1 files changed, 157 insertions, 152 deletions

diff --git a/net/sunrpc/xprtrdma/svc_rdma.c b/net/sunrpc/xprtrdma/svc_rdma.c
index c1b6270262c2..415c0310101f 100644
--- a/net/sunrpc/xprtrdma/svc_rdma.c
+++ b/net/sunrpc/xprtrdma/svc_rdma.c
@@ -1,4 +1,6 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
 /*
+ * Copyright (c) 2015-2018 Oracle.  All rights reserved.
  * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -38,8 +40,7 @@
  *
  * Author: Tom Tucker <tom@opengridcomputing.com>
  */
-#include <linux/module.h>
-#include <linux/init.h>
+
 #include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/sysctl.h>
@@ -47,74 +48,61 @@
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/sched.h>
 #include <linux/sunrpc/svc_rdma.h>
-#include "xprt_rdma.h"
 
 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
 
 /* RPC/RDMA parameters */
-unsigned int svcrdma_ord = RPCRDMA_ORD;
+unsigned int svcrdma_ord = 16;	/* historical default */
 static unsigned int min_ord = 1;
-static unsigned int max_ord = 4096;
+static unsigned int max_ord = 255;
 unsigned int svcrdma_max_requests = RPCRDMA_MAX_REQUESTS;
+unsigned int svcrdma_max_bc_requests = RPCRDMA_MAX_BC_REQUESTS;
 static unsigned int min_max_requests = 4;
 static unsigned int max_max_requests = 16384;
-unsigned int svcrdma_max_req_size = RPCRDMA_MAX_REQ_SIZE;
-static unsigned int min_max_inline = 4096;
-static unsigned int max_max_inline = 65536;
-
-atomic_t rdma_stat_recv;
-atomic_t rdma_stat_read;
-atomic_t rdma_stat_write;
-atomic_t rdma_stat_sq_starve;
-atomic_t rdma_stat_rq_starve;
-atomic_t rdma_stat_rq_poll;
-atomic_t rdma_stat_rq_prod;
-atomic_t rdma_stat_sq_poll;
-atomic_t rdma_stat_sq_prod;
+unsigned int svcrdma_max_req_size = RPCRDMA_DEF_INLINE_THRESH;
+static unsigned int min_max_inline = RPCRDMA_DEF_INLINE_THRESH;
+static unsigned int max_max_inline = RPCRDMA_MAX_INLINE_THRESH;
+static unsigned int svcrdma_stat_unused;
+static unsigned int zero;
 
-/* Temporary NFS request map and context caches */
-struct kmem_cache *svc_rdma_map_cachep;
-struct kmem_cache *svc_rdma_ctxt_cachep;
+struct percpu_counter svcrdma_stat_read;
+struct percpu_counter svcrdma_stat_recv;
+struct percpu_counter svcrdma_stat_sq_starve;
+struct percpu_counter svcrdma_stat_write;
 
-struct workqueue_struct *svc_rdma_wq;
+enum {
+	SVCRDMA_COUNTER_BUFSIZ	= sizeof(unsigned long long),
+};
 
-/*
- * This function implements reading and resetting an atomic_t stat
- * variable through read/write to a proc file. Any write to the file
- * resets the associated statistic to zero. Any read returns it's
- * current value.
- */
-static int read_reset_stat(struct ctl_table *table, int write,
-			   void __user *buffer, size_t *lenp,
-			   loff_t *ppos)
+static int svcrdma_counter_handler(const struct ctl_table *table, int write,
+				   void *buffer, size_t *lenp, loff_t *ppos)
 {
-	atomic_t *stat = (atomic_t *)table->data;
+	struct percpu_counter *stat = (struct percpu_counter *)table->data;
+	char tmp[SVCRDMA_COUNTER_BUFSIZ + 1];
+	int len;
 
-	if (!stat)
-		return -EINVAL;
+	if (write) {
+		percpu_counter_set(stat, 0);
+		return 0;
+	}
 
