CIFS: Respect negotiated MaxMpxCount
authorPavel Shilovsky <piastry@etersoft.ru>
Tue, 20 Mar 2012 09:55:09 +0000 (12:55 +0300)
committerSteve French <stevef@smf-gateway.(none)>
Tue, 20 Mar 2012 15:17:40 +0000 (10:17 -0500)
Some servers sets this value less than 50 that was hardcoded and
we lost the connection if when we exceed this limit. Fix this by
respecting this value - not sending more than the server allows.

Cc: stable@kernel.org
Reviewed-by: Jeff Layton <jlayton@samba.org>
Signed-off-by: Pavel Shilovsky <piastry@etersoft.ru>
Signed-off-by: Steve French <stevef@smf-gateway.(none)>
fs/cifs/cifsfs.c
fs/cifs/cifsglob.h
fs/cifs/cifssmb.c
fs/cifs/connect.c
fs/cifs/dir.c
fs/cifs/file.c
fs/cifs/transport.c

index b1fd382..6ee1cb4 100644 (file)
@@ -76,7 +76,7 @@ MODULE_PARM_DESC(cifs_min_small, "Small network buffers in pool. Default: 30 "
 unsigned int cifs_max_pending = CIFS_MAX_REQ;
 module_param(cifs_max_pending, int, 0444);
 MODULE_PARM_DESC(cifs_max_pending, "Simultaneous requests to server. "
-                                  "Default: 50 Range: 2 to 256");
+                                  "Default: 32767 Range: 2 to 32767.");
 unsigned short echo_retries = 5;
 module_param(echo_retries, ushort, 0644);
 MODULE_PARM_DESC(echo_retries, "Number of echo attempts before giving up and "
@@ -1116,9 +1116,9 @@ init_cifs(void)
        if (cifs_max_pending < 2) {
                cifs_max_pending = 2;
                cFYI(1, "cifs_max_pending set to min of 2");
-       } else if (cifs_max_pending > 256) {
-               cifs_max_pending = 256;
-               cFYI(1, "cifs_max_pending set to max of 256");
+       } else if (cifs_max_pending > CIFS_MAX_REQ) {
+               cifs_max_pending = CIFS_MAX_REQ;
+               cFYI(1, "cifs_max_pending set to max of %u", CIFS_MAX_REQ);
        }
 
        rc = cifs_fscache_register();
index 76e7d8b..d47d20a 100644 (file)
 
 /*
  * MAX_REQ is the maximum number of requests that WE will send
- * on one socket concurrently. It also matches the most common
- * value of max multiplex returned by servers.  We may
- * eventually want to use the negotiated value (in case
- * future servers can handle more) when we are more confident that
- * we will not have problems oveloading the socket with pending
- * write data.
+ * on one socket concurrently.
  */
-#define CIFS_MAX_REQ 50
+#define CIFS_MAX_REQ 32767
 
 #define RFC1001_NAME_LEN 15
 #define RFC1001_NAME_LEN_WITH_NULL (RFC1001_NAME_LEN + 1)
@@ -263,6 +258,7 @@ struct TCP_Server_Info {
        bool session_estab; /* mark when very first sess is established */
        u16 dialect; /* dialect index that server chose */
        enum securityEnum secType;
+       bool oplocks:1; /* enable oplocks */
        unsigned int maxReq;    /* Clients should submit no more */
        /* than maxReq distinct unanswered SMBs to the server when using  */
        /* multiplexed reads or writes */
index 8b7794c..cd66b76 100644 (file)
@@ -458,7 +458,10 @@ CIFSSMBNegotiate(unsigned int xid, struct cifs_ses *ses)
                        goto neg_err_exit;
                }
                server->sec_mode = (__u8)le16_to_cpu(rsp->SecurityMode);
-               server->maxReq = le16_to_cpu(rsp->MaxMpxCount);
+               server->maxReq = min_t(unsigned int,
+                                      le16_to_cpu(rsp->MaxMpxCount),
+                                      cifs_max_pending);
+               server->oplocks = server->maxReq > 1 ? enable_oplocks : false;
                server->maxBuf = le16_to_cpu(rsp->MaxBufSize);
                server->max_vcs = le16_to_cpu(rsp->MaxNumberVcs);
                /* even though we do not use raw we might as well set this
@@ -564,7 +567,9 @@ CIFSSMBNegotiate(unsigned int xid, struct cifs_ses *ses)
 
