Only check user pages rather than all bytes in-between, for known-size pointers

Add validation to check for missing files in exec() args

See merge request lab2425_autumn/pintos_22!36

src/filesys/file.h

@@ -4,6 +4,9 @@
 #include "filesys/off_t.h"
 #include <stdbool.h>
 
+/* The maximum length of a file name in PintOS. */
+#define FNAME_MAX_LEN 14
+
 struct inode;
 
 /* Opening and closing files. */

src/lib/kernel/list.c

@@ -170,6 +170,9 @@ list_insert (struct list_elem *before, struct list_elem *elem)
 {
   ASSERT (is_interior (before) || is_tail (before));
   ASSERT (elem != NULL);
+  // Sanity checks to prevent (some) loop lists
+  ASSERT (before != elem);
+  ASSERT (before->prev != elem);
 
   elem->prev = before->prev;
   elem->next = before;

src/lib/user/syscall.c

@@ -166,7 +166,7 @@ mkdir (const char *dir)
 }
 
 bool
-readdir (int fd, char name[READDIR_MAX_LEN + 1]) 
+readdir (int fd, char name[FNAME_MAX_LEN + 1]) 
 {
   return syscall2 (SYS_READDIR, fd, name);
 }

src/lib/user/syscall.h

@@ -3,6 +3,7 @@
 
 #include <stdbool.h>
 #include <debug.h>
+#include "../../filesys/file.h"
 
 /* Process identifier. */
 typedef int pid_t;
@@ -12,9 +13,6 @@ typedef int pid_t;
 typedef int mapid_t;
 #define MAP_FAILED ((mapid_t) -1)
 
-/* Maximum characters in a filename written by readdir(). */
-#define READDIR_MAX_LEN 14
-
 /* Typical return values from main() and arguments to exit(). */
 #define EXIT_SUCCESS 0          /* Successful execution. */
 #define EXIT_FAILURE 1          /* Unsuccessful execution. */
@@ -41,7 +39,7 @@ void munmap (mapid_t);
 /* Task 4 only. */
 bool chdir (const char *dir);
 bool mkdir (const char *dir);
-bool readdir (int fd, char name[READDIR_MAX_LEN + 1]);
+bool readdir (int fd, char name[FNAME_MAX_LEN + 1]);
 bool isdir (int fd);
 int inumber (int fd);
 

src/threads/synch.c

@@ -113,28 +113,33 @@ void
 sema_up (struct semaphore *sema)
 {
   enum intr_level old_level;
+  bool thread_unblocked = false; /* Flag to track if any thread was woken up. */
 
   ASSERT (sema != NULL);
 
   old_level = intr_disable ();
   if (!list_empty (&sema->waiters))
-    {
-      /* Enforces wake-up of the highest priority thread waiting for the
-         semaphore. */
-      struct list_elem *e = list_max (&sema->waiters, priority_less, NULL);
-      list_remove (e);
-      thread_unblock (list_entry (e, struct thread, elem));
-    }
+  {
+    /* Enforces wake-up of the highest priority thread waiting for the
+       semaphore. */
+    struct list_elem *e = list_max (&sema->waiters, priority_less, NULL);
+    list_remove (e);
+    thread_unblock (list_entry (e, struct thread, elem));
+    thread_unblocked = true;
+  }
   sema->value++;
   intr_set_level (old_level);
 
   /* Yields the CPU in case the thread that has been woken up has a higher
      priority that the current running thread, including the case when called
      within an interrupt handler. */
-  if (intr_context ())
-    intr_yield_on_return ();
-  else
-    thread_yield ();
+  if (thread_unblocked)
+  {
+    if (intr_context ())
+      intr_yield_on_return ();
+    else
+      thread_yield ();
+  }
 }
 
 static void sema_test_helper (void *sema_);
@@ -336,6 +341,7 @@ lock_release (struct lock *lock)
      released, transfer the remaining orphaned donors to its donor list. */
   if (max_donor != NULL)
     {
+      list_remove (&max_donor->donor_elem);
       while (!list_empty (&orphan_list))
         list_push_back (&max_donor->donors_list, list_pop_front (&orphan_list));
     }
@@ -347,7 +353,6 @@ lock_release (struct lock *lock)
 
   lock->holder = NULL;
   sema_up (&lock->semaphore);
-  thread_yield ();
 }
 
 /* Returns true if the current thread holds LOCK, false

src/threads/thread.c

@@ -1,5 +1,6 @@
 #include "threads/thread.h"
 #include <debug.h>
+#include <hash.h>
 #include <stddef.h>
 #include <random.h>
 #include <stdio.h>
@@ -9,12 +10,14 @@
 #include "threads/flags.h"
 #include "threads/interrupt.h"
 #include "threads/intr-stubs.h"
+#include "threads/malloc.h"
 #include "threads/palloc.h"
 #include "threads/switch.h"
 #include "threads/synch.h"
 #include "threads/vaddr.h"
 #ifdef USERPROG
 #include "userprog/process.h"
+#include "userprog/syscall.h"
 #endif
 
