[PATCH v6] rsockets: fine grained interception mechanism for rsocket preloading

[PATCH v6] rsockets: fine grained interception mechanism for rsocket preloading

[Sreedhar Kodali]

[-- Attachment #1: Type: text/plain, Size: 8061 bytes --]

Note: Minor modification to the latest change set from Sean
       Also attached the patch file for convenience

 From: Sreedhar Kodali <srkodali-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Date:   Fri Sep 26 12:29:00 2014 +0530

     By default the R-Sockets pre-loading library intercepts all
     the stream and datagram sockets belonging to a launched
     program processes and threads.

     However, distributed application and database servers may
     require fine grained interception to ensure that only the
     processes which are listening for remote connections on the
     RDMA transport need to be enabled with RDMA while remaining
     can continue to use TCP as before.  This allows proper
     communication happening between various server components locally.

     A configuration file based mechanism is introduced to facilitate
     this fine grained interception mechanism.  As part of preload
     initialization, the configuration file is scanned and an
     in-memory record store is created with all the entries found.
     When a request is made to intercept a socket, its attributes
     are cross checked with stored records to see whether we
     should proceed with rsocket switch over.

     Note: Right now, the fine grained interception mechanism is
     enabled only for newly created sockets.  Going forward,
     this can be extened to select connections based on the
     specified host/IP addresses and ports as well.

     "preload_config" is the name of the configuration file which
     should exist in the default configuration location
     (usually the full path to this configuration file is:
     <install-root>/etc/rdma/rsocket/preload_config)
     of an installed rsocket library.

     The sample format for this configuration file is shown below:

     Signed-off-by: Sreedhar Kodali <srkodali-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
     Reviewed-by: Pradeep Satyanarayana <pradeeps-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
     Signed-off-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
     ---
     I made several adjustments to the submitted patch.  Please
     verify that these work for you.  Changes from v5:

     - Simplified input file format slightly.
     - Removed typedef
     - Rename entryp and removed unneeded variable
     - Replaced token parsing with sscanf
     - Added wildcard support for the program name
     - Wildcard values now stored as 0
     - Enhanced checks for domain, type, and protocol strings
     - Fixed realloc error handling
     - Simplified free_config implementation
     - Added support for app passing in 0 for protocol

     Additional minor change:
     - Compare only first n chars of program name

diff --git a/src/preload.c b/src/preload.c
index fb2149b..1e62a06 100644
--- a/src/preload.c
+++ b/src/preload.c
@@ -50,6 +50,9 @@
  #include <netinet/tcp.h>
  #include <unistd.h>
  #include <semaphore.h>
+#include <ctype.h>
+#include <stdlib.h>
+#include <stdio.h>

  #include <rdma/rdma_cma.h>
  #include <rdma/rdma_verbs.h>
@@ -122,6 +125,135 @@ struct fd_info {
         atomic_t refcnt;
  };

