Firefox for eCS (OS/2) 24.8.1 Beta2 using very high CPU.

[ivan]

The latest firefox has the same problem as the previous version, namely very high CPU usage and plugin-container.exe tending to stay open even when firefox is closed.

It did this on my 4 core and 6 core processor machines as well as our dual 6 core processor test server, pushing CPU usage to well over 90% on all cores.

Note I said DID, it now behaves itself and is working as I expect it to for the last 7 hours.

To fix this problem you need to go to about:config and look for dom.ipc.plugins.enabled and toggle true to false - you can also toggle any other true statements in that dom.ipc group to false.

Since firefox 17 also uses plugin-container.exe this fix should work for those versions as well.
 

[Dave Yeo]

Out of curiousity, what plugins do you have installed? Enter about:plugins in the URL bar.
BTW for 17.0.11 I made your fix the default but Dmik only disables plug-container.exe for Flash

[ivan]

Hi Dave,

The only plugin on any of these machines is flash - which doesn't work - cloned mozprofile.

I did notice that the flash plugin was set to false but the plugin-container.exe still loaded and swamped the CPU.  With all of the true config options toggled to false there is no problem.

Next week I will let it loose in the office and see what response it gets.

BTW, what is the plugin-container.exe actually for and does it even have a purpose in firefox for OS/2?
 

[Doug Bissett]

I think you must have a problem caused by something other than Firefox and FLASH. I made a WarpIn installer for firefox-24.8.1.en-US.os2.zip, and installed it that way (now at HOBBES and NETLABS), but that should be identical to installing using the ZIP file to a clean folder. I do use RUN! to start Firefox.

I am not seeing what you describe, and I have not changed any of the defaults. FLASH works fine, for FLASH that will accept flash_ecs_0_4_2_GA.exe and odin32bin-0_8_9.wpi (from the eCS betazone). It seems that some web sites don't like our version of FLASH, and insist that it needs to be updated, so they don't even try to play the FLASH (nothing new there). The only other plugin, that I have, is from IcedTea-Web_1-3b.wpi.

I am using an Asus M3A78-EM motherboard, with quad core Phenom processor. It has 5 GB of memory (the usual amount available to eCS, and the rest is used by the QSINIT RAMDISK). I am currently using eCS 2.1, but I am pretty sure that it also works with eCS 2.2. Everything is updated to the latest (I think).

[Dave Yeo]

Hi Ivan. did you double check that no other plugins are loaded by entering about:plugins? As Doug mentioned, the default should work for Flash. I had a problem with the awget plugin also loading plugin-container.exe and so had to disable it for everything.
Plugin-container.exe is for running some plugins out of process, so Flash runs in a separate process and if it crashes, as it does on all platforms, it just takes out the tab it is running in rather then the whole browser.

[ivan]

Hi Dave,  about:plugins just shows shockwave flash (it is the old version from the days when flash did work v9.0).

Just to see if there was/is anything odd in the systems I have here I replaced the hard disks on the Asus M4A78LT-M with the 4 core AMD Phenom processor with a new 500GB Hitachi disk.  I then installed OS/2 (no eCS here - our clients don't use it) CP2.  With the basic OS/2 running I added the necessary DLLs, the latest ODIN, XWP and warpin.  After adding the set mozilla_home and set moz_plugin_path to config.sys and rebooting I then unzipped firefox-24-8-1 into its own directory

Starting firefox from a command window and going to a known PITA site (Daily Mail) there are spikes on the CPU monitor for each core going to about 80%, exit firefox plugin-container still running evidenced by the command window not opening. 

Restart firefox, open about:config and toggle dom.ipc.plugins.enabled to false.  Open daily mail page, a few 30% spikes as page loads and viewing is normal.  Exit firefox, command window opens almost immediately. 

For me setting the dom.ipc.plugins.enabled to false works and that is all I ask.  As far as our clients are concerned flash is a potential attack vector on their systems so turning this off is 'a good thing' as far as they are concerned.  For me, if I want to watch a flash something I download it and use VLC.

[Doug Bissett]

Hi Dave,  about:plugins just shows shockwave flash (it is the old version from the days when flash did work v9.0).

Well, that explains why it doesn't work. It probably explains the spikes, and the high CPU usage too. Either update FLASH, get rid of FLASH, or use an older version of FF that will support the antique version that probably won't work anyway, in a lot of cases.

