diff --git a/kernel.h b/kernel.h
index 6de9112..f336a57 100644
--- a/kernel.h
+++ b/kernel.h
@@ -10,6 +10,9 @@
 // scheduler settings
 #define TIME_SLICE_LEN 10          // number of cpu time ticks per slice
 
+// set size of allocated stack for user processes
+#define USER_STACK_SIZE (1 << 12)
+
 // init function
 extern __attribute__((__noreturn__)) void init();
 
diff --git a/kinclude/ecall.c b/kinclude/ecall.c
index a49e089..e54d478 100644
--- a/kinclude/ecall.c
+++ b/kinclude/ecall.c
@@ -73,7 +73,26 @@ optional_int ecall_handle_join(int* args, ProcessControlBlock* pcb)
 
 optional_int ecall_handle_kill(int* args, ProcessControlBlock* pcb)
 {
-    return (optional_int) { .error = EINVAL };
+    int pid = args[0];
+    ProcessControlBlock* target = process_from_pid(pid);
+
+    // return error if no process has that id
+    if (target == NULL)
+        return (optional_int) { .error = ESRCH };
+
+    // return success if the process is dead
+    if (target == PROC_DEAD)
+        return (optional_int) { .value = 1 };
+
+    // kill target by marking it dead
+    target->status = PROC_DEAD;
+    target->exit_code = -10;    // set unique exit code
+
+    // call process destructor
+    destroy_process(pcb);
+
+    // return success
+    return (optional_int) { .value = 1 };
 }
 
 optional_int ecall_handle_exit(int* args, ProcessControlBlock* pcb)
@@ -92,8 +111,8 @@ optional_int ecall_handle_exit(int* args, ProcessControlBlock* pcb)
         dbgln(msg, 34);
     }
 
-    // recursively kill all child processes
-    kill_child_processes(pcb);
+    // call process destructor
+    destroy_process(pcb);
 
     return (optional_int) { .value = 0 };
 }
@@ -183,6 +202,7 @@ void init_ecall_table()
 
 void handle_exception(int ecode, int mtval)
 {
+    //TODO: handle exceptions well
     dbgln("exception encountered!", 17);
     __asm__ ("ebreak");
 }
diff --git a/kinclude/ktypes.h b/kinclude/ktypes.h
index 10835eb..1ac5e0b 100644
--- a/kinclude/ktypes.h
+++ b/kinclude/ktypes.h
@@ -42,6 +42,7 @@ struct ProcessControlBlock {
     unsigned long long int asleep_until;
     // parent
     ProcessControlBlock* parent;
+    void* stack_top;
 };
 
 enum pcb_struct_registers {
@@ -88,6 +89,8 @@ typedef struct loaded_binary {
         type value;                         \
     } optional_ ## type
 
+// has_value and has_error are not dependent on the value type
+// therefore we can define them as macros
 #define has_value(optional) (optional.error == 0)
 #define has_error(optional) (!has_value(optional))
 
@@ -104,4 +107,42 @@ CreateOptionalOfType(size_t);
 CreateOptionalOfType(pcbptr);
 CreateOptionalOfType(voidptr);
 
+/*
+ * Stacks
+ *
+ * stacks allow pushing and popping elements off it.
+ */
+
+// we currently only need a voidptr stack, if we need more we can either move them to a separate file
+// or just type cast the pointers afterwards
+struct voidptr_stack {
+    void** data;
+    int pos;
+    int size;
+};
+
+inline void* voidptr_stack_pop(struct voidptr_stack* stack)
+{
+    if (stack->pos == -1)
+        return NULL;
+    return stack->data[stack->pos--];
+}
+
+inline int voidptr_stack_push(struct voidptr_stack* stack, void* ptr)
+{
+    if (stack->pos + 1 > stack->size)
+        return 0;
+
+    stack->pos += 1;
+    stack->data[stack->pos] = ptr;
+    return 1;
+}
+
+inline void voidptr_stack_new(struct voidptr_stack* stack, void** data, int size)
+{
+    stack->pos = -1;
+    stack->size = size;
+    stack->data = data;
+}
+
 #endif
diff --git a/kinclude/malloc.c b/kinclude/malloc.c
index d38a3ac..89f1997 100644
--- a/kinclude/malloc.c
+++ b/kinclude/malloc.c
@@ -1,26 +1,58 @@
+#include "../kernel.h"
 #include "malloc.h"
 #include "ecall.h"
+#include "io.h"
 
-malloc_info global_malloc_info = { 0 };
-void* allocate_memory_end;
+// information about the systems memory layout is stored here
+static malloc_info global_malloc_info = { 0 };
+// this pointer points to the end of unused memory
+static void* allocate_memory_end;
+// this stack holds currently unused program stacks
+// it is a stack in the sense that you can push and pop variables
+static struct voidptr_stack unused_stacks;
 
