LU-9120 lnet: set retry count from lnetctl

Added an lnetctl command to set the retry_count from userspace.

lnetctl set retry_count [0|>0]

0 - turns off retries in the system
>0 - number of retries.

Test-Parameters: forbuildonly
Signed-off-by: Amir Shehata <ashehata@whamcloud.com>
Change-Id: I2fd5c88a91590195cfdad52e6d177619ccbbc840
Reviewed-on: https://review.whamcloud.com/32777
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Sonia Sharma <sharmaso@whamcloud.com>
Tested-by: Jenkins

diff --git a/lnet/utils/lnetconfig/liblnetconfig.c b/lnet/utils/lnetconfig/liblnetconfig.c
index 32ed598..4ae90e8 100644 (file)
--- a/lnet/utils/lnetconfig/liblnetconfig.c
+++ b/lnet/utils/lnetconfig/liblnetconfig.c
@@ -2429,6 +2429,28 @@ int ioctl_set_value(__u32 val, int ioc, char *name,
        return rc;
 }
 
+int lustre_lnet_config_retry_count(int count, int seq_no, struct cYAML **err_rc)
+{
+       int rc = LUSTRE_CFG_RC_NO_ERR;
+       char err_str[LNET_MAX_STR_LEN];
+       char val[LNET_MAX_STR_LEN];
+
+       snprintf(err_str, sizeof(err_str), "\"success\"");
+
+       snprintf(val, sizeof(val), "%d", count);
+
+       rc = write_sysfs_file(modparam_path, "lnet_retry_count", val,
+                             1, strlen(val) + 1);
+       if (rc)
+               snprintf(err_str, sizeof(err_str),
+                        "\"cannot configure retry count: %s\"",
+                        strerror(errno));
+
+       cYAML_build_error(rc, seq_no, ADD_CMD, "retry_count", err_str, err_rc);
+
+       return rc;
+}
+
 int lustre_lnet_config_max_intf(int max, int seq_no, struct cYAML **err_rc)
 {
        int rc = LUSTRE_CFG_RC_NO_ERR;
@@ -3187,6 +3209,31 @@ static int ioctl_show_global_values(int ioc, int seq_no, char *name,
                                       data.sv_value, show_rc, err_rc, l_errno);
 }
 
+int lustre_lnet_show_retry_count(int seq_no, struct cYAML **show_rc,
+                                struct cYAML **err_rc)
+{
+       int rc = LUSTRE_CFG_RC_OUT_OF_MEM;
+       char val[LNET_MAX_STR_LEN];
+       int retry_count = -1, l_errno = 0;
+       char err_str[LNET_MAX_STR_LEN];
+
+       snprintf(err_str, sizeof(err_str), "\"out of memory\"");
+
+       rc = read_sysfs_file(modparam_path, "lnet_retry_count", val,
+                            1, sizeof(val));
+       if (rc) {
+               l_errno = -errno;
+               snprintf(err_str, sizeof(err_str),
+                        "\"cannot get retry count: %d\"", rc);
+       } else {
+               retry_count = atoi(val);
+       }
+
+       return build_global_yaml_entry(err_str, sizeof(err_str), seq_no,
+                                      "retry_count", retry_count, show_rc,
+                                      err_rc, l_errno);
+}
+
 int lustre_lnet_show_max_intf(int seq_no, struct cYAML **show_rc,
                              struct cYAML **err_rc)
 {
@@ -4083,7 +4130,7 @@ static int handle_yaml_config_global_settings(struct cYAML *tree,
                                              struct cYAML **show_rc,
                                              struct cYAML **err_rc)
 {
-       struct cYAML *max_intf, *numa, *discovery, *seq_no;
+       struct cYAML *max_intf, *numa, *discovery, *retry, *seq_no;
        int rc = 0;
 
        seq_no = cYAML_get_object_item(tree, "seq_no");
@@ -4108,6 +4155,13 @@ static int handle_yaml_config_global_settings(struct cYAML *tree,
                                                        : -1,
                                                  err_rc);
 
+       retry = cYAML_get_object_item(tree, "retry_count");
+       if (retry)
+               rc = lustre_lnet_config_retry_count(retry->cy_valueint,
+                                                   seq_no ? seq_no->cy_valueint
+                                                       : -1,
+                                                   err_rc);
+
        return rc;
 }
 
@@ -4148,7 +4202,7 @@ static int handle_yaml_show_global_settings(struct cYAML *tree,
                                            struct cYAML **show_rc,
                                            struct cYAML **err_rc)
 {
-       struct cYAML *max_intf, *numa, *discovery, *seq_no;
+       struct cYAML *max_intf, *numa, *discovery, *retry, *seq_no;
        int rc = 0;
 
        seq_no = cYAML_get_object_item(tree, "seq_no");
@@ -4170,6 +4224,13 @@ static int handle_yaml_show_global_settings(struct cYAML *tree,
                                                        : -1,
                                                show_rc, err_rc);
 
