Re: [PATCH v2] huge_memory: return -EINVAL in folio split functions when THP is disabled

[Kiryl Shutsemau <kirill@shutemov.name>]

On Fri, Sep 05, 2025 at 10:13:05AM +0200, Pankaj Raghav wrote:
> On 9/4/25 17:26, Kiryl Shutsemau wrote:
> > On Thu, Sep 04, 2025 at 11:51:29AM +0200, Pankaj Raghav (Samsung) wrote:
> >> From: Pankaj Raghav <p.raghav@samsung.com>
> >>
> >> split_huge_page_to_list_[to_order](), split_huge_page() and
> >> try_folio_split() return 0 on success and error codes on failure.
> >>
> >> When THP is disabled, these functions return 0 indicating success even
> >> though an error code should be returned as it is not possible to split a
> >> folio when THP is disabled.
> >>
> >> Make all these functions return -EINVAL to indicate failure instead of
> >> 0. As large folios depend on CONFIG_THP, issue warning as this function
> >> should not be called without a large folio.
> >>
> >> Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
> > 
> > Acked-by: Kiryl Shutsemau <kas@kernel.org>
> > 
> > Just curious, did you give BUILD_BUG() a try?
> > 
> 
> Yes, I tried BUILD_BUG() but it actually fails the build because truncate_inode_partial_folio()
> calls try_folio_split().
> 
> It won't be called in runtime because there is a check for large folios before we call this
> function. So a runtime warning is better in this case.

If truncate_inode_partial_folio() shouldn't ever be calling on THP=n
kernel, it also can be BUILD_BUG(). Just saying :P

-- 
  Kiryl Shutsemau / Kirill A. Shutemov