media_build and Terratec Cinergy T Black.

media_build and Terratec Cinergy T Black.

[Hin-Tak Leung]

Firstly, the downloaded linux-media.tar.bz2 has some simple typos, missing 3 brackets:

(please feel free to add my signed-off though this is trivial).

-------------
--- linux/drivers/media/video/v4l2-compat-ioctl32.c~	2012-07-06 04:45:17.000000000 +0100
+++ linux/drivers/media/video/v4l2-compat-ioctl32.c	2012-07-06 07:35:54.166512279 +0100
@@ -387,7 +387,7 @@
 		get_user(kp->index, &up->index) ||
 		get_user(kp->type, &up->type) ||
 		get_user(kp->flags, &up->flags) ||
-		get_user(kp->memory, &up->memory)
+		get_user(kp->memory, &up->memory))
 			return -EFAULT;
 
 	if (V4L2_TYPE_IS_OUTPUT(kp->type))
@@ -471,7 +471,7 @@
 		put_user(kp->index, &up->index) ||
 		put_user(kp->type, &up->type) ||
 		put_user(kp->flags, &up->flags) ||
-		put_user(kp->memory, &up->memory)
+		put_user(kp->memory, &up->memory))
 			return -EFAULT;
 
 	if (put_user(kp->bytesused, &up->bytesused) ||
@@ -481,7 +481,7 @@
 		copy_to_user(&up->timecode, &kp->timecode, sizeof(struct v4l2_timecode)) ||
 		put_user(kp->sequence, &up->sequence) ||
 		put_user(kp->reserved2, &up->reserved2) ||
-		put_user(kp->reserved, &up->reserved)
+		put_user(kp->reserved, &up->reserved))
 			return -EFAULT;
 
 	if (V4L2_TYPE_IS_MULTIPLANAR(kp->type)) {
------------

A few comments & issues:

- don't realy like the build script trying to clone media_tree, etc had hard-coded remotes - if(0) out'ed the whole git-on-the-fly block inside to make it use an existing symlink'ed checkout - consider offering something similar?

- $ lsdvb seems to be doing garbage:(fedora 17's)

usb (-1975381336:62 64848224:32767) on PCI Domain:-1965359032 Bus:62 Device:64848416 Function:32767
	DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek RTL2832 (DVB-T)) 
		 FE_OFDM Fmin=174MHz Fmax=862MHz

lsdvb on mercury is only marginally better with the PCI zero's, but the other numbers swapped:

usb (62:-1975379912 32767:-348245472) on PCI Domain:0 Bus:0 Device:0 Function:0
	DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek RTL2832 (DVB-T)) 
		 FE_OFDM Fmin=174MHz Fmax=862MHz

- 'scandvb' segfault at the end on its own.


- "scandvb /usr/share/dvb/dvb-t/uk-SandyHeath" (supposedly where I am) got a few "WARNING: >>> tuning failed!!!" and no list.

- 'w_scan -G -c GB'
  have a few curious
WARNING: received garbage data: crc = 0xcc93876c; expected crc = 0xb81bb6c4

return a list of 26, with entries like (which seems to be vaguely correct):

BBC ONE;(null):522000:B8C23D0G32M64T8Y0:T:27500:101=2:102,106=eng:0:0:4173:9018:4173:0:100

So I just put it in ~/.mplayer:channels.conf

Took me a while to figure out that mplayer wants:

mplayer 'dvb://BBC ONE;(null)'

rather than anything else - curious about the ';(null)' part.

--------
Playing dvb://BBC ONE;(null).
dvb_tune Freq: 522000
ERROR IN SETTING DMX_FILTER 9018 for fd 4: ERRNO: 22ERROR, COULDN'T SET CHANNEL  8: Failed to open dvb://BBC ONE;(null).
----------

At this point I am lost :-).

Re: media_build and Terratec Cinergy T Black.

[Hin-Tak Leung]

One more comment: blowing away /lib/modules/`uname -r`/kernel/drivers/ seems drastic?

The compat-wireless people uses some clever tricks to get modprobe to preferentially load /lib/modules/`uname -r`/updates first. This way you can just remove that, run 'depmod -a' and go back to stock distro kernel behavior.

