

Originally Posted by kbyrd

Well, I didn't plan to document it so this may not work for everyone. I may have done something here or there that I forgot to mention. If I reinstall from scratch, I'll keep better notes.

See the earlier post I'm quoting for context...

Well, while I didn't install from scratch. I did to a fresh install of Ubuntu Jaunty (9.04) 64-bit, then copied various config files over. It's not even a full install. It is WAY stripped down, about as stripped down as an Ubuntu install can be. I did debootstrap to a chroot then built things up until everything I needed worked.

Related to Sage, I can say that I ONLY installed the 32-bit 1.6 JRE. No mad, faad, or anything else. Basic functionality (tv recording/playback, playback of a few random but supported files with common codecs, mp3 playing, jpg image viewing) all worked. So, I don't appear to need the extra libraries. It may be they're required for transcoding or gathering information from file formats (my library is built, so I can imagine Sage doesn't actually need to read the mp3 files), but it may be that it's all built into Sage now and nothing external is required. I really wish Narflex would weigh in so we would know for sure.

[jpappas]

Hey all,

After a couple weeks post migration from old 32-bit server to new 64-bit (OpenSuse), I have nailed down most of the niggling issues.

My last issue is that the 2nd tuner on the Dual PVR-500 occasionally quits with a `kernel: ivtv1: DMA TIMEOUT 00000001 0` reflected in the logs. I have a request open with support, but in the mean time, any ideas?

Background:

I run all LVM+RAID disks, so created a new partition of 2GB for /opt/sagetv. In that directory, I have the sagetv_<version> directories, 32-bit jre_<version>, backup, archive, and a downloads directory. I use softlinks to point server -> sagetv_latest and jre -> jre_latest, that way I just run a cp+tar and repoint link to upgrade. I did have to modify my `startsage` script to set up the jre variables for the different path to use (rather than the devalt ENV for the 64-bit JRE that is OS installed and managed).

After various discussions with support, I am running SageTV 6.5.14 and JRE 1.6u14.

I have a PVR-500 running capture, and I used the HD-PVR thread to update the drivers and DVB support stuff to get rid of most of the issues related to the 64-bit OS and 32-bit SageTV app, but I cannot keep the second tuner running for any sustained length of time (failure times vary, but a couple of days is the longest so far, running with video1 unconfigured in sage leads to weeks of stability). There is some discussion over at Gossamer related to this issue, but seemingly no fix. I am running 64-bit kernel 2.6.27.23-0.1-default on an A-bit AMD/nVidia board. The card worked in the other system, so this may be one of the problems that "cannot be overcome" referenced by Hans.

Do I need a second SageTV license to run a "network encoder": ie move the PVR-500 to a different system?

Thanks!
john

[bcjenkins]

You would need one to run as a network tuner. Have you done a driver pull from the dvb mercurial? I wouldn't use any patches in the thread. The current mercurial has what you need to go.

B

[jpappas]

Originally Posted by bcjenkins

You would need one to run as a network tuner. Have you done a driver pull from the dvb mercurial? I wouldn't use any patches in the thread. The current mercurial has what you need to go.

B

I ran the steps that you documented in the HD-PVR thread and removed the OpenSUSE provided IVTV RPMS (Driver and Firmware).

The ivtv driver did not load, as the firmware was now missing. So it seems that the firmware is not included when I did the make install from mercurial, so I am wondering what the correct method to grab the latest firmware files is, as there may have been a firmware change between the latest and the 1.3 firmware files provided by the OpenSuse repos (I had to install the ivtv-firmware RPM so that the mercurial ivtv driver would load.

Thoughts?

[bcjenkins]

Take a look here: http://www.ivtvdriver.org/index.php/Firmware_versions

The firmware files are actually in the Windows driver packages and you can grab the latest from Hauppauge's website. Installation instructions are here: http://www.ivtvdriver.org/index.php/Firmware

I would assume that the Suse package has the latest firmware, but not the latest drivers. However, it wouldn't hurt to check. Would you post a dmesg output?

B

[jpappas]

Originally Posted by bcjenkins

Take a look here: http://www.ivtvdriver.org/index.php/Firmware_versions

The firmware files are actually in the Windows driver packages and you can grab the latest from Hauppauge's website. Installation instructions are here: http://www.ivtvdriver.org/index.php/Firmware

I would assume that the Suse package has the latest firmware, but not the latest drivers. However, it wouldn't hurt to check. Would you post a dmesg output?

dmesg output is TXT attachment

OpenSUSE has 2 ivtv packages, one with the driver and one with the firmware. I have removed the driver package and kept the firmware package.

