On May 9, 10:58am, yamt@mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
-- Subject: Re: kern/30154

| > | i think that now you have the same problem for ui_slock.
| > 
| > Well, if we don't lock then there is a problem with getting inconsistent
| > values. How should we fix this? Defer it until later?
| 
| yes, it's a possible solution.
| 
| alternatives are:
| 	1. block unsafe interrupts when grabbing the lock.
| 	2. use atomic operations.  (needs to introduce MD/MI interface)
| 
| i think #1 is the easiest.

I guess we'll need 2 in the long run... What are unsafe interrupts in
this context though? Is splsoftnet() good enough?

christos