Revert "revert "Revert "mm: remove __GFP_NO_KSWAPD""" and associated damage

This reverts commits a50915394f1fc02c2861d3b7ce7014788aa5066e and
d7c3b937bdf45f0b844400b7bf6fd3ed50bac604.

This is a revert of a revert of a revert.  In addition, it reverts the
even older i915 change to stop using the __GFP_NO_KSWAPD flag due to the
original commits in linux-next.

It turns out that the original patch really was bogus, and that the
original revert was the correct thing to do after all.  We thought we
had fixed the problem, and then reverted the revert, but the problem
really is fundamental: waking up kswapd simply isn't the right thing to
do, and direct reclaim sometimes simply _is_ the right thing to do.

When certain allocations fail, we simply should try some direct reclaim,
and if that fails, fail the allocation.  That's the right thing to do
for THP allocations, which can easily fail, and the GPU allocations want
to do that too.

So starting kswapd is sometimes simply wrong, and removing the flag that
said "don't start kswapd" was a mistake.  Let's hope we never revisit
this mistake again - and certainly not this many times ;)

Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

2 files changed, 7 insertions, 5 deletions

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 107f09befe92..9b285da4449b 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1796,7 +1796,7 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 	 */
 	mapping = obj->base.filp->f_path.dentry->d_inode->i_mapping;
 	gfp = mapping_gfp_mask(mapping);
-	gfp |= __GFP_NORETRY | __GFP_NOWARN;
+	gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
 	gfp &= ~(__GFP_IO | __GFP_WAIT);
 	for_each_sg(st->sgl, sg, page_count, i) {
 		page = shmem_read_mapping_page_gfp(mapping, i, gfp);
@@ -1809,7 +1809,7 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 			 * our own buffer, now let the real VM do its job and
 			 * go down in flames if truly OOM.
 			 */
-			gfp &= ~(__GFP_NORETRY | __GFP_NOWARN);
+			gfp &= ~(__GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD);
 			gfp |= __GFP_IO | __GFP_WAIT;
 
 			i915_gem_shrink_all(dev_priv);
@@ -1817,7 +1817,7 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 			if (IS_ERR(page))
 				goto err_pages;
 
-			gfp |= __GFP_NORETRY | __GFP_NOWARN;
+			gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
 			gfp &= ~(__GFP_IO | __GFP_WAIT);
 		}
 
diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index 374c46dff7dd..ec794a72975d 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -1077,7 +1077,8 @@ EXPORT_SYMBOL_GPL(mtd_writev);
  * until the request succeeds or until the allocation size falls below
  * the system page size. This attempts to make sure it does not adversely
  * impact system performance, so when allocating more than one page, we
- * ask the memory allocator to avoid re-trying.
+ * ask the memory allocator to avoid re-trying, swapping, writing back
+ * or performing I/O.
  *
  * Note, this function also makes sure that the allocated buffer is aligned to
  * the MTD device's min. I/O unit, i.e. the "mtd->writesize" value.
@@ -1091,7 +1092,8 @@ EXPORT_SYMBOL_GPL(mtd_writev);
  */
 void *mtd_kmalloc_up_to(const struct mtd_info *mtd, size_t *size)
 {
-	gfp_t flags = __GFP_NOWARN | __GFP_WAIT | __GFP_NORETRY;
+	gfp_t flags = __GFP_NOWARN | __GFP_WAIT |
+		       __GFP_NORETRY | __GFP_NO_KSWAPD;
 	size_t min_alloc = max_t(size_t, mtd->writesize, PAGE_SIZE);
 	void *kbuf;