impementation of sched_{setparam, getparam, setscheduler, getscheduler, get_priority_min, get_priority_max, rr_get_interval} system calls

arch/x86/kernel/include/syscall_list.h

@@ -71,6 +71,13 @@ SYSCALL_HANDLED(128, rt_sigtimedwait)
 SYSCALL_HANDLED(129, rt_sigqueueinfo)
 SYSCALL_HANDLED(130, rt_sigsuspend)
 SYSCALL_HANDLED(131, sigaltstack)
+SYSCALL_HANDLED(142, sched_setparam)
+SYSCALL_HANDLED(143, sched_getparam)
+SYSCALL_HANDLED(144, sched_setscheduler)
+SYSCALL_HANDLED(145, sched_getscheduler)
+SYSCALL_HANDLED(146, sched_get_priority_max)
+SYSCALL_HANDLED(147, sched_get_priority_min)
+SYSCALL_HANDLED(148, sched_rr_get_interval)
 SYSCALL_HANDLED(149, mlock)
 SYSCALL_HANDLED(150, munlock)
 SYSCALL_HANDLED(158, arch_prctl)

executer/kernel/syscall.c

@@ -1505,6 +1505,8 @@ fail:
 	return error;
 }
 
+#define SCHED_CHECK_SAME_OWNER        0x01
+#define SCHED_CHECK_ROOT              0x02
 
 int __do_in_kernel_syscall(ihk_os_t os, struct mcctrl_channel *c, struct syscall_request *sc)
 {
@@ -1592,6 +1594,71 @@ int __do_in_kernel_syscall(ihk_os_t os, struct mcctrl_channel *c, struct syscall
 		error = writecore(os, sc->args[1], sc->args[0]);
 		ret = 0;
 		break;
+	
+	case __NR_sched_setparam: {
+
+		switch (sc->args[0]) {
+			
+			case SCHED_CHECK_SAME_OWNER: {
+				const struct cred *cred = current_cred();
+				const struct cred *pcred;
+				bool match;
+				struct task_struct *p;
+				int pid = sc->args[1];
+				
+				rcu_read_lock();
+				p = pid_task(find_get_pid(pid), PIDTYPE_PID);
+				if (!p) {
+					rcu_read_unlock();
+					ret = -ESRCH;
+					goto sched_setparam_out;
+				}
+				rcu_read_unlock();
+
+				rcu_read_lock();
+				pcred = __task_cred(p);
+#if LINUX_VERSION_CODE > KERNEL_VERSION(3,4,0)
+				match = (uid_eq(cred->euid, pcred->euid) ||
+					 uid_eq(cred->euid, pcred->uid));
+#else
+				match = ((cred->euid == pcred->euid) ||
+						(cred->euid == pcred->uid));
+#endif
+				rcu_read_unlock();
+				
+				if (match) {
+					ret = 0;
+				}
+				else {
+					ret = -EPERM;
+				}
+				
+				break;
+			}
+
+			case SCHED_CHECK_ROOT: {
+				const struct cred *cred = current_cred();
+				bool match;
+				
+#if LINUX_VERSION_CODE > KERNEL_VERSION(3,4,0)
+				match = uid_eq(cred->euid, GLOBAL_ROOT_UID);
+#else
+				match = (cred->euid == 0);
+#endif
+				if (match) {
+					ret = 0;
+				}
+				else {
+					ret = -EPERM;
+				}
+				
+				break;
+			}
+		}
+			
+sched_setparam_out:
+		break;
+	}
 
 	default:
 		error = -ENOSYS;

