Thread (51 messages) 51 messages, 6 authors, 2025-07-25

Re: [PATCH v9 09/14] khugepaged: avoid unnecessary mTHP collapse attempts

From: Baolin Wang <baolin.wang@linux.alibaba.com>
Date: 2025-07-18 02:14:58
Also in: linux-doc, linux-mm, lkml


On 2025/7/14 08:32, Nico Pache wrote:
quoted hunk ↗ jump to hunk
There are cases where, if an attempted collapse fails, all subsequent
orders are guaranteed to also fail. Avoid these collapse attempts by
bailing out early.

Signed-off-by: Nico Pache <npache@redhat.com>
---
  mm/khugepaged.c | 17 +++++++++++++++++
  1 file changed, 17 insertions(+)
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index a701d9f0f158..7a9c4edf0e23 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1367,6 +1367,23 @@ static int collapse_scan_bitmap(struct mm_struct *mm, unsigned long address,
  				collapsed += (1 << order);
  				continue;
  			}
After doing more testing, I think you need to add the following changes 
after patch 8.

Because when collapsing mTHP, if we encounter a PTE-mapped large folio 
within the PMD range, we should continue scanning to complete that PMD, 
in case there is another mTHP that can be collapsed within that PMD range.

+                       if (ret == SCAN_PTE_MAPPED_HUGEPAGE)
+                               continue;
+			/*
+			 * Some ret values indicate all lower order will also
+			 * fail, dont trying to collapse smaller orders
+			 */
+			if (ret == SCAN_EXCEED_NONE_PTE ||
+				ret == SCAN_EXCEED_SWAP_PTE ||
+				ret == SCAN_EXCEED_SHARED_PTE ||
+				ret == SCAN_PTE_NON_PRESENT ||
+				ret == SCAN_PTE_UFFD_WP ||
+				ret == SCAN_ALLOC_HUGE_PAGE_FAIL ||
+				ret == SCAN_CGROUP_CHARGE_FAIL ||
+				ret == SCAN_COPY_MC ||
+				ret == SCAN_PAGE_LOCK ||
+				ret == SCAN_PAGE_COUNT)
+				goto next;
+			else
Nit: the 'else' statement can be dropped.
+				break;
  		}
  
  next:
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help