+struct config_entry {
+       char *name;
+       int domain;
+       int type;
+       int protocol;
+};
+
+static struct config_entry *config;
+static int config_cnt;
+extern char *program_invocation_short_name;
+
+
+static void free_config(void)
+{
+       while (config_cnt)
+               free(config[--config_cnt].name);
+
+       free(config);
+}
+
+/*
+ * Config file format:
+ * # Starting '#' indicates comment
+ * # wild card values are supported using '*'
+ * # domain - *, INET, INET6, IB
+ * # type - *, STREAM, DGRAM
+ * # protocol - *, TCP, UDP
+ * program_name domain type protocol
+ */
+static void scan_config(void)
+{
+       struct config_entry *new_config;
+       FILE *fp;
+       char line[120], prog[64], dom[16], type[16], proto[16];
+
+       fp = fopen(RS_CONF_DIR "/preload_config", "r");
+       if (!fp)
+               return;
+
+       while (fgets(line, sizeof(line), fp)) {
+               if (line[0] == '#')
+                       continue;
+
+               if (sscanf(line, "%64s%16s%16s%16s", prog, dom, type, 
proto) != 4)
+                       continue;
+
+               new_config = realloc(config, (config_cnt + 1) *
+                                            sizeof(struct 
config_entry));
+               if (!new_config)
+                       break;
+
+               config = new_config;
+               memset(&config[config_cnt], 0, sizeof(struct 
config_entry));
+
+               if (!strcasecmp(dom, "INET") ||
+                   !strcasecmp(dom, "AF_INET") ||
+                   !strcasecmp(dom, "PF_INET")) {
+                       config[config_cnt].domain = AF_INET;
+               } else if (!strcasecmp(dom, "INET6") ||
+                          !strcasecmp(dom, "AF_INET6") ||
+                          !strcasecmp(dom, "PF_INET6")) {
+                       config[config_cnt].domain = AF_INET6;
+               } else if (!strcasecmp(dom, "IB") ||
+                          !strcasecmp(dom, "AF_IB") ||
+                          !strcasecmp(dom, "PF_IB")) {
+                       config[config_cnt].domain = AF_IB;
+               } else if (strcmp(dom, "*")) {
+                       continue;
+               }
+
+               if (!strcasecmp(type, "STREAM") ||
+                   !strcasecmp(type, "SOCK_STREAM")) {
+                       config[config_cnt].type = SOCK_STREAM;
+               } else if (!strcasecmp(type, "DGRAM") ||
+                          !strcasecmp(type, "SOCK_DGRAM")) {
+                       config[config_cnt].type = SOCK_DGRAM;
+               } else if (strcmp(type, "*")) {
+                       continue;
+               }
+
+               if (!strcasecmp(proto, "TCP") ||
+                   !strcasecmp(proto, "IPPROTO_TCP")) {
+                       config[config_cnt].protocol = IPPROTO_TCP;
+               } else if (!strcasecmp(proto, "UDP") ||
+                          !strcasecmp(proto, "IPPROTO_UDP")) {
+                       config[config_cnt].protocol = IPPROTO_UDP;
+               } else if (strcmp(proto, "*")) {
+                       continue;
+               }
+
+               if (strcmp(prog, "*")) {
+                   if (!(config[config_cnt].name = strdup(prog)))
+                           continue;
+               }
+
+               config_cnt++;
+       }
+
+       fclose(fp);
+       if (config_cnt)
+               atexit(free_config);
+}
+
+static int intercept_socket(int domain, int type, int protocol)
+{
+       int i;
+
+       if (!config_cnt)
+               return 1;
+
+       if (!protocol) {
+               if (type == SOCK_STREAM)
+                       protocol = IPPROTO_TCP;
+               else if (type == SOCK_DGRAM)
+                       protocol = IPPROTO_UDP;
+       }
+
+       for (i = 0; i < config_cnt; i++) {
+               if ((!config[i].name ||
+                    !strncasecmp(config[i].name, 
program_invocation_short_name, strlen(config[i].name))) &&
+                   (!config[i].domain || config[i].domain == domain) &&
+                   (!config[i].type || config[i].type == type) &&
+                   (!config[i].protocol || config[i].protocol == 
protocol))
+                       return 1;
+       }
+
+       return 0;
+}
+
  static int fd_open(void)
  {
         struct fd_info *fdi;
@@ -308,6 +440,7 @@ static void init_preload(void)
         rs.fcntl = dlsym(RTLD_DEFAULT, "rfcntl");

         getenv_options();
+       scan_config();
         init = 1;
  out:
         pthread_mutex_unlock(&mut);
@@ -404,10 +537,11 @@ int socket(int domain, int type, int protocol)
         static __thread int recursive;
         int index, ret;

-       if (recursive)
+       init_preload();
+
+       if (recursive || !intercept_socket(domain, type, protocol))
                 goto real;

-       init_preload();
         index = fd_open();
         if (index < 0)
                 return index;

[-- Attachment #2: librdmacm_src_rsocket_fine_grained_20140926.patch1 --]
[-- Type: text/plain, Size: 6605 bytes --]

commit 5ad8c6892e56e7e7c8cd3a1ac5702568a4a3847e
Author: Sreedhar Kodali <srkodali@linux.vnet.ibm.com>
Date:   Fri Sep 26 12:29:00 2014 +0530

    By default the R-Sockets pre-loading library intercepts all
    the stream and datagram sockets belonging to a launched
    program processes and threads.
    