+void stash_stack(void* stack_top)
+{
+    if (voidptr_stack_push(&unused_stacks, stack_top) == 0) {
+        // if this happens, we know we messed up and created more stacks than we
+        // could ever have processes running simultaneously!
+        // we probably have an erro in retrieving stashed stacks?
+        dbgln("cannot stash stack, no space left! This should never ever happen!", 65);
+    }
+}
 
 void malloc_init(malloc_info* given_info)
 {
     global_malloc_info = *given_info;
     allocate_memory_end = given_info->allocate_memory_end;
+    allocate_memory_end = (void*) (((int) allocate_memory_end) - PROCESS_COUNT);
+    voidptr_stack_new(&unused_stacks, (void**) allocate_memory_end, PROCESS_COUNT);
 }
 
 // allocate stack and return a pointer to the *end* of the allocated region
 // this doesn't reuse stack from functions which exited
-optional_voidptr malloc_stack(size_t size)
+optional_voidptr malloc_stack()
 {
-    void* new_alloc_end = (void*) (((int) allocate_memory_end) - size);
+    // try to reuse a stack top
+    void* stack_top = voidptr_stack_pop(&unused_stacks);
+
+    if (stack_top != NULL) {
+        return (optional_voidptr) { .value = stack_top };
+    }
+
+    // otherwise allocate a new stack at the end of available memory
+    void* new_alloc_end = (void*) (((int) allocate_memory_end) - USER_STACK_SIZE);
 
+    // check if we ran out of space
     if (new_alloc_end < global_malloc_info.allocate_memory_start)
         return (optional_voidptr) { .error = ENOMEM };
-    void* stack_top = allocate_memory_end;
 
+    stack_top = allocate_memory_end;
     allocate_memory_end = new_alloc_end;
     return (optional_voidptr) { .value = stack_top };
 }
+
+void free_stack(void* stack_top)
+{
+    stash_stack(stack_top);
+}
diff --git a/kinclude/malloc.h b/kinclude/malloc.h
index ee1d650..c48bf85 100644
--- a/kinclude/malloc.h
+++ b/kinclude/malloc.h
@@ -8,7 +8,8 @@ typedef struct malloc_info {
     void* allocate_memory_start;
 } malloc_info;
 
-optional_voidptr malloc_stack(size_t size);
+optional_voidptr malloc_stack();
+void free_stack(void* stack_top);
 
 void malloc_init(malloc_info* info);
 
diff --git a/kinclude/sched.c b/kinclude/sched.c
index 7bde96c..7ccf9e4 100644
--- a/kinclude/sched.c
+++ b/kinclude/sched.c
@@ -229,6 +229,7 @@ optional_pcbptr create_new_process(loaded_binary* bin, int stack_size)
     pcb->binary = bin;
     pcb->parent = NULL;
     pcb->asleep_until = 0;
+    pcb->stack_top = stack_top_or_err.value;
     // zero out registers
     memset(0, pcb->regs, pcb->regs + 31);
     // load stack top into stack pointer register
@@ -273,6 +274,7 @@ optional_pcbptr create_new_thread(ProcessControlBlock* parent, void* entrypoint,
     pcb->binary = parent->binary;
     pcb->parent = parent;
     pcb->asleep_until = 0;
+    pcb->stack_top = stack_top_or_err.value;
     // zero out registers
     memset(0, pcb->regs, pcb->regs + 31);
     // set return address to global thread finalizer
@@ -293,7 +295,8 @@ void kill_child_processes(ProcessControlBlock* pcb)
 {
     for (int i = 0; i < PROCESS_COUNT; i++) {
         ProcessControlBlock* proc = processes + i;
-        if (proc->parent != pcb)
+        // if this is not a child process or already exited
+        if (proc->parent != pcb || proc->status == PROC_DEAD)
             continue;
 
         proc->status = PROC_DEAD;
@@ -301,3 +304,13 @@ void kill_child_processes(ProcessControlBlock* pcb)
         kill_child_processes(proc);
     }
 }
+
+void destroy_process(ProcessControlBlock* pcb)
+{
+    // kill child processes
+    kill_child_processes(pcb);
+    // free allocated stack
+    free_stack(pcb->stack_top);
+    // make sure the thread is not rescheduled
+    pcb->status = PROC_DEAD;
+}
diff --git a/kinclude/sched.h b/kinclude/sched.h
index 03f0ed6..16ed363 100644
--- a/kinclude/sched.h
+++ b/kinclude/sched.h
@@ -21,5 +21,6 @@ void mark_ecall_entry();
 // process creation / destruction
 optional_pcbptr create_new_process(loaded_binary*, int);
 optional_pcbptr create_new_thread(ProcessControlBlock*, void*, void*, int);
+void destroy_process(ProcessControlBlock* pcb);
 void kill_child_processes(ProcessControlBlock* pcb);
 #endif