[v2] tmpfs: fault in smaller chunks if large folio allocation not allowed

Commit Message

Kefeng Wang Sept. 20, 2024, 2:36 p.m. UTC

The tmpfs supports large folio, but there is some configurable options
to enable/disable large folio allocation, and for huge=within_size,
large folio only allowabled if it fully within i_size, so there is
performance issue when perform write without large folio, the issue is
similar to commit 4e527d5841e2 ("iomap: fault in smaller chunks for
non-large folio mappings").

Fix it by checking whether it allows large folio allocation or not
before perform write.

Fixes: 9aac777aaf94 ("filemap: Convert generic_perform_write() to support large folios")
Signed-off-by: Kefeng Wang <[e-mail]>
---
v2:
- Don't use IOCB flags

 mm/filemap.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

Patch

diff --git a/mm/filemap.c b/mm/filemap.c
index 3e46ca45e13d..b33f260fa32f 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -4126,13 +4126,28 @@  generic_file_direct_write(struct kiocb *iocb, struct iov_iter *from)
 }
 EXPORT_SYMBOL(generic_file_direct_write);
 
+static size_t generic_mapping_max_folio_size(struct address_space *mapping,
+					     loff_t pos)
+{
+	struct inode *inode = mapping->host;
+	pgoff_t index = pos >> PAGE_SHIFT;
+
+	if (!shmem_mapping(mapping))
+		goto out;
+
+	if (!shmem_allowable_huge_orders(inode, NULL, index, 0, false))
+		return PAGE_SIZE;
+out:
+	return mapping_max_folio_size(mapping);
+}
+
 ssize_t generic_perform_write(struct kiocb *iocb, struct iov_iter *i)
 {
 	struct file *file = iocb->ki_filp;
 	loff_t pos = iocb->ki_pos;
 	struct address_space *mapping = file->f_mapping;
 	const struct address_space_operations *a_ops = mapping->a_ops;
-	size_t chunk = mapping_max_folio_size(mapping);
+	size_t chunk = generic_mapping_max_folio_size(mapping, pos);
 	long status = 0;
 	ssize_t written = 0;