Skip to contentSkip to conversation

net: Don't return pfmemalloc pages to the page pool.

PATCH In review Jonathan Lemon 2 participants 4 messages Dec 20, 2018 – Jan 7, 2019

Subsystems (3)
Networking Network drivers Page pool
lore
Share & tools
lore

Jonathan Lemon

author
#0 lore raw

PATCH 1/2 net: Don't return pfmemalloc pages to the page pool.

Return pfmemalloc pages back to the page allocator, instead of holding them in the page pool.

Signed-off-by Jonathan Lemon <[email protected]>
---
 net/core/page_pool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/core/page_pool.c +1 -1
index 43a932cb609b..364b893be66f 100644
@@ -233,7 +233,7 @@ void __page_pool_put_page(struct page_pool *pool,
 	 *
 	 * refcnt == 1 means page_pool owns page, and can recycle it.
 	 */
-	if (likely(page_ref_count(page) == 1)) {
+	if (likely(page_ref_count(page) == 1 && !page_is_pfmemalloc(page))) {
 		/* Read barrier done in page_ref_count / READ_ONCE */
 
 		if (allow_direct && in_serving_softirq())
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 2/2 net: Use the __page_pool_return_page API

Use the __page_pool_return_page API instead of re-rolling our own.

Signed-off-by Jonathan Lemon <[email protected]>
---
 net/core/page_pool.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
net/core/page_pool.c +1 -2
index 364b893be66f..091007ff14a3 100644
@@ -259,8 +259,7 @@ void __page_pool_put_page(struct page_pool *pool,
 	 * doing refcnt based recycle tricks, meaning another process
 	 * will be invoking put_page.
 	 */
-	__page_pool_clean_page(pool, page);
-	put_page(page);
+	__page_pool_return_page(pool, page);
 }
 EXPORT_SYMBOL(__page_pool_put_page);
Signature · 1 line
2.17.1
#2 lore raw
In reply to Jonathan Lemon
Quoted 13 lines · On Thu, 20 Dec 2018 14:21:32 -0800 Jonathan Lemon <[email protected]>…

Return pfmemalloc pages back to the page allocator, instead of holding them in the page pool.

Signed-off-by Jonathan Lemon <[email protected]>
---
 net/core/page_pool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/core/page_pool.c +0 -0
index 43a932cb609b..364b893be66f 100644

net/core/page_pool.c void __page_pool_put_page(struct page_pool *pool, · line 236

Quoted 5 lines
net/core/page_pool.c +1 -1
@@ -233,7 +233,7 @@ void __page_pool_put_page(struct page_pool *pool,
 	 *
 	 * refcnt == 1 means page_pool owns page, and can recycle it.
 	 */
-	if (likely(page_ref_count(page) == 1)) {
+	if (likely(page_ref_count(page) == 1 && !page_is_pfmemalloc(page))) {

I took at closer look at the page_pool issue recycling pages from emergency reserve (pfmemalloc), and it actually cannot happen, because page_pool does not use the __GFP_MEMALLOC gfp_t flag. Thus, page_pool are not allowed to get pages from the emergency reserve in the first place (unless ksoftirqd current->flags have PF_MEMALLOC, which I don't think it have).

See: page_pool_dev_alloc_pages() compared to __dev_alloc_pages().

The doc for:
/* %__GFP_MEMALLOC allows access to all memory. This should only be used when
* the caller guarantees the allocation will allow more memory to be freed
* very shortly e.g. process exiting or swapping. Users either should
* be the MM or co-ordinating closely with the VM (e.g. swap over NFS). */

With that desc, I don't understand why we actually allow dev_alloc_pages() to get emergency reserve (pfmemalloc) pages, as we store these in an RX-ring queue (usual size 512-1024) that isn't used until N-packets later... even if used as a signal to network stack, to free other resources, this happens at a later point-in-time, not "very shortly".

Signature · 4 lines
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer
author
#3 lore raw
Quoted 21 lines · Jesper Dangaard Brouer
Quoted 18 lines · On Thu, 20 Dec 2018 14:21:32 -0800 Jonathan Lemon <[email protected]>…

Return pfmemalloc pages back to the page allocator, instead of holding them in the page pool.

Signed-off-by Jonathan Lemon <[email protected]>
---
 net/core/page_pool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/core/page_pool.c +1 -1
index 43a932cb609b..364b893be66f 100644
@@ -233,7 +233,7 @@ void __page_pool_put_page(struct page_pool *pool,
 	 *
 	 * refcnt == 1 means page_pool owns page, and can recycle it.
 	 */
-	if (likely(page_ref_count(page) == 1)) {
+	if (likely(page_ref_count(page) == 1 && !page_is_pfmemalloc(page)))
Quoted 8 lines

net/core/page_pool.c void __page_pool_put_page(struct page_pool *pool, · line 237

Quoted 1 line
net/core/page_pool.c +0 -0
 {

I took at closer look at the page_pool issue recycling pages from emergency reserve (pfmemalloc), and it actually cannot happen, because page_pool does not use the __GFP_MEMALLOC gfp_t flag. Thus, page_pool are not allowed to get pages from the emergency reserve in the first place (unless ksoftirqd current->flags have PF_MEMALLOC, which I don't think it have).

page_pool_dev_alloc_pages() sets GFP_ATOMIC | __GFP_NOWARN. If the page pool really doesn't want to allow emergency reserves, then shouldn't it set __GFP_NOMEMALLOC as well?

Quoted 22 lines

See: page_pool_dev_alloc_pages() compared to __dev_alloc_pages().

The doc for:
/* %__GFP_MEMALLOC allows access to all memory. This should only be used when
* the caller guarantees the allocation will allow more memory to be
freed
* very shortly e.g. process exiting or swapping. Users either should
* be the MM or co-ordinating closely with the VM (e.g. swap over
NFS).
*/

With that desc, I don't understand why we actually allow dev_alloc_pages() to get emergency reserve (pfmemalloc) pages, as we store these in an RX-ring queue (usual size 512-1024) that isn't used until N-packets later... even if used as a signal to network stack, to free other resources, this happens at a later point-in-time, not "very shortly".

Signature · 4 lines
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer