Kernel: Add queue syscall and improve TCP

8 files changed, 284 insertions, 36 deletions

diff --git a/kernel/Makefile b/kernel/Makefile
index a82983b..e9ef843 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -1,6 +1,6 @@
 CC="i686-sb-gcc"
 AS="i686-sb-as"
-OBJ = arch/i386/boot.o init/kernel.o cpu/gdt.o cpu/reload_gdt.o cpu/idt.o cpu/io.o libc/stdio/print.o drivers/keyboard.o log.o drivers/pit.o libc/string/memcpy.o libc/string/strlen.o libc/string/memcmp.o drivers/ata.o libc/string/memset.o cpu/syscall.o  read_eip.o libc/exit/assert.o process.o libc/string/strcpy.o arch/i386/mmu.o kmalloc.o fs/ext2.o fs/vfs.o fs/devfs.o cpu/spinlock.o random.o libc/string/strcmp.o crypto/ChaCha20/chacha20.o crypto/SHA1/sha1.o fs/tmpfs.o libc/string/isequal.o drivers/pst.o kubsan.o drivers/serial.o socket.o poll.o fs/fifo.o hashmap/hashmap.o fs/shm.o elf.o ksbrk.o sched/scheduler.o libc/string/copy.o drivers/mouse.o libc/string/strlcpy.o libc/string/strcat.o drivers/vbe.o drivers/pci.o drivers/rtl8139.o network/ethernet.o network/arp.o network/bytes.o network/ipv4.o network/udp.o math.o signal.o network/tcp.o drivers/ahci.o crypto/xoshiro256plusplus/xoshiro256plusplus.o arch/i386/interrupts.o cpu/isr.o lib/stack.o lib/buffered_write.o lib/list.o cpu/arch_inst.o cpu/int_syscall.o lib/ringbuffer.o lib/relist.o arch/i386/tsc.o arch/i386/asm_tsc.o drivers/cmos.o timer.o
+OBJ = arch/i386/boot.o init/kernel.o cpu/gdt.o cpu/reload_gdt.o cpu/idt.o cpu/io.o libc/stdio/print.o drivers/keyboard.o log.o drivers/pit.o libc/string/memcpy.o libc/string/strlen.o libc/string/memcmp.o drivers/ata.o libc/string/memset.o cpu/syscall.o  read_eip.o libc/exit/assert.o process.o libc/string/strcpy.o arch/i386/mmu.o kmalloc.o fs/ext2.o fs/vfs.o fs/devfs.o cpu/spinlock.o random.o libc/string/strcmp.o crypto/ChaCha20/chacha20.o crypto/SHA1/sha1.o fs/tmpfs.o libc/string/isequal.o drivers/pst.o kubsan.o drivers/serial.o socket.o poll.o fs/fifo.o hashmap/hashmap.o fs/shm.o elf.o ksbrk.o sched/scheduler.o libc/string/copy.o drivers/mouse.o libc/string/strlcpy.o libc/string/strcat.o drivers/vbe.o drivers/pci.o drivers/rtl8139.o network/ethernet.o network/arp.o network/bytes.o network/ipv4.o network/udp.o math.o signal.o network/tcp.o drivers/ahci.o crypto/xoshiro256plusplus/xoshiro256plusplus.o arch/i386/interrupts.o cpu/isr.o lib/stack.o lib/buffered_write.o lib/list.o cpu/arch_inst.o cpu/int_syscall.o lib/ringbuffer.o lib/relist.o arch/i386/tsc.o arch/i386/asm_tsc.o drivers/cmos.o timer.o queue.o
 CFLAGS = -std=c99 -O0 -fsanitize=vla-bound,shift-exponent,pointer-overflow,shift,signed-integer-overflow,bounds -ggdb -ffreestanding -Wall -Wextra -Wno-int-conversion -Wno-unused-parameter -Werror -mgeneral-regs-only -Wimplicit-fallthrough -I./libc/include/ -I. -Wno-pointer-sign -DKERNEL
 LDFLAGS=
 INCLUDE=-I./includes/ -I../include/ -I./libc/include/
diff --git a/kernel/cpu/syscall.c b/kernel/cpu/syscall.c
index 9c3dde0..0dd2356 100644
--- a/kernel/cpu/syscall.c
+++ b/kernel/cpu/syscall.c
@@ -16,6 +16,7 @@
 #include <network/tcp.h>
 #include <network/udp.h>
 #include <poll.h>
+#include <queue.h>
 #include <random.h>
 #include <socket.h>
 #include <string.h>
@@ -553,29 +554,67 @@ int syscall_fcntl(int fd, int cmd, int arg) {
   return 0;
 }
 
