Merge pull request #72 from chu11/redfishpower_timeout

redfishpower: allow timeout to be set by device script

etc/devices/redfishpower-cray-r272z30.dev

@@ -10,11 +10,9 @@
 #
 # - This device specification was tested on a Cray with Gigabyte R272-Z30.
 #
-# - If using a slightly different model or BMC firmware version, its
-#   possible the URI paths will be different.  A good starting point
-#   is to run the following command to discover the correct paths:
-#
-#   curl -s -u USER:PASS -k -X GET https://pnode1/redfish/v1/Systems
+# - CAUTION: If you intend to use this file as the basis for a different
+#   Redfish system, read the section "UPDATING REDFISHPOWER DEVICE FILES"
+#   in redfishpower(8).
 #
 specification "redfishpower-cray-r272z30" {
 	timeout 	60
@@ -33,6 +31,8 @@ specification "redfishpower-cray-r272z30" {
 		expect "redfishpower> "
 		send "setcyclepath redfish/v1/Systems/Self/Actions/ComputerSystem.Reset {\"ResetType\":\"ForceRestart\"}\n"
 		expect "redfishpower> "
+		send "settimeout 60\n"
+		expect "redfishpower> "
 	}
 	script logout {
 		send "quit\n"

etc/devices/redfishpower-cray-windom.dev

@@ -18,11 +18,9 @@
 #   attached to a blade's service processor.  Some experimentation may
 #   be needed to determine which node is attached to which suffix.
 #
-# - If using a slightly different model or BMC firmware version, its
-#   possible the URI paths will be different.  A good starting point
-#   is to run the following command to discover the correct paths:
-#
-#   curl -s -u USER:PASS -k -X GET https://pnode1/redfish/v1/Systems
+# - CAUTION: If you intend to use this file as the basis for a different
+#   Redfish system, read the section "UPDATING REDFISHPOWER DEVICE FILES"
+#   in redfishpower(8).
 #
 specification "redfishpower-cray-windom-node0" {
 	timeout 	60
@@ -39,6 +37,8 @@ specification "redfishpower-cray-windom-node0" {
 		expect "redfishpower> "
 		send "setoffpath redfish/v1/Systems/Node0/Actions/ComputerSystem.Reset {\"ResetType\":\"ForceOff\"}\n"
 		expect "redfishpower> "
+		send "settimeout 60\n"
+		expect "redfishpower> "
 	}
 	script logout {
 		send "quit\n"
@@ -82,6 +82,8 @@ specification "redfishpower-cray-windom-node1" {
 		expect "redfishpower> "
 		send "setoffpath redfish/v1/Systems/Node1/Actions/ComputerSystem.Reset {\"ResetType\":\"ForceOff\"}\n"
 		expect "redfishpower> "
+		send "settimeout 60\n"
+		expect "redfishpower> "
 	}
 	script logout {
 		send "quit\n"

etc/devices/redfishpower-supermicro.dev

@@ -10,11 +10,9 @@
 #
 # - This device specification was tested on a Supermicro H12DSG-O-CPU.
 #
-# - If using a slightly different model or BMC firmware version, its
-#   possible the URI paths will be different.  A good starting point
-#   is to run the following command to discover the correct paths:
-#
-#   curl -s -u USER:PASS -k -X GET https://pnode1/redfish/v1/Systems/1
+# - CAUTION: If you intend to use this file as the basis for a different
+#   Redfish system, read the section "UPDATING REDFISHPOWER DEVICE FILES"
+#   in redfishpower(8).
 #
 specification "redfishpower-supermicro" {
 	timeout 	60
@@ -33,6 +31,8 @@ specification "redfishpower-supermicro" {
 		expect "redfishpower> "
 		send "setcyclepath redfish/v1/Systems/1/Actions/ComputerSystem.Reset {\"ResetType\":\"ForceRestart\"}\n"
 		expect "redfishpower> "
+		send "settimeout 60\n"
+		expect "redfishpower> "
 	}
 	script logout {
 		send "quit\n"

man/redfishpower.8.in

@@ -75,6 +75,9 @@ Set path and optional post data to turn off node.
 .I "setcyclepath <path> [data]"
 Set path and optional post data to turn cycle node.
 .TP
+.I "settimeout <seconds>"
+Set command timeout in seconds.
+.TP
 .I "stat [nodes]"
 Get power status of all nodes or specified subset of nodes.
 .TP
@@ -87,6 +90,65 @@ Turn off all nodes or specified subset of nodes.  Will return "ok" after confirm
 .I "cycle [nodes]"
 Turn off all nodes or specified subset of nodes.
 
