NFS: Fix a deadlock involving nfs_release_folio()

Wang Zhaolong reports a deadlock involving NFSv4.1 state recovery
waiting on kthreadd, which is attempting to reclaim memory by calling
nfs_release_folio(). The latter cannot make progress due to state
recovery being needed.

It seems that the only safe thing to do here is to kick off a writeback
of the folio, without waiting for completion, or else kicking off an
asynchronous commit.

Reported-by: Wang Zhaolong <wangzhaolong@huaweicloud.com>
Fixes: 96780ca55e3c ("NFS: fix up nfs_release_folio() to try to release the page")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

1 files changed, 33 insertions, 0 deletions

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 336c510f3750..bf412455e8ed 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -2025,6 +2025,39 @@ int nfs_wb_folio_cancel(struct inode *inode, struct folio *folio)
 }
 
 /**
+ * nfs_wb_folio_reclaim - Write back all requests on one page
+ * @inode: pointer to page
+ * @folio: pointer to folio
+ *
+ * Assumes that the folio has been locked by the caller
+ */
+int nfs_wb_folio_reclaim(struct inode *inode, struct folio *folio)
+{
+	loff_t range_start = folio_pos(folio);
+	size_t len = folio_size(folio);
+	struct writeback_control wbc = {
+		.sync_mode = WB_SYNC_ALL,
+		.nr_to_write = 0,
+		.range_start = range_start,
+		.range_end = range_start + len - 1,
+		.for_sync = 1,
+	};
+	int ret;
+
+	if (folio_test_writeback(folio))
+		return -EBUSY;
+	if (folio_clear_dirty_for_io(folio)) {
+		trace_nfs_writeback_folio_reclaim(inode, range_start, len);
+		ret = nfs_writepage_locked(folio, &wbc);
+		trace_nfs_writeback_folio_reclaim_done(inode, range_start, len,
+						       ret);
+		return ret;
+	}
+	nfs_commit_inode(inode, 0);
+	return 0;
+}
+
+/**
  * nfs_wb_folio - Write back all requests on one page
  * @inode: pointer to page
  * @folio: pointer to folio