 /* Random value for struct thread's `magic' member.
@@ -68,6 +71,7 @@ static void kernel_thread (thread_func *, void *aux);
 static void idle (void *aux UNUSED);
 static struct thread *running_thread (void);
 static struct thread *next_thread_to_run (void);
+static void init_process_result (struct thread *t);
 static void init_thread (struct thread *, const char *name, int nice,
                          int priority, fp32_t recent_cpu);
 static bool is_thread (struct thread *) UNUSED;
@@ -110,6 +114,7 @@ thread_init (void)
                initial_thread_recent_cpu);
   initial_thread->status = THREAD_RUNNING;
   initial_thread->tid = allocate_tid ();
+  initial_thread->result = NULL; /* Main thread cannot be waited for. */
 }
 
 /* Starts preemptive thread scheduling by enabling interrupts.
@@ -236,6 +241,11 @@ thread_create (const char *name, int priority,
   struct thread *parent_thread = thread_current ();
   init_thread (t, name, parent_thread->nice, priority, parent_thread->recent_cpu);
   tid = t->tid = allocate_tid ();
+  init_process_result (t);
+
+  #ifdef USERPROG
+    hash_init (&t->open_files, fd_hash, fd_less, NULL);
+  #endif
 
   /* Prepare thread for first run by initializing its stack.
      Do this atomically so intermediate values for the 'stack' 
@@ -259,6 +269,10 @@ thread_create (const char *name, int priority,
 
   intr_set_level (old_level);
 
+  /* No need to synchronise child_results since it is only ever accessed by one
+     thread. By the nature of increasing TIDs, this list is ordered. */
+  list_push_back (&parent_thread->child_results, &t->result->elem);
+
   /* Add to run queue. */
   thread_unblock (t);
 
@@ -632,6 +646,18 @@ is_thread (struct thread *t)
   return t != NULL && t->magic == THREAD_MAGIC;
 }
 
+/* Allocate and initialise a process result for given thread. */
+static void
+init_process_result (struct thread *t)
+{
+  struct process_result *result = malloc (sizeof (struct process_result));
+  result->tid = t->tid;
+  result->exit_status = t->exit_status;
+  lock_init (&result->lock);
+  sema_init (&result->sema, 0);
+  t->result = result;
+}
+
 /* Does basic initialization of T as a blocked thread named
    NAME. */
 static void
@@ -659,6 +685,9 @@ init_thread (struct thread *t, const char *name, int nice, int priority,
   t->recent_cpu = recent_cpu;
   t->priority = t->base_priority;
 
+  t->exit_status = -1;
+  list_init (&t->child_results);
+
   old_level = intr_disable ();
   list_push_back (&all_list, &t->allelem);
   intr_set_level (old_level);

src/threads/thread.h

@@ -2,8 +2,11 @@
 #define THREADS_THREAD_H
 
 #include <debug.h>
+#include <hash.h>
 #include <list.h>
 #include <stdint.h>
+#include <stdbool.h>
+#include "threads/synch.h"
 #include "threads/fixed-point.h"
 
 /* States in a thread's life cycle. */
@@ -29,6 +32,18 @@ typedef int tid_t;
 #define NICE_DEFAULT 0 /* Default niceness. */
 #define NICE_MAX 20    /* Highest niceness. */
 
+/* A process result, synchronised between parent and child. */
+struct process_result
+{
+  tid_t tid;             /* The tid of the child process. */
+  int exit_status;       /* The exit status of the child process. Initially set
+                            to -1, then to exit_status when child dies. */
+  struct lock lock;      /* Lock the exit_status and sema. */
+  struct semaphore sema; /* Semaphore to signal the parent that the exit_status
+                            has been set. */
+  struct list_elem elem; /* List element for the parent's children list. */
+};
+
 /* A kernel thread or user process.
 
    Each thread structure is stored in its own 4 kB page.  The
@@ -108,13 +123,21 @@ struct thread
     int nice;                           /* Nice value for this thread */
     fp32_t recent_cpu;                  /* Amount of time this process received */
 
+    /* Process wait properties. */
+    struct process_result *result;      /* Result of the process. */
+    struct list child_results;          /* List of children's of this thread
+                                           process results. */
+    struct file *exec_file;             /* Thread's currently running file */
+
     /* Shared between thread.c and synch.c. */
     struct list_elem elem;              /* List element. */
 
+    int exit_status;                    /* Exit Status: 0 = successful exit. */
 
 #ifdef USERPROG
     /* Owned by userprog/process.c. */
     uint32_t *pagedir;                  /* Page directory. */
+    struct hash open_files;             /* Hash Table of FD -> Struct File */
 #endif
 
     /* Owned by thread.c. */

src/userprog/exception.c

@@ -145,6 +145,14 @@ page_fault (struct intr_frame *f)
   write = (f->error_code & PF_W) != 0;
   user = (f->error_code & PF_U) != 0;
 
