LCDProc 0.5.4-dev

fmertz

Folks,

I would like to kick off the effort to bring the LED support into the driver again. We have some support already, but only for the box I own (the X-Core-e). I was hoping folks with the other models could run a command to help me identify the EXACT ICH we need to code for. Best I can figure, this command is already in pfSense and should be run as root:

pciconf -r pci0:31:0 0:256

This command reads the PCI configuration area (256 bytes) for the Low Pin count (LPC) device. The LPC device does GPIO, and can control the LEDs. Based on the exact device id, I can look up the spec, and find out the offset for GPIO base register, etc.

I would like the output of the command, for the X-Core and X-Peak models. The key is the first 8 digits, the last 4 being 8086, Intel's vendor ID. Thanks.

stephenw10

X-Peak:


[2.0.1-RELEASE][root@pfsense.fire.box]/root(7): pciconf -r pci0:31:0 0:256
25a18086 0280000f 06010002 00800000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000401 00000000 00000000 00000000
00000000 00000000 00000481 00000010
050a0c0b 000000d0 09808080 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
000054d5 00000000 00000000 00000000
00000220 00000000 0000000d 00000300
00000000 00000000 05415555 00000000
00000000 00000000 00000000 00000000
00002186 00000f02 00000004 00000000
c0000000 34040000 00112233 45670291
00e40000 00000000 00020f66 00010000
ffffffff

I'll have to try your driver without priority setting and see what happens.

Steve

Edit: You were correct. The driver with priority removed can be set to nice 20. I am testing now with several scrolling screens to see if I can still access the box.

mdima

Hello everybody,
I am making some tests to improve the stability of the package from the "client side". Until now from what I read all the tries have been made on the binary package and the driver, I think that maybe also a little help from the client can solve some problems.

I am testing this changes on my boxes and I find no problems, so I would like to share this changes with you.

The changes are:

Added a 20ms delay between each command sent from the client to LCDproc.
Better managed errors. Now the client resets the error counter every successful communication session with LCDproc (before was a global counter). The error counter is managed inside the client (lcdproc_client.php).
Because of the above change, now the "client script" (lcdclient.sh) do not cycle anymore.

I hope at least some of the problems will be solved… I wait for your feedback. The new version is XXX.0.9.1.

Thanks,
Michele

mdima

… and we are with 0.9.2.

I didn't realize that there were some clients pending, that with the new error counter management could work behind. So now all the lcdproc_client.php processes are killed during the package resync.

Sorry for the people that was already upgrading do 0.9.1...

stephenw10

Looks good. :)

Now the error counter is in the php script (much better) why bother having lcdclient.sh at all?
Just call the php client from the rc file directly.

Also I have been running fmertz's driver he removed real time priority from with Nice level 20. Doing this allows the box to remain responsive even when some error event occours. Since the LCD is really of no importance compared to the firewall functions it seems better to run it at minimum priority. E.g.


$start .= "\t/usr/bin/nice -20 /usr/local/sbin/LCDd -c ". LCDPROC_CONFIG ."\n";
$start .= "\t/usr/bin/nice -20 ". LCDPROC_CLIENT ." &\n";

Steve

mdima

@stephenw10:

Looks good. :)

Thanks!

@stephenw10:

Now the error counter is in the php script (much better) why bother having lcdclient.sh at all?
Just call the php client from the rc file directly.

makes sense…

@stephenw10:

Also I have been running fmertz's driver he removed real time priority from with Nice level 20. Doing this allows the box to remain responsive even when some error event occours. Since the LCD is really of no importance compared to the firewall functions it seems better to run it at minimum priority. E.g.
$start .= "\t/usr/bin/nice -20 /usr/local/sbin/LCDd -c ". LCDPROC_CONFIG ."\n";
$start .= "\t/usr/bin/nice -20 ". LCDPROC_CLIENT ." &\n";

also this makes sense. Just we have to consider if this influences negatively the client or LCDd to refresh the data or communicate with the panel, but it's just a matter of tuning…

fmertz

Folks, the SDEC driver for Fireboxes is now officially part of the upstream lcdproc project! I received confirmation this morning that my code submission was committed. I guess it should come to pfSense as part of the package when the project leaders decide the current development branch is stable enough for release.

Now, on to support for the LEDs…

mdima

That's great! So the next version of LCDproc will have the SDEC driver… fmerz, consider that the compiling option in pfSense is "WITH_USB=true" only, I don't know if with this option the SDEC driver will be compiled...

BTW, how's going with the new package version? I found out it's more stable, but anyway still give some problems... what do you think?

Thanks,
Michele

stephenw10

Excellent work! ;D

Steve

fmertz

@stephenw10:

X-Peak:


[2.0.1-RELEASE][root@pfsense.fire.box]/root(7): pciconf -r pci0:31:0 0:256
25a18086

Device ID 25a1 is a 6300ESB, data sheet: http://ark.intel.com/products/27663/Intel-6300ESB-IO-Controller

For the X-Peak, the LEDs are on GPIO pins 40 and 41. This is part of the second set of pins, so there no blink support in hardware. We already knew this…

Anyone with an X-Core?

tix

@fmertz:

I would like the output of the command, for the X-Core and X-Peak models. The key is the first 8 digits, the last 4 being 8086, Intel's vendor ID. Thanks.

