Puppy Linux Discussion Forum

[rjbrewer]

About ath5k:

http://murga-linux.com/puppy/viewtopic.php?search_id=763769231&t=47267

[JustGreg]

Thank you jjbrewer for point to the post by tempestuous. I had found it before.

But, I did find a solution! I added the kernel boot parameter, irqpoll to the kernel line. This eliminates the IRQ #11 disabled messages and the sound works normally. An alternate solution is to use acpi=off, but, this causes an incomplete power shutdown. The system only halts and one has to press the power button to completely turn off the system.

I found the information on irqpoll in Ubuntu post on the same problem. There are kernel boot parameters that do many things. The only way to find out about them is to read the kernel documentation.

For Puppy 4.3.1, my suggestion is the message from initrd.gz be change from "IRQ #11 disabled" to "IRQ #11 disabled try acpi=off kernel boot parameter". The "acpi=off" kernel boot parameter is documented on Puppy Linux wiki and other sources. The irqpoll parameter is not documented. This would help people when they encountered the problem. Pressing the power button to turn off the system at shutdown is a reasonable work around.

This whole problem is the result of Windows 2000 plug and pray, opps, I mean plug and play scheme. It uses irq 11 as general interrupt for a host of devices. I hope this helps.

31 January. I did some more research into kernel parameters. I found and tested acpi=noirq. This appears to be a better way to solve this rather than irqpoll.

[rjbrewer]

JustGreg:

To have it poweroff cleanly; add acpi=force, to the end of the
kernel line in /boot/grub/menu.lst.

[ecomoney]

Thanks technosaurus, but ecopup has been modified to use a "eco_save.2fs", so this cant be the reason. There was no pup_save.2fs on the hard disk at all when I booted it.

[JustGreg]

Thanks again, for the advice and help, jdbrewer. I tried the boot parameter acpi=force and it did not work. I got the same with IRQ #11 being disabled. Only boot parameters acpi=off and irqpoll seem to work. But, I do thank you for help.

It seems the best approach is acpi=noirq

[rerwin]

The module/firmware loader is executed by multiple concurrent processes, but utilizes its resources as if run by processes one-at-a-time. Multi-threading requires precautions to avoid concurrent access and updating of files, which occurs in the loader. The modem detection and initialization processes are subject to these vulnerabilities, and can function unreliably as a consequence. Furthermore, the module override and preference functions omit some capabilities needed to select modules predictably. The attached packages correct these and other issues.

The fix is provided in two packages, so that the modem fixes can be omitted if users do not have dialup modems, and so that the packages can be updated independently. The packages are:
- pup_event_backend_modprobe and its new resource-serialization service
- modem "fix pack" that reliably integrates the Conexant drivers and upgrades wireless modem coverage, as well as usb_modeswitching.

These packages are intended for application to puppy 4.3.1 and to be incorporated into puppy 4.4 and all future versions of Puppy Linux. I also provide a new "3G-pupdial-13" package containing the modem updates for puppies 4.1.2 and 4.2.1 (but not the modprobe fix). Together, they increase Puppy's robustness, especially in unusual configurations. They were developed on my 400 MHz Aptiva, which exaggerates the impact of multi-threading and long-running firmware installations (e.g., the Conexant drivers). Most users should notice little difference, although those experiencing erratic modem or network-module detection should see that corrected. I also tested on a dual 466 MHZ PC that froze with the DGC USB modem plugged in, before I installed the packages; with them installed, the modem behaved normally, and no freeze!

The packages contain many code changes, so could cause some regression errors. Please install them to fresh Puppy installations or those that have been backed up, in case we get an ugly surprise. I have worked on this for three months, including extensive testing, to make it solid. Please report (via PM, to avoid hijacking the thread) any regression behavior you encounter. Barry and technosaurus might wait a while before integrating them, to allow users to "shake it out."

I will give more details in the next posting.
Richard

UPDATE 2/20/2010: The new versions correct the lock function and change how usb_modeswitch is used, so that modem initialization waits for it. Be sure to uninstall the previous versions and the modemprobe_erase-1 fix before installing these packages.. I recommend these packages especially for anyone running on a multi-processor or multi-core-processor system, to avoid the risk of a "hang" during initialization..