+int syscall_queue_create(void) {
+  return queue_create();
+}
+
+int syscall_queue_mod_entries(int fd, int flag, struct queue_entry *entries,
+                              int num_entries) {
+  return queue_mod_entries(fd, flag, entries, num_entries);
+}
+
+int syscall_queue_wait(int fd, struct queue_entry *events, int num_events) {
+  return queue_wait(fd, events, num_events);
+}
+
 int (*syscall_functions[])() = {
-    (void(*))syscall_open,          (void(*))syscall_read,
-    (void(*))syscall_write,         (void(*))syscall_pread,
-    (void(*))syscall_pwrite,        (void(*))syscall_fork,
-    (void(*))syscall_exec,          (void(*))syscall_getpid,
-    (void(*))syscall_exit,          (void(*))syscall_wait,
-    (void(*))syscall_brk,           (void(*))syscall_sbrk,
-    (void(*))syscall_pipe,          (void(*))syscall_dup2,
-    (void(*))syscall_close,         (void(*))syscall_openpty,
-    (void(*))syscall_poll,          (void(*))syscall_mmap,
-    (void(*))syscall_accept,        (void(*))syscall_bind,
-    (void(*))syscall_socket,        (void(*))syscall_shm_open,
-    (void(*))syscall_ftruncate,     (void(*))syscall_fstat,
-    (void(*))syscall_msleep,        (void(*))syscall_uptime,
-    (void(*))syscall_mkdir,         (void(*))syscall_recvfrom,
-    (void(*))syscall_sendto,        (void(*))syscall_kill,
-    (void(*))syscall_sigaction,     (void(*))syscall_chdir,
-    (void(*))syscall_getcwd,        (void(*))syscall_isatty,
-    (void(*))syscall_randomfill,    (void(*))syscall_munmap,
-    (void(*))syscall_open_process,  (void(*))syscall_lseek,
-    (void(*))syscall_connect,       (void(*))syscall_setsockopt,
-    (void(*))syscall_getpeername,   (void(*))syscall_fcntl,
+    (void(*))syscall_open,
+    (void(*))syscall_read,
+    (void(*))syscall_write,
+    (void(*))syscall_pread,
+    (void(*))syscall_pwrite,
+    (void(*))syscall_fork,
+    (void(*))syscall_exec,
+    (void(*))syscall_getpid,
+    (void(*))syscall_exit,
+    (void(*))syscall_wait,
+    (void(*))syscall_brk,
+    (void(*))syscall_sbrk,
+    (void(*))syscall_pipe,
+    (void(*))syscall_dup2,
+    (void(*))syscall_close,
+    (void(*))syscall_openpty,
+    (void(*))syscall_poll,
+    (void(*))syscall_mmap,
+    (void(*))syscall_accept,
+    (void(*))syscall_bind,
+    (void(*))syscall_socket,
+    (void(*))syscall_shm_open,
+    (void(*))syscall_ftruncate,
+    (void(*))syscall_fstat,
+    (void(*))syscall_msleep,
+    (void(*))syscall_uptime,
+    (void(*))syscall_mkdir,
+    (void(*))syscall_recvfrom,
+    (void(*))syscall_sendto,
+    (void(*))syscall_kill,
+    (void(*))syscall_sigaction,
+    (void(*))syscall_chdir,
+    (void(*))syscall_getcwd,
+    (void(*))syscall_isatty,
+    (void(*))syscall_randomfill,
+    (void(*))syscall_munmap,
+    (void(*))syscall_open_process,
+    (void(*))syscall_lseek,
+    (void(*))syscall_connect,
+    (void(*))syscall_setsockopt,
+    (void(*))syscall_getpeername,
+    (void(*))syscall_fcntl,
     (void(*))syscall_clock_gettime,
+    (void(*))syscall_queue_create,
+    (void(*))syscall_queue_mod_entries,
+    (void(*))syscall_queue_wait,
+
 };
 
 void int_syscall(reg_t *r);
diff --git a/kernel/network/tcp.c b/kernel/network/tcp.c
index a2df3cd..8a56dd1 100644
--- a/kernel/network/tcp.c
+++ b/kernel/network/tcp.c
@@ -136,7 +136,10 @@ void tcp_close_connection(struct TcpConnection *con) {
     return;
   }
   if (TCP_STATE_ESTABLISHED == con->state) {
-    tcp_send_empty_payload(con, FIN);
+    // FIXME:
+    // Book says it should be FIN but observed network traffic says it
+    // should be FIN|ACK?
+    tcp_send_empty_payload(con, FIN | ACK);
     con->state = TCP_STATE_FIN_WAIT1;
     return;
   }
@@ -192,6 +195,10 @@ int send_tcp_packet(struct TcpConnection *con, const u8 *payload,
   memcpy(send_buffer, &header, sizeof(header));
   memcpy(send_buffer + sizeof(header), payload, payload_length);
 
