Re: kern/54818: 9.0_RC1 pagedaemon spins

[Michael van Elst <mlelstv%serpens.de@localhost>]

The following reply was made to PR kern/54818; it has been noted by GNATS.

From: Michael van Elst <mlelstv%serpens.de@localhost>
To: Havard Eidnes <he%uninett.no@localhost>
Cc: gnats-bugs%netbsd.org@localhost, ad%netbsd.org@localhost, tsutsui%ceres.dti.ne.jp@localhost
Subject: Re: kern/54818: 9.0_RC1 pagedaemon spins
Date: Mon, 2 Mar 2020 16:10:27 +0100

 On Mon, Mar 02, 2020 at 03:34:00PM +0100, Havard Eidnes wrote:
 > 
 > However...  This leaves two questions which IMO needs validation:
 > 
 > 1) Is there "enough" free memory?
 > 2) Is there a kernel VA shortage?
 > 
 > Is there any information I can collect to answer one or both of
 > these?  Is just "vmstat -s" enough (only answers #1?), or is more
 > needed?
 
 Once kicked, the pagedaemon keeps running until free memory ("pages
 free") exceeds reserved kernel memory ("reserve kernel pages") and there
 is no kernel VA shortage.
 
 Of course it might be kicked again immediately if some memory allocation
 still fails. this is counted ("times daemon wokeup").
 
 
 I use a local patch that reports when VA is starved and that
 slows down the busy loop somewhat.
 
 http://ftp.netbsd.org/pub/NetBSD/misc/mlelstv/uvm_pdaemon.patch
 
 But I haven't reproduced the issue so far.
 
 
 Greetings,
 -- 
                                 Michael van Elst
 Internet: mlelstv%serpens.de@localhost
                                 "A potential Snark may lurk in every tree."