Waking from sleep - one core goes 100% load!

[benko]

I have this problem for a long time, today I decided to kill windows, and put a completely fresh setup. But the problem still persist.

I believe the problem lies in GPU drivers, not sure whatever it is an Intel or Nvidia, but one of these is a problem for me.

So, when I installed fresh windows the sleep option was dimmed, so I couldn't get to sleep, as I installed GPU drivers, sleep option was enabled, but than there is the problem described above. No, it's not a windows issue, same problem is with Windows 8!

Looks like "System Interupts" is using 18% CPU - whatever that is, it's using 1 core.

 

[benko]

Anybody

 

[cluberti]

Note that this is actually something seen with network drivers that don't actually support RSS properly - when the computer resumes from a sleep event (sleep or hibernate), one CPU can get completely used by the driver (which normally can spread the load across at least 4 CPUs to do RSS scheduling). If you disable the network cards in the machine (start > ncpa.cpl) and pull the network cable (if it's wired), does the problem go away?

 

[benko]

Note that this is actually something seen with network drivers that don't actually support RSS properly - when the computer resumes from a sleep event (sleep or hibernate), one CPU can get completely used by the driver (which normally can spread the load across at least 4 CPUs to do RSS scheduling). If you disable the network cards in the machine (start > ncpa.cpl) and pull the network cable (if it's wired), does the problem go away?

a-a, just tried, does not help

 

[cluberti]

OK, a more definitive test - if you disable the network cards and reboot to boot without network connectivity, does the problem still occur if you sleep and then wake? If it still occurs, then we may need to suggest some additional debugging.

 

[benko]

Still same

 

[cluberti]

OK, so I think we can be fairly certain it's not the network driver at this point. Next question - are you willing to crash your machine "for science"?

 

[benko]

Always

Whatever you want, I will do

 

[cluberti]

Unfortunately, you've uploaded a corrupted dump (probably because you split it). I can't get anything from it, so I'll need for you to upload it again as a single file. If it's too big, I can create some space on my FTP for you.

 

[benko]

Hmm, I just downloaded it and tested, it extracts just fine!

 

[cluberti]

Yes, but it can't be opened and debugged with the debugger , which is a bit more important.

Sent from my SGH-i917 using Board Express

 

[benko]

Ok, here it is


MEMORY.rar

 

[benko]

Any news on this, is it ok now ?

 

[cluberti]

Still can't make this load properly - something's up with the data in this dump. I hate to ask, but can you take another? This one still only has user-mode context data in it.

 

[benko]

Maybe I am doing smth wrong ?

Here is new one

MEMORYdump.rar

 

[benko]

Using Process explorer I found that Interrupts is the one I should me watching!

Searching a little bit I have found that problem could be in ATA Channel, that it can be switched to PIO mode, but on mine, it says that DMA is enabled!

Also, when waking from sleep I hear that sound like when you remove USB from PC, even thought I haven't removed anything...

Here are the errors I get in event log

1. The server could not bind to the transport \Device\NetBT_Tcpip_{82C32301-772F-42A6-B96C-C167143DB45A} because another computer on the network has the same name.  The server could not start.
2. The name "BENKO-PC       :20" could not be registered on the interface with IP address 192.168.1.81. The computer with the IP address 192.168.1.80 did not allow the name to be claimed by this computer.
3. The name "BENKO-PC       :0" could not be registered on the interface with IP address 192.168.1.81. The computer with the IP address 192.168.1.80 did not allow the name to be claimed by this computer.

For first source is server, for 2nd and 3rd source is NetBT

 

[cluberti]

Well, interrupts (and the system process) would indicate a driver issue or a system configuration issue, but it's unfortunately not easy to tell from a process explorer snapshot. As I said initially, my suspicion is a networking issue, but all of the "easy" tests sort of dispel that.

Let me think about what to do here - dumping it hasn't been successful, but more info is good. Let's see if I can come up with something that'll spill the beans, so to speak, that we can do here.

 

[benko]

Well, interrupts (and the system process) would indicate a driver issue or a system configuration issue, but it's unfortunately not easy to tell from a process explorer snapshot. As I said initially, my suspicion is a networking issue, but all of the "easy" tests sort of dispel that.

Let me think about what to do here - dumping it hasn't been successful, but more info is good. Let's see if I can come up with something that'll spill the beans, so to speak, that we can do here.

So you still weren't able to debug my dump, or you managed to but it didn't help ?

I highly doubt it is network since with it disabled issues still remained.

Thanks for your time and help!

 

[cluberti]

Unfortunately, when your machine crashes and dumps, it only seems to save the user-mode portion of data (the kernel-mode stacks and such are not included). This probably has something to do with your issue, and in a customer scenario I'd just set up a kernel debugger and crash cart, but that's difficult over the internet .