[PATCH 04/13] i915: Convert shmem_sg_free_table() to use a folio_batch

From: Matthew Wilcox (Oracle)
Date: Wed Jun 21 2023 - 12:47:22 EST


Remove a few hidden compound_head() calls by converting the returned
page to a folio once and using the folio APIs. We also only increment
the refcount on the folio once instead of once for each page. Ideally,
we would have a for_each_sgt_folio macro, but until then this will do.

Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 55 +++++++++++++----------
1 file changed, 31 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
index 33d5d5178103..8f1633c3fb93 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -19,13 +19,13 @@
#include "i915_trace.h"

/*
- * Move pages to appropriate lru and release the pagevec, decrementing the
- * ref count of those pages.
+ * Move folios to appropriate lru and release the batch, decrementing the
+ * ref count of those folios.
*/
-static void check_release_pagevec(struct pagevec *pvec)
+static void check_release_folio_batch(struct folio_batch *fbatch)
{
- check_move_unevictable_pages(pvec);
- __pagevec_release(pvec);
+ check_move_unevictable_folios(fbatch);
+ __folio_batch_release(fbatch);
cond_resched();
}

@@ -33,24 +33,29 @@ void shmem_sg_free_table(struct sg_table *st, struct address_space *mapping,
bool dirty, bool backup)
{
struct sgt_iter sgt_iter;
- struct pagevec pvec;
+ struct folio_batch fbatch;
+ struct folio *last = NULL;
struct page *page;

mapping_clear_unevictable(mapping);

- pagevec_init(&pvec);
+ folio_batch_init(&fbatch);
for_each_sgt_page(page, sgt_iter, st) {
- if (dirty)
- set_page_dirty(page);
+ struct folio *folio = page_folio(page);

+ if (folio == last)
+ continue;
+ last = folio;
+ if (dirty)
+ folio_mark_dirty(folio);
if (backup)
- mark_page_accessed(page);
+ folio_mark_accessed(folio);

- if (!pagevec_add(&pvec, page))
- check_release_pagevec(&pvec);
+ if (!folio_batch_add(&fbatch, folio))
+ check_release_folio_batch(&fbatch);
}
- if (pagevec_count(&pvec))
- check_release_pagevec(&pvec);
+ if (fbatch.nr)
+ check_release_folio_batch(&fbatch);

sg_free_table(st);
}
@@ -63,8 +68,7 @@ int shmem_sg_alloc_table(struct drm_i915_private *i915, struct sg_table *st,
unsigned int page_count; /* restricted by sg_alloc_table */
unsigned long i;
struct scatterlist *sg;
- struct page *page;
- unsigned long last_pfn = 0; /* suppress gcc warning */
+ unsigned long next_pfn = 0; /* suppress gcc warning */
gfp_t noreclaim;
int ret;

@@ -95,6 +99,7 @@ int shmem_sg_alloc_table(struct drm_i915_private *i915, struct sg_table *st,
sg = st->sgl;
st->nents = 0;
for (i = 0; i < page_count; i++) {
+ struct folio *folio;
const unsigned int shrink[] = {
I915_SHRINK_BOUND | I915_SHRINK_UNBOUND,
0,
@@ -103,12 +108,12 @@ int shmem_sg_alloc_table(struct drm_i915_private *i915, struct sg_table *st,

do {
cond_resched();
- page = shmem_read_mapping_page_gfp(mapping, i, gfp);
- if (!IS_ERR(page))
+ folio = shmem_read_folio_gfp(mapping, i, gfp);
+ if (!IS_ERR(folio))
break;

if (!*s) {
- ret = PTR_ERR(page);
+ ret = PTR_ERR(folio);
goto err_sg;
}

@@ -147,19 +152,21 @@ int shmem_sg_alloc_table(struct drm_i915_private *i915, struct sg_table *st,

if (!i ||
sg->length >= max_segment ||
- page_to_pfn(page) != last_pfn + 1) {
+ folio_pfn(folio) != next_pfn) {
if (i)
sg = sg_next(sg);

st->nents++;
- sg_set_page(sg, page, PAGE_SIZE, 0);
+ sg_set_folio(sg, folio, folio_size(folio), 0);
} else {
- sg->length += PAGE_SIZE;
+ /* XXX: could overflow? */
+ sg->length += folio_size(folio);
}
- last_pfn = page_to_pfn(page);
+ next_pfn = folio_pfn(folio) + folio_nr_pages(folio);
+ i += folio_nr_pages(folio) - 1;

/* Check that the i965g/gm workaround works. */
- GEM_BUG_ON(gfp & __GFP_DMA32 && last_pfn >= 0x00100000UL);
+ GEM_BUG_ON(gfp & __GFP_DMA32 && next_pfn >= 0x00100000UL);
}
if (sg) /* loop terminated early; short sg table */
sg_mark_end(sg);
--
2.39.2