On Thu, Aug 15, 2024 at 12:04 PM Yu Zhao <yuzhao@xxxxxxxxxx> wrote:
>
> On Thu, Aug 15, 2024 at 8:41 AM Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
> >
> >
> >
> > On 2024/8/12 5:21, Yu Zhao wrote:
> > > With alloc_contig_range() and free_contig_range() supporting large
> > > folios, CMA can allocate and free large folios too, by
> > > cma_alloc_folio() and cma_release().
> > >
> > > Signed-off-by: Yu Zhao <yuzhao@xxxxxxxxxx>
> > > ---
> > > include/linux/cma.h | 1 +
> > > mm/cma.c | 47 ++++++++++++++++++++++++++++++---------------
> > > 2 files changed, 33 insertions(+), 15 deletions(-)
> > >
> > > diff --git a/include/linux/cma.h b/include/linux/cma.h
> > > index 9db877506ea8..086553fbda73 100644
> > > --- a/include/linux/cma.h
> > > +++ b/include/linux/cma.h
> > > @@ -46,6 +46,7 @@ extern int cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
> > > struct cma **res_cma);
> > > extern struct page *cma_alloc(struct cma *cma, unsigned long count, unsigned int align,
> > > bool no_warn);
> > > +extern struct folio *cma_alloc_folio(struct cma *cma, int order, gfp_t gfp);
> > > extern bool cma_pages_valid(struct cma *cma, const struct page *pages, unsigned long count);
> > > extern bool cma_release(struct cma *cma, const struct page *pages, unsigned long count);
> > >
> > > diff --git a/mm/cma.c b/mm/cma.c
> > > index 95d6950e177b..46feb06db8e7 100644
> > > --- a/mm/cma.c
> > > +++ b/mm/cma.c
> > > @@ -403,18 +403,8 @@ static void cma_debug_show_areas(struct cma *cma)
> > > spin_unlock_irq(&cma->lock);
> > > }
> > >
> > > -/**
> > > - * cma_alloc() - allocate pages from contiguous area
> > > - * @cma: Contiguous memory region for which the allocation is performed.
> > > - * @count: Requested number of pages.
> > > - * @align: Requested alignment of pages (in PAGE_SIZE order).
> > > - * @no_warn: Avoid printing message about failed allocation
> > > - *
> > > - * This function allocates part of contiguous memory on specific
> > > - * contiguous memory area.
> > > - */
> > > -struct page *cma_alloc(struct cma *cma, unsigned long count,
> > > - unsigned int align, bool no_warn)
> > > +static struct page *__cma_alloc(struct cma *cma, unsigned long count,
> > > + unsigned int align, gfp_t gfp)
> > > {
> > > unsigned long mask, offset;
> > > unsigned long pfn = -1;
> > > @@ -463,8 +453,7 @@ struct page *cma_alloc(struct cma *cma, unsigned long count,
> > >
> > > pfn = cma->base_pfn + (bitmap_no << cma->order_per_bit);
> > > mutex_lock(&cma_mutex);
> > > - ret = alloc_contig_range(pfn, pfn + count, MIGRATE_CMA,
> > > - GFP_KERNEL | (no_warn ? __GFP_NOWARN : 0));
> > > + ret = alloc_contig_range(pfn, pfn + count, MIGRATE_CMA, gfp);
> > > mutex_unlock(&cma_mutex);
> > > if (ret == 0) {
> > > page = pfn_to_page(pfn);
> > > @@ -494,7 +483,7 @@ struct page *cma_alloc(struct cma *cma, unsigned long count,
> > > page_kasan_tag_reset(nth_page(page, i));
> > > }
> > >
> > > - if (ret && !no_warn) {
> > > + if (ret && !(gfp & __GFP_NOWARN)) {
> > > pr_err_ratelimited("%s: %s: alloc failed, req-size: %lu pages, ret: %d\n",
> > > __func__, cma->name, count, ret);
> > > cma_debug_show_areas(cma);
> > > @@ -513,6 +502,34 @@ struct page *cma_alloc(struct cma *cma, unsigned long count,
> > > return page;
> > > }
> > >
> > > +/**
> > > + * cma_alloc() - allocate pages from contiguous area
> > > + * @cma: Contiguous memory region for which the allocation is performed.
> > > + * @count: Requested number of pages.
> > > + * @align: Requested alignment of pages (in PAGE_SIZE order).
> > > + * @no_warn: Avoid printing message about failed allocation
> > > + *
> > > + * This function allocates part of contiguous memory on specific
> > > + * contiguous memory area.
> > > + */
> > > +struct page *cma_alloc(struct cma *cma, unsigned long count,
> > > + unsigned int align, bool no_warn)
> > > +{
> > > + return __cma_alloc(cma, count, align, GFP_KERNEL | (no_warn ? __GFP_NOWARN : 0));
> > > +}
> > > +
> > > +struct folio *cma_alloc_folio(struct cma *cma, int order, gfp_t gfp)
> > > +{
> > > + struct page *page;
> > > +
> > > + if (WARN_ON(order && !(gfp | __GFP_COMP)))
> > > + return NULL;
> > > +
> > > + page = __cma_alloc(cma, 1 << order, order, gfp);
> > > +
> > > + return page ? page_folio(page) : NULL;
> >
> > We don't set large_rmappable for cma alloc folio, which is not consistent
> > with other folio allocation, eg folio_alloc/folio_alloc_mpol(),
> > there is no issue for HugeTLB folio, and for HugeTLB folio must without
> > large_rmappable, but once we use it for mTHP/THP, it need some extra
> > handle, maybe we set large_rmappable here, and clear it in
> > init_new_hugetlb_folio()?
>
> I want to hear what Matthew thinks about this.
>
> My opinion is that we don't want to couple largely rmappable (or
> deferred splittable) with __GFP_COMP, and for that matter, with large
> folios, because the former are specific to THPs whereas the latter can
> potentially work for most types of high order allocations.
>
> Again, IMO, if we want to seriously answer the question of
> Can we get rid of non-compound multi-page allocations? [1]
> then we should start planning on decouple large rmappable from the
> generic folio allocation API.
>
> [1] https://lpc.events/event/18/sessions/184/#20240920
Also along the similar lines, Usama is trying to add
PG_partially_mapped [1], which I have explicitly asked him not to
introduce that flag to hugeTLB, unless there are good reasons (none
ATM).
[1] https://lore.kernel.org/CAOUHufbmgwZwzUuHVvEDMqPGcsxE2hEreRZ4PhK5yz27GdK-Tw@xxxxxxxxxxxxxx/