+       retry = cYAML_get_object_item(tree, "retry_count");
+       if (retry)
+               rc = lustre_lnet_show_retry_count(seq_no ? seq_no->cy_valueint
+                                                       : -1,
+                                                 show_rc, err_rc);
+
+
        return rc;
 }
 

diff --git a/lnet/utils/lnetconfig/liblnetconfig.h b/lnet/utils/lnetconfig/liblnetconfig.h
index ce005d1..55348a0 100644 (file)
--- a/lnet/utils/lnetconfig/liblnetconfig.h
+++ b/lnet/utils/lnetconfig/liblnetconfig.h
@@ -225,6 +225,29 @@ int lustre_lnet_show_numa_range(int seq_no, struct cYAML **show_rc,
                                struct cYAML **err_rc);
 
 /*
+ * lustre_lnet_config_retry_count
+ *   sets the maximum number of retries to resend a message
+ *
+ *   count - maximum value to configure
+ *   seq_no - sequence number of the request
+ *   err_rc - [OUT] struct cYAML tree describing the error. Freed by
+ *   caller
+ */
+int lustre_lnet_config_retry_count(int count, int seq_no, struct cYAML **err_rc);
+
+/*
+ * lustre_lnet_show_retry_count
+ *    show current maximum number of retries in the system
+ *
+ *   seq_no - sequence number of the request
+ *   show_rc - [OUT] struct cYAML tree containing retry count info
+ *   err_rc - [OUT] struct cYAML tree describing the error. Freed by
+ *   caller
+ */
+int lustre_lnet_show_retry_count(int seq_no, struct cYAML **show_rc,
+                                struct cYAML **err_rc);
+
+/*
  * lustre_lnet_config_max_intf
  *   Sets the maximum number of interfaces per node. this tunable is
  *   primarily useful for sanity checks prior to allocating memory.

diff --git a/lnet/utils/lnetctl.c b/lnet/utils/lnetctl.c
index e14d43e..27ba97d 100644 (file)
--- a/lnet/utils/lnetctl.c
+++ b/lnet/utils/lnetctl.c
@@ -53,6 +53,7 @@ static int jt_set_tiny(int argc, char **argv);
 static int jt_set_small(int argc, char **argv);
 static int jt_set_large(int argc, char **argv);
 static int jt_set_numa(int argc, char **argv);
+static int jt_set_retry_count(int argc, char **argv);
 static int jt_add_peer_nid(int argc, char **argv);
 static int jt_del_peer_nid(int argc, char **argv);
 static int jt_set_max_intf(int argc, char **argv);
@@ -174,6 +175,9 @@ command_t set_cmds[] = {
        {"discovery", jt_set_discovery, 0, "enable/disable peer discovery\n"
         "\t0 - disable peer discovery\n"
         "\t1 - enable peer discovery (default)\n"},
+       {"retry_count", jt_set_retry_count, 0, "number of retries\n"
+        "\t0 - turn of retries\n"
+        "\t>0 - number of retries\n"},
        { 0, 0, 0, NULL }
 };
 
@@ -328,6 +332,34 @@ static int jt_set_numa(int argc, char **argv)
        return rc;
 }
 
+static int jt_set_retry_count(int argc, char **argv)
+{
+       long int value;
+       int rc;
+       struct cYAML *err_rc = NULL;
+
+       rc = check_cmd(set_cmds, "set", "retry_count", 2, argc, argv);
+       if (rc)
+               return rc;
+
+       rc = parse_long(argv[1], &value);
+       if (rc != 0) {
+               cYAML_build_error(-1, -1, "parser", "set",
+                                 "cannot parse retry_count value", &err_rc);
+               cYAML_print_tree2file(stderr, err_rc);
+               cYAML_free_tree(err_rc);
+               return -1;
+       }
+
+       rc = lustre_lnet_config_retry_count(value, -1, &err_rc);
+       if (rc != LUSTRE_CFG_RC_NO_ERR)
+               cYAML_print_tree2file(stderr, err_rc);
+
+       cYAML_free_tree(err_rc);
+
+       return rc;
+}
+
 static int jt_set_discovery(int argc, char **argv)
 {
        long int value;
@@ -980,6 +1012,12 @@ static int jt_show_global(int argc, char **argv)
                goto out;
        }
 
+       rc = lustre_lnet_show_retry_count(-1, &show_rc, &err_rc);
+       if (rc != LUSTRE_CFG_RC_NO_ERR) {
+               cYAML_print_tree2file(stderr, err_rc);
+               goto out;
+       }
+
        if (show_rc)
                cYAML_print_tree(show_rc);