+  if (header.flags & ACK) {
+    con->should_send_ack = 0;
+  }
+
   tcp_send(con, send_buffer, send_len, con->snd_nxt, payload_length);
 
   con->snd_nxt += payload_length;
@@ -290,6 +297,7 @@ void handle_tcp(ipv4_t src_ip, ipv4_t dst_ip, const u8 *payload,
     if (FIN & flags) {
       tcp_send_empty_payload(con, ACK);
       con->state = TCP_STATE_CLOSE_WAIT;
+      tcp_strip_connection(con);
       break;
     }
     if (tcp_payload_length > 0) {
@@ -330,6 +338,8 @@ void handle_tcp(ipv4_t src_ip, ipv4_t dst_ip, const u8 *payload,
   case TCP_STATE_FIN_WAIT2: {
     if (FIN & flags) {
       tcp_send_empty_payload(con, ACK);
+      con->state = TCP_STATE_CLOSED;
+      tcp_destroy_connection(con);
     }
     break;
   }
diff --git a/kernel/queue.c b/kernel/queue.c
new file mode 100644
index 0000000..6d4209d
--- /dev/null
+++ b/kernel/queue.c
@@ -0,0 +1,159 @@
+#include <assert.h>
+#include <fs/vfs.h>
+#include <queue.h>
+
+#define OBJECT_QUEUE 0x1337
+
+void queue_close(vfs_fd_t *inode) {
+  // TODO
+  (void)inode;
+}
+
+int queue_get_entries(struct queue_list *list, struct queue_entry *events,
+                      int num_events) {
+  int rc = 0;
+  for (int i = 0; rc < num_events; i++) {
+    struct queue_entry *entry;
+    int end;
+    if (!relist_get(&list->entries, i, (void **)&entry, &end)) {
+      if (end) {
+        break;
+      }
+      continue;
+    }
+    if (0 == entry->listen) {
+      continue;
+    }
+    vfs_fd_t *ptr = get_vfs_fd(entry->fd, list->process);
+    if (!ptr) {
+      continue;
+    }
+    int should_add = 0;
+    if (QUEUE_WAIT_READ & entry->listen) {
+      if (ptr->inode->_has_data) {
+        if (ptr->inode->_has_data(ptr->inode)) {
+          should_add = 1;
+        }
+      }
+    }
+    if (QUEUE_WAIT_WRITE & entry->listen) {
+      if (ptr->inode->_can_write) {
+        if (ptr->inode->_can_write(ptr->inode)) {
+          should_add = 1;
+        }
+      }
+    }
+    if (should_add) {
+      if (events) {
+        memcpy(events + rc, entry, sizeof(struct queue_entry));
+      }
+      rc++;
+    }
+  }
+  return rc;
+}
+
+int queue_has_data(vfs_inode_t *inode) {
+  if (1 == queue_get_entries(inode->internal_object, NULL, 1)) {
+    return 1;
+  }
+  return 0;
+}
+
+int queue_create(void) {
+  struct queue_list *list = kmalloc(sizeof(struct queue_list));
+  assert(list);
+  relist_init(&list->entries);
+  list->process = current_task;
+
+  vfs_inode_t *inode =
+      vfs_create_inode(0, 0, queue_has_data, NULL, 1 /*is_open*/, OBJECT_QUEUE,
+                       list /*internal_object*/, 0, NULL, NULL, NULL, NULL,
+                       queue_close, NULL, NULL /*get_vm_object*/, NULL, NULL,
+                       NULL /*send_signal*/, NULL /*connect*/);
+  assert(inode);
+  return vfs_create_fd(0, 0, 0, inode, NULL);
+}
+
+int queue_mod_entries(int fd, int flag, struct queue_entry *entries,
+                      int num_entries) {
+  vfs_fd_t *fd_ptr = get_vfs_fd(fd, NULL);
+  assert(fd_ptr);
+  assert(OBJECT_QUEUE == fd_ptr->inode->internal_object_type);
+  struct queue_list *list = fd_ptr->inode->internal_object;
+  if (QUEUE_MOD_ADD == flag) {
+    int i = 0;
+    for (; i < num_entries; i++) {
+      struct queue_entry *copy = kmalloc(sizeof(struct queue_entry));
+      if (!copy) {
+        break;
+      }
+      memcpy(copy, entries + i, sizeof(struct queue_entry));