X-Core (x700)


/root(1): pciconf -r pci0:31:0 0:256
24408086 0280000f 06010005 00800000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00004001 00000010 00000000 00000000
00000000 00000000 00004081 00000010
09060b0c 000000d0 0a058003 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00005475 00000000 00000000 00000000
00000200 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00004004 00000000 00000000
00002002 00001f02 00000004 00000000
c0000010 14050000 00112233 45670291
017c000f 00000000 00000f47 00000200
ffffffff

I'm using the WGXepc script and it works flawlessly!

fmertz

@tix:

X-Core (x700)


/root(1): pciconf -r pci0:31:0 0:256
24408086

2440 is an 82801BA, Intel ICH2, datasheet here: http://www.intel.com/content/dam/doc/datasheet/82801ba-i-o-controller-hub-2-82801bam-i-o-controller-hub-2-mobile-datasheet.pdf

Somehow the existing WGXepc code does not seem to line up with the spec…

mdima

Hi,
how's going with the latest (0.9.2) package?

Right now I am working on this:

I divided / 2 the number of commands sent to LCDproc every cycle. So if before were sent 10 commands, now only 5
I wrote a little better code for error handling
I slowed down a bit the scrolling (just a little bit)
Simplified the script to launch. There's no more lcdproc_client.sh.

Then I will:
4) add and test the "nice" command to the programs running
5) I will add a "top value" for waiting to the next cycle of 10 seconds. So it will be sure that the communication won't timeout

What else?

Ciao,
Michele

stephenw10

@fmertz:

Somehow the existing WGXepc code does not seem to line up with the spec…

Hmm, in what way?
There was a problem because the GPIO base is set to a non-standard value on the X-core.
On the other boxes it is at 0x480, which I believe is the standard value where as the X-Core is at 0x4080.

Steve

fmertz

@stephenw10:

@fmertz:

Somehow the existing WGXepc code does not seem to line up with the spec…

Hmm, in what way?
There was a problem because the GPIO base is set to a non-standard value on the X-core.
On the other boxes it is at 0x480, which I believe is the standard value where as the X-Core is at 0x4080.

Steve

The way I read it, this particular device ID is an ICH2, and has the GPIO base port stored at offset 0x58 (page 9-1). Fine. The value at that offset happens to be 4081, which, masked off is 4080 (this was listed in the pciconf command, and is already in your code). Fine. In your code, you use port 4080 + 0x0f. Offset 15 in the GPIO area is for GPIO level, and bits control pins. Fine. The blink register is at offset 0x18 (32 bit), so I am not sure how it can blink with just the level register, or how it is not possible to turn the LEDs off…

For the X-Core-e, the Level port controls on/off, the Blink port controls the, well, blinking, and it all can be turned off with putting the bit to zero.

stephenw10

I had to re-read the thread to remind myself what happened. I agree with you it doesn't make any sense! In fact you can see, here, where I was expecting the registers to work as you describe above but the experimental results show otherwise.

One thing that is different about the X-core is that you can set the LED state in the BIOS (though I've never tried it myself). And you have a choice of red or green and fast or slow flash. There is no fast/slow blink described in the documentation yet it is clearly available. :-\

Steve

tix

Hello

I've tried the latest .92 package and my x-core will still become unresponsive at the 10 hour mark.

I'm out of ideas on what the issue could be… :(

mdima

@tix:

Hello

I've tried the latest .92 package and my x-core will still become unresponsive at the 10 hour mark.

I'm out of ideas on what the issue could be… :(

Hi Tix,
Thank you for the feedback. I am trying a new version with a lot of changes, if it passes my tests I will publish it… Cross your fingers...

Ciao,
Michele

fmertz

@mdima:

@tix:

Hello

I've tried the latest .92 package and my x-core will still become unresponsive at the 10 hour mark.

I'm out of ideas on what the issue could be… :(

Hi Tix,
Thank you for the feedback. I am trying a new version with a lot of changes, if it passes my tests I will publish it… Cross your fingers...

Ciao,
Michele

Shot in the dark: is the client reading the responses back from LCDd? The release notes mentioned something about ignoring them causing unpleasant behavior …

Cino

I'm noticing a bug that I haven't seen since before the re-write of the package. If my second WAN does down (3G wireless), and comes back up; lcdclient seems to lose the connection to LCDd. In the past I would have pages of errors in the system log. but i dont have any this time. I do want to say this started with the last changes.

Still want to see if I can reproduce this on the fly (maybe tin cup around the 3G Wireless USB stick) but wanted to report it

Edit: I think this may have been a fluke… my 3g went down a couple of times last night and no issues with lcdproc