+  /* Kernel page fault is further handled by the kernel itself. */
+  if (!user)
+    {
+      f->eip = (void *)f->eax;
+      f->eax = 0xffffffff;
+      return;
+    }
+
   /* To implement virtual memory, delete the rest of the function
      body, and replace it with code that brings in the page to
      which fault_addr refers. */

src/userprog/process.c

@@ -1,6 +1,7 @@
 #include "userprog/process.h"
 #include <debug.h>
 #include <inttypes.h>
+#include <list.h>
 #include <round.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -14,57 +15,138 @@
 #include "threads/flags.h"
 #include "threads/init.h"
 #include "threads/interrupt.h"
+#include "threads/synch.h"
 #include "threads/palloc.h"
+#include "threads/malloc.h"
 #include "threads/thread.h"
 #include "threads/vaddr.h"
+#include "threads/synch.h"
+#include "devices/timer.h"
+
+/* Defines the native number of bytes processed by the processor
+   (for the purposes of alignment). */
+#define WORD_SIZE 4
+
+/* Keeps track of the position of pointers to user program arguments
+   within a linked list. */
+struct arg_elem
+  {
+    char* arg;
+    struct list_elem elem;
+  };
+
+/* Holds the data required to be passed from a kernel thread to a thread
+   that executes process_start for the purpose of starting a user process. */
+struct process_start_data
+  {
+    char *cmd; /* Pointer to a copy of the command used to execute the process.
+                  Allocated a page that must be freed by process_start. */
+    char *cmd_saveptr; /* Value pointed to by 'saveptr' argument used by
+                          successive calls to strtok_r to split 'cmd' into
+                          tokens while maintaining state. */
+    char file_name[FNAME_MAX_LEN + 1]; /* Name of the file of the process to
+                                      be started. */
+  };
 
 static thread_func start_process NO_RETURN;
 static bool load (const char *cmdline, void (**eip) (void), void **esp);
 
-/* Starts a new thread running a user program loaded from
-   FILENAME.  The new thread may be scheduled (and may even exit)
+/* Starts a new thread running a user program executed via
+   CMD.  The new thread may be scheduled (and may even exit)
    before process_execute() returns.  Returns the new process's
    thread id, or TID_ERROR if the thread cannot be created. */
 tid_t
-process_execute (const char *file_name) 
+process_execute (const char *cmd) 
 {
-  char *fn_copy;
+  char *cmd_copy;
   tid_t tid;
+  
+  struct process_start_data *data = malloc (sizeof (struct process_start_data));
+  if (data == NULL)
+    {
+      return TID_ERROR;
+    }
 
-  /* Make a copy of FILE_NAME.
+  /* Make a copy of command.
      Otherwise there's a race between the caller and load(). */
-  fn_copy = palloc_get_page (0);
-  if (fn_copy == NULL)
+  cmd_copy = palloc_get_page (0);
+  if (cmd_copy == NULL)
     return TID_ERROR;
-  strlcpy (fn_copy, file_name, PGSIZE);
 
-  /* Create a new thread to execute FILE_NAME. */
-  tid = thread_create (file_name, PRI_DEFAULT, start_process, fn_copy);
+  /* Imposing implicit limit that the command line arguments
+    including the user program name fit within a single page. */
+  strlcpy (cmd_copy, cmd, PGSIZE);
+
+  /* Retrieve first argument of command, which is the file name
+     of the process. */
+  char *file_name = strtok_r (cmd_copy, " ", &data->cmd_saveptr);
+
+/* Validates that the current file to be executed is a valid file */
+if (filesys_open (file_name) == NULL)
+    return TID_ERROR;
+
+  /* Create a new thread to execute the command, by initializing
+     it running the function 'start_process' with the appropriate
+     arguments. For details of arguments, see 'start_process'. */
+  data->cmd = cmd_copy;
+  strlcpy (data->file_name, file_name, FNAME_MAX_LEN + 1);
+
+  tid = thread_create (file_name, PRI_DEFAULT, start_process, data);
   if (tid == TID_ERROR)
-    palloc_free_page (fn_copy); 
+    palloc_free_page (cmd_copy); 
   return tid;
 }
 
-/* A thread function that loads a user process and starts it
-   running. */
+static bool install_page (void *upage, void *kpage, bool writable);
+static bool process_init_stack (char *cmd_saveptr, void **esp, char *file_name);
+static void *push_to_stack (void **esp, void *data, size_t data_size);
+#define push_var_to_stack(esp, var) (push_to_stack (esp, &var, sizeof (var)))
+
+/* Make the current thread execute 'cmd', passing in a copy of the
+   command string used for processing, the saveptr used by strtok_r
+   (in order to further tokenize the same command and retrieve its
+   arguments), as well as the name of the file being executed. This
+   involves loading the specified file and starting it running. */
 static void
-start_process (void *file_name_)
+start_process (void *proc_start_data)
 {
-  char *file_name = file_name_;
   struct intr_frame if_;
   bool success;
 
+  struct process_start_data *data = proc_start_data;
+
   /* Initialize interrupt frame and load executable. */
   memset (&if_, 0, sizeof if_);
   if_.gs = if_.fs = if_.es = if_.ds = if_.ss = SEL_UDSEG;
   if_.cs = SEL_UCSEG;
   if_.eflags = FLAG_IF | FLAG_MBS;
-  success = load (file_name, &if_.eip, &if_.esp);
+  success = load (data->file_name, &if_.eip, &if_.esp);
 