Also it might be useful/quicker not to build the whole 500+ kernel modules...

--- On Fri, 6/7/12, Hin-Tak Leung <htl10@users.sourceforge.net> wrote:

> From: Hin-Tak Leung <htl10@users.sourceforge.net>
> Subject: media_build and Terratec Cinergy T Black.
> To: "Antti Palosaari" <crope@iki.fi>, mchehab@redhat.com
> Cc: linux-media@vger.kernel.org
> Date: Friday, 6 July, 2012, 11:54
> Firstly, the downloaded
> linux-media.tar.bz2 has some simple typos, missing 3
> brackets:
> 
> (please feel free to add my signed-off though this is
> trivial).
> 
> -------------
> ---
> linux/drivers/media/video/v4l2-compat-ioctl32.c~   
> 2012-07-06 04:45:17.000000000 +0100
> +++
> linux/drivers/media/video/v4l2-compat-ioctl32.c   
> 2012-07-06 07:35:54.166512279 +0100
> @@ -387,7 +387,7 @@
>         
> get_user(kp->index, &up->index) ||
>          get_user(kp->type,
> &up->type) ||
>         
> get_user(kp->flags, &up->flags) ||
> -       
> get_user(kp->memory, &up->memory)
> +       
> get_user(kp->memory, &up->memory))
>             
> return -EFAULT;
>  
>      if (V4L2_TYPE_IS_OUTPUT(kp->type))
> @@ -471,7 +471,7 @@
>         
> put_user(kp->index, &up->index) ||
>          put_user(kp->type,
> &up->type) ||
>         
> put_user(kp->flags, &up->flags) ||
> -       
> put_user(kp->memory, &up->memory)
> +       
> put_user(kp->memory, &up->memory))
>             
> return -EFAULT;
>  
>      if (put_user(kp->bytesused,
> &up->bytesused) ||
> @@ -481,7 +481,7 @@
>         
> copy_to_user(&up->timecode, &kp->timecode,
> sizeof(struct v4l2_timecode)) ||
>         
> put_user(kp->sequence, &up->sequence) ||
>         
> put_user(kp->reserved2, &up->reserved2) ||
> -       
> put_user(kp->reserved, &up->reserved)
> +       
> put_user(kp->reserved, &up->reserved))
>             
> return -EFAULT;
>  
>      if
> (V4L2_TYPE_IS_MULTIPLANAR(kp->type)) {
> ------------
> 
> A few comments & issues:
> 
> - don't realy like the build script trying to clone
> media_tree, etc had hard-coded remotes - if(0) out'ed the
> whole git-on-the-fly block inside to make it use an existing
> symlink'ed checkout - consider offering something similar?
> 
> - $ lsdvb seems to be doing garbage:(fedora 17's)
> 
> usb (-1975381336:62 64848224:32767) on PCI
> Domain:-1965359032 Bus:62 Device:64848416 Function:32767
>     DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek
> RTL2832 (DVB-T)) 
>          FE_OFDM
> Fmin=174MHz Fmax=862MHz
> 
> lsdvb on mercury is only marginally better with the PCI
> zero's, but the other numbers swapped:
> 
> usb (62:-1975379912 32767:-348245472) on PCI Domain:0 Bus:0
> Device:0 Function:0
>     DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek
> RTL2832 (DVB-T)) 
>          FE_OFDM
> Fmin=174MHz Fmax=862MHz
> 
> - 'scandvb' segfault at the end on its own.
> 
> 
> - "scandvb /usr/share/dvb/dvb-t/uk-SandyHeath" (supposedly
> where I am) got a few "WARNING: >>> tuning
> failed!!!" and no list.
> 
> - 'w_scan -G -c GB'
>   have a few curious
> WARNING: received garbage data: crc = 0xcc93876c; expected
> crc = 0xb81bb6c4
> 
> return a list of 26, with entries like (which seems to be
> vaguely correct):
> 
> BBC
> ONE;(null):522000:B8C23D0G32M64T8Y0:T:27500:101=2:102,106=eng:0:0:4173:9018:4173:0:100
> 
> So I just put it in ~/.mplayer:channels.conf
> 
> Took me a while to figure out that mplayer wants:
> 
> mplayer 'dvb://BBC ONE;(null)'
> 
> rather than anything else - curious about the ';(null)'
> part.
> 
> --------
> Playing dvb://BBC ONE;(null).
> dvb_tune Freq: 522000
> ERROR IN SETTING DMX_FILTER 9018 for fd 4: ERRNO: 22ERROR,
> COULDN'T SET CHANNEL  8: Failed to open dvb://BBC
> ONE;(null).
> ----------
> 
> At this point I am lost :-).
> 
> 
> 
> 
> 
> 
> 

Re: media_build and Terratec Cinergy T Black.

