lkml.org 
[lkml]   [2020]   [Mar]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRE: [PATCH] mm/vmscan: add vm_swappiness configuration knobs
Date
On Wed, 11 Mar 2020, David Rientjes wrote:

> On Wed, 11 Mar 2020, Ivan Teterevkov wrote:
>
> > This patch adds a couple of knobs:
> >
> > - The configuration option (CONFIG_VM_SWAPPINESS).
> > - The command line parameter (vm_swappiness).
> >
> > The default value is preserved, but now defined by CONFIG_VM_SWAPPINESS.
> >
> > Historically, the default swappiness is set to the well-known value 60,
> > and this works well for the majority of cases. The vm_swappiness is also
> > exposed as the kernel parameter that can be changed at runtime too, e.g.
> > with sysctl.
> >
> > This approach might not suit well some configurations, e.g. systemd-based
> > distros, where systemd is put in charge of the cgroup controllers,
> > including the memory one. In such cases, the default swappiness 60
> > is copied across the cgroup subtrees early at startup, when systemd
> > is arranging the slices for its services, before the sysctl.conf
> > or tmpfiles.d/*.conf changes are applied.
> >
>
> Seems like something that can be fully handled by an initscript that would
> set the sysctl and then iterate the memcg hierarchy propagating the
> non-default value. I don't think that's too much of an ask if userspace
> wants to manipulate the swappiness value.
>

This is exactly what I'm trying to avoid: in some distros there is no way
to tackle the configuration early enough, e.g. in systemd-based systems
the systemd is the process that starts first and arranges memcg in a way
it's configured, but unfortunately, it doesn't offer the swappiness knob.

There could be a script to iterate the memcg later, but there would be a
race condition with the system entity that's put in charge of the memcg
because the configuration can't be changed atomically, e.g. a possible
script could iterate the memcg tree and update each memory.swappiness
while systemd is creating another slice or scope subtree.

> Or maybe we can be more clever: have memcg->swappiness store -1 by default
> unless it is changed by the user explicitly and then have
> mem_cgroup_swappiness() return vm_swappiness for this value. If the user
> overwrites it, it's intended.
>

Does it mean that -1 would become a reference to the vm_swappiness
or the parent's memory.swappiness? It sounds interesting and if so then
it would address my issues with the swappiness but would also change
the existing memcg behaviour: if the referred-to value changed, would
the memory.swappiness backed by -1 also change?

> So there are a couple options here but I don't think one of them is to add
> a new config option or kernel command line option.
>

The vm_swappiness starts its lifespan in the kernel and thus
why not to facilitate it with a simple "constructor" there?

> > One could run a script to traverse the cgroup trees later and set the
> > desired memory.swappiness individually in each occurrence when the runtime
> > is set up, but this would require some amount of work to implement
> > properly. Instead, why not set the default swappiness as early as possible?
> >
> > Signed-off-by: Ivan Teterevkov <ivan.teterevkov@nutanix.com>
> > ---
> > .../admin-guide/kernel-parameters.txt | 4 ++++
> > mm/Kconfig | 10 ++++++++
> > mm/vmscan.c | 24 ++++++++++++++++++-
> > 3 files changed, 37 insertions(+), 1 deletion(-)
> >
> > diff --git a/Documentation/admin-guide/kernel-parameters.txt
> b/Documentation/admin-guide/kernel-parameters.txt
> > index c07815d230bc..5d54a4303522 100644
> > --- a/Documentation/admin-guide/kernel-parameters.txt
> > +++ b/Documentation/admin-guide/kernel-parameters.txt
> > @@ -5317,6 +5317,10 @@
> > P Enable page structure init time poisoning
> > - Disable all of the above options
> >
> > + vm_swappiness= [KNL]
> > + Sets the default vm_swappiness.
> > + Ranges from 0 to 100, the default value is 60.
> > +
> > vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an
> exact
> > size of <nn>. This can be used to increase the
> > minimum size (128MB on x86). It can also be used to
> diff --git a/mm/Kconfig b/mm/Kconfig index ab80933be65f..ec59c19e578e
> 100644
> > --- a/mm/Kconfig
> > +++ b/mm/Kconfig
> > @@ -739,4 +739,14 @@ config ARCH_HAS_HUGEPD config
> MAPPING_DIRTY_HELPERS
> > bool
> >
> > +config VM_SWAPPINESS
> > + int "Default memory swappiness"
> > + default 60
> > + range 0 100
> > + help
> > + Sets the default vm_swappiness, that could be changed later
> > + in the runtime, e.g. kernel command line, sysctl, etc.
> > +
> > + Higher value means more swappy. Historically, defaults to 60.
> > +
> > endmenu
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 876370565455..7d2d3550f698 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -163,7 +163,29 @@ struct scan_control {
> > /*
> > * From 0 .. 100. Higher means more swappy.
> > */
> > -int vm_swappiness = 60;
> > +int vm_swappiness = CONFIG_VM_SWAPPINESS;
> > +
> > +static int __init swappiness_cmdline(char *str) {
> > + int val, err;
> > +
> > + if (!str)
> > + return -EINVAL;
> > +
> > + err = kstrtoint(str, 10, &val);
> > + if (err)
> > + return -EINVAL;
> > +
> > + if (val < 0 || val > 100)
> > + return -EINVAL;
> > +
> > + vm_swappiness = val;
> > +
> > + return 0;
> > +}
> > +
> > +early_param("vm_swappiness", swappiness_cmdline);
> > +
> > /*
> > * The total number of pages which are beyond the high watermark within all
> > * zones.

\
 
 \ /
  Last update: 2020-03-12 14:23    [W:0.223 / U:0.076 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site