2025-02-SystemProgramming/notes/10.md

# Signals and Nonlocal Jumps

## Shell

Linux Process Hierachy: all is rooted from `init` or `systemd` (pid = 1).

A shell is an application that runs programs on behalf of the user.

* `sh` Original Unix shell
* `csh` BSD Unix C Shell
* `bash` GNU Bourne-Again Shell (default Linux shell)


## Signals

All signal is a small message that notifies a process that an event has occured in the system.

Signal is sent by kernel.

Signal is identified by small integer ID (1-30).

Only information in a signal is **its ID** and **the fact that it arrived**.

| ID  | name    | Default Action | Corresponding Event                 |
| --- | ------- | -------------- | ----------------------------------- |
| 2   | SIGINT  | terminate      | user types `Ctrl+C`                 |
| 9   | SIGKILL | terminate      | kill program                        |
| 11  | SIGSEGV | terminate      | segmentation violation              |
| 14  | SIGALRM | terminate      | timer expired                       |
| 17  | SIGCHLD | ignore         | child process stopped or terminated |

Sending a signal is essentially **updating some state** in the context of the destination process.
Kernel sends a signal for one of the following reasons:

A destination process receives a signal when it is forced by the kernel and reacts some way to the signal.
* Ignore
* Terminate (with optional core-dump)
* Catch
  * executing a user-level handler function called **signal handler**.

### Pending & Blocking

A signal is **pending** if sent but not yet received. There can be at most one pending signal of any particular type because signals are **not queued**. Signal is managed by bitmask. So subsequent signals of same type is discarded.
* sets `k` bit when delivered
* clears `k` bit when received
 
Blocking signals: a process can block the receipt of certain signals. It is not received until unblocked. It is also managed by bitmask.
* can be set and cleared by using `sigprocmask`

### Receiving

Suppose kernel is returning from an exception handler and is ready to pass control to process `p`
Kernel computes `pnb = pending(p) & ~blocked(p)`.
* if `pnb == 0`, no unblocked pending signals for `p`, so just return to `p` normally.
* otherwise, choose least nonzero bit `k` in `pnb` and force process `p` to receive signal `k`
  * The receipt of the signal triggers action by `p`
  * Repeat for all nonzero `k`

### Default Action & Signal Handler

Each signal type has a predefined default action:
* terminates
* stops until restarted by a SIGCONT signal
* ignores

```c
handler_t * signal(int signum, handler_t *handler);
```

* `SIG_IGN` ignore
* `SIG_DFL` revert to default action
* handler function pointer

### Blocking and Unblocking

* Implicit blocking
Kernel blocks any pending signals of type currently being handled.
* Explicit blocking `sigprocmask()`

Supporting functions:
* `sigemptyset()` create empty set
* `sigfillset()` add every signal number to set
* `sigaddset()` add signal number to set
* `sigdelset()` delete signal number from set

```c
sigset_t mask, prev_mask;
sigemptyset(&mask);
sigaddset(&mask, SIGINT);

sigprocmaksk(SIG_BLOCK, &mask, &prev_mask);
// critical section
sigprocmask(SIG_SETMASK, &prev_mask, NULL);
```

### Safe Signal Handling

Handlers are concurrent with main program and share global data structures. This can lead to troubles.

Some guidlines help to avoid troubles:

0. Keep you handlers as simple as possible
1. Call only **async-signal-safe** functions in your handlers
2. Save and restore `errno` on entry and exit
3. Protect accesses to shared data structure by temporarily blocking all signals
4. Declare global variables as `volatile`
5. Declare global flags as `volatile sig_atomic_t`

#### Async-Signal-Safety

A function satisfying either two conditions(below) is **Async-Signal-Safety function**:
* **reentrant**: not modifying global data (or static data), only use local data(stack frame).
* **non-interruptible**: blocking singnals during modifying global data.

POSIX guarantees that the 117 functions to be async-signal-safe, including:
* `_exit`, `write`, `wait`, `waitpid`, `sleep`, `kill`

**IMPORTANT**: `printf`, `malloc`, `sprintf`, `eixt` are **NOT** async-signal-safe.

### Correct Signal Handling Example for reaping multiple childs

```c
void child_handler(int sig) {
  int olderrno = errno; pid_t pid;
  while ((pid = wait(NULL)) > 0) {
    ccount --;
    sio_puts("Handler reaped child ");
    sio_putl((long) pid);
    sio_puts(" \n");
  }
  if (errno != ECHILD) {
    sio_error("wait error");
  }
  errno = olderrno;
}

```

### Portable Signal Handling

Different UNIX systems have different signal handling sematnics.

So `sigaction` is introduced to provide a portable interface for signal handling.

```c
handler_t * Signal(int signum, handler_t *handler) {
  struct sigaction action, old_action;
  action.sa_handler = handler;
  sigemptyset(&action.sa_mask); // block sigs of type being handled
  action.sa_flags = SA_RESTART; // restart syscalls if possible
  if (sigaction(signum, &action, &old_action) < 0)
    unix_error("Signal error");
  return (old_action.sa_handler);
}
```

### Concurrent Flows to lead Races

```c
void handler(int sig) {
  int olderrno = errno;
  pid_t pid;
  while ((pid = waitpid(-1, NULL, 0)) > 0) {
    deletejob(pid);
  }
  if (errno != ECHILD) sio_error("waitpid error");
  errno = olderrno;
}

int main() {
  while (1) {
    if ((pid = Fork()) == 0) {
      execve("/bin/date", argv, NULL);
    }
  }
  addjob(pid);
}
```

In above example, the ecf flow can be executed before the deletejob; thus it cannot be deleted.
Therefore, we need to synchronize the two concurrent flows.

```c
void handler(int sig) {
  int olderrno = errno;

  sigset_t mask_all, prev_all;
  pid_t pid;

  sigfillset(&mask_all);
  while ((pid = waitpid(-1, NULL, 0)) > 0) {
    sigprocmask(SIG_BLOCK, &mask_all, &prev_all);
    deletejob(pid);
    sigprocmask(SIG_SETMASK, &prev_all, NULL);
  }
  if (errno != ECHILD) sio_error("waitpid error");
  errno = olderrno;
}

int main() {
  pid_t pid;
  sigset_t mask_all, prev_one;

  sigfillset(&mask_all);
  signal(SIGCHLD, handler);

  while (1) {
    sigprocmask(SIG_BLOCK, &mask_all, &prev_one); // Block SIGCHLD
    if ((pid = Fork()) == 0) {
      sigprocmask(SIG_SETMASK, &prev_one, NULL); // Unblock SIGCHLD
      execve("/bin/date", argv, NULL);
    }
    sigprocmask(SIG_BLOCK, &mask_one, NULL);
    addjob(pid);
    sigprocmask(SIG_SETMASK, &prev_one, NULL);
  }
}
```