[Antti Palosaari]

On 07/06/2012 01:54 PM, Hin-Tak Leung wrote:
> - $ lsdvb seems to be doing garbage:(fedora 17's)
>
> usb (-1975381336:62 64848224:32767) on PCI Domain:-1965359032 Bus:62 Device:64848416 Function:32767
> 	DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek RTL2832 (DVB-T))
> 		 FE_OFDM Fmin=174MHz Fmax=862MHz
>
> lsdvb on mercury is only marginally better with the PCI zero's, but the other numbers swapped:
>
> usb (62:-1975379912 32767:-348245472) on PCI Domain:0 Bus:0 Device:0 Function:0
> 	DEVICE:0 ADAPTER:0 FRONTEND:0 (Realtek RTL2832 (DVB-T))
> 		 FE_OFDM Fmin=174MHz Fmax=862MHz

I was aware of that tool but didn't know it lists USB devices too.
Someone should fix it working properly for USB devices.

> - 'scandvb' segfault at the end on its own.

I didn't see that.

> - "scandvb /usr/share/dvb/dvb-t/uk-SandyHeath" (supposedly where I am) got a few "WARNING: >>> tuning failed!!!" and no list.
>
> - 'w_scan -G -c GB'
>    have a few curious
> WARNING: received garbage data: crc = 0xcc93876c; expected crc = 0xb81bb6c4
>
> return a list of 26, with entries like (which seems to be vaguely correct):
>
> BBC ONE;(null):522000:B8C23D0G32M64T8Y0:T:27500:101=2:102,106=eng:0:0:4173:9018:4173:0:100

Both scandvb and w_scan works here, same device used. I suspect your 
signal is just simply too weak for reception. Small antenna coming with 
those DVB sticks is not suitable unless you are living very near 
transmitter. Try to connect it roof antenna. One thing that helps a lot 
is to attach small bundled antenna to outside window.

There is both dvbscan and scandvb in Fedora dvb-apps. It is not clear 
for me why two similar looking tools. Anyhow it is just scandvb which I 
found working one.


> So I just put it in ~/.mplayer:channels.conf
>
> Took me a while to figure out that mplayer wants:
>
> mplayer 'dvb://BBC ONE;(null)'
>
> rather than anything else - curious about the ';(null)' part.
>
> --------
> Playing dvb://BBC ONE;(null).
> dvb_tune Freq: 522000
> ERROR IN SETTING DMX_FILTER 9018 for fd 4: ERRNO: 22ERROR, COULDN'T SET CHANNEL  8: Failed to open dvb://BBC ONE;(null).
> ----------

Typical channels.conf entry looks like that:
MTV3:714000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_2_3:FEC_AUTO:QAM_64:TRANSMISSION_MODE_8K:GUARD_INTERVAL_1_8:HIERARCHY_NONE:305:561:49

And tuning to that channel using mplayer:
mplayer dvb://MTV3

However, I prefer VLC. Just open channels.conf to VLC and should show 
playlist. Totem does not work anymore. at least stream used here in 
Finland. It went broken when they changed from playbin to playbin2 which 
is really shame as it is default video player for Gnome desktop.


> At this point I am lost :-).

Not big surprise unfortunately :/

Unfortunately desktop integration is currently poor and most users are 
coming from the HTPC.

regards
Antti


-- 
http://palosaari.fi/

Re: media_build and Terratec Cinergy T Black.