+.SH "UPDATING REDFISHPOWER DEVICE FILES"
+.LP
+Users may often need to modify
+.B redfishpower
+device files for other systems, as URI paths may be slightly different.
+.LP
+To find the URI paths on your system, a good starting point is to
+use
+.B curl(1)
+to discover the correct paths at the following standard location.
+Note that the output can be easier to read if piped to
+.B jq(1),
+but that is optional.
+.PP
+.B # curl -s -u USER:PASS -k -X GET https://<node>/redfish/v1/Systems | jq
+.LP
+Information in the response will point you on how to dig further.
+Several paths the authors have seen include:
+.PP
+.B # curl -s -u USER:PASS -k -X GET https://<node>/redfish/v1/Systems/1
+.PP
+.B # curl -s -u USER:PASS -k -X GET https://<node>/redfish/v1/Systems/Node0
+.PP
+.B # curl -s -u USER:PASS -k -X GET https://<node>/redfish/v1/Systems/Self
+.LP
+To discover if \fIcycle\fR is supported, a path similar to the following might
+be found:
+.PP
+.B # curl -s -u USER:PASS -k -X GET https://<node>/redfish/v1/Systems/Node0/ResetActionInfo
+.LP
+If \fIcycle\fR (i.e. \fIForceRestart\fR) is not available, the
+\fIcycle_ranged\fR script could be replaced with.
+.PP
+.nf
+      script cycle_ranged {
+              send "off %s\\n"
+              expect "redfishpower> "
+              send "on %s\\n"
+              expect "redfishpower> "
+      }
+.fi
+.LP
+Internally within
+.B redfishpower,
+an \fIon\fR or \fIoff\fR command will not return until the
+\fIon\fR/\fIoff\fR is confirmed to complete.  i.e. \fIstat\fR returns
+the expected status.
+.LP
+This can sometimes take awhile and on some systems may push the
+timeout limit specified in the device file (typically 60 seconds).  If
+necessary, it should be increased at the top of the specification file
+(via \fItimeout\fR) and the login section of device file (via
+\fIsettimeout\fR).
+.LP
+Note that if \fIcycle_ranged\fR is scripted as an \fIoff\fR followed by an
+\fIon\fR, the
+.B powerman
+timeout should account for the combined time of both.
+
 .SH "FILES"
 @X_SBINDIR@/redfishpower
 .br

src/redfishpower/redfishpower.c

@@ -20,8 +20,10 @@
 #include <jansson.h>
 #include <unistd.h>
 #include <sys/select.h>
+#include <limits.h>
 #include <sys/time.h>
 #include <ctype.h>
+#include <errno.h>
 #include <assert.h>
 
 #include "xtypes.h"
@@ -46,7 +48,7 @@ static char *cyclepostdata = NULL;
 
 /* in seconds */
 #define MESSAGE_TIMEOUT            10
-#define CMD_TIMEOUT                60
+#define CMD_TIMEOUT_DEFAULT        60
 
 /* Per documentation, wait incremental time then proceed if timeout < 0 */
 #define INCREMENTAL_WAIT           500
@@ -109,6 +111,8 @@ static struct option longopts[] = {
         {0,0,0,0},
 };
 
+static time_t cmd_timeout = CMD_TIMEOUT_DEFAULT;
+
 void help(void)
 {
     printf("Valid commands are:\n");
@@ -118,6 +122,7 @@ void help(void)
     printf("  setonpath url [data]\n");
     printf("  setoffpath url [data]\n");
     printf("  setcyclepath url [data]\n");
+    printf("  settimeout seconds\n");
     printf("  stat [nodes]\n");
     printf("  on [nodes]\n");
     printf("  off [nodes]\n");
@@ -218,7 +223,10 @@ static struct powermsg *powermsg_create(CURLM *mh,
     else
         gettimeofday(&pm->start, NULL);
 
-    pm->timeout.tv_sec = pm->start.tv_sec + CMD_TIMEOUT;
+    if (cmd_timeout > (LONG_MAX - pm->start.tv_sec))
+        err_exit(FALSE, "cmd_timeout overflow");
+
+    pm->timeout.tv_sec = pm->start.tv_sec + cmd_timeout;
     pm->timeout.tv_usec = pm->start.tv_usec;
 
     if (delay_usec) {
@@ -589,6 +597,22 @@ static void setpowerpath(char **av, char **path, char **postdata)
         (*postdata) = xstrdup(av[1]);
 }
 
+static void settimeout(char **av)
+{
+  if (av[0]) {
+      char *endptr;
+      long tmp;
+
+      errno = 0;
+      tmp = strtol (av[0], &endptr, 10);
+      if (errno
+          || endptr[0] != '\0'
+          || tmp <= 0)
+        printf("invalid timeout specified\n");
+      cmd_timeout = tmp;
+  }
+}
+
 static void process_cmd(List activecmds, CURLM *mh, char **av, int *exitflag)
 {
     if (av[0] != NULL) {
@@ -608,6 +632,8 @@ static void process_cmd(List activecmds, CURLM *mh, char **av, int *exitflag)
             setpowerpath(av + 1, &offpath, &offpostdata);
         else if (strcmp(av[0], "setcyclepath") == 0)
             setpowerpath(av + 1, &cyclepath, &cyclepostdata);
+        else if (strcmp(av[0], "settimeout") == 0)
+            settimeout(av + 1);
         else if (strcmp(av[0], "stat") == 0)
             stat_cmd(activecmds, mh, av + 1);
         else if (strcmp(av[0], "on") == 0)

test/redfishpower.c

@@ -220,7 +220,8 @@ _prompt_loop(void)
                    || !strcmp(buf, "setstatpath")
                    || !strcmp(buf, "setonpath")
                    || !strcmp(buf, "setoffpath")
-                   || !strcmp(buf, "setcyclepath")) {
+                   || !strcmp(buf, "setcyclepath")
+                   || !strcmp(buf, "settimeout")) {
             /* do nothing with config, just accept */
             ;
         } else if (!strcmp(buf, "stat")) {