kernel/include/prio.h

@@ -0,0 +1,60 @@
+#ifndef _SCHED_PRIO_H
+#define _SCHED_PRIO_H
+
+#define MAX_NICE	19
+#define MIN_NICE	-20
+#define NICE_WIDTH	(MAX_NICE - MIN_NICE + 1)
+
+/*
+ * Priority of a process goes from 0..MAX_PRIO-1, valid RT
+ * priority is 0..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH
+ * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority
+ * values are inverted: lower p->prio value means higher priority.
+ *
+ * The MAX_USER_RT_PRIO value allows the actual maximum
+ * RT priority to be separate from the value exported to
+ * user-space.  This allows kernel threads to set their
+ * priority to a value higher than any user task. Note:
+ * MAX_RT_PRIO must not be smaller than MAX_USER_RT_PRIO.
+ */
+
+#define MAX_USER_RT_PRIO	100
+#define MAX_RT_PRIO		MAX_USER_RT_PRIO
+
+#define MAX_PRIO		(MAX_RT_PRIO + NICE_WIDTH)
+#define DEFAULT_PRIO		(MAX_RT_PRIO + NICE_WIDTH / 2)
+
+/*
+ * Convert user-nice values [ -20 ... 0 ... 19 ]
+ * to static priority [ MAX_RT_PRIO..MAX_PRIO-1 ],
+ * and back.
+ */
+#define NICE_TO_PRIO(nice)	((nice) + DEFAULT_PRIO)
+#define PRIO_TO_NICE(prio)	((prio) - DEFAULT_PRIO)
+
+/*
+ * 'User priority' is the nice value converted to something we
+ * can work with better when scaling various scheduler parameters,
+ * it's a [ 0 ... 39 ] range.
+ */
+#define USER_PRIO(p)		((p)-MAX_RT_PRIO)
+#define TASK_USER_PRIO(p)	USER_PRIO((p)->static_prio)
+#define MAX_USER_PRIO		(USER_PRIO(MAX_PRIO))
+
+/*
+ * Convert nice value [19,-20] to rlimit style value [1,40].
+ */
+static inline long nice_to_rlimit(long nice)
+{
+	return (MAX_NICE - nice + 1);
+}
+
+/*
+ * Convert rlimit style value [1,40] to nice value [-20, 19].
+ */
+static inline long rlimit_to_nice(long prio)
+{
+	return (MAX_NICE - prio + 1);
+}
+
+#endif /* _SCHED_PRIO_H */

kernel/include/process.h

@@ -316,6 +316,29 @@ struct fork_tree_node {
 void hold_fork_tree_node(struct fork_tree_node *ftn);
 void release_fork_tree_node(struct fork_tree_node *ftn);
 
+/*
+ * Scheduling policies
+ */
+#define SCHED_NORMAL		0
+#define SCHED_FIFO		1
+#define SCHED_RR		2
+#define SCHED_BATCH		3
+/* SCHED_ISO: reserved but not implemented yet */
+#define SCHED_IDLE		5
+#define SCHED_DEADLINE		6
+
+/* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
+#define SCHED_RESET_ON_FORK     0x40000000
+
+/*
+ * For the sched_{set,get}attr() calls
+ */
+#define SCHED_FLAG_RESET_ON_FORK	0x01
+
+struct sched_param {
+	int sched_priority;
+};
+
 struct process {
 	int cpu_id;
 
@@ -327,6 +350,8 @@ struct process {
 	
 	// Runqueue list entry
 	struct list_head sched_list;  
+	int sched_policy;
+	struct sched_param sched_param;
 	
 	ihk_spinlock_t spin_sleep_lock;
 	int spin_sleep;

kernel/process.c

@@ -154,6 +154,8 @@ struct process *create_process(unsigned long user_pc)
 		}
 	}
 
+	proc->sched_policy = SCHED_NORMAL;
+
 	proc->sighandler = kmalloc(sizeof(struct sig_handler), IHK_MC_AP_NOWAIT);
 	if(!proc->sighandler){
 		goto err_free_process;
@@ -252,6 +254,9 @@ struct process *clone_process(struct process *org, unsigned long pc,
 
 	init_fork_tree_node(proc->ftn, org->ftn, proc);
 
+	proc->sched_policy = org->sched_policy;
+	proc->sched_param.sched_priority = org->sched_param.sched_priority;
+
 	/* clone signal handlers */
 	if (clone_flags & CLONE_SIGHAND) {
 		proc->sigstack.ss_sp = NULL;

kernel/syscall.c

@@ -46,6 +46,7 @@
 #include <kmalloc.h>
 #include <memobj.h>
 #include <shm.h>
+#include <prio.h>
 
 /* Headers taken from kitten LWK */
 #include <lwk/stddef.h>
@@ -1364,8 +1365,9 @@ SYSCALL_DECLARE(getppid)
 		if (proc->ftn->ppid_parent)
 			pid = proc->ftn->ppid_parent->pid;
 	} else {
-		if (proc->ftn->parent)
+		if (proc->ftn->parent) {
 			pid = proc->ftn->parent->pid;
+		}
 	}
 	ihk_mc_spinlock_unlock_noirq(&proc->ftn->lock);
 	return pid;
@@ -3202,6 +3204,272 @@ SYSCALL_DECLARE(ptrace)
 	return error;
 }
 