UPDATE 3/2/2010: Replaced the packages with minor updates. Adds analog modem script utilization of the new way to verify a module is loaded, all modem script use the "protect --modcheck" function. re-implements prevention of invalid dialup probe results after modem replacement. These are added for completeness and consistency in the modem init scripts. The "protect log" entries time field is corrected. Since there is little functional change, there is no need to reinstall the packages -- but new downloads of them will contain the latest developments. I expect to post a delta dotpet later in this thread, mainly for Barry to use to update his copy.

UPDATE 4/6/2010: Replaced the packages -- pup_event to support added-module loading, modem-fix to upgrade to the "automated" feature of usb_modeswitch, to eliminate the need for the multiple, editable rules files (51-) and to expand the models supported. The directory, /etc/usb_modeswitch.d, contains individual configuration files and replaces the old usb_modeswitch.conf global configuration file. To address cases where wireless modems (ttyUSB?) need a device name other than ttyUSB0, there is a fix to find the two most appropriate names, select one, and allow the user to change to the other in pupdial (with the "ttyUSB-" button, the name showing the alternate name to try). As this has had minimal user-testing, please report whether it works.

NOTE: These two packages contain all of my modem and modprobe-serialization fixes to be applied to puppy 4.3.1 -- except for a package for "mwave" modems. It was left out because of the size of that package and because it is rarely needed. However, it can be downloaded from this link (to the download on page 17): http://www.murga-linux.com/puppy/viewtopic.php?mode=attach&id=26909

UPDATE 4/15/10: Attached separate package for the usb_modeswitch binary for inclusion in projects, so it is identified in the package lists.

UPDATE 5/16/10: I have attempted to complete the scope of Puppy's support of ttyUSBn wireless modems, by adding the Qualcom (qcserial) and TI (ti_usb_3410_5052) initialization scripts and minor supporting changes. I have also corrected the device name for PCMCIA modems that relate to ttyIPWpn devices, using the ipwireless driver module. The attachment can be installed onto fix_pack 6, puppeee, quirky, wary, and lupu and anything else built with Woof as of this date.

UPDATE 5/28/10: I have updated the modem_ti_qc_ipwi_updates dotpet to add a minor correction to the martian firmware, for inclusion in Woof. I also added a file specific to Virgin (Australia) connections, to insert the "-chap" option automatically. If you do not need this option, there is no need to replace "-1".

[rerwin]

Some new dialup modem features to note:
1. The Riptide sound-and-modem card is supported so that the modem part is disabled, because the driver cannot handle it. This allows addition of another modem, without necessitating a complicated user workaround.

2. An attempt is made to support Vodaphone modems not detected in the normal way. Smokey01, who has been working with me via PMs, reports that this now works!.

3. All three Conexant drivers were tested concurrently, so that a PCI and USB modem can co-habit without interference, although installation of both PCI types (HCF, HSF) concurrently seems to cause the HSF modem to lock up Puppy when attempting a connection - better to avoid that combination.

4. Some built-in modems are associated with HSF, ALSA and SmartLink drivers, defaulting to the wrong driver. The HSF driver gets detected but is later found to not be a Conexant modem. In this case, Puppy avoids detecting the HSF and sets things up so that the ALSA driver is detected on the next boot-up.

5. A new udev rule supports some phone modems with hardware IDS 15eb:0001. This is incomplete and requires user feedback for further development. See comments within file /etc/udev/rules.d/50-modprobe-usbserial.rules. Some of the modems are:
- Haier D1200P USB Modem
- BSNL Clarity USB phone modem

6. USB-modem mode switching is upgraded to utilize the automated/integrated feature that discriminates better among modem models and improves the detection of actual mode switching. The old (51-) rules files (and editing of them) are eliminated. Modems not already supported may be added to rules files 51-usb_modeswitch-puppy.rules and 50-modprobe-option.rules.

7. Handling of device names ttyUSB0+ is changed to find the two most appropriate such names and to allow the user to change names with the pupdial "ttyUSB-" (Test) button. The name shown is the name that will be tested and should be different from the name of the "detected" device (assuming two are found).