+      if (!relist_add(&list->entries, copy, NULL)) {
+        kfree(copy);
+        break;
+      }
+    }
+    return i;
+  } else if (QUEUE_MOD_CHANGE == flag) {
+    int changes = 0;
+    for (int i = 0; changes < num_entries; i++) {
+      struct queue_entry *entry;
+      int end;
+      if (!relist_get(&list->entries, i, (void **)&entry, &end)) {
+        if (end) {
+          break;
+        }
+        continue;
+      }
+      for (int j = 0; j < num_entries; j++) {
+        if (entry->fd == entries[j].fd) {
+          entry->listen = entries[j].listen;
+          entry->data_type = entries[j].data_type;
+          entry->data = entries[j].data;
+          changes++;
+          break;
+        }
+      }
+    }
+    return changes;
+  } else if (QUEUE_MOD_DELETE == flag) {
+    int changes = 0;
+    for (int i = 0; changes < num_entries; i++) {
+      struct queue_entry *entry;
+      int end;
+      if (!relist_get(&list->entries, i, (void **)&entry, &end)) {
+        if (end) {
+          break;
+        }
+        continue;
+      }
+      for (int j = 0; j < num_entries; j++) {
+        if (entry->fd == entries[j].fd) {
+          relist_remove(&list->entries, i);
+          kfree(entry);
+          break;
+        }
+      }
+    }
+    return changes;
+  } else {
+    assert(0);
+  }
+  return 0;
+}
+
+int queue_wait(int fd, struct queue_entry *events, int num_events) {
+  vfs_fd_t *fd_ptr = get_vfs_fd(fd, NULL);
+  assert(fd_ptr);
+  assert(OBJECT_QUEUE == fd_ptr->inode->internal_object_type);
+  struct queue_list *list = fd_ptr->inode->internal_object;
+  int rc = queue_get_entries(list, events, num_events);
+  if (0 == rc) {
+    list_add(&current_task->read_list, fd_ptr->inode, NULL);
+    switch_task();
+    list_reset(&current_task->read_list);
+    rc = queue_get_entries(list, events, num_events);
+  }
+  return rc;
+}
diff --git a/kernel/queue.h b/kernel/queue.h
new file mode 100644
index 0000000..0885f16
--- /dev/null
+++ b/kernel/queue.h
@@ -0,0 +1,29 @@
+#include <lib/relist.h>
+#include <sched/scheduler.h>
+#include <typedefs.h>
+
+#define QUEUE_WAIT_READ (1 << 0)
+#define QUEUE_WAIT_WRITE (1 << 1)
+
+#define QUEUE_MOD_ADD 0
+#define QUEUE_MOD_CHANGE 1
+#define QUEUE_MOD_DELETE 2
+
+struct queue_entry {
+  int fd;
+  u8 listen;
+  int data_type;
+  void *data;
+};
+
+struct queue_list {
+  struct relist entries; // Store this as a binary tree(red black trees
+                         // are used by epoll) as file
+                         // descriptors will be used for lookups
+  process_t *process;
+};
+
+int queue_create(void);
+int queue_mod_entries(int fd, int flag, struct queue_entry *entries,
+                      int num_entries);
+int queue_wait(int fd, struct queue_entry *events, int num_events);
diff --git a/kernel/sched/scheduler.c b/kernel/sched/scheduler.c
index a77e1e0..d04b6e7 100644
--- a/kernel/sched/scheduler.c
+++ b/kernel/sched/scheduler.c
@@ -441,6 +441,7 @@ process_t *next_task(process_t *s) {
       c = ready_queue;
     }
     if (s == c) {
+      ms_time = timer_get_uptime();
       // wait_for_interrupt();
     }
     if (c->sleep_until > ms_time) {
diff --git a/kernel/socket.c b/kernel/socket.c
index 6398d33..fbccac5 100644
--- a/kernel/socket.c
+++ b/kernel/socket.c
@@ -37,8 +37,6 @@ void gen_ipv4(ipv4_t *ip, u8 i1, u8 i2, u8 i3, u8 i4) {
 }
 