        /* one byte, so no need to convert this or EncryptionKeyLen from
           little endian */
-       server->maxReq = le16_to_cpu(pSMBr->MaxMpxCount);
+       server->maxReq = min_t(unsigned int, le16_to_cpu(pSMBr->MaxMpxCount),
+                              cifs_max_pending);
+       server->oplocks = server->maxReq > 1 ? enable_oplocks : false;
        /* probably no need to store and check maxvcs */
        server->maxBuf = le32_to_cpu(pSMBr->MaxBufferSize);
        server->max_rw = le32_to_cpu(pSMBr->MaxRawSize);
index 602f77c..03f71fb 100644 (file)
@@ -642,14 +642,10 @@ static void clean_demultiplex_info(struct TCP_Server_Info *server)
        spin_unlock(&GlobalMid_Lock);
        wake_up_all(&server->response_q);
 
-       /*
-        * Check if we have blocked requests that need to free. Note that
-        * cifs_max_pending is normally 50, but can be set at module install
-        * time to as little as two.
-        */
+       /* Check if we have blocked requests that need to free. */
        spin_lock(&GlobalMid_Lock);
-       if (atomic_read(&server->inFlight) >= cifs_max_pending)
-               atomic_set(&server->inFlight, cifs_max_pending - 1);
+       if (atomic_read(&server->inFlight) >= server->maxReq)
+               atomic_set(&server->inFlight, server->maxReq - 1);
        /*
         * We do not want to set the max_pending too low or we could end up
         * with the counter going negative.
@@ -1910,6 +1906,7 @@ cifs_get_tcp_session(struct smb_vol *volume_info)
        tcp_ses->noautotune = volume_info->noautotune;
        tcp_ses->tcp_nodelay = volume_info->sockopt_tcp_nodelay;
        atomic_set(&tcp_ses->inFlight, 0);
+       tcp_ses->maxReq = 1; /* enough to send negotiate request */
        init_waitqueue_head(&tcp_ses->response_q);
        init_waitqueue_head(&tcp_ses->request_q);
        INIT_LIST_HEAD(&tcp_ses->pending_mid_q);
index bc7e244..d172c8e 100644 (file)
@@ -171,7 +171,7 @@ cifs_create(struct inode *inode, struct dentry *direntry, umode_t mode,
        }
        tcon = tlink_tcon(tlink);
 
-       if (enable_oplocks)
+       if (tcon->ses->server->oplocks)
                oplock = REQ_OPLOCK;
 
        if (nd)
@@ -492,7 +492,7 @@ cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
 {
        int xid;
        int rc = 0; /* to get around spurious gcc warning, set to zero here */
-       __u32 oplock = enable_oplocks ? REQ_OPLOCK : 0;
+       __u32 oplock;
        __u16 fileHandle = 0;
        bool posix_open = false;
        struct cifs_sb_info *cifs_sb;
@@ -518,6 +518,8 @@ cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
        }
        pTcon = tlink_tcon(tlink);
 
+       oplock = pTcon->ses->server->oplocks ? REQ_OPLOCK : 0;
+
        /*
         * Don't allow the separator character in a path component.
         * The VFS will not allow "/", but "\" is allowed by posix.
index 8e02dbd..159fcc5 100644 (file)
@@ -380,7 +380,7 @@ int cifs_open(struct inode *inode, struct file *file)
        cFYI(1, "inode = 0x%p file flags are 0x%x for %s",
                 inode, file->f_flags, full_path);
 
-       if (enable_oplocks)
+       if (tcon->ses->server->oplocks)
                oplock = REQ_OPLOCK;
        else
                oplock = 0;
@@ -505,7 +505,7 @@ static int cifs_reopen_file(struct cifsFileInfo *pCifsFile, bool can_flush)
        cFYI(1, "inode = 0x%p file flags 0x%x for %s",
                 inode, pCifsFile->f_flags, full_path);
 
-       if (enable_oplocks)
+       if (tcon->ses->server->oplocks)
                oplock = REQ_OPLOCK;
        else
                oplock = 0;
index 0cc9584..99a27cf 100644 (file)
@@ -265,12 +265,12 @@ static int wait_for_free_request(struct TCP_Server_Info *server,
 
        spin_lock(&GlobalMid_Lock);
        while (1) {
-               if (atomic_read(&server->inFlight) >= cifs_max_pending) {
+               if (atomic_read(&server->inFlight) >= server->maxReq) {
                        spin_unlock(&GlobalMid_Lock);
                        cifs_num_waiters_inc(server);
                        wait_event(server->request_q,
                                   atomic_read(&server->inFlight)
-                                    < cifs_max_pending);
+                                    < server->maxReq);
                        cifs_num_waiters_dec(server);
                        spin_lock(&GlobalMid_Lock);
                } else {