On Tue, Feb 13, 2024 at 12:26 AM Johannes Weiner <hannes@xxxxxxxxxxx> wrote: > > On Tue, Feb 13, 2024 at 12:09:26AM -0800, Yosry Ahmed wrote: > > On Tue, Feb 13, 2024 at 12:06 AM Michal Hocko <mhocko@xxxxxxxx> wrote: > > > > > > On Tue 13-02-24 08:01:37, Narcis Garcia wrote: > > > > El 12/2/24 a les 23:37, Yosry Ahmed ha escrit: > > > > > On Sun, Feb 11, 2024 at 1:33 AM Narcis Garcia <debianlists@xxxxxxxxx> wrote: > > > > > > > > > > > > Hello, > > > > > > > > > > > > I read this warning on Linux 6.1 boot: > > > > > > The swapaccount= commandline option is deprecated. Please report your > > > > > > usecase to linux-mm@xxxxxxxxx if you depend on this functionality. > > > > > > > > > > > > And I want to know if this impacts on LinuX Containers (LXC) swap > > > > > > accounting. > > > > > > > > > > swapaccount= was deprecated in the sense that it is always on now > > > > > (i.e. you cannot do swapaccount=0 anymore). I believe you are using > > > > > swapaccount=1 and want to make sure that swap accounting will still > > > > > work with v6.1, right? > > > > > > > > > > This is the second question we have had about this message [1]. I > > > > > think we need to clarify the warning (and perhaps backport to stable > > > > > kernels). > > > > > > > > > > [1]https://lore.kernel.org/linux-mm/2593755.Xrv9ptJ69j@sinistra.local/ > > > > > > > > > > > > Thank you for answering this question, and sure this clarification is needed > > > > both in documentation and warning message. > > > > I've found deprecation message was commited on 26 Sep 2022. > > > > From what Linux version release, this deprecation is included? > > > > > > > > I suggest this for setup_swap_account() code: > > > > pr_warn_once("The swapaccount= commandline option is deprecated. " > > > > "Swap accounting is always on. " > > > > "Please report your usecase to linux-mm@xxxxxxxxx if you " > > > > "depend on this functionality.\n"); > > > > > > Would it make more sense to warn only if somebody tries to disable the > > > swap accounting? > > > --- > > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c > > > index e4c8735e7c85..dec75eedf634 100644 > > > --- a/mm/memcontrol.c > > > +++ b/mm/memcontrol.c > > > @@ -7936,7 +7936,10 @@ bool mem_cgroup_swap_full(struct folio *folio) > > > > > > static int __init setup_swap_account(char *s) > > > { > > > - pr_warn_once("The swapaccount= commandline option is deprecated. " > > > + bool res; > > > + > > > + if (kstrtobool(s, &res) || !res) > > > + pr_warn_once("The swapaccount= commandline option is deprecated. " > > > "Please report your usecase to linux-mm@xxxxxxxxx if you " > > > "depend on this functionality.\n"); > > > return 1; > > > -- > > > > Yes, this makes sense to me, but I still think explicitly specifying > > that swap accounting is always on now would make things even clearer. > > I think it's okay if it only warns when somebody requests disabling, > since the confusion won't happen anymore for folks still explicitly > requesting the enablement. I did add a blurb about what the preferred > method to control swap behavior is, though. You did s/swapaccount=/swapaccount=0/g, and I think that's good enough.
