; https://github.com/bobrik/linux/pull/2
From 1910e1157eb4b455b877a813f0a1b786dcbf799c Mon Sep 17 00:00:00 2001
From: Ivan Babrou <ibobrik@gmail.com>
Date: Tue, 27 Nov 2018 14:37:25 -0800
Subject: [PATCH] xfs: add a sysctl to disable memory reclaim participation

XFS may try to flush dirty inodes in reclaim and it slows things down
considerably, especially in high page cache and slow disk environment.

This sysctl allows to exclude XFS from kswapd and direct reclaim.

See: https://marc.info/?t=154345187200003
diff --git a/Documentation/admin-guide/xfs.rst b/Documentation/admin-guide/xfs.rst
index 3b9b5c149f322..b33a4822f879c 100644
--- a/Documentation/admin-guide/xfs.rst
+++ b/Documentation/admin-guide/xfs.rst
@@ -331,6 +331,12 @@ The following sysctls are available for the XFS filesystem:
 	is to control the rate at which the allocator moves between
 	allocation groups when allocating extents for new files.
 
+  fs.xfs.memory_reclaim		(Min: 0  Default: 2  Max: 2)
+	Set memory reclaim strategy:
+	0: no inode reclaim (background reclaim is still enabled)
+	1: async inode reclaim of clean inodes only
+	2: sync inode reclaim (includes synchronous writes)
+
 Deprecated Sysctls
 ==================
 
diff --git a/fs/xfs/xfs_globals.c b/fs/xfs/xfs_globals.c
index 3e1cc3001bcbf..2927cc2b12b57 100644
--- a/fs/xfs/xfs_globals.c
+++ b/fs/xfs/xfs_globals.c
@@ -43,6 +43,7 @@ xfs_param_t xfs_params = {
 	.fstrm_timer	= {	1,		30*100,		3600*100},
 	.eofb_timer	= {	1,		300,		3600*24},
 	.cowb_timer	= {	1,		1800,		3600*24},
+	.memory_reclaim	= {	0,		2,		2,	},
 };
 
 struct xfs_globals xfs_globals = {
diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 544b5211221cd..54fcf1af7e3f5 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -1380,7 +1380,11 @@ xfs_reclaim_inodes_nr(
 	xfs_reclaim_work_queue(mp);
 	xfs_ail_push_all(mp->m_ail);
 
-	return xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK | SYNC_WAIT, &nr_to_scan);
+	int flags = SYNC_TRYLOCK;
+	if (xfs_memory_reclaim == XFS_MEMORY_RECLAIM_SYNC)
+		flags |= SYNC_WAIT;
+
+	return xfs_reclaim_inodes_ag(mp, flags, &nr_to_scan);
 }
 
 /*
diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h
index bff4d85e54984..2547c0d85e5a1 100644
--- a/fs/xfs/xfs_icache.h
+++ b/fs/xfs/xfs_icache.h
@@ -54,6 +54,12 @@ struct xfs_eofblocks {
  */
 #define XFS_AGITER_INEW_WAIT	0x1	/* wait on new inodes */
 
+enum {
+	XFS_MEMORY_RECLAIM_NONE = 0,
+	XFS_MEMORY_RECLAIM_ASYNC,
+	XFS_MEMORY_RECLAIM_SYNC,
+};
+
 int xfs_iget(struct xfs_mount *mp, struct xfs_trans *tp, xfs_ino_t ino,
 	     uint flags, uint lock_flags, xfs_inode_t **ipp);
 
diff --git a/fs/xfs/xfs_linux.h b/fs/xfs/xfs_linux.h
index dcd1292664b34..3cddbffaa8cb4 100644
--- a/fs/xfs/xfs_linux.h
+++ b/fs/xfs/xfs_linux.h
@@ -110,6 +110,7 @@ typedef __u32			xfs_nlink_t;
 #define xfs_fstrm_centisecs	xfs_params.fstrm_timer.val
 #define xfs_eofb_secs		xfs_params.eofb_timer.val
 #define xfs_cowb_secs		xfs_params.cowb_timer.val
+#define xfs_memory_reclaim	xfs_params.memory_reclaim.val
 
 #define current_cpu()		(raw_smp_processor_id())
 #define current_pid()		(current->pid)
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 0b0282d2f011c..232e3706a0a60 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1828,6 +1828,8 @@ xfs_fs_nr_cached_objects(
 	/* Paranoia: catch incorrect calls during mount setup or teardown */
 	if (WARN_ON_ONCE(!sb->s_fs_info))
 		return 0;
+	if (xfs_memory_reclaim == XFS_MEMORY_RECLAIM_NONE)
+		return 0;
 	return xfs_reclaim_inodes_count(XFS_M(sb));
 }
 
diff --git a/fs/xfs/xfs_sysctl.c b/fs/xfs/xfs_sysctl.c
index afe1f66aaa698..ccf0d1759c215 100644
--- a/fs/xfs/xfs_sysctl.c
+++ b/fs/xfs/xfs_sysctl.c
@@ -193,6 +193,15 @@ static struct ctl_table xfs_table[] = {
 		.extra1		= &xfs_params.cowb_timer.min,
 		.extra2		= &xfs_params.cowb_timer.max,
 	},
+	{
+		.procname	= "memory_reclaim",
+		.data		= &xfs_params.memory_reclaim.val,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &xfs_params.memory_reclaim.min,
+		.extra2		= &xfs_params.memory_reclaim.max,
+	},
 	/* please keep this the last entry */
 #ifdef CONFIG_PROC_FS
 	{
diff --git a/fs/xfs/xfs_sysctl.h b/fs/xfs/xfs_sysctl.h
index 82afee005140a..eaf3addd486e7 100644
--- a/fs/xfs/xfs_sysctl.h
+++ b/fs/xfs/xfs_sysctl.h
@@ -49,6 +49,7 @@ typedef struct xfs_param {
 	xfs_sysctl_val_t fstrm_timer;	/* Filestream dir-AG assoc'n timeout. */
 	xfs_sysctl_val_t eofb_timer;	/* Interval between eofb scan wakeups */
 	xfs_sysctl_val_t cowb_timer;	/* Interval between cowb scan wakeups */
+	xfs_sysctl_val_t memory_reclaim;/* Memory reclaim policy. */
 } xfs_param_t;
 
 /*
@@ -89,6 +90,7 @@ enum {
 	XFS_ROTORSTEP = 20,
 	XFS_INHERIT_NODFRG = 21,
 	XFS_FILESTREAM_TIMER = 22,
+	XFS_MEMORY_RECLAIM = 23,
 };
 
 extern xfs_param_t	xfs_params;