 void tcp_remove_connection(struct TcpConnection *con) {
-  // TODO: It should also be freed but I am unsure if a inode might
-  // still have a pointer(that should not be the case)
   for (int i = 0;; i++) {
     struct TcpConnection *c;
     int end;
@@ -50,6 +48,7 @@ void tcp_remove_connection(struct TcpConnection *con) {
     }
     if (c == con) {
       relist_remove(&open_tcp_connections, i);
+      kfree(con);
       break;
     }
   }
@@ -72,8 +71,6 @@ struct TcpConnection *tcp_connect_to_listen(ipv4_t src_ip, u16 src_port,
     if (TCP_STATE_LISTEN != c->state) {
       continue;
     }
-    kprintf("c->incoming_port: %d\n", c->incoming_port);
-    kprintf("dst_port: %d\n", dst_port);
     if (c->incoming_port == dst_port) {
       struct TcpConnection *new_connection =
           kmalloc(sizeof(struct TcpConnection));
@@ -99,8 +96,9 @@ struct TcpConnection *tcp_connect_to_listen(ipv4_t src_ip, u16 src_port,
 
       new_connection->snd_wnd =
           ringbuffer_unused(&new_connection->outgoing_buffer);
-      assert(relist_add(&open_tcp_connections, new_connection, NULL));
-      assert(list_add(&c->incoming_connections, new_connection, NULL));
+      u32 index;
+      assert(relist_add(&open_tcp_connections, new_connection, &index));
+      assert(relist_add(&c->incoming_connections, new_connection, NULL));
       return new_connection;
     }
   }
@@ -228,11 +226,18 @@ int tcp_sync_buffer(struct TcpConnection *con) {
   }
 }
 
+void tcp_strip_connection(struct TcpConnection *con) {
+  ringbuffer_free(&con->incoming_buffer);
+  ringbuffer_free(&con->outgoing_buffer);
+}
+
 void tcp_close(vfs_fd_t *fd) {
   struct TcpConnection *con = fd->inode->internal_object;
   assert(con);
   tcp_sync_buffer(con);
-
+  if (TCP_STATE_ESTABLISHED == con->state) {
+    tcp_strip_connection(con);
+  }
   tcp_close_connection(con);
 }
 
@@ -603,15 +608,19 @@ struct TcpConnection *tcp_get_incoming_connection(struct TcpConnection *con,
                                                   int remove) {
   for (int i = 0;; i++) {
     struct TcpConnection *c;
-    if (!list_get(&con->incoming_connections, i, (void **)&c)) {
-      break;
+    int end;
+    if (!relist_get(&con->incoming_connections, i, (void **)&c, &end)) {
+      if (end) {
+        break;
+      }
+      continue;
     }
     if (!c) {
       continue;
     }
     if (TCP_STATE_ESTABLISHED == c->state) {
       if (remove) {
-        assert(list_set(&con->incoming_connections, i, NULL));
+        assert(relist_remove(&con->incoming_connections, i));
       }
       return c;
     }
@@ -676,7 +685,7 @@ int bind(int sockfd, const struct sockaddr *addr, socklen_t addrlen) {
 
     // TODO: Move this to listen()
     con->state = TCP_STATE_LISTEN;
-    list_init(&con->incoming_connections);
+    relist_init(&con->incoming_connections);
     fd->inode->_has_data = tcp_has_incoming_connection;
     assert(relist_add(&open_tcp_connections, con, NULL));
     return 0;
diff --git a/kernel/socket.h b/kernel/socket.h
index bba63c5..35074a1 100644
--- a/kernel/socket.h
+++ b/kernel/socket.h
@@ -5,8 +5,8 @@ typedef int socklen_t;
 #include <fs/fifo.h>
 #include <fs/vfs.h>
 #include <lib/buffered_write.h>
-#include <lib/relist.h>
 #include <lib/list.h>
+#include <lib/relist.h>
 #include <lib/ringbuffer.h>
 #include <lib/stack.h>
 #include <stddef.h>
@@ -71,7 +71,7 @@ struct TcpConnection {
   u32 snd_max;
   u32 snd_wnd;
 
-  struct list incoming_connections;
+  struct relist incoming_connections;
 };
 
 struct TcpConnection *tcp_find_connection(ipv4_t src_ip, u16 src_port,
@@ -150,4 +150,5 @@ int setsockopt(int socket, int level, int option_name, const void *option_value,
 void tcp_remove_connection(struct TcpConnection *con);
 void tcp_flush_acks(void);
 void tcp_flush_buffers(void);
+void tcp_strip_connection(struct TcpConnection *con);
 #endif