   /* If load failed, quit. */
-  palloc_free_page (file_name);
   if (!success) 
-    thread_exit ();
+    {
+      palloc_free_page (data->cmd);
+      goto fail;
+    }
+
+  /* Initialize user process stack and free page used to store the
+     command that executed the process. */
+  success = process_init_stack (data->cmd_saveptr, &if_.esp, data->file_name);
+  palloc_free_page (data->cmd);
+  
+  /* If stack initialization failed, free resources and quit. */
+  if (!success)
+    {
+      process_exit ();
+      goto fail;
+    }
+
+  /* NOTE: Currently, the file being executed is closed in load () and then 
+           reopened here. Because load is an exported public function, this
+           might be necessary. */
+  struct file *exec_file = filesys_open (data->file_name);
+  thread_current ()->exec_file = exec_file;
+  file_deny_write (exec_file);
 
   /* Start the user process by simulating a return from an
      interrupt, implemented by intr_exit (in
@@ -74,6 +156,109 @@ start_process (void *file_name_)
      and jump to it. */
   asm volatile ("movl %0, %%esp; jmp intr_exit" : : "g" (&if_) : "memory");
   NOT_REACHED ();
+
+/* If starting the process failed, free its common resources and exit. */
+ fail:
+  free (data);
+  thread_exit ();
+}
+
+/* Helper function that initializes the stack of a newly created 
+   user process. Returns true if successful, false otherwise. */
+static bool
+process_init_stack (char *cmd_saveptr, void **esp, char *file_name)
+  {
+    /* Load command line argument *data* to user process stack.
+       This can't cause overflow due to enforcing that the size of
+       command line input must fit in a page. Also keep track
+       of pointers to the argument data within a linked list. */
+    struct list arg_list;
+    list_init (&arg_list);
+
+    char *arg = file_name;
+    int arg_count = 0;
+    while (arg != NULL)
+      {
+        push_to_stack (esp, arg, (strlen (arg) + 1) * sizeof (char));
+        
+        struct arg_elem *arg_elem = malloc (sizeof (struct arg_elem));
+        if (arg_elem == NULL)
+          {
+            printf("ERROR: Couldn't allocate argument pointer memory for %s!\n",
+                   thread_current ()->name);
+            return false;
+          }
+
+        arg_elem->arg = *esp;
+        list_push_front (&arg_list, &arg_elem->elem);
+
+        arg_count++;
+        arg = strtok_r (NULL, " ", &cmd_saveptr);
+      }
+
+    /* Calculate the remaining number of bytes that need to be written
+       to the user process stack in order to check for possible overflow. */
+    size_t align_size = ((unsigned int) *esp % WORD_SIZE) * sizeof (uint8_t);
+    size_t argv_data_size = (arg_count + 1) * sizeof (char *);
+    size_t argv_size = sizeof (char **);
+    size_t argc_size = sizeof (int);
+    size_t return_addr_size = sizeof (void *);
+    size_t remaining_size = align_size + argv_data_size + argv_size + argc_size
+                            + return_addr_size;
+
+    /* If pushing the rest of the data required for the stack would cause
+       overflow, allocate an extra page that is contiguous within the
+       virtual address space (below the current address range). */ 
+    if (PHYS_BASE - *esp + remaining_size > PGSIZE)
+      {
+        uint8_t *kpage = palloc_get_page (PAL_USER | PAL_ZERO);
+        if (!install_page (((uint8_t *) PHYS_BASE) - PGSIZE * 2, kpage, true))
+          return false;
+      }
+
+    /* Align stack pointer to word size before pushing argv elements for
+       performance. */
+    *esp -= align_size;
+    
+    /* Push a null pointer sentinel inside argv. */
+    char *null_sentinel = NULL;
+    push_var_to_stack (esp, null_sentinel); 
+
+    /* Push pointers to process arguments from argument linked list */
+    struct list_elem *e = list_begin (&arg_list);
+    struct list_elem *tail = list_tail (&arg_list);
+    while (e != tail)
+      {
+        struct arg_elem *arg_elem = list_entry (e, struct arg_elem, elem);
+        
+        push_var_to_stack(esp, arg_elem->arg);
+        
+        e = list_next (e);
+        free (arg_elem);
+      }
+
+    /* Push pointer to the start of argv array. */
+    char **argv = *esp;
+    push_var_to_stack(esp, argv);
+
+    /* Push the number of arguments to the stack. */
+    push_var_to_stack(esp, arg_count);
+
+    /* Push fake return address (null pointer). */
+    push_var_to_stack (esp, null_sentinel); 
+
+    return true;
+ }
+
+/* Helper function that pushes the first 'data_size' bytes stored
+   in the address '*data' into the stack given a pointer to the
+   stack pointer '**esp'. */
+static void *
+push_to_stack (void **esp, void *data, size_t data_size)
+{
+  *esp -= data_size;
+  memcpy (*esp, data, data_size);
+  return *esp;
 }
 
