Date: Thu, 18 Aug 2022 02:17:58 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Gleb Smirnoff <glebius@freebsd.org> Cc: src-committers@freebsd.org, dev-commits-src-all@freebsd.org, dev-commits-src-main@freebsd.org Subject: Re: git: 81a34d374ed6 - main - protosw: retire pr_drain and use EVENTHANDLER(9) directly Message-ID: <Yv13JpzLftsx3UBV@kib.kiev.ua> In-Reply-To: <Yv1nKXi77zMU89Aw@FreeBSD.org> References: <202208171852.27HIqL67013112@gitrepo.freebsd.org> <Yv1hOLxDo8rDMzfR@kib.kiev.ua> <Yv1nKXi77zMU89Aw@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Aug 17, 2022 at 03:09:45PM -0700, Gleb Smirnoff wrote: > On Thu, Aug 18, 2022 at 12:44:24AM +0300, Konstantin Belousov wrote: > K> > protosw: retire pr_drain and use EVENTHANDLER(9) directly > K> > > K> > The method was called for two different conditions: 1) the VM layer is > K> > low on pages or 2) one of UMA zones of mbuf allocator exhausted. > K> > This change 2) into a new event handler, but all affected network > K> > subsystems modified to subscribe to both, so this change shall not > K> > bring functional changes under different low memory situations. > K> > > K> > There were three subsystems still using pr_drain: TCP, SCTP and frag6. > K> > The latter had its protosw entry for the only reason to register its > K> > pr_drain method. > K> > > K> > Reviewed by: tuexen, melifaro > K> > Differential revision: https://reviews.freebsd.org/D36164 > K> ... > K> > K> > diff --git a/sys/vm/vm_pageout.h b/sys/vm/vm_pageout.h > K> > index 82ba3c81ef1b..63d31dc1d135 100644 > K> > --- a/sys/vm/vm_pageout.h > K> > +++ b/sys/vm/vm_pageout.h > K> > @@ -87,6 +87,7 @@ extern int vm_pageout_page_count; > K> > */ > K> > #define VM_LOW_KMEM 0x01 > K> > #define VM_LOW_PAGES 0x02 > K> > +#define VM_LOW_MBUFS 0x04 > K> > K> This is very weird, to put it mildly. > > As said in the commit message the old code called the pr_drain for two > very different events and new code just preserves old behavior. I didn't > create the weirdness, I exposed it :) I'm open to redesigning this. > This probably needs to be done separately for every protocol and tested > in real life conditions that indeed trigger these events. I still do not understand, from your reply, why do you need to have mbuf-related condition to be defined in the pageout daemon' namespace. It seems that the problem comes from the IMO mis-guided attempt to mix vm_lowmem and mbuf_lowmem events.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Yv13JpzLftsx3UBV>