e1000e: Fix incorrect debug warning

Doing 'WARN_ON(preempt_count())' was horribly horribly wrong, and would
cause tons of warnings at bootup if PREEMPT was enabled because the
initcalls currently run with the kernel lock, which increments the
preempt count.

At the same time, the warning was also insufficient, since it didn't
check that interrupts were enabled.

The proper debug function to use for something that can sleep and wants
a warning if it's called in the wrong context is 'might_sleep()'.

Reported-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/net/e1000e/ich8lan.c

@@ -396,7 +396,7 @@ static s32 e1000_acquire_swflag_ich8lan(struct e1000_hw *hw)
 	u32 extcnf_ctrl;
 	u32 timeout = PHY_CFG_TIMEOUT;
 
-	WARN_ON(preempt_count());
+	might_sleep();
 
 	if (!mutex_trylock(&nvm_mutex)) {
 		WARN(1, KERN_ERR "e1000e mutex contention. Owned by pid %d\n",