[Hin-Tak Leung]

--- On Fri, 6/7/12, Antti Palosaari <crope@iki.fi> wrote:

> On 07/06/2012 01:54 PM, Hin-Tak Leung
> wrote:
> > - $ lsdvb seems to be doing garbage:(fedora 17's)
> > 
> > usb (-1975381336:62 64848224:32767) on PCI
> Domain:-1965359032 Bus:62 Device:64848416 Function:32767
> >     DEVICE:0 ADAPTER:0 FRONTEND:0
> (Realtek RTL2832 (DVB-T))
> >    
>      FE_OFDM Fmin=174MHz
> Fmax=862MHz
> > 
> > lsdvb on mercury is only marginally better with the PCI
> zero's, but the other numbers swapped:
> > 
> > usb (62:-1975379912 32767:-348245472) on PCI Domain:0
> Bus:0 Device:0 Function:0
> >     DEVICE:0 ADAPTER:0 FRONTEND:0
> (Realtek RTL2832 (DVB-T))
> >    
>      FE_OFDM Fmin=174MHz
> Fmax=862MHz
> 
> I was aware of that tool but didn't know it lists USB
> devices too.
> Someone should fix it working properly for USB devices.

The mercury repository is on linuxtv, so presumably one of you can do it :-). It is wierd that (1) those numbers are swapped between fc17's vs upstream, (2) the numbers also change between runs....
 
> > - 'scandvb' segfault at the end on its own.
> 
> I didn't see that.

This is fc17's - it does so in a string function (v*printf) - probably easy to fix if/when I get the debuginfo package, if it isn't fixed upstream already.

> > - "scandvb /usr/share/dvb/dvb-t/uk-SandyHeath"
> (supposedly where I am) got a few "WARNING: >>>
> tuning failed!!!" and no list.

This is where it gets confusing - /usr/share/dvb/dvb-t/* and  /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* have similiar names, but different content, and no explanation which is which. The latter seems to be in the correct format you wrote below, but does not work.

One thing would be nice to do would be write up a description of those formats, and say which is which.

> > - 'w_scan -G -c GB'
> >    have a few curious
> > WARNING: received garbage data: crc = 0xcc93876c;
> expected crc = 0xb81bb6c4
> > 
> > return a list of 26, with entries like (which seems to
> be vaguely correct):
> > 
> > BBC
> ONE;(null):522000:B8C23D0G32M64T8Y0:T:27500:101=2:102,106=eng:0:0:4173:9018:4173:0:100

w_scan is the only one which seems to be able to scan, but its output format isn't correct.

Any ideas about the garbage data message?

> Both scandvb and w_scan works here, same device used. I
> suspect your signal is just simply too weak for reception.
> Small antenna coming with those DVB sticks is not suitable
> unless you are living very near transmitter. Try to connect
> it roof antenna. One thing that helps a lot is to attach
> small bundled antenna to outside window.

I was using my satellite dish as antenna, BTW. Nothing with the small antenna.

> There is both dvbscan and scandvb in Fedora dvb-apps. It is
> not clear for me why two similar looking tools. Anyhow it is
> just scandvb which I found working one.

I just found a dvbv5-scan on my harddisk (fc17) also, and dvbscan is in locate.db but gone. Apparently one might be 'scan' but too confusing and got its name changed during packaging.


> > So I just put it in ~/.mplayer:channels.conf
> > 
> > Took me a while to figure out that mplayer wants:
> > 
> > mplayer 'dvb://BBC ONE;(null)'
> > 
> > rather than anything else - curious about the ';(null)'
> part.
> > 
> > --------
> > Playing dvb://BBC ONE;(null).
> > dvb_tune Freq: 522000
> > ERROR IN SETTING DMX_FILTER 9018 for fd 4: ERRNO:
> 22ERROR, COULDN'T SET CHANNEL  8: Failed to open
> dvb://BBC ONE;(null).
> > ----------
> 
> Typical channels.conf entry looks like that:
> MTV3:714000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_2_3:FEC_AUTO:QAM_64:TRANSMISSION_MODE_8K:GUARD_INTERVAL_1_8:HIERARCHY_NONE:305:561:49
> 
> And tuning to that channel using mplayer:
> mplayer dvb://MTV3

Well that at least clear up something - I tried this form (from  /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* ) but did not get anything either - the error message seemed worse so I didn't go further. I guess I should try getting w_scan to do this form.

