[linux-cifs-client] Devolpment problem

[James Roper]

I've been stuck on this problem for quite a while now...

In trying to develop an async readpages operation, my current implementation 
of cifs_readpages seems to leave cifs in a bad state.  A single call to it, 
lets say I copy a 40KB file:
cp cifsmountpoint/test .
works fine.  The file is copied perfectly and the system is still usable after 
that (ie I can do anything other than interact with the cifs filesystem).  
However, lets say a few seconds after that I do something else that interacts 
with the cifs module, eg
ls cifsmountpoint
the computer freezes, and I have to reboot.

So my conclusion is that my cifs_readpages routine isn't cleaning up properly.  
However I have no idea what it's not cleaning up that could possibly freeze 
the system, whether it's the page cache list given to it, the cifs buffer, 
the xid, midq, tcp semaphore, or what.  According to syslog, the xid is 
freed, I've already experienced and got rid of the lovely bug output when you 
don't free all the cache pages.  And the rest of those things should at worst 
only cause the program to hang or cause memory leaks, not freeze the OS.

Any ideas of what could cause that?

Thanks,

James