All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ross Lagerwall <ross.lagerwall-Sxgqhf6Nn4DQT0dZR+AlfA@public.gmane.org>
To: <linux-cifs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
Cc: Steve French <sfrench-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org>,
	Ross Lagerwall
	<ross.lagerwall-Sxgqhf6Nn4DQT0dZR+AlfA@public.gmane.org>
Subject: [PATCH] cifs: Limit the overall credit acquired
Date: Tue, 20 Sep 2016 13:37:13 +0100	[thread overview]
Message-ID: <1474375033-19895-1-git-send-email-ross.lagerwall@citrix.com> (raw)

The kernel client requests 2 credits for many operations even though
they only use 1 credit (presumably to build up a buffer of credit).
Some servers seem to give the client as much credit as is requested.  In
this case, the amount of credit the client has continues increasing to
the point where (server->credits * MAX_BUFFER_SIZE) overflows in
smb2_wait_mtu_credits().

Fix this by throttling the credit requests if an set limit is reached.
For async requests where the credit charge may be > 1, request as much
credit as what is charged.
The limit is chosen somewhat arbitrarily. The Windows client
defaults to 128 credits, the Windows server allows clients up to
512 credits, and the NetApp server does not limit clients at all.
Choose a high enough value such that the client shouldn't limit
performance.

This behavior was seen with a NetApp filer (NetApp Release 9.0RC2).

Signed-off-by: Ross Lagerwall <ross.lagerwall-Sxgqhf6Nn4DQT0dZR+AlfA@public.gmane.org>
---
 fs/cifs/smb2glob.h | 10 ++++++++++
 fs/cifs/smb2pdu.c  | 18 +++++++++++++++++-
 2 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/smb2glob.h b/fs/cifs/smb2glob.h
index 0ffa180..0e1d788 100644
--- a/fs/cifs/smb2glob.h
+++ b/fs/cifs/smb2glob.h
@@ -61,4 +61,14 @@
 /* Maximum buffer size value we can send with 1 credit */
 #define SMB2_MAX_BUFFER_SIZE 65536
 
+/*
+ * Maximum number of credits to keep available.
+ * This value is chosen somewhat arbitrarily. The Windows client
+ * defaults to 128 credits, the Windows server allows clients up to
+ * 512 credits, and the NetApp server does not limit clients at all.
+ * Choose a high enough value such that the client shouldn't limit
+ * performance.
+ */
+#define SMB2_MAX_CREDITS_AVAILABLE 1024
+
 #endif	/* _SMB2_GLOB_H */
diff --git a/fs/cifs/smb2pdu.c b/fs/cifs/smb2pdu.c
index 29e06db..967a790 100644
--- a/fs/cifs/smb2pdu.c
+++ b/fs/cifs/smb2pdu.c
@@ -100,7 +100,21 @@ smb2_hdr_assemble(struct smb2_hdr *hdr, __le16 smb2_cmd /* command */ ,
 	hdr->ProtocolId = SMB2_PROTO_NUMBER;
 	hdr->StructureSize = cpu_to_le16(64);
 	hdr->Command = smb2_cmd;
-	hdr->CreditRequest = cpu_to_le16(2); /* BB make this dynamic */
+	if (tcon && tcon->ses && tcon->ses->server) {
+		struct TCP_Server_Info *server = tcon->ses->server;
+
+		spin_lock(&server->req_lock);
+		/* Request up to 2 credits but don't go over the limit. */
+		if (server->credits >= SMB2_MAX_CREDITS_AVAILABLE)
+			hdr->CreditRequest = cpu_to_le16(0);
+		else
+			hdr->CreditRequest = cpu_to_le16(
+				min_t(int, SMB2_MAX_CREDITS_AVAILABLE -
+						server->credits, 2));
+		spin_unlock(&server->req_lock);
+	} else {
+		hdr->CreditRequest = cpu_to_le16(2);
+	}
 	hdr->ProcessId = cpu_to_le32((__u16)current->tgid);
 
 	if (!tcon)
@@ -2057,6 +2071,7 @@ smb2_async_readv(struct cifs_readdata *rdata)
 	if (rdata->credits) {
 		buf->CreditCharge = cpu_to_le16(DIV_ROUND_UP(rdata->bytes,
 						SMB2_MAX_BUFFER_SIZE));
+		buf->CreditRequest = buf->CreditCharge;
 		spin_lock(&server->req_lock);
 		server->credits += rdata->credits -
 						le16_to_cpu(buf->CreditCharge);
@@ -2243,6 +2258,7 @@ smb2_async_writev(struct cifs_writedata *wdata,
 	if (wdata->credits) {
 		req->hdr.CreditCharge = cpu_to_le16(DIV_ROUND_UP(wdata->bytes,
 						    SMB2_MAX_BUFFER_SIZE));
+		req->hdr.CreditRequest = req->hdr.CreditCharge;
 		spin_lock(&server->req_lock);
 		server->credits += wdata->credits -
 					le16_to_cpu(req->hdr.CreditCharge);
-- 
2.7.4

             reply	other threads:[~2016-09-20 12:37 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-09-20 12:37 Ross Lagerwall [this message]
     [not found] ` <1474375033-19895-1-git-send-email-ross.lagerwall-Sxgqhf6Nn4DQT0dZR+AlfA@public.gmane.org>
2016-09-20 20:00   ` [PATCH] cifs: Limit the overall credit acquired Steve French

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1474375033-19895-1-git-send-email-ross.lagerwall@citrix.com \
    --to=ross.lagerwall-sxgqhf6nn4dqt0dzr+alfa@public.gmane.org \
    --cc=linux-cifs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=sfrench-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.