> However, I prefer VLC. Just open channels.conf to VLC and
> should show playlist. Totem does not work anymore. at least
> stream used here in Finland. It went broken when they
> changed from playbin to playbin2 which is really shame as it
> is default video player for Gnome desktop.
> 
> 
> > At this point I am lost :-).
> 
> Not big surprise unfortunately :/
> 
> Unfortunately desktop integration is currently poor and most
> users are coming from the HTPC.

There seems to be at least two channels.conf formats (one for mplayer/vlc/gstreamer, one for vdr?), and unfortunately both seems to have the same name conventionally, but different content. I can't find documentation about either, or even examples :-).

I am also surprised by the checksum failed message with w_scan, and that it does not change output format with adding options, etc.

bugs in dvbscan/scan, dvb-apps(Re: media_build and Terratec Cinergy T Black.)

[Hin-Tak Leung]

--- On Fri, 6/7/12, Hin-Tak Leung <htl10@users.sourceforge.net> wrote:

<snipped>
> > > - 'scandvb' segfault at the end on its own.
> > 
> > I didn't see that.
> 
> This is fc17's - it does so in a string function (v*printf)
> - probably easy to fix if/when I get the debuginfo package,
> if it isn't fixed upstream already.

I got hold of the debuginfo, and found the bug - it is in bad_usage() in dvb-apps/util/scan/scan.c, line 2583 (case 2). The format strings contains two %s, but that statement only supply one:

   fprintf (stderr, usage, pname);

Who should I report this to? (not very good with mercury at the moment...).

running scandvb on its own is supposed to result in an input file listing then print usage; on fc17 it segfaults while printing usage; on a mercury source build, it fills the 2nd %s with some random other string from the program itself.

<snipped>
> > There is both dvbscan and scandvb in Fedora dvb-apps.
> It is
> > not clear for me why two similar looking tools. Anyhow
> it is
> > just scandvb which I found working one.
> 
> I just found a dvbv5-scan on my harddisk (fc17) also, and
> dvbscan is in locate.db but gone. Apparently one might be
> 'scan' but too confusing and got its name changed during
> packaging.

Found the reason how/why it was gone while I was looking for debuginfo (see change log entry):
http://koji.fedoraproject.org/koji/buildinfo?buildID=327654

The fedora packager withdrew dvbscan: "drop dvbscan as it's obsolete/broken" - changelog June 26.

success! (Re: media_build and Terratec Cinergy T Black.)

[Hin-Tak Leung]

--- On Fri, 6/7/12, Hin-Tak Leung <htl10@users.sourceforge.net> wrote:

<snipped>
> > Typical channels.conf entry looks like that:
> >
> MTV3:714000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_2_3:FEC_AUTO:QAM_64:TRANSMISSION_MODE_8K:GUARD_INTERVAL_1_8:HIERARCHY_NONE:305:561:49
> > 
> > And tuning to that channel using mplayer:
> > mplayer dvb://MTV3
> 
> Well that at least clear up something - I tried this form
> (from 
> /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* ) but
> did not get anything either - the error message seemed worse
> so I didn't go further. I guess I should try getting w_scan
> to do this form.

<snipped>
> There seems to be at least two channels.conf formats (one
> for mplayer/vlc/gstreamer, one for vdr?), and unfortunately
> both seems to have the same name conventionally, but
> different content. I can't find documentation about either,
> or even examples :-).

Apparently it was just me not reading the manual/options properly. There are 3 formats - gstreamer, vdr and mplayer's! I thought it was just vdr vs everything else. It was also confusing that mplayer did not complain about it.

scandvb still does not work at all, nor those sample config files under /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* , or use those (tried about 6 different transmitter config various distance away). It is also true that I have poor reception: all the BBC* channels seems to work okay - about 18 channels - but none of the non-BBC stations. Also they are not there on a 2nd scan, so I guess they are weaker.

So obviously there are a lot of rough edges. I also think it is a bit stupid for w_scan not to offer writing *all* the formats, since all the information should be there after a scan. It takes about 8 minutes to do a full scan. It would be more logical to generate all the channels.conf formats on one scan, and let the user throw away the ones he does not need, if it takes that long to do a full scan.

unload/unplugging (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Hin-Tak Leung]

BTW, I tried just pulling the USB stick out while mplayer is running. Strangely enough mplayer did not notice it gone and kept going for some 5 to 10 seconds. Probably buffering?

