Subject: Re: 1.2 features, again
To: Jason Thorpe <firstname.lastname@example.org>
From: Greg Hudson <email@example.com>
Date: 06/25/1996 22:29:46
> Not to my knowledge, no. I think Greg Hudson had something for
> this, but it turned out to cause deadlocks.
No. What I had a fix for was that if you mmap() a file MAP_PRIVATE,
and then msync() the memory region, the shadow objects for that region
do *not* get looked at and you might still get out-of-date cached
data. This is why you can't fix the vi problem, for example, with a
simple msync(). (Of course, I don't know if there's any reason why vi
couldn't do a MAP_SHARED.)
The fix was to walk the shadow chain of each object during an msync(),
but that apparently caused deadlocks.
I don't know anything much about the swap leakage problem.