Change to monitor command line and python evaluation process

ML/eval_proc.py

@@ -2,82 +2,101 @@
 
 import socket
 import torch
+import sys
 import torch.nn as nn
 import numpy as np 
 
-PATH= "model_NN50_20.pth"
+PATH= "/home/sgundo/Desktop/Tesi/OS_FaultMonitor/ML/"
 len_index = 20
 
 class BinaryClassification(nn.Module):
-    def __init__(self):
-        super(BinaryClassification, self).__init__()
+	def __init__(self):
+		super(BinaryClassification, self).__init__()
 
-        self.features = nn.Sequential(
-              nn.Linear(len_index-1, 32),
-              nn.BatchNorm1d(32),
-              nn.Dropout(0.5), #50 % probability 
-              nn.LeakyReLU(),
-              torch.nn.Linear(32, 64),
-              nn.BatchNorm1d(64),
-              torch.nn.Dropout(0.2), #20% probability
-              torch.nn.LeakyReLU(),
-              torch.nn.Linear(64, 22),
-              nn.BatchNorm1d(22),
-              torch.nn.LeakyReLU())
-        self.classifier = torch.nn.Linear(22, 2)
-        
-    def forward(self, x):
-        x = self.features(x)
-        x = self.classifier(x)
-        return x
+		self.features = nn.Sequential(
+			  nn.Linear(len_index-1, 32),
+			  nn.BatchNorm1d(32),
+			  nn.Dropout(0.5), #50 % probability 
+			  nn.LeakyReLU(),
+			  torch.nn.Linear(32, 64),
+			  nn.BatchNorm1d(64),
+			  torch.nn.Dropout(0.2), #20% probability
+			  torch.nn.LeakyReLU(),
+			  torch.nn.Linear(64, 22),
+			  nn.BatchNorm1d(22),
+			  torch.nn.LeakyReLU())
+		self.classifier = torch.nn.Linear(22, 2)
+		
+	def forward(self, x):
+		x = self.features(x)
+		x = self.classifier(x)
+		return x
 
 def eval(feat):
-        X = np.array(feat)
-        X = torch.from_numpy(X).float()
-        X = X.unsqueeze(0)
-        y = model(X)
-        y = torch.log_softmax(y, dim=1)
-        _, y = torch.max(y, dim = 1)   
-    	
-        return y.item()
+		X = np.array(feat)
+		X = torch.from_numpy(X).float()
+		X = X.unsqueeze(0)
+		y = model(X)
+		y = torch.log_softmax(y, dim=1)
+		_, y = torch.max(y, dim = 1)   
+		
+		return y.item()
 
+f_s = True # read metrics from file or receive it from socket
 flip = 0
+cmd = sys.argv[1]
+bi = sys.argv[2]
 model = BinaryClassification()
-model.load_state_dict(torch.load(PATH, map_location=torch.device('cpu')))
+model.load_state_dict(torch.load(PATH+bi+".pth", map_location=torch.device('cpu')))
 model.eval()
 HOST = '127.0.0.1'  # Standard loopback interface address (localhost)
 PORT = 2325        # Port to listen on (non-privileged ports are > 1023)
 i = 0
-num_feat = 19
-with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-    s.bind((HOST, PORT))
-    s.listen()
-    conn, addr = s.accept()
-    with conn:
-        print('Connected by', addr)
-        while True:
-            i = 0
-            feat = list()
-            for i in range(0, num_feat+1):
-                data = conn.recv(4)
-                num = int.from_bytes(data, byteorder='little', signed=False)
-                if i==0:
-                    bina=num
-                    continue
-                feat.append(num)
-            if not data:
-                break
-            print("Evaluate process bin: "+str(bina))
-            result = eval(feat);
-            if flip == 0:
-                result=1
-                flip=1
-            else:
-                flip=0
-                result=0
-            if result == 0:
-            	conn.sendall(b"0")
-            	print("Sent not fault valuation")
-            else:
-            	conn.sendall(b"1")
-            	print("Sent fault valuation")
+num_feat = 7
+
+if f_s:
+	with open(cmd+".out", "r") as file:
+		first_line = file.readline()
+		for last_line in file:
+			pass
+		print (last_line)
+		attr = last_line.split()
+		attr = attr[3:]
+		print(attr)
+		attr_int = [int(i) for i in attr]
+		if len(attr_int) == 7:
+			eval(attr_int)
+
+else:
+	with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+		s.bind((HOST, PORT))
+		s.listen()
+		conn, addr = s.accept()
+		with conn:
+			print('Connected by', addr)
+			while True:
+				i = 0
+				feat = list()
+				for i in range(0, num_feat+1):
+					data = conn.recv(4)
+					num = int.from_bytes(data, byteorder='little', signed=False)
+					if i==0:
+						bina=num
+						continue
+					feat.append(num)
+				if not data:
+					break
+				print("Evaluate process bin: "+str(bina))
+				result = eval(feat)
+				if flip == 0:
+					result=1
+					flip=1
+				else:
+					flip=0
+					result=0
+				if result == 0:
+					conn.sendall(b"0")
+					print("Sent not fault valuation")
+				else:
+					conn.sendall(b"1")
+					print("Sent fault valuation")