The only sign about it is two lines in dmesg (other than the usual usb messages about device being unplug).

[227690.953311] rtl2832: i2c rd failed=-19 reg=01 len=1
[227710.818089] usb 1-2: dvb_usbv2: streaming_ctrl() failed=-19

I also have quite a few :

[224773.229293] DVB: adapter 0 frontend 0 frequency 2 out of range (174000000..862000000)

This seems to come from running w_scan.

The kernel seems happy while having the device physically pulled out. But the kernel module does not like to be unloaded (modprobe -r) while mplayer is running, so we need to fix that.

Re: unload/unplugging (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Antti Palosaari]

Hello Hin-Tak

On 07/07/2012 05:26 AM, Hin-Tak Leung wrote:
> BTW, I tried just pulling the USB stick out while mplayer is running. Strangely enough mplayer did not notice it gone and kept going for some 5 to 10 seconds. Probably buffering?

yes

> The only sign about it is two lines in dmesg (other than the usual usb messages about device being unplug).
>
> [227690.953311] rtl2832: i2c rd failed=-19 reg=01 len=1
> [227710.818089] usb 1-2: dvb_usbv2: streaming_ctrl() failed=-19

Jul  7 12:40:42 localhost kernel: [  906.030829] usb 2-2: USB 
disconnect, device number 4
Jul  7 12:40:42 localhost kernel: [  906.035518] rtl2832: i2c rd 
failed=-19 reg=01 len=1

Next application kills (I think so) and closes file nodes => device is 
unloaded:

Jul  7 12:40:49 localhost kernel: [  912.751437] usb 2-2: dvb_usbv2: 
streaming_ctrl() failed=-19
Jul  7 12:40:49 localhost kernel: [  912.751981] usb 2-2: dvb_usbv2: 
'Terratec Cinergy T Stick Black' successfully deinitialized and disconnected


> I also have quite a few :
>
> [224773.229293] DVB: adapter 0 frontend 0 frequency 2 out of range (174000000..862000000)
>
> This seems to come from running w_scan.

yes, those warnings are coming when application request illegal 
frequency. Setting frequency as a 2 Hz is something totally wrong, wild 
guess, it is some other value set accidentally as frequency.


> The kernel seems happy while having the device physically pulled out. But the kernel module does not like to be unloaded (modprobe -r) while mplayer is running, so we need to fix that.

Yep, seems to refuse unload. I suspect it is refused since there is 
ongoing USB transmission as it streams video. But should we allow that? 
And is removing open device nodes OK as applications holds those?

regards
Antti


-- 
http://palosaari.fi/

Re: unload/unplugging (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Hin-Tak Leung]

--- On Sat, 7/7/12, Antti Palosaari <crope@iki.fi> wrote:

<snipped>
> > I also have quite a few :
> > 
> > [224773.229293] DVB: adapter 0 frontend 0 frequency 2
> out of range (174000000..862000000)
> > 
> > This seems to come from running w_scan.
> 
> yes, those warnings are coming when application request
> illegal frequency. Setting frequency as a 2 Hz is something
> totally wrong, wild guess, it is some other value set
> accidentally as frequency.

I am thinking either w_scan is doing something it should not, in which case we should inform its author to have this looked at, or the message does not need to be there?

> > The kernel seems happy while having the device
> physically pulled out. But the kernel module does not like
> to be unloaded (modprobe -r) while mplayer is running, so we
> need to fix that.
> 
> Yep, seems to refuse unload. I suspect it is refused since
> there is ongoing USB transmission as it streams video. But
> should we allow that? And is removing open device nodes OK
> as applications holds those?

I am thinking about suspend/resume, the poorman's way, which is to unload/reload. One interesting thing to try would be to pause but not quit the application - either just press pause, or say, 'gdb <mplayerbinary> <pid>', and see if 'modprobe -r' can be made to work under that sort of condition, if it isn't already.