-	if (write)
-		atomic_set(stat, 0);
-	else {
-		char str_buf[32];
-		char *data;
-		int len = snprintf(str_buf, 32, "%d\n", atomic_read(stat));
-		if (len >= 32)
-			return -EFAULT;
-		len = strlen(str_buf);
-		if (*ppos > len) {
-			*lenp = 0;
-			return 0;
-		}
-		data = &str_buf[*ppos];
-		len -= *ppos;
-		if (len > *lenp)
-			len = *lenp;
-		if (len && copy_to_user(buffer, str_buf, len))
-			return -EFAULT;
-		*lenp = len;
-		*ppos += len;
+	len = snprintf(tmp, SVCRDMA_COUNTER_BUFSIZ, "%lld\n",
+		       percpu_counter_sum_positive(stat));
+	if (len >= SVCRDMA_COUNTER_BUFSIZ)
+		return -EFAULT;
+	len = strlen(tmp);
+	if (*ppos > len) {
+		*lenp = 0;
+		return 0;
 	}
+	len -= *ppos;
+	if (len > *lenp)
+		len = *lenp;
+	if (len)
+		memcpy(buffer, tmp, len);
+	*lenp = len;
+	*ppos += len;
+
 	return 0;
 }
 
@@ -150,153 +138,170 @@ static struct ctl_table svcrdma_parm_table[] = {
 
 	{
 		.procname	= "rdma_stat_read",
-		.data		= &rdma_stat_read,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_read,
+		.maxlen		= SVCRDMA_COUNTER_BUFSIZ,
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= svcrdma_counter_handler,
 	},
 	{
 		.procname	= "rdma_stat_recv",
-		.data		= &rdma_stat_recv,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_recv,
+		.maxlen		= SVCRDMA_COUNTER_BUFSIZ,
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= svcrdma_counter_handler,
 	},
 	{
 		.procname	= "rdma_stat_write",
-		.data		= &rdma_stat_write,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_write,
+		.maxlen		= SVCRDMA_COUNTER_BUFSIZ,
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= svcrdma_counter_handler,
 	},
 	{
 		.procname	= "rdma_stat_sq_starve",
-		.data		= &rdma_stat_sq_starve,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_sq_starve,
+		.maxlen		= SVCRDMA_COUNTER_BUFSIZ,
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= svcrdma_counter_handler,
 	},
 	{
 		.procname	= "rdma_stat_rq_starve",
-		.data		= &rdma_stat_rq_starve,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_unused,
+		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &zero,
 	},
 	{
 		.procname	= "rdma_stat_rq_poll",
-		.data		= &rdma_stat_rq_poll,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_unused,
+		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &zero,
 	},
 	{
 		.procname	= "rdma_stat_rq_prod",
-		.data		= &rdma_stat_rq_prod,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_unused,
+		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &zero,
 	},
 	{
 		.procname	= "rdma_stat_sq_poll",
-		.data		= &rdma_stat_sq_poll,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_unused,
+		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &zero,
 	},
 	{
 		.procname	= "rdma_stat_sq_prod",
-		.data		= &rdma_stat_sq_prod,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_unused,
+		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &zero,
 	},
-	{ },
 };
 
-static struct ctl_table svcrdma_table[] = {
-	{
-		.procname	= "svc_rdma",
-		.mode		= 0555,
-		.child		= svcrdma_parm_table
-	},
-	{ },
-};
+static void svc_rdma_proc_cleanup(void)
+{
+	if (!svcrdma_table_header)
+		return;
+	unregister_sysctl_table(svcrdma_table_header);
+	svcrdma_table_header = NULL;
 
-static struct ctl_table svcrdma_root_table[] = {
-	{
-		.procname	= "sunrpc",
-		.mode		= 0555,
-		.child		= svcrdma_table
-	},
-	{ },
-};
+	percpu_counter_destroy(&svcrdma_stat_write);
+	percpu_counter_destroy(&svcrdma_stat_sq_starve);
+	percpu_counter_destroy(&svcrdma_stat_recv);
+	percpu_counter_destroy(&svcrdma_stat_read);
+}
+
+static int svc_rdma_proc_init(void)
+{
+	int rc;
+
+	if (svcrdma_table_header)
+		return 0;
+
+	rc = percpu_counter_init(&svcrdma_stat_read, 0, GFP_KERNEL);
+	if (rc)
+		goto err;
+	rc = percpu_counter_init(&svcrdma_stat_recv, 0, GFP_KERNEL);
+	if (rc)
+		goto err_read;
+	rc = percpu_counter_init(&svcrdma_stat_sq_starve, 0, GFP_KERNEL);
+	if (rc)
+		goto err_recv;
+	rc = percpu_counter_init(&svcrdma_stat_write, 0, GFP_KERNEL);
+	if (rc)
+		goto err_sq;
+
+	svcrdma_table_header = register_sysctl("sunrpc/svc_rdma",
+					       svcrdma_parm_table);
+	if (!svcrdma_table_header)
+		goto err_write;
+
+	return 0;
+
+err_write:
+	rc = -ENOMEM;
+	percpu_counter_destroy(&svcrdma_stat_write);
+err_sq:
+	percpu_counter_destroy(&svcrdma_stat_sq_starve);
+err_recv:
+	percpu_counter_destroy(&svcrdma_stat_recv);
+err_read:
+	percpu_counter_destroy(&svcrdma_stat_read);
+err:
+	return rc;
+}
+
+struct workqueue_struct *svcrdma_wq;
 