cmd/monitor

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Command line utility to start a program and signal PID to kernel monitor
+# Command line utility to start a program and monitor it
 cmd=""
 i="0"
 
@@ -8,14 +8,14 @@ while getopts ":b:" opt; do
     b )
       target=$OPTARG
       if [ "$target" == "bitcount" ];then
-				bin=0
-			elif [ "$target" == "basicmath" ];then
-				bin=1
-			elif [ "$target" =="qsort" ];then
-				bin=2
-			else
-				bin=-1
-			fi
+			bin=0
+		elif [ "$target" == "basicmath" ];then
+			bin=1
+		elif [ "$target" =="qsort" ];then
+			bin=2
+		else
+			bin=-1
+		fi
       ;;
     \? )
       echo "Invalid option: $OPTARG" 1>&2
@@ -39,8 +39,11 @@ do
 
 done
 echo "comando lanciato $cmd"
-nohup $cmd &
-pid=$!
-echo $pid
-echo "monitor lanciato"
-echo "$pid $bin" > /proc/monitor
+
+conf="instr_retired,unhalted_ref_cycles_fixed,cycles,instr,llc_references,llc_misses,branch_instr_retired"
+nohup pmctrack -T 1 -c $conf $cmd > $cmd.out
+python3 /home/sgundo/Desktop/Tesi/OS_FaultMonitor/ML/eval_proc.py $cmd $target
+# pid=$!
+# echo $pid
+# echo "monitor lanciato"
+#echo "$pid $bin" > /proc/monitor

kernel_module/Makefile

@@ -2,5 +2,7 @@ obj-m += monitor.o
 monitor-objs := src/monitor.o src/connt.o src/m_pfile.o src/f_list.o
 all:
 	make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
+install:  
+	make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules_install
 clean:
 	make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean

kernel_module/src/f_list.c

@@ -27,20 +27,26 @@ static LIST_HEAD(f_list);
 static unsigned int* get_randoms(unsigned int id){
 	int i = 0;
 	unsigned int *fts = NULL;
-	uint64_t value;
+	uint64_t value=0;
 
 	fts = kzalloc(NUM*sizeof(unsigned int), GFP_KERNEL);
 	if(fts == NULL) return NULL;
 
-
+	//pmctrack_hwmon_export_device(NULL, 0, NULL);
 	for(i=0;i<NUM;i++){
-		//get_random_bytes(&fts[i], sizeof(unsigned int));		
-		pmcs_get_current_metric_value(get_pid_task(find_get_pid(id),PIDTYPE_PID),\
-		 				0, &value);
+		value = -1;
+		//get_random_bytes(&fts[i], sizeof(unsigned int));
+		struct task_struct* task = get_pid_task(find_get_pid(id), PIDTYPE_PID);
+		if(task == NULL){ 
+			pr_err("ERROR: task not found\n");
+			return NULL;
+		}
+		pmcs_get_current_metric_value(task, 0, &value);
+		if(value != -1)
+			pr_info("PMC value: %llu", value);
 		fts[i] = (unsigned int) value;
-
 	}
-	
+
 	return fts;
 }
 

kernel_module/src/monitor.c

@@ -15,6 +15,7 @@
 MODULE_AUTHOR("Simone Dutto");
 MODULE_DESCRIPTION("module to evaluate faulty proc");
 MODULE_LICENSE("GPL");
+MODULE_SOFTDEP("pre: mchw_intel_core");
 
 static void proc_eval_handler(struct work_struct *w);
 static void feat_updater(struct work_struct *w);