+/* We do not have actual scheduling classes so we just make sure we store
+ * policies and priorities in a POSIX/Linux complaint manner */
+static int setscheduler(struct process *proc, int policy, struct sched_param *param)
+{
+	if ((policy == SCHED_FIFO || policy == SCHED_RR) &&
+		((param->sched_priority < 1) ||
+		 (param->sched_priority > MAX_USER_RT_PRIO - 1))) {
+		return -EINVAL;
+	}
+	
+	if ((policy == SCHED_NORMAL || policy == SCHED_BATCH || policy == SCHED_IDLE) &&
+		(param->sched_priority != 0)) {
+		return -EINVAL;
+	}
+
+	memcpy(&proc->sched_param, param, sizeof(*param));
+	proc->sched_policy = policy;
+
+	return 0;
+}
+
+#define SCHED_CHECK_SAME_OWNER        0x01
+#define SCHED_CHECK_ROOT              0x02
+
+SYSCALL_DECLARE(sched_setparam)
+{
+	int retval = 0;
+	int pid = (int)ihk_mc_syscall_arg0(ctx);
+	struct sched_param *uparam = (struct sched_param *)ihk_mc_syscall_arg1(ctx);
+	struct sched_param param;
+	struct process *proc = cpu_local_var(current);
+	unsigned long irqstate = 0;
+	ihk_spinlock_t *lock;
+	
+	ihk_mc_user_context_t ctx1;
+	struct syscall_request request1 IHK_DMA_ALIGN;
+
+	dkprintf("sched_setparam: pid: %d, uparam: 0x%lx\n", pid, uparam);
+
+	if (!uparam || pid < 0) {
+		return -EINVAL;
+	}
+
+	if (pid == 0)
+		pid = proc->ftn->pid;
+
+	if (proc->ftn->pid != pid) {
+		proc = findthread_and_lock(pid, pid, &lock, &irqstate);
+		if (!proc) {
+			return -ESRCH;
+		}
+		process_unlock(lock, irqstate);
+		
+		/* Ask Linux about ownership.. */
+		request1.number = __NR_sched_setparam;
+		request1.args[0] = SCHED_CHECK_SAME_OWNER;
+		request1.args[1] = pid;
+
+		retval = do_syscall(&request1, &ctx1, ihk_mc_get_processor_id(), 0);
+		if (retval != 0) {
+			return retval;
+		}
+	}
+
+	retval = copy_from_user(proc, &param, uparam, sizeof(param));
+	if (retval < 0) {
+		return -EFAULT;
+	}
+
+	return setscheduler(proc, proc->sched_policy, &param);
+}
+
+SYSCALL_DECLARE(sched_getparam)
+{
+	int retval = 0;
+	int pid = (int)ihk_mc_syscall_arg0(ctx);
+	struct sched_param *param = (struct sched_param *)ihk_mc_syscall_arg1(ctx);
+	struct process *proc = cpu_local_var(current);
+	unsigned long irqstate = 0;
+	ihk_spinlock_t *lock;
+
+	if (!param || pid < 0) {
+		return -EINVAL;
+	}
+
+	if (pid == 0)
+		pid = proc->ftn->pid;
+
+	if (proc->ftn->pid != pid) {
+		proc = findthread_and_lock(pid, pid, &lock, &irqstate);
+		if (!proc) {
+			return -ESRCH;
+		}
+		process_unlock(lock, irqstate);
+	}
+	
+	retval = copy_to_user(proc, param, &proc->sched_param, sizeof(*param)) ? -EFAULT : 0;
+	
+	return retval;
+}
+
+SYSCALL_DECLARE(sched_setscheduler)
+{
+	int retval;
+	int pid = (int)ihk_mc_syscall_arg0(ctx);
+	int policy = ihk_mc_syscall_arg1(ctx);
+	struct sched_param *uparam = (struct sched_param *)ihk_mc_syscall_arg2(ctx);
+	struct sched_param param;
+	struct process *proc = cpu_local_var(current);
+	unsigned long irqstate = 0;
+	ihk_spinlock_t *lock;
+	
+	ihk_mc_user_context_t ctx1;
+	struct syscall_request request1 IHK_DMA_ALIGN;
+	
+	if (!uparam || pid < 0) {
+		return -EINVAL;
+	}
+	