 void svc_rdma_cleanup(void)
 {
-	dprintk("SVCRDMA Module Removed, deregister RPC RDMA transport\n");
-	destroy_workqueue(svc_rdma_wq);
-	if (svcrdma_table_header) {
-		unregister_sysctl_table(svcrdma_table_header);
-		svcrdma_table_header = NULL;
-	}
 	svc_unreg_xprt_class(&svc_rdma_class);
-	kmem_cache_destroy(svc_rdma_map_cachep);
-	kmem_cache_destroy(svc_rdma_ctxt_cachep);
+	svc_rdma_proc_cleanup();
+	if (svcrdma_wq) {
+		struct workqueue_struct *wq = svcrdma_wq;
+
+		svcrdma_wq = NULL;
+		destroy_workqueue(wq);
+	}
+
+	dprintk("SVCRDMA Module Removed, deregister RPC RDMA transport\n");
 }
 
 int svc_rdma_init(void)
 {
-	dprintk("SVCRDMA Module Init, register RPC RDMA transport\n");
-	dprintk("\tsvcrdma_ord      : %d\n", svcrdma_ord);
-	dprintk("\tmax_requests     : %d\n", svcrdma_max_requests);
-	dprintk("\tsq_depth         : %d\n",
-		svcrdma_max_requests * RPCRDMA_SQ_DEPTH_MULT);
-	dprintk("\tmax_inline       : %d\n", svcrdma_max_req_size);
+	struct workqueue_struct *wq;
+	int rc;
 
-	svc_rdma_wq = alloc_workqueue("svc_rdma", 0, 0);
-	if (!svc_rdma_wq)
+	wq = alloc_workqueue("svcrdma", WQ_UNBOUND, 0);
+	if (!wq)
 		return -ENOMEM;
 
-	if (!svcrdma_table_header)
-		svcrdma_table_header =
-			register_sysctl_table(svcrdma_root_table);
-
-	/* Create the temporary map cache */
-	svc_rdma_map_cachep = kmem_cache_create("svc_rdma_map_cache",
-						sizeof(struct svc_rdma_req_map),
-						0,
-						SLAB_HWCACHE_ALIGN,
-						NULL);
-	if (!svc_rdma_map_cachep) {
-		printk(KERN_INFO "Could not allocate map cache.\n");
-		goto err0;
+	rc = svc_rdma_proc_init();
+	if (rc) {
+		destroy_workqueue(wq);
+		return rc;
 	}
 
-	/* Create the temporary context cache */
-	svc_rdma_ctxt_cachep =
-		kmem_cache_create("svc_rdma_ctxt_cache",
-				  sizeof(struct svc_rdma_op_ctxt),
-				  0,
-				  SLAB_HWCACHE_ALIGN,
-				  NULL);
-	if (!svc_rdma_ctxt_cachep) {
-		printk(KERN_INFO "Could not allocate WR ctxt cache.\n");
-		goto err1;
-	}
-
-	/* Register RDMA with the SVC transport switch */
+	svcrdma_wq = wq;
 	svc_reg_xprt_class(&svc_rdma_class);
+
+	dprintk("SVCRDMA Module Init, register RPC RDMA transport\n");
+	dprintk("\tsvcrdma_ord      : %d\n", svcrdma_ord);
+	dprintk("\tmax_requests     : %u\n", svcrdma_max_requests);
+	dprintk("\tmax_bc_requests  : %u\n", svcrdma_max_bc_requests);
+	dprintk("\tmax_inline       : %d\n", svcrdma_max_req_size);
 	return 0;
- err1:
-	kmem_cache_destroy(svc_rdma_map_cachep);
- err0:
-	unregister_sysctl_table(svcrdma_table_header);
-	destroy_workqueue(svc_rdma_wq);
-	return -ENOMEM;
 }
-MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>");
-MODULE_DESCRIPTION("SVC RDMA Transport");
-MODULE_LICENSE("Dual BSD/GPL");
-module_init(svc_rdma_init);
-module_exit(svc_rdma_cleanup);