The module-loading controls for overrides and preferences have been extended, to minimize addition of special coding in backend_modprobe, keeping most substitutions visible in the MODULESCONFIG file:
- PCI overrides can now include a module name of "(none)", which means to avoid loading a module for a particular hardware ID, more precise blacklisting. (This is used to disable the Riptide modem.)
- Preferences can be a bar-separated series of substitute modules (e.g., usb_storage:option|sierra). The result depends on the modules available for a particular hardware ID. It indicates priority in case multiple substitutes are present. (This is used for hybrid modems to ensure both usb_storage and a modem driver are loaded.)
- A dynamically added set of preferences can be activated by software. (This is used to reverse preferences for sound-card modems supported by either an ALSA or HSF module, when the HSF driver indicates that it is wrong for the modem.)

To demonstrate what the modprobe package does, here is its log for my Aptiva with an HSF PCI modem and a DGC USB modem installed. Note that it takes about 12 extra seconds before the firmware is loaded! Now Puppy waits until the firmware is loaded before attempting to select/detect a modem.

[smokey01]

rerwin,

I thought I should do one more test.

I booted puppy with pfix=ram
installed the modem_fix_pack_to_431-1.pet file
plugged in the modem
ran pupdial
Modem was found
modified the parameters as per screenshot.
Works great.

This setup is for Vodafone prepaid in Australia.

Smokey

[sindi]

I tried viewnior, inklite, mtpaint, seamonkey, then downloaded gpicview and fotoxx and gqview which don't do it either. Or links2 (graphical).

p?m is a standard linux image format, viewed by zgv and I think xli, xv, not available as pet packages. Used by netpbm, which also has a small viewer. gimp would take 20 min to download. What else might I try?

Thanks for setting up a default pdf viewer in 4.3.1. And gif jpeg tiff.

[nassausky]

There are alot of threads about puppy hanging either during an unexpected power loss by user intervention or electrical issues including the reboot command (which seems irregular that we should be using an odd command wmreboot from the terminal session).

This brought out another bug which I saw USB keyboard is able to select the booting menu.lst option yet the USB driver is disabled during this unexpected or forced shutdown/reboot procedure.

There are some ideas listed in the below threads and I hope this gets fixed because it's a major issue with installing this operating system on clients machines with remote control. Heavens forbid I lose control when a client is way too far to have physical access to the machine. Ouch the thought hurts.. lol

Puppy 4.1 Boot Hangs
http://www.murga-linux.com/puppy/viewtopic.php?t=34738

Puppy Freezes after 'reboot' from Terminal session window
http://www.murga-linux.com/puppy/viewtopic.php?t=21726

Make Puppy recover automatically from improper shutdowns
http://www.murga-linux.com/puppy/viewtopic.php?t=47668

reboot command locks up puppy
http://www.murga-linux.com/puppy/viewtopic.php?t=52149

[ecomoney]

Yes, an important bug to get fixed! I know that feeling nassausky

Another one to report...although Abiword in 4.3.1 is capable of opening openoffice .odt files, they are not "associated" with abiword, so clicking on them in ROX does not open them automatically in Abiword. This is something to do with "Mime Types" I believe.

[lvds]

Hi everyone,

While making some tests this week I noticed that PFIX=nox have
no effect. I was willing to set a boot loader to get me fast into a
linux console but puppy always brings me to graphical manager
to choose from xorg-xvesa and etc...

Don't know if someone else noticed it already

Best regards
Laurent

[Abnormalter]

[MeKino]

Hi,

As much as I like most of what puppy can do there are a couple of bugs which prevent it being the operating system of choice.

1. vlc media player crashes when trying to play mkv files. I can play these with no problems in ubuntu. I need vlc player because I can use it to record tv programmes.

2. Abiword is incredibly SLOW! I thought that being in ram it would be just the opposite. Also, clicking on "create & modify styles" causes abiword to crash. (On a related note, maybe, Gnumeric does not suffer from the slowness of abiword)

I am using puppy4.3.1 and have checked the file integrity so I'm sure that everything is installed ok.

Kino

[lvds]

...mmh and PFILE=no has no effect neither, puppy keeps on asking
me if I want to save a file or no

Best regards
Laurent