I have installed the v4l-dvb mercurial driver via instructions.

As per the second link, I downloaded, diff'd, and md5summed the appropriate firmware files from both the ivtv bundle and the opensuse rpm. The md5sums don't match what listed as correct on the first link, but they do match what came from the ivtv firmware bundle.

ghost:/opt/sagetv/downloads # wget http://dl.ivtvdriver.org/ivtv/firmware/ivtv-firmware.tar.gz
--2009-07-01 10:09:50--  http://dl.ivtvdriver.org/ivtv/firmware/ivtv-firmware.tar.gz
Resolving dl.ivtvdriver.org... 130.133.35.29
Connecting to dl.ivtvdriver.org|130.133.35.29|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 125526 (123K) [application/x-gzip]
Saving to: `ivtv-firmware.tar.gz'

100%[==========================================================>] 125,526      133K/s   in 0.9s

2009-07-01 10:09:52 (133 KB/s) - `ivtv-firmware.tar.gz' saved [125526/125526]

ghost:/opt/sagetv/downloads # tar xzvf ivtv-firmware.tar.gz
license-end-user.txt
license-oemihvisv.txt
v4l-cx2341x-dec.fw
v4l-cx2341x-enc.fw
v4l-cx2341x-init.mpg
v4l-cx25840.fw
v4l-pvrusb2-24xxx-01.fw
v4l-pvrusb2-29xxx-01.fw
ghost:/opt/sagetv/downloads # diff v4l-cx2341x-enc.fw /lib/firmware/v4l-cx2341x-enc.fw
ghost:/opt/sagetv/downloads # diff v4l-cx2341x-dec.fw /lib/firmware/v4l-cx2341x-dec.fw
ghost:/opt/sagetv/downloads # diff v4l-cx25840.fw /lib/firmware/v4l-cx25840.fw
ghost:/opt/sagetv/downloads # md5sum v4l-*
305dba74bbe5905447add8883f3ecb68  v4l-cx2341x-dec.fw
9b39b3d3bba1ce2da40f82ef0c50ef48  v4l-cx2341x-enc.fw
0661f8b2693fe3123e6234557353eacc  v4l-cx2341x-init.mpg
b3704908fd058485f3ef136941b2e513  v4l-cx25840.fw
ghost:/opt/sagetv/downloads # cd /lib/firmware
ghost:/lib/firmware # md5sum v4l-*
305dba74bbe5905447add8883f3ecb68  v4l-cx2341x-dec.fw
9b39b3d3bba1ce2da40f82ef0c50ef48  v4l-cx2341x-enc.fw
0661f8b2693fe3123e6234557353eacc  v4l-cx2341x-init.mpg
b3704908fd058485f3ef136941b2e513  v4l-cx25840.fw

Does not appear to be a difference. I did not grab from the direct Haup driver files, but I can if that seems to be necessary.

Thanks for all your help!
John

[bcjenkins]

A few more pieces of info:

http://www.gossamer-threads.com/lists/ivtv/users/37744

http://www.mail-archive.com/ivtv-use.../msg08910.html

And: http://www.mail-archive.com/ivtv-use.../msg08912.html

Hans is the driver maintainer and creator (I think)

You can always send an email to the linux-media list as well. You don't have to join to do so.

Good Luck!

B

[jpappas]

Yeah, testing is a pain. It is unclear if the latency_timer setting takes effect immediately. It is also unclear where the ivtv-ctrl debugging info goes. I am only running the 2.6.27 kernel, and the ivtv-utils v1.4 require 2.6.28, so I am trying the v1.3 utils.

Update: I am kicking around the idea to revamp the system anyway, so I will be starting a new thread to get some input (will probably only use the PVR-500 for manual analog transcodes from VCR or such)

Update 2, RE Latency_timer settings: In my research, I have learned that the setpci command is immediate but not persistent (and in hex, not decimal), and only applies to classical PCI slots (PCIe/AGP does not have a setting, so lspci shows a latency time of 0 for those devices. I have learned that the default is set by the Mobo manufacturer (usual 32) and IVTV resets to a 64 minimum. In my case the 64 was not high enough, so I added setpci commands to my startsage scripts (will probably add to rc.local since it not really related to sagetv software) using a setting of latency_timer=80 (128 decimal). Since I have a PCI SATA controller, I set that to c0 (192--so that it can handle larger transactions and gets a bigger share of the PCI bandwidth, thus not "losing" to the encoders). Again, this only applies to the classic PCI bus, so PVR-XXX and HVR-1600 type encoders are affected.

I have been running stable and dual-analog for 3 days with the new settings so fingers crossed....