 /* Waits for thread TID to die and returns its exit status. 
@@ -88,7 +273,39 @@ start_process (void *file_name_)
 int
 process_wait (tid_t child_tid UNUSED) 
 {
-  return -1;
+  struct process_result *child_result = NULL;
+  struct list_elem *e;
+  struct thread *cur = thread_current ();
+  for (e = list_begin (&cur->child_results);
+       e != list_end (&cur->child_results); e = list_next (e))
+    {
+      struct process_result *result
+          = list_entry (e, struct process_result, elem);
+      if (result->tid == child_tid)
+        {
+          child_result = result;
+          break;
+        }
+      /* List is ordered, allowing us to break early. */
+      else if (result->tid > child_tid)
+        break;
+    }
+  if (child_result == NULL)
+    return -1;
+  /* Wait for child to die. */
+  sema_down (&child_result->sema);
+  /* We need lock release in process_exit, so we need to acquire (and possibly
+     wait) for it here to ensure we don't free the lock memory before it is
+     released in process_exit. */
+  lock_acquire (&child_result->lock);
+  /* To prevent waiting for child twice, remove it from the list.
+     No need to use lock since this is the only thread with access to
+     the struct process_result now. */
+  list_remove (&child_result->elem);
+  int exit_status = child_result->exit_status;
+  lock_release (&child_result->lock);
+  free (child_result);
+  return exit_status;
 }
 
 /* Free the current process's resources. */
@@ -98,6 +315,51 @@ process_exit (void)
   struct thread *cur = thread_current ();
   uint32_t *pd;
 
+  printf ("%s: exit(%d)\n", cur->name, cur->exit_status);
+  file_close (cur->exec_file);
+
+  /* Update process result. */
+  if (cur->result != NULL)
+    {
+      lock_acquire (&cur->result->lock);
+      cur->result->exit_status = cur->exit_status;
+      /* Parent has died, child has to free the struct process_result * */
+      if (sema_try_down (&cur->result->sema))
+        {
+          lock_release (&cur->result->lock);
+          free (cur->result);
+        }
+      /* Parent is still alive and will be the one to free the
+         struct process_result *, and may be waiting so call sema_up */
+      else
+        {
+          sema_up (&cur->result->sema);
+          lock_release (&cur->result->lock);
+        }
+    }
+
+  /* Free child process results or signal parent's death. */
+  struct list_elem *e;
+  for (e = list_begin (&cur->child_results);
+       e != list_end (&cur->child_results); e = list_next (e))
+    {
+      struct process_result *result
+          = list_entry (e, struct process_result, elem);
+      lock_acquire (&result->lock);
+      /* Child has died (and was not waited for). Free the result. */
+      if (sema_try_down (&result->sema))
+        {
+          lock_release (&result->lock);
+          free (result);
+        }
+      /* Child is still alive, signal via sema that parent has died. */
+      else
+        {
+          sema_up (&result->sema);
+          lock_release (&result->lock);
+        }
+    }
+
   /* Destroy the current process's page directory and switch back
      to the kernel-only page directory. */
   pd = cur->pagedir;
@@ -318,8 +580,6 @@ load (const char *file_name, void (**eip) (void), void **esp)
 
 /* load() helpers. */
 
-static bool install_page (void *upage, void *kpage, bool writable);
-
 /* Checks whether PHDR describes a valid, loadable segment in
    FILE and returns true if so, false otherwise. */
 static bool

src/userprog/syscall.c

@@ -1,20 +1,482 @@
 #include "userprog/syscall.h"
-#include <stdio.h>
-#include <syscall-nr.h>
+#include "devices/shutdown.h"
+#include "devices/input.h"
+#include "filesys/file.h"
+#include "filesys/filesys.h"
+#include "threads/vaddr.h"
 #include "threads/interrupt.h"
+#include "threads/malloc.h"
 #include "threads/thread.h"
+#include "threads/synch.h"
+#include "userprog/process.h"
+#include "userprog/pagedir.h"
+#include <stdio.h>
+#include <stdbool.h>
+#include <syscall-nr.h>
+
+#define MAX_SYSCALL_ARGS 3
+
+static struct lock filesys_lock;
+static unsigned fd_counter = MIN_USER_FD;
+
+struct open_file
+  {
+    int fd;                     /* File Descriptor / Identifier */
+    struct file *file;          /* Pointer to the associated file */
+    struct hash_elem elem;      /* elem for a hash table */
+  };
 
 static void syscall_handler (struct intr_frame *);
 