Re: unload/unplugging (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Antti Palosaari]

On 07/07/2012 01:10 PM, Hin-Tak Leung wrote:
> --- On Sat, 7/7/12, Antti Palosaari <crope@iki.fi> wrote:
>
> <snipped>
>>> I also have quite a few :
>>>
>>> [224773.229293] DVB: adapter 0 frontend 0 frequency 2
>> out of range (174000000..862000000)
>>>
>>> This seems to come from running w_scan.
>>
>> yes, those warnings are coming when application request
>> illegal frequency. Setting frequency as a 2 Hz is something
>> totally wrong, wild guess, it is some other value set
>> accidentally as frequency.
>
> I am thinking either w_scan is doing something it should not, in which case we should inform its author to have this looked at, or the message does not need to be there?

As scandvb and all the other applications are able to set desired 
parameters without that error it must be w_scan issue.

And personally I don't care whole warning, returning some error code 
(which is likely -EINVAL) should be enough. It is not error situation in 
the mean of Kernel or device error - it is just user error as user tries 
to set unsupported frequency.

>>> The kernel seems happy while having the device
>> physically pulled out. But the kernel module does not like
>> to be unloaded (modprobe -r) while mplayer is running, so we
>> need to fix that.
>>
>> Yep, seems to refuse unload. I suspect it is refused since
>> there is ongoing USB transmission as it streams video. But
>> should we allow that? And is removing open device nodes OK
>> as applications holds those?
>
> I am thinking about suspend/resume, the poorman's way, which is to unload/reload. One interesting thing to try would be to pause but not quit the application - either just press pause, or say, 'gdb <mplayerbinary> <pid>', and see if 'modprobe -r' can be made to work under that sort of condition, if it isn't already.

hmm, what is that kind of suspend/resume?
Is that different what is now implemented?

regards
Antti

-- 
http://palosaari.fi/

Re: unload/unplugging (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Hin-Tak Leung]

--- On Sat, 7/7/12, Antti Palosaari <crope@iki.fi> wrote:

<snipped>
> > I am thinking either w_scan is doing something it
> should not, in which case we should inform its author to
> have this looked at, or the message does not need to be
> there?
> 
> As scandvb and all the other applications are able to set
> desired 
> parameters without that error it must be w_scan issue.

But scandvb does not work at all in my case (or rather, the bundled tuning files were wrong/out-dated).

> And personally I don't care whole warning, returning some
> error code 
> (which is likely -EINVAL) should be enough. It is not error
> situation in 
> the mean of Kernel or device error - it is just user error
> as user tries 
> to set unsupported frequency.

It is possibly just too much information in klog/dmesg .

> >>> The kernel seems happy while having the device
> >> physically pulled out. But the kernel module does
> not like
> >> to be unloaded (modprobe -r) while mplayer is
> running, so we
> >> need to fix that.
> >>
> >> Yep, seems to refuse unload. I suspect it is
> refused since
> >> there is ongoing USB transmission as it streams
> video. But
> >> should we allow that? And is removing open device
> nodes OK
> >> as applications holds those?
> >
> > I am thinking about suspend/resume, the poorman's way,
> which is to unload/reload. One interesting thing to try
> would be to pause but not quit the application - either just
> press pause, or say, 'gdb <mplayerbinary>
> <pid>', and see if 'modprobe -r' can be made to work
> under that sort of condition, if it isn't already.
> 
> hmm, what is that kind of suspend/resume?
> Is that different what is now implemented?

In the good old days, for drivers which does not suspend/hibernate well, one can add a file /etc/pm/config.d/myfile containing SUSPEND_MODULES="modname" to get them unloaded on suspend. Does it work with mplayer/vlc running or pausing?

Re: success! (Re: media_build and Terratec Cinergy T Black.)

