Re: [PATCH v2 08/10] ceph: convert page cleanup loop in writepages_finish() to folios

[Matthew Wilcox <willy@infradead.org>]

On Wed, Aug 05, 2026 at 02:33:34AM +0300, Tal Zussman wrote:
> >  		for (j = 0; j < num_pages; j++) {
> > +			struct folio *folio;
> > +
> >  			fscrypt_finalize_bounce_page(&osd_data->pages[j]);
> > -			page = osd_data->pages[j];
> > -			BUG_ON(!page);
> > -			WARN_ON(!PageUptodate(page));
> > +			folio = page_folio(osd_data->pages[j]);
> > +			BUG_ON(!folio);
> > +			WARN_ON(!folio_test_uptodate(folio));
> 
> Sashiko complains:
> 
> "Since page_folio() invokes _compound_head() which dereferences the page
> pointer, won't this cause a NULL pointer dereference before reaching the
> BUG_ON(!folio) check if osd_data->pages[j] is actually NULL?
> The previous code checked the page pointer before any dereference occurred.
> Would it be safer to check the array element for NULL before passing it to
> page_folio()?"
> 
> The BUG_ON can just be removed, as fscrypt_finalize_bounce_page() cannot take
> or set its argument to NULL as constructed, so the check was unnecessary
> to begin with.

Heh, Sashiko beat me to it ;-)

I'd agree, the BUG_ON was always unnecessary, and I was going to
recommend just deleting it.