Re: git: 457b940bfb6a - main - tuning.7: add more explanation about swap (over-)accounting

Hi!

Attached a small fix.

On Tue, Jan 13, 2026 at 3:04 PM Konstantin Belousov <kib@freebsd.org> wrote:
>
> The branch main has been updated by kib:
>
> URL: https://cgit.FreeBSD.org/src/commit/?id=457b940bfb6a228af1281f357b627ecf1e26fea5
>
> commit 457b940bfb6a228af1281f357b627ecf1e26fea5
> Author:     Konstantin Belousov <kib@FreeBSD.org>
> AuthorDate: 2026-01-12 03:54:58 +0000
> Commit:     Konstantin Belousov <kib@FreeBSD.org>
> CommitDate: 2026-01-13 14:03:14 +0000
>
>     tuning.7: add more explanation about swap (over-)accounting
>
>     Sponsored by:   The FreeBSD Foundation
>     MFC after:      1 week
>     Differential revision:  https://reviews.freebsd.org/D54572
> ---
>  share/man/man7/tuning.7 | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
>
> diff --git a/share/man/man7/tuning.7 b/share/man/man7/tuning.7
> index 44c427c4559d..a3709ce82b8c 100644
> --- a/share/man/man7/tuning.7
> +++ b/share/man/man7/tuning.7
> @@ -222,6 +222,21 @@ and
>  .Va vm.stats.vm.v_wire_count
>  sysctls, respectively).
>  .Pp
> +Due to the architecture of the
> +.Fx
> +virtual memory subsystem, the use of copy on write (CoW) anonymous
> +memory, e.g. on
> +.Xr fork 2 ,
> +causes swap reservation for all three regions (VM objects),
> +as in the original pre-fork mapping, and its copies in
> +the parent and child, instead of only two.
> +Eventually the subsystem tries to optimize the internal layout
> +of the tracking for CoW and often removes (collapses) no longer
> +needed backing objects, re-assigning its pages and swap
> +reservations to the copies.
> +Coolapsing frees the swap reserve, but it is not guaranteed
> +to happen.
> +.Pp
>  The
>  .Va kern.ipc.maxpipekva
>  loader tunable is used to set a hard limit on the
>