+	if (policy != SCHED_DEADLINE &&
+			policy != SCHED_FIFO && policy != SCHED_RR &&
+			policy != SCHED_NORMAL && policy != SCHED_BATCH &&
+			policy != SCHED_IDLE) {
+		return -EINVAL;
+	}
+
+	if (policy != SCHED_NORMAL) {
+		
+		/* Ask Linux about permissions */
+		request1.number = __NR_sched_setparam;
+		request1.args[0] = SCHED_CHECK_ROOT;
+
+		retval = do_syscall(&request1, &ctx1, ihk_mc_get_processor_id(), 0);
+		if (retval != 0) {
+			return retval;
+		}
+	}
+	
+	retval = copy_from_user(proc, &param, uparam, sizeof(param));
+	if (retval < 0) {
+		return -EFAULT;
+	}
+
+	if (pid == 0)
+		pid = proc->ftn->pid;
+
+	if (proc->ftn->pid != pid) {
+		proc = findthread_and_lock(pid, pid, &lock, &irqstate);
+		if (!proc) {
+			return -ESRCH;
+		}
+		process_unlock(lock, irqstate);
+		
+		/* Ask Linux about ownership.. */
+		request1.number = __NR_sched_setparam;
+		request1.args[0] = SCHED_CHECK_SAME_OWNER;
+		request1.args[1] = pid;
+
+		retval = do_syscall(&request1, &ctx1, ihk_mc_get_processor_id(), 0);
+		if (retval != 0) {
+			return retval;
+		}
+	}
+
+	return setscheduler(proc, policy, &param);
+}
+
+SYSCALL_DECLARE(sched_getscheduler)
+{
+	int pid = (int)ihk_mc_syscall_arg0(ctx);
+	struct process *proc = cpu_local_var(current);
+	unsigned long irqstate = 0;
+	ihk_spinlock_t *lock;
+
+	if (pid < 0) {
+		return -EINVAL;
+	}
+
+	if (pid == 0)
+		pid = proc->ftn->pid;
+
+	if (proc->ftn->pid != pid) {
+		proc = findthread_and_lock(pid, pid, &lock, &irqstate);
+		if (!proc) {
+			return -ESRCH;
+		}
+		process_unlock(lock, irqstate);
+	}
+
+	return proc->sched_policy;
+}
+
+SYSCALL_DECLARE(sched_get_priority_max)
+{
+	int ret = -EINVAL;
+	int policy = ihk_mc_syscall_arg0(ctx);
+
+	switch (policy) {
+		case SCHED_FIFO:
+		case SCHED_RR:
+			ret = MAX_USER_RT_PRIO - 1;
+			break;
+		case SCHED_DEADLINE:
+		case SCHED_NORMAL:
+		case SCHED_BATCH:
+		case SCHED_IDLE:
+			ret = 0;
+			break;
+	}
+	return ret;
+}
+
+SYSCALL_DECLARE(sched_get_priority_min)
+{
+	int ret = -EINVAL;
+	int policy = ihk_mc_syscall_arg0(ctx);
+
+	switch (policy) {
+		case SCHED_FIFO:
+		case SCHED_RR:
+			ret = 1;
+			break;
+		case SCHED_DEADLINE:
+		case SCHED_NORMAL:
+		case SCHED_BATCH:
+		case SCHED_IDLE:
+			ret = 0;
+	}
+	return ret;
+}
+
+SYSCALL_DECLARE(sched_rr_get_interval)
+{
+	int pid = ihk_mc_syscall_arg0(ctx);
+	struct timespec *utime = (struct timespec *)ihk_mc_syscall_arg1(ctx);
+	struct timespec t;
+	struct process *proc = cpu_local_var(current);
+	unsigned long irqstate = 0;
+	ihk_spinlock_t *lock;
+	int retval = 0;
+
+	if (pid < 0) 
+		return -EINVAL;
+
+	if (pid == 0)
+		pid = proc->ftn->pid;
+
+	if (proc->ftn->pid != pid) {
+		proc = findthread_and_lock(pid, pid, &lock, &irqstate);
+		if (!proc) {
+			return -ESRCH;
+		}
+		process_unlock(lock, irqstate);
+	}
+	
+	t.tv_sec = 0;
+	t.tv_nsec = 0;
+	if (proc->sched_policy == SCHED_RR) {
+		t.tv_nsec = 10000;
+	}
+	
+	retval = copy_to_user(proc, utime, &t, sizeof(t)) ? -EFAULT : 0;
+	
+	return retval;
+}
+
 #define MIN2(x,y) (x) < (y) ? (x) : (y)
 SYSCALL_DECLARE(sched_setaffinity)
 {