On Fri, 2022-03-25 at 13:24 -0600, Jonathan Corbet wrote:
> Dylan Yudaken <dylany@xxxxxx> writes:
>
> > ignore-gfp-wait and ignore-gfp-highmem defaults are actually true
> > (Y) in
> > both failslab and fail_page_alloc, not false as the docs suggest.
> > See
> > page_alloc.c:3762 and failslab.c:13
> >
> > At the same time use 'Y' instead of '1' in the example scripts just
> > for
> > consistency. (though 1 would work)
> >
> > Signed-off-by: Dylan Yudaken <dylany@xxxxxx>
> > ---
> > Documentation/fault-injection/fault-injection.rst | 10 +++++-----
> > 1 file changed, 5 insertions(+), 5 deletions(-)
> >
> > diff --git a/Documentation/fault-injection/fault-injection.rst
> > b/Documentation/fault-injection/fault-injection.rst
> > index 4a25c5eb6f07..8b4e1ce26d01 100644
> > --- a/Documentation/fault-injection/fault-injection.rst
> > +++ b/Documentation/fault-injection/fault-injection.rst
> > @@ -132,7 +132,7 @@ configuration of fault-injection capabilities.
> >
> > Format: { 'Y' | 'N' }
> >
> > - default is 'N', setting it to 'Y' won't inject failures
> > into
> > + default is 'Y', setting it to 'Y' won't inject failures
> > into
> > highmem/user allocations.
>
> So the associated help text should perhaps have its sense changed too
> -
> what happens if you set it to "N"? The help text would be more
> useful,
> IMO, if it read something like "set to 'N' if you want errors
> injected
> into GFP_HIGHUSER allocations" (or whatever GFP flags actually apply
> here).
>
> > - /sys/kernel/debug/failslab/ignore-gfp-wait:
> > @@ -140,7 +140,7 @@ configuration of fault-injection capabilities.
> >
> > Format: { 'Y' | 'N' }
> >
> > - default is 'N', setting it to 'Y' will inject failures
> > + default is 'Y', setting it to 'Y' will inject failures
> > only into non-sleep allocations (GFP_ATOMIC allocations).
>
> This is a little better in general, but still would be improved by
> telling the reader what they get if they change the default.
Makes sense. I'll submit v2 with that but it might be a couple weeks
when I get time (and I first have to understand what they actually get
when changing the default).
>
> Thanks,
>
> jon