[guzzi]

The Flash isn't trouble free either... It causes my system to reboot like if I did a CAD. The latest version doesn't always do that, but reboot may still occur even after a flash video has played. And it's not something that can be debugged(

[Dave Yeo]

Hi Guzzi, that almost sounds like a hardware problem. Possibly heat. Have you looked in your computer lately for dust bunnies? Ran a memory checker?

[guzzi]

Hi Guzzi, that almost sounds like a hardware problem. Possibly heat. Have you looked in your computer lately for dust bunnies? Ran a memory checker?

Yes, did all that, even replaced the memory just to be sure. There are only two events that will trigger such a reboot. Flash will always do it, and very very seldom a combination of firefox being up very long, several other programs running and openoffice opened and closed a few times might trigger a reboot. It might be  a hardware problem, but not 'faulty' hardware, I think rather something specific to my hardware that causes the kernel to panic when certain memory addresses are used. Some form of memory corruption perhaps. I should try if with Panorama instead of Snap the problem goes away but unfortunately I can't set the proper resolution with panorama (it will not 'see' all the resolutions the monitor supports as if it can't properly read the ddc information). I'm just guessing, of course. If I would log via com port I might get some useful debugging info for someone knowledgeable to look at, but since I am the only one who has reported this issue I am loath to have have scarce manpower spent on this issue.

[Jan-Erik Lärka]

Hello,

both my AMD Phenom Quad Core computer and my Sony Vaio Intel i3 reboot like that.
Everything goes black instantly and then BIOS info show up as it reboot.
It happen with JAVA applications such as JDownloader as well.

[Dave Yeo]

What is your VIRTUALADDRESSLIMIT set to? If higher then 1536 you might try lowering it, if lower raising it.
Snap could also be the problem if you're using more then 1 core as parts aren't SMP safe depending on how your BIOS sets things up.

[Doug Bissett]

Everything goes black instantly and then BIOS info show up as it reboot.

Actually, I saw something similar (it just went black, no reboot until I did CAD twice), once, when I downloaded a nightly version of what got released in firefox-24.8.1.en-US.os2.zip. At the time, I didn't relate it to FF, because FF had been closed for about 20 minutes. Later, I was wondering if FF had something to so with it because that was the only thing that had changed for a few days. After I rebooted, the "official" release was available before I tried it again, so I got rid of the nightly, and have been using 24.8.1 (how can you tell if it is Beta2?) since then (a few days), with no trouble like that (yet). I have had a couple of cases where other programs (Open Office) would not start, until I closed FF.

I have not seen the high usage problem. I do use SNAP with a quad core AMD Phenom processor, and I have used FLASH a couple of times. VIRTUALADDRESSLIMIT is set to 1536. I also have AdBlock and FLASHBLOCK installed. FLASHBLOCK may be the answer, because it prevents trying to run a lot of FLASH content concurrently.

I am also using the QSINIT loader, which may be setting the video properly, before eCS gets booted. I also use the QSINIT RAM disk, which makes use of the memory above what eCS can normally use, and I use "SET SHELLHANDLESINC=40", which may make some difference (20 is default).

[guzzi]

Hello,

both my AMD Phenom Quad Core computer and my Sony Vaio Intel i3 reboot like that.
Everything goes black instantly and then BIOS info show up as it reboot.
It happen with JAVA applications such as JDownloader as well.

Does it do a 'clean' reboot, like CAD does, i.e. the filesystem is clean on reboot like here?

[Andy Willis]

Hello,

both my AMD Phenom Quad Core computer and my Sony Vaio Intel i3 reboot like that.
Everything goes black instantly and then BIOS info show up as it reboot.
It happen with JAVA applications such as JDownloader as well.

Does it do a 'clean' reboot, like CAD does, i.e. the filesystem is clean on reboot like here?

I can replicate the reboot with apps such as ae.exe, newview.exe and I think tcpconfig.  These were created with Sybyl or WDSybyl (tcpconfig was with the latter but I am not sure if I had it reboot or just hang the system hard which I have had the previous two do as well).  I mostly see it with ae.exe and it started some years ago (maybe when I started using the SMP kernel).  I tried cleaning the fans, changing memory sticks, etc.  Finally I swapped the HDD into another machine and it still had the problem.  Later I put a clean eCS installation on the new machine with a new HDD and still had the problem.  The other day I put a clean installation on yet another machine and and had the problem on it as well.  I opened a ticket some months ago now.  I may be able to open ae.exe 30 times with no problem or it may reboot two launches in a row.  newview I have less problems with but then I don't launch it very often so it may be just as often.