[Antti Palosaari]

On 07/07/2012 05:11 AM, Hin-Tak Leung wrote:
> --- On Fri, 6/7/12, Hin-Tak Leung <htl10@users.sourceforge.net> wrote:
>
> <snipped>
>>> Typical channels.conf entry looks like that:
>>>
>> MTV3:714000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_2_3:FEC_AUTO:QAM_64:TRANSMISSION_MODE_8K:GUARD_INTERVAL_1_8:HIERARCHY_NONE:305:561:49
>>>
>>> And tuning to that channel using mplayer:
>>> mplayer dvb://MTV3
>>
>> Well that at least clear up something - I tried this form
>> (from
>> /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* ) but
>> did not get anything either - the error message seemed worse
>> so I didn't go further. I guess I should try getting w_scan
>> to do this form.
>
> <snipped>
>> There seems to be at least two channels.conf formats (one
>> for mplayer/vlc/gstreamer, one for vdr?), and unfortunately
>> both seems to have the same name conventionally, but
>> different content. I can't find documentation about either,
>> or even examples :-).
>
> Apparently it was just me not reading the manual/options properly. There are 3 formats - gstreamer, vdr and mplayer's! I thought it was just vdr vs everything else. It was also confusing that mplayer did not complain about it.
>
> scandvb still does not work at all, nor those sample config files under /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* , or use those (tried about 6 different transmitter config various distance away). It is also true that I have poor reception: all the BBC* channels seems to work okay - about 18 channels - but none of the non-BBC stations. Also they are not there on a 2nd scan, so I guess they are weaker.

UK is a little bit special case as it uses different channel raster. I 
just looked and there seems to be "auto-With167kHzOffsets" which could 
be used. It scans all the possible channels using auto parameter detection.

> So obviously there are a lot of rough edges. I also think it is a bit stupid for w_scan not to offer writing *all* the formats, since all the information should be there after a scan. It takes about 8 minutes to do a full scan. It would be more logical to generate all the channels.conf formats on one scan, and let the user throw away the ones he does not need, if it takes that long to do a full scan.

Yes there is. There is no many desktop applications supporting DVB and 
even less applications having own channels scanner.

regards
Antti

-- 
http://palosaari.fi/

channel scanning (Re: success! (Re: media_build and Terratec Cinergy T Black.))

[Hin-Tak Leung]

--- On Sat, 7/7/12, Antti Palosaari <crope@iki.fi> wrote:

<snipped>
> > scandvb still does not work at all, nor those sample
> config files under
> /usr/share/doc/dvb-apps-1.1.2/channels-conf/dvb-t/* , or use
> those (tried about 6 different transmitter config various
> distance away). It is also true that I have poor reception:
> all the BBC* channels seems to work okay - about 18 channels
> - but none of the non-BBC stations. Also they are not there
> on a 2nd scan, so I guess they are weaker.
> 
> UK is a little bit special case as it uses different channel
> raster. I 
> just looked and there seems to be "auto-With167kHzOffsets"
> which could 
> be used. It scans all the possible channels using auto
> parameter detection.

Those files (dvb-t/uk-{regions}) are not in the dvb-apps hg repo, so I don't know where they came from, but considered that they don't work for me at all (and w_scan does find some), their contents are either wrong or outdated.

The fedora package changelog possibly says tuning files are kept elsewhere and were last updated over 2 years ago - I know it is probably cliche to assume every *.fi person knows every other *.fi person, but do you? :-)

* Sat Jun 05 2010 Ville Skyttä <ville.skytta@iki.fi> - 1.1.1-22
- Patch to fix dvbnet -h crash (#597604).
- Update tuning files to 20100605.

> > So obviously there are a lot of rough edges. I also
> think it is a bit stupid for w_scan not to offer writing
> *all* the formats, since all the information should be there
> after a scan. It takes about 8 minutes to do a full scan. It
> would be more logical to generate all the channels.conf
> formats on one scan, and let the user throw away the ones he
> does not need, if it takes that long to do a full scan.
> 
> Yes there is. There is no many desktop applications
> supporting DVB and 
> even less applications having own channels scanner.

With a full scan taking 8 minutes (and quick "assisted" scan based on presets not working at all), channel-scanning can't really be modularized and built into an application. Actually even w_scan needs assistance about which *country*. 
Also a typical user might like to try a few different applications. I use mplayer-derivatives - mplayer, gmplayer, gnome-mplayer, dragonplayer - mostly because of its widest/earliest codec support, but I am not against using another if it offers more DVB-centric user-interface. I really think w_scan should by default write all variants of channels.conf formats, or a master-form convertable to others, just to save the 8 minutes of re-scan. (being able to convert quickly, or for application to standardize would be nice, but seeing as the formats have different info, that idea does not seem likely).