+/* A syscall_function is a function that receives up to 3 arguments, the
+   arguments to the functions are either ints or pointers taking up to 32 bits
+   in size. */
+typedef uintptr_t (*syscall_function) (uintptr_t, uintptr_t, uintptr_t);
+
+/* System call function prototypes */
+static void syscall_halt (void);
+static void syscall_exit (int status);
+static pid_t syscall_exec (const char *cmd_line);
+static int syscall_wait (pid_t pid);
+static bool syscall_create (const char *file, unsigned initial_size);
+static bool syscall_remove (const char *file);
+static int syscall_open (const char *file);
+static int syscall_filesize (int fd);
+static int syscall_read (int fd, void *buffer, unsigned size);
+static int syscall_write (int fd, const void *buffer, unsigned size);
+static void syscall_seek (int fd, unsigned position);
+static unsigned syscall_tell (int fd);
+static void syscall_close (int fd);
+
+static struct open_file *fd_get_file (int fd);
+static void validate_user_pointer (const void *ptr, size_t size,
+                                   bool check_write);
+static void validate_user_string (const char *str, bool check_write);
+static int get_user (const uint8_t *);
+static bool put_user (uint8_t *, uint8_t);
+
+/* A struct defining a syscall_function pointer along with its arity. */
+typedef struct
+  {
+    syscall_function function;    /* Function pointer. */
+    int arity;                    /* Number of arguments of the function. */
+  } syscall_arguments;
+
+/* A look-up table mapping numbers to system call functions with their number of
+   arguments. */
+static const syscall_arguments syscall_lookup[] =
+{
+  [SYS_HALT] = {(syscall_function) syscall_halt, 0},
+  [SYS_EXIT] = {(syscall_function) syscall_exit, 1},
+  [SYS_EXEC] = {(syscall_function) syscall_exec, 1},
+  [SYS_WAIT] = {(syscall_function) syscall_wait, 1},
+  [SYS_CREATE] = {(syscall_function) syscall_create, 2},
+  [SYS_REMOVE] = {(syscall_function) syscall_remove, 1},
+  [SYS_OPEN] = {(syscall_function) syscall_open, 1},
+  [SYS_FILESIZE] = {(syscall_function) syscall_filesize, 1},
+  [SYS_READ] = {(syscall_function) syscall_read, 3},
+  [SYS_WRITE] = {(syscall_function) syscall_write, 3},
+  [SYS_SEEK] = {(syscall_function) syscall_seek, 2},
+  [SYS_TELL] = {(syscall_function) syscall_tell, 1},
+  [SYS_CLOSE] = {(syscall_function) syscall_close, 1},
+};
+
+/* The number of syscall functions (i.e, number of elements) within the
+   syscall_lookup table. */
+static const int LOOKUP_SIZE
+  = sizeof (syscall_lookup) / sizeof (syscall_arguments);
+
+
+/* Initialises the syscall handling system, as well as a global lock to
+   synchronise all file access between processes. */
 void
 syscall_init (void) 
 {
   intr_register_int (0x30, 3, INTR_ON, syscall_handler, "syscall");
+  lock_init (&filesys_lock);
+}
+
+/* Function that takes a interrupt frame containing a syscall and its args.
+   Validates the arguments and pointers before calling the relevant
+   high-level system call function, storing its output (if any) in f->eax */
+static void
+syscall_handler (struct intr_frame *f)
+{
+  /* First, read the system call number from the stack. */
+  validate_user_pointer (f->esp, sizeof (uintptr_t), false);
+  uintptr_t syscall_number = *(int *)f->esp;
+
+  /* Ensures the number corresponds to a system call that can be handled. */
+  if (syscall_number >= LOOKUP_SIZE)
+    thread_exit ();
+
+  syscall_arguments syscall = syscall_lookup[syscall_number];
+
+  /* Next, read and copy the arguments from the stack pointer. */
+  validate_user_pointer (f->esp + sizeof (uintptr_t),
+                         syscall.arity * sizeof (uintptr_t), false);
+  uintptr_t args[MAX_SYSCALL_ARGS] = { 0 };
+  for (int i = 0; i < syscall.arity && i < MAX_SYSCALL_ARGS; i++)
+    args[i] = *(uintptr_t *)(f->esp + sizeof (uintptr_t) * (i + 1));
+
+  /* Call the function that handles this system call with the arguments. When
+     there is a return value it is stored in f->eax. */
+  f->eax = syscall.function (args[0], args[1], args[2]);
+}
+
+/* Called upon a "halt" syscall, resulting in a complete shutdown of the
+   process, via shutdown_power_off (); */
+static void
+syscall_halt (void)
+{
+  shutdown_power_off ();
 }
 
