Is it possible to show what file a process i waiting for?

[JanN]

I have a hung process.
It's showing state "D" in ps auxw
lsof shows all open files/socket
But how can I tell which one it is hanging for?

The only suggestion I have is using a debugger.
Can this information be collected from the kernel using some tool?

[Steven E. Protter]

Shalom,

The best utility for this is lsof.

It has a number of options to show what resouces are associated with a process.

SEP

[Ivan Ferreira]

You can use strace or ltrace to view what is doing your process.

[JanN]

Well - with LSOF i didn't find any flag to have it show me WHAT file/socket the process is hanging on.
I might have to re-read the man page or read the source...

When it comes to strace/ltrace, I don't know when this happens (not reproducable) and I cant run this process with strace all the time.
And when it have hung it stays hung!

/JanN

[Jaime Bolanos Rojas.]

JanN,

This is one of the places where you can find the software:

http://freshmeat.net/projects/lsof/

Regards,

Jaime.

[Ivan Ferreira]

You can wait to the process to hang and then run strace -p PID to attach to the process, there is no need to start the process with strace.

[Tom Robinson_1]

Processes in state 'D' are waiting on a 'slow' device. Essentially this means that the process has given control to the hardware and has gone to sleep. State 'D' is an 'un-interruptible' state. No amount of killing will have any effect until the hardware returns control to the process. The process can still ignore signals of course, except SIGKILL, which should pend on the return of control to the process. Processes in state 'D' should eventually wake up but if they don't you'll have to reboot.

lsof will show files but if your process is waiting on hardware then you'll just have to wait.

[JanN]

So far - the answers I got didn't really help me.

Ivan: I can attach strace on the fly - but then I've already missed the interesting system call that hung the process.

Steven et.al.: I already have lsof and know it is a handy tool. But I can't seem to find the answer to my question with it.

Tom: You're explaining it well - but I already knew that.

Questions are:
1. Can lsof (or any other tool) tell me (on the already hung process) what system call or file descriptor it is waiting for?

2. I cant use gdb since I can't signal the process. (Uninteruptable sleep) So I take it the only possible way is to poke into process or kernel memory somehow to find out. How can this be done?

Or is this a dead end? Must I either "crash" the whole system and look through the system dump or trouble shoot from another direction?

/JanN