    However, distributed application and database servers may
    require fine grained interception to ensure that only the
    processes which are listening for remote connections on the
    RDMA transport need to be enabled with RDMA while remaining
    can continue to use TCP as before.  This allows proper
    communication happening between various server components locally.
    
    A configuration file based mechanism is introduced to facilitate
    this fine grained interception mechanism.  As part of preload
    initialization, the configuration file is scanned and an
    in-memory record store is created with all the entries found.
    When a request is made to intercept a socket, its attributes
    are cross checked with stored records to see whether we
    should proceed with rsocket switch over.
    
    Note: Right now, the fine grained interception mechanism is
    enabled only for newly created sockets.  Going forward,
    this can be extened to select connections based on the
    specified host/IP addresses and ports as well.
    
    "preload_config" is the name of the configuration file which
    should exist in the default configuration location
    (usually the full path to this configuration file is:
    <install-root>/etc/rdma/rsocket/preload_config)
    of an installed rsocket library.
    
    The sample format for this configuration file is shown below:
    
    Signed-off-by: Sreedhar Kodali <srkodali@linux.vnet.ibm.com>
    Reviewed-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com>
    Signed-off-by: Sean Hefty <sean.hefty@intel.com>
    ---
    I made several adjustments to the submitted patch.  Please
    verify that these work for you.  Changes from v5:
    
    - Simplified input file format slightly.
    - Removed typedef
    - Rename entryp and removed unneeded variable
    - Replaced token parsing with sscanf
    - Added wildcard support for the program name
    - Wildcard values now stored as 0
    - Enhanced checks for domain, type, and protocol strings
    - Fixed realloc error handling
    - Simplified free_config implementation
    - Added support for app passing in 0 for protocol
    
    Additional minor change:
    - Compare only first n chars of program name

diff --git a/src/preload.c b/src/preload.c
index fb2149b..1e62a06 100644
--- a/src/preload.c
+++ b/src/preload.c
@@ -50,6 +50,9 @@
 #include <netinet/tcp.h>
 #include <unistd.h>
 #include <semaphore.h>
+#include <ctype.h>
+#include <stdlib.h>
+#include <stdio.h>
 
 #include <rdma/rdma_cma.h>
 #include <rdma/rdma_verbs.h>
@@ -122,6 +125,135 @@ struct fd_info {
 	atomic_t refcnt;
 };
 
