Analyzing wired memory?

[Robert N. M. Watson]

On 8 Feb 2011, at 10:37, Alan Cox wrote:

>> John and I have occasionally talked about making procstat -v work on the kernel; conceivably it could also export a wired page count for mappings where it makes sense.  Ideally procstat would drill in a bit and allow you to see things at least at the granularty of "this page range was allocated to UMA".
> 
> I would certainly have found this useful on a few occasions, and would gladly help out with implementing it.  For example, it would help us in understanding the kmem_map fragmentation caused by ZFS.  That said, I'm not sure how you will represent the case where UMA allocates physical memory directly and uses the direct map to access it.

I agree -- in some sense, my proposal is related to, but somewhat orthogonal to, the problem Ivan is running into.

We could also have a more detailed but UMA-specific inspection tool; I have a uma dumping tool in src/tools somewhere that dumps zone information, buckets, etc, for UMA types. I've used this in the past to explore mbuf alloc/free behaviour, multi-CPU imbalances in allocation, over-caching, etc. It may be time to revisit that, decide how to make it a bit more user/developer-friendly, and make it a formally supported tool as part of vmstat (which involves adding some new sysctls as well).

Robert