 static void
-syscall_handler (struct intr_frame *f UNUSED) 
+syscall_exit (int status)
 {
-  printf ("system call!\n");
+  /* Sets exit_status of the thread to status. thread_exit () will call
+     process_exit () if user programs are allowed. */
+  thread_current ()->exit_status = status;
   thread_exit ();
 }
+
+/* Executes a given command with the relevant args, by calling process_execute.
+   Acquires the filesystem lock as process_execute accesses the file system.
+   Returns PID for the process that is running the CMD_LINE
+ */
+static pid_t
+syscall_exec (const char *cmd_line)
+{
+  validate_user_string (cmd_line, false);
+
+  lock_acquire (&filesys_lock);
+  pid_t pid = process_execute(cmd_line);
+  lock_release (&filesys_lock);
+
+  return pid;
+}
+
+/* Handles the syscall of wait. Effectively a wrapper for process_wait as the
+   necessary validation and such all happens in process_wait anyway. */
+static int
+syscall_wait (pid_t pid)
+{
+  return process_wait (pid);
+}
+
+/* Handles the syscall for file creation. First validates the user file
+   pointer. Acquires the file system lock to prevent synchronisation issues,
+   and then uses FILESYS_CREATE to create the file, returning the same status */
+static bool
+syscall_create (const char *file UNUSED, unsigned initial_size UNUSED)
+{
+  validate_user_string (file, false);
+
+  lock_acquire (&filesys_lock);
+  bool status = filesys_create (file, initial_size);
+  lock_release (&filesys_lock);
+
+  return status;
+}
+
+/* Handles the syscall for file removal. First validates the user file pointer.
+   Acquires the file system lock to prevent synchronisation issues, and then
+   uses FILESYS_REMOVE to remove the file, returning the same success status */
+static bool
+syscall_remove (const char *file)
+{
+  validate_user_string (file, false);
+
+  lock_acquire (&filesys_lock);
+  bool status = filesys_remove (file);
+  lock_release (&filesys_lock);
+
+  return status;
+}
+
+/* Handles the syscall for opening a file connection. First, validates the file
+   pointer. Then it acquires a lock for the file system, in order to open the
+   connection without synchronisation issues. It then maps a new fd to this file
+   in the hash table before returning the fd. */
+static int
+syscall_open (const char *file)
+{
+  validate_user_string (file, false);
+
+  lock_acquire (&filesys_lock);
+  struct file *ptr = filesys_open (file);
+  lock_release (&filesys_lock);
+  if (ptr == NULL)
+    return -1;
+
+  /* Allocate space for a struct representing a mapping from an FD to a struct
+     file. */
+  struct open_file *file_info
+    = (struct open_file*) malloc (sizeof (struct open_file));
+  if (file_info == NULL)
+    return -1;
+
+  /* Populate the above struct, with a unique FD and the current open file */
+  file_info->fd = fd_counter++;
+  file_info->file = ptr;
+
+  /* Add the new FD->file mapping to the hashtable for the current thread */
+  hash_insert (&thread_current ()->open_files, &file_info->elem);
+
+  /* Return the new FD */
+  return file_info->fd;
+}
+
+/* Handles the syscall for getting a file's size. Converts a provided FD into
+   the asssociated file struct. Acquire the lock for the filesystem and use
+   FILE_LENGTH to calculate the length for return. */
+static int
+syscall_filesize (int fd)
+{
+  struct open_file *file_info = fd_get_file (fd);
+  if (file_info == NULL)
+      return -1;
+
+  lock_acquire (&filesys_lock);
+  int bytes = file_length (file_info->file);
+  lock_release (&filesys_lock);
+
+  return bytes;
+}
+
+/* Handles the syscall for reading SIZE bytes from a file referenced by FD.
+   If the FD references the console, use input_getc (), otherwise convert the
+   FD to its associated file struct, acquire the filesystem lock, read up to
+   SIZE bytes and then return the number of bytes read.*/
+static int
+syscall_read (int fd, void *buffer, unsigned size)
+{
+  /* Only console (fd = 0) or other files, not including STDOUT, (fd > 1) are
+     allowed. */
+  if (fd < 0 || fd == STDOUT_FILENO)
+    return -1;
+
+  validate_user_pointer (buffer, size, true);
+
+  if (fd == STDIN_FILENO)
+    {
+      /* Reading from the console. */
+      char *write_buffer = buffer;
+      for (int i = 0; i < size; i++)
+        write_buffer[i] = input_getc ();
+
+      return size;
+    }
+  else
+    {
+      /* Reading from a file. */
+      struct open_file *file_info = fd_get_file (fd);
+      if (file_info == NULL)
+        return -1;
+
+      lock_acquire (&filesys_lock);
+      int bytes_written = file_read (file_info->file, buffer, size);
+      lock_release (&filesys_lock);
+      return bytes_written;
+    }
+}
+
+/* Handles the syscall for writing SIZE bytes to a file referenced by FD.
+   If the FD references the console, use put_buf (), otherwise convert the
+   FD to its associated file struct, acquire the filesystem lock, write up to
+   SIZE bytes and then return the number of bytes written.*/
+static int
+syscall_write (int fd, const void *buffer, unsigned size)
+{
+  /* Only console (fd = 1) or other files, not including STDIN, (fd > 1) are
+     allowed. */
+  if (fd <= 0)
+    return 0;
+
+  validate_user_pointer (buffer, size, false);
+
+  if (fd == STDOUT_FILENO)
+    {
+      /* Writing to the console. */
+      putbuf (buffer, size);
+      return size;
+    }
+  else
+    {
+      /* Writing to a file. */
+      struct open_file *file_info = fd_get_file (fd);
+      if (file_info == NULL)
