Network lockups if low memory condition

[whose]

Hi,

I didnt check yet, if this is a platform specific issue, so I post it here.

I experienced network lockups when I played around a bit with the pager. I filled up my RAM Disk until the pager kicks in. After that, most programs work still fine (although a bit slow ), but as soon as I try to load any web page using the network connection, my machine locks up after short time. This is independend of the browser used. Any browser is able to load a certain (and small) amount of data, then the transmission stops and after a dozen seconds the system freezes completely. Hard reset needed.

I suspect RoadShow being the bad boy, as its possible to load web page from disk without any problems (I filled up my RAM Disk with e mere GB of data, and loading a web page from disk still works, although a bit slow ).

Anybody here able to reproduce it? I use a MicroA1 with onboard ethernet, but I would like to know if this happens for other machines, too.

[ChrisH]

I was under the impression the VM paging was broken by Update 3, such that OS4 tends to lock-up as soon as it runs out of real memory. So this may be nothing to do with Roadshow...

[whose]

Hmm... maybe the lockups will "wait" several minutes then... I didnt notice any lockups here, as long as I dont do any network/usb transfers.

I recheck now, and I do several actions with "more than filled up" RAM Disk (256MB physical RAM, RAM Disk holding a mere 400MB of data), still no lockups. Even compiling a small project using the latest CodeBench works fine (although a bit slow ). Ok, CodeBench has problems with low memory condition, too ("corruption of projects dictionary" requester, Grim on end), but its beta...

Even Wordworth7 is running fine, including document loading and printing (!) via a1parallel.device, although all the conditions above still apply (I do realtime testing atm., using a windows machine to post here). Added MUI-OWB (without retrieving any data from my network, that is), system is still running stable. I quit all the programs now, and the pager/memory subsystem gives back free 80MB RAM and system is still running fine. Seems that it works ok. I do another test atm., running GLQuake... still RAM Disk is filled with more data than physical memory available, Quake runs fine. Playing MP3 from a directory located in RAM Disk works, too.

Well, I dont think that the lockups I experienced originate from the pager itself

As Im no betatester yet, I dont know much about problems of the pager at all, but for me it works amazingly good (and quite fast, program startup times are near to normal, just a bit slower). Except for network traffic, which will cause _reproducable_ lockups with low memory condition. If my machine would have been a bit more stable with IDE (onboard), I would check for the network device (onboard) by replacing it with a Realtek one, but as IDE (onboard) is _utterly_ unstable, I dont think that it would help much in finding the cause of the lockups... thats why I asked for other people trying the pager with their machines (ideally their machines should be no MicroA1).

Maybe I should add that first I experienced problems with the OS itself, too. But this was due to incorrect preparation of my SWAP partition. After I rechecked and fixed it, the pager works fine and the system is working quite good.

Edit: Oh... forgot mentioning that I use a SII0680 as the onboard-IDE replacement.

[ssolie]

I was under the impression the VM paging was broken by Update 3, such that OS4 tends to lock-up as soon as it runs out of real memory. So this may be nothing to do with Roadshow...

That is incorrect. The paging was not broken by update 3 but rather fixed by update 3. If it seemed to "work" before it didn't because it was broken.

[ChrisH]

That is incorrect. The paging was not broken by update 3 but rather fixed by update 3. If it seemed to "work" before it didn't because it was broken.

I'm afraid that's not making any sense to me: If I over-fill the RAM disk with large files, then OS4 locks-up. Others also reported finding the same. Are you saying that OS4 does not work with it's own VM paging system?

[whose]

That is incorrect. The paging was not broken by update 3 but rather fixed by update 3. If it seemed to "work" before it didn't because it was broken.

I'm afraid that's not making any sense to me: If I over-fill the RAM disk with large files, then OS4 locks-up. Others also reported finding the same. Are you saying that OS4 does not work with it's own VM paging system?

The question is: What means "large files"? Larger than the physical memory? As I said, I expirienced no problems so far, except the network lockups.

[Spectre660]

Maybe this is an issue with the ram disk rather than VM paging.

I was under the impression the VM paging was broken by Update 3, such that OS4 tends to lock-up as soon as it runs out of real memory. So this may be nothing to do with Roadshow...

That is incorrect. The paging was not broken by update 3 but rather fixed by update 3. If it seemed to "work" before it didn't because it was broken.

[ssolie]

That is incorrect. The paging was not broken by update 3 but rather fixed by update 3. If it seemed to "work" before it didn't because it was broken.

I'm afraid that's not making any sense to me: If I over-fill the RAM disk with large files, then OS4 locks-up. Others also reported finding the same. Are you saying that OS4 does not work with it's own VM paging system?

All I am saying that paging was not broken by update 3 so you heard wrong.

[ChrisH]

All I am saying that paging was not broken by update 3 so you heard wrong.

I am not at all sure what you are now trying to say: When I said that Update 3 broke the VM system, it was based upon my own experience & that of others. I did not "hear" anything. Me & others might be wrong in claiming the VM system was broken by Update 3 (maybe the Ram Disk handler is broken?), but that doesn't seem to be what you are saying.

edit: @whose
Sorry, but I thought this was relevant see my first post in the thread.

[whose]

Ermh... would you please be so kind to either test, what I suggested, or open another thread to discuss your misunderstandings? Thanks in advance.