+struct config_entry {
+	char *name;
+	int domain;
+	int type;
+	int protocol;
+};
+
+static struct config_entry *config;
+static int config_cnt;
+extern char *program_invocation_short_name;
+
+
+static void free_config(void)
+{
+	while (config_cnt)
+		free(config[--config_cnt].name);
+
+	free(config);
+}
+
+/*
+ * Config file format:
+ * # Starting '#' indicates comment
+ * # wild card values are supported using '*'
+ * # domain - *, INET, INET6, IB
+ * # type - *, STREAM, DGRAM
+ * # protocol - *, TCP, UDP
+ * program_name domain type protocol
+ */
+static void scan_config(void)
+{
+	struct config_entry *new_config;
+	FILE *fp;
+	char line[120], prog[64], dom[16], type[16], proto[16];
+
+	fp = fopen(RS_CONF_DIR "/preload_config", "r");
+	if (!fp)
+		return;
+
+	while (fgets(line, sizeof(line), fp)) {
+		if (line[0] == '#')
+			continue;
+
+		if (sscanf(line, "%64s%16s%16s%16s", prog, dom, type, proto) != 4)
+			continue;
+
+		new_config = realloc(config, (config_cnt + 1) *
+					     sizeof(struct config_entry));
+		if (!new_config)
+			break;
+
+		config = new_config;
+		memset(&config[config_cnt], 0, sizeof(struct config_entry));
+
+		if (!strcasecmp(dom, "INET") ||
+		    !strcasecmp(dom, "AF_INET") ||
+		    !strcasecmp(dom, "PF_INET")) {
+			config[config_cnt].domain = AF_INET;
+		} else if (!strcasecmp(dom, "INET6") ||
+			   !strcasecmp(dom, "AF_INET6") ||
+			   !strcasecmp(dom, "PF_INET6")) {
+			config[config_cnt].domain = AF_INET6;
+		} else if (!strcasecmp(dom, "IB") ||
+			   !strcasecmp(dom, "AF_IB") ||
+			   !strcasecmp(dom, "PF_IB")) {
+			config[config_cnt].domain = AF_IB;
+		} else if (strcmp(dom, "*")) {
+			continue;
+		}
+
+		if (!strcasecmp(type, "STREAM") ||
+		    !strcasecmp(type, "SOCK_STREAM")) {
+			config[config_cnt].type = SOCK_STREAM;
+		} else if (!strcasecmp(type, "DGRAM") ||
+			   !strcasecmp(type, "SOCK_DGRAM")) {
+			config[config_cnt].type = SOCK_DGRAM;
+		} else if (strcmp(type, "*")) {
+			continue;
+		}
+
+		if (!strcasecmp(proto, "TCP") ||
+		    !strcasecmp(proto, "IPPROTO_TCP")) {
+			config[config_cnt].protocol = IPPROTO_TCP;
+		} else if (!strcasecmp(proto, "UDP") ||
+			   !strcasecmp(proto, "IPPROTO_UDP")) {
+			config[config_cnt].protocol = IPPROTO_UDP;
+		} else if (strcmp(proto, "*")) {
+			continue;
+		}
+
+		if (strcmp(prog, "*")) {
+		    if (!(config[config_cnt].name = strdup(prog)))
+			    continue;
+		}
+
+		config_cnt++;
+	}
+
+	fclose(fp);
+	if (config_cnt)
+		atexit(free_config);
+}
+
+static int intercept_socket(int domain, int type, int protocol)
+{
+	int i;
+
+	if (!config_cnt)
+		return 1;
+
+	if (!protocol) {
+		if (type == SOCK_STREAM)
+			protocol = IPPROTO_TCP;
+		else if (type == SOCK_DGRAM)
+			protocol = IPPROTO_UDP;
+	}
+
+	for (i = 0; i < config_cnt; i++) {
+		if ((!config[i].name ||
+		     !strncasecmp(config[i].name, program_invocation_short_name, strlen(config[i].name))) &&
+		    (!config[i].domain || config[i].domain == domain) &&
+		    (!config[i].type || config[i].type == type) &&
+		    (!config[i].protocol || config[i].protocol == protocol))
+			return 1;
+	}
+
+	return 0;
+}
+
 static int fd_open(void)
 {
 	struct fd_info *fdi;
@@ -308,6 +440,7 @@ static void init_preload(void)
 	rs.fcntl = dlsym(RTLD_DEFAULT, "rfcntl");
 
 	getenv_options();
+	scan_config();
 	init = 1;
 out:
 	pthread_mutex_unlock(&mut);
@@ -404,10 +537,11 @@ int socket(int domain, int type, int protocol)
 	static __thread int recursive;
 	int index, ret;
 
-	if (recursive)
+	init_preload();
+
+	if (recursive || !intercept_socket(domain, type, protocol))
 		goto real;
 
-	init_preload();
 	index = fd_open();
 	if (index < 0)
 		return index;

RE: [PATCH v6] rsockets: fine grained interception mechanism for rsocket preloading

[Hefty, Sean]

Thanks - all updates should be applied.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html