+        return 0;
+
+      lock_acquire (&filesys_lock);
+      int bytes = file_write (file_info->file, buffer, size);
+      lock_release (&filesys_lock);
+
+      return bytes;
+    }
+}
+
+/* Handles the syscall for seeking to POSITION bytes in a file referenced by
+   FD. Converts the FD to its associated file struct, acquires the filesystem
+   lock and then uses file_seek to adjust the cursor to a specific position in
+   the file.*/
+static void
+syscall_seek (int fd, unsigned position)
+{
+  struct open_file *file_info = fd_get_file (fd);
+  if (file_info != NULL)
+  {
+    lock_acquire (&filesys_lock);
+    file_seek (file_info->file, position);
+    lock_release (&filesys_lock);
+  }
+}
+
+/* Handles the syscall for returning the next byte in a file referenced by
+   FD. Converts the FD to its associated file struct, acquires the filesystem
+   lock and then uses file_tell to read the next byte.*/
+static unsigned
+syscall_tell (int fd)
+{
+  struct open_file *file_info = fd_get_file (fd);
+  if (file_info == NULL)
+    return 0;
+
+  lock_acquire (&filesys_lock);
+  unsigned pos = file_tell (file_info->file);
+  lock_release (&filesys_lock);
+
+  return pos;
+}
+
+/* Handles the syscall for closing a connection to a file. Converts the FD to
+   its associated file struct. If it exists, it removes it from the hash table,
+   acquires the filesystem lock, and uses file_close to close the connection.*/
+static void
+syscall_close (int fd)
+{
+  struct open_file *file_info = fd_get_file (fd);
+  if (file_info != NULL)
+    {
+      hash_delete (&thread_current ()->open_files, &file_info->elem);
+
+      lock_acquire (&filesys_lock);
+      file_close (file_info->file);
+      lock_release (&filesys_lock);
+      free (file_info);
+    }
+}
+
+/* Hashing function needed for the open_file table. Returns a hash for an entry,
+   based on its FD. */
+unsigned
+fd_hash (const struct hash_elem *element, void *aux UNUSED)
+{
+  return hash_int (hash_entry (element, struct open_file, elem)->fd);
+}
+
+/* Comparator function for the open_file table. Compares two entries based on
+   the FDs. */
+bool
+fd_less (const struct hash_elem *a_, const struct hash_elem *b_,
+        void *aux UNUSED)
+{
+  struct open_file *a = hash_entry (a_, struct open_file, elem);
+  struct open_file *b = hash_entry (b_, struct open_file, elem);
+
+  return a->fd < b->fd;
+}
+
+/* Gets a file from its descriptor (FD number). If there is no file with the fd
+   FD it returns NULL. */
+static struct open_file *
+fd_get_file (int fd)
+{
+  /* We have to set up a fake open_file in order to be able to search the hash
+     table. See hash.h. */
+  struct open_file fake_file_info;
+  fake_file_info.fd = fd;
+
+  struct hash_elem *e
+    = hash_find (&thread_current ()->open_files, &fake_file_info.elem);
+
+  if (e == NULL)
+      return NULL;
+
+  return hash_entry (e, struct open_file, elem);
+}
+
+/* Validates if a block of memory starting at PTR and of size SIZE bytes is
+   fully contained within valid user virtual memory. thread_exit () if the
+   memory is invalid.
+   If the size is 0, the function does no checks and returns PTR. */
+static void
+validate_user_pointer (const void *ptr, size_t size, bool check_write)
+{
+  if (size == 0)
+    return;
+  /* ptr < ptr + size - 1, so sufficient to check that (ptr + size -1) is a
+     valid user virtual memory address. */
+  void *last = ptr + size - 1;
+  if (!is_user_vaddr (last))
+    thread_exit ();
+  ptr = pg_round_down (ptr);
+  while (ptr <= last)
+    {
+      int result;
+      /* Check read access to pointer. */
+      if ((result = get_user (ptr)) == -1)
+        thread_exit ();
+      /* Check write access to pointer (if required). */
+      if (check_write && !put_user (ptr, result))
+        thread_exit ();
+      ptr += PGSIZE;
+    }
+}
+
+/* Validates of a C-string starting at ptr is fully contained within valid
+   user virtual memory. thread_exit () if the memory is invalid. */
+static void
+validate_user_string (const char *ptr, bool check_write)
+{
+  while (true)
+    {
+      if (!is_user_vaddr (ptr))
+        thread_exit ();
+      int result;
+      if ((result = get_user ((const uint8_t *)ptr)) == -1)
+        thread_exit ();
+      if (check_write && !put_user ((uint8_t *)ptr, result))
+        thread_exit ();
+      if (*ptr == '\0')
+        return;
+      ptr++;
+    }
+}
+
+/* PROVIDED BY SPEC.
+   Reads a byte at user virtual address UADDR.
+   UADDR must be below PHYS_BASE.
+   Returns the byte value if successful, -1 if a segfault occurred. */
+static int
+get_user (const uint8_t *uaddr)
+{
+  int result;
+  asm ("movl $1f, %0; movzbl %1, %0; 1:" : "=&a"(result) : "m"(*uaddr));
+  return result;
+}
+
+/* PROVIDED BY SPEC.
+   Writes BYTE to user address UDST.
+   UDST must be below PHYS_BASE.
+   Returns true if successful, false if a segfault occurred. */
+static bool
+put_user (uint8_t *udst, uint8_t byte)
+{
+  int error_code;
+  asm ("movl $1f, %0; movb %b2, %1; 1:"
+       : "=&a"(error_code), "=m"(*udst)
+       : "q"(byte));
+  return error_code != -1;
+}

src/userprog/syscall.h

@@ -1,6 +1,15 @@
 #ifndef USERPROG_SYSCALL_H
 #define USERPROG_SYSCALL_H
 
+#include <hash.h>
+
+#define MIN_USER_FD 2
+
+typedef int pid_t;
+
 void syscall_init (void);
 
+unsigned fd_hash (const struct hash_elem *element, void *aux);
+bool fd_less (const struct hash_elem *a, const struct hash_elem *b, void *aux);
+
 #endif /* userprog/syscall.h */