Eduard,

I have also seen this problem with both Red Hat and stock kernels.  I've
got some patches That I'm testing to try and fix the problem for me, but
it takes a while for my setup to reproduce the problem.  How easy is it
for you to trigger?

I've attached two patches for 2.4.22 that I'm currently testing.  The
first (nfsiput.patch) is to fix the deadlocked processes you see by
moving the iput() call to happen after the waited on request is
unlocked.

But that simple fix is incomplete.  It creates another problem that
generates a BUG() in iput() as now instead of deadlocking, nfs can
continue on and end up calling iput() on a cleared inode.  The second
patch (nfsigrab.patch) checks for a failure when calling igrab() and
aborts an async write and instead converts it to a sync write which I
believe should fix the problem.  But, I haven't triggered it yet so this
patch is effectively untested.

If someone who knows nfs/vfs better could sanity check what I'm doing,
it would be appreciated.

David Jeffery