cluster-devel.redhat.com archive mirror
 help / color / mirror / Atom feed
* [Cluster-devel] [PATCH] gfs2: Fix should_fault_in_pages() logic
@ 2022-03-16 14:24 Andreas Gruenbacher
  0 siblings, 0 replies; only message in thread
From: Andreas Gruenbacher @ 2022-03-16 14:24 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Fix the fault-in window size logic:
* Use a maximum window size of 1 MiB instead of BIO_MAX_VECS * PAGE_SIZE.
  The previous window size was always one page because the pages variable
  was accidentally being redefined in should_fault_in_pages().
* Account for page overlap (unaligned iterators).
* The nr_dirtied heuristic for guessing when there might be memory
  pressure often results in very window sizes.  Don't let nr_dirtied
  drop below 8 pages (as btrfs does).
* Compute the window size in units of bytes, not pages.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 fs/gfs2/file.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index a743a9eb602f..ac0665a70d46 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -775,8 +775,7 @@ static inline bool should_fault_in_pages(ssize_t ret, struct iov_iter *i,
 					 size_t *window_size)
 {
 	size_t count = iov_iter_count(i);
-	char __user *p;
-	int pages = 1;
+	size_t size, offs;
 
 	if (likely(!count))
 		return false;
@@ -785,18 +784,20 @@ static inline bool should_fault_in_pages(ssize_t ret, struct iov_iter *i,
 	if (!iter_is_iovec(i))
 		return false;
 
+	size = PAGE_SIZE;
+	offs = offset_in_page(i->iov[0].iov_base + i->iov_offset);
 	if (*prev_count != count || !*window_size) {
-		int pages, nr_dirtied;
+		int nr_dirtied;
 
-		pages = min_t(int, BIO_MAX_VECS, DIV_ROUND_UP(count, PAGE_SIZE));
+		size = ALIGN(offs + count, PAGE_SIZE);
+		size = min_t(size_t, size, SZ_1M);
 		nr_dirtied = max(current->nr_dirtied_pause -
-				 current->nr_dirtied, 1);
-		pages = min(pages, nr_dirtied);
+				 current->nr_dirtied, 8);
+		size = min(size, nr_dirtied << PAGE_SHIFT);
 	}
 
 	*prev_count = count;
-	p = i->iov[0].iov_base + i->iov_offset;
-	*window_size = (size_t)PAGE_SIZE * pages - offset_in_page(p);
+	*window_size = size - offs;
 	return true;
 }
 
-- 
2.33.1


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2022-03-16 14:24 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-03-16 14:24 [Cluster-devel] [PATCH] gfs2: Fix should_fault_in_pages() logic Andreas Gruenbacher

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).