When we detect consecutive allocation of pages swap them to avoid accidentally freeing them as huge page.
v2: use swap v3: check if it's really the first allocated page
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_page_alloc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index c82d94cbbabc..7c4d4edce0ba 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -875,7 +875,7 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, #endif struct list_head plist; struct page *p = NULL; - unsigned count; + unsigned count, first; int r;
/* No pool for cached pages */ @@ -914,6 +914,7 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, } #endif
+ first = i; while (npages) { p = alloc_page(gfp_flags); if (!p) { @@ -921,6 +922,10 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, return -ENOMEM; }
+ /* Swap the pages if we detect consecutive order */ + if (i > first && pages[i - 1] == p - 1) + swap(p, pages[i - 1]); + pages[i++] = p; --npages; }
When we detect consecutive allocation of pages swap them to avoid accidentally freeing them as huge page.
v2: use swap v3: check if it's really the first allocated page
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_page_alloc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index 7c4d4edce0ba..da8a50f7c8fe 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -954,8 +954,13 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, r = ttm_page_pool_get_pages(pool, &plist, flags, cstate, npages - count, 0);
- list_for_each_entry(p, &plist, lru) + first = count; + list_for_each_entry(p, &plist, lru) { + /* Swap the pages if we detect consecutive order */ + if (count > first && pages[count - 1] == p - 1) + swap(p, pages[count - 1]); pages[count++] = p; + }
if (r) { /* If there is any pages in the list put them back to
Series is: Reviewed-by: Roger He Hongbo.He@amd.com
Thanks Roger(Hongbo.He) -----Original Message----- From: Christian König [mailto:ckoenig.leichtzumerken@gmail.com] Sent: Monday, December 04, 2017 8:46 PM To: dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org; He, Roger Hongbo.He@amd.com Subject: [PATCH 2/2] drm/ttm: swap consecutive allocated pooled pages v3
When we detect consecutive allocation of pages swap them to avoid accidentally freeing them as huge page.
v2: use swap v3: check if it's really the first allocated page
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_page_alloc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index 7c4d4edce0ba..da8a50f7c8fe 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -954,8 +954,13 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, r = ttm_page_pool_get_pages(pool, &plist, flags, cstate, npages - count, 0);
- list_for_each_entry(p, &plist, lru) + first = count; + list_for_each_entry(p, &plist, lru) { + /* Swap the pages if we detect consecutive order */ + if (count > first && pages[count - 1] == p - 1) + swap(p, pages[count - 1]); pages[count++] = p; + }
if (r) { /* If there is any pages in the list put them back to -- 2.11.0
Hold on, only Patch 1 is Reviewed-by: Roger He Hongbo.He@amd.com.
For Patch 2: + list_for_each_entry(p, &plist, lru) { + /* Swap the pages if we detect consecutive order */ + if (count > first && pages[count - 1] == p - 1) { + swap(p, pages[count - 1]); // swap seems breaking the plist, after that iteration plist will not work. } pages[count++] = temp; + }
-----Original Message----- From: dri-devel [mailto:dri-devel-bounces@lists.freedesktop.org] On Behalf Of He, Roger Sent: Tuesday, December 05, 2017 10:01 AM To: Christian König ckoenig.leichtzumerken@gmail.com; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org Subject: RE: [PATCH 2/2] drm/ttm: swap consecutive allocated pooled pages v3
Series is: Reviewed-by: Roger He Hongbo.He@amd.com
Thanks Roger(Hongbo.He) -----Original Message----- From: Christian König [mailto:ckoenig.leichtzumerken@gmail.com] Sent: Monday, December 04, 2017 8:46 PM To: dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org; He, Roger Hongbo.He@amd.com Subject: [PATCH 2/2] drm/ttm: swap consecutive allocated pooled pages v3
When we detect consecutive allocation of pages swap them to avoid accidentally freeing them as huge page.
v2: use swap v3: check if it's really the first allocated page
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_page_alloc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index 7c4d4edce0ba..da8a50f7c8fe 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -954,8 +954,13 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, r = ttm_page_pool_get_pages(pool, &plist, flags, cstate, npages - count, 0);
- list_for_each_entry(p, &plist, lru) + first = count; + list_for_each_entry(p, &plist, lru) { + /* Swap the pages if we detect consecutive order */ + if (count > first && pages[count - 1] == p - 1) + swap(p, pages[count - 1]); pages[count++] = p; + }
if (r) { /* If there is any pages in the list put them back to -- 2.11.0
_______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On 2017-12-04 01:46 PM, Christian König wrote:
When we detect consecutive allocation of pages swap them to avoid accidentally freeing them as huge page.
v2: use swap v3: check if it's really the first allocated page
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_page_alloc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index c82d94cbbabc..7c4d4edce0ba 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -875,7 +875,7 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, #endif struct list_head plist; struct page *p = NULL;
- unsigned count;
unsigned count, first; int r;
/* No pool for cached pages */
@@ -914,6 +914,7 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, } #endif
first = i;@@ -921,6 +922,10 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, return -ENOMEM; }
/* Swap the pages if we detect consecutive order */if (i > first && pages[i - 1] == p - 1)swap(p, pages[i - 1]);pages[i++] = p; --npages;
Both patches are
Reviewed-by: Michel Dänzer michel.daenzer@amd.com
dri-devel@lists.freedesktop.org
-
Christian König -
He, Roger -
Michel Dänzer