Re: [PATCH 3/5] perf_counter: rework ioctl()s

[Corey Ashford <cjashfor@linux.vnet.ibm.com>]

Paul Mackerras wrote:
> Corey Ashford writes:
> 
>> Another hypothesis is that the old PAPI code would open, mmap, close, 
>> then reopen, and mmap again.  I wonder if because I wasn't munmap'ing 
>> before the close, that I got some strange behavior from the kernel. 
> 
> In general an mmap will keep a reference to the file it is mapping,
> which means that if you do an mmap on a fd and then close the fd, the
> file stays open until the region is munmapped.  So your old PAPI code
> would have ended up with more counters active than you expected, if
> you thought closing the fd would have destroyed the counter.
> 
> That is, you don't have to munmap before close, but you do have to do
> both close and munmap to destroy the counter.


That makes perfect sense.  I'll bet that's what the problem was.  I think this 
issue can be closed.  Thank you Paul M. and Peter Z. for looking into this.

- Corey