Re: [PATCH v2 15/19] nptl: Use tidlock when accessing TID on pthread_getname_np

[Florian Weimer via Libc-alpha <libc-alpha@sourceware.org>]

* Adhemerval Zanella:

> +  /* Block all signal, since the lock is recursive and used on pthread_cancel
> +     (which should be async-signal-safe).  */
> +  sigset_t oldmask;
> +  __libc_signal_block_all (&oldmask);
> +  lll_lock (pd->tidlock, LLL_PRIVATE);
>  
> -  int fd = __open64_nocancel (fname, O_RDONLY);
> -  if (fd == -1)
> -    return errno;
> +  char fname[sizeof ("/proc/self/task//comm" ) + INT_BUFSIZE_BOUND (pid_t)];
> +  __snprintf (fname, sizeof (fname), "/proc/self/task/%d/comm", pd->tid);
>  
>    int res = 0;
> -  ssize_t n = TEMP_FAILURE_RETRY (__read_nocancel (fd, buf, len));
> -  if (n < 0)
> -    res = errno;
> -  else
> +  int fd = __open64_nocancel (fname, O_RDONLY);
> +  if (fd != -1)
>      {
> -      if (buf[n - 1] == '\n')
> -	buf[n - 1] = '\0';
> -      else if (n == len)
> -	res = ERANGE;
> +      ssize_t n = TEMP_FAILURE_RETRY (__read_nocancel (fd, buf, len));
> +      if (n > 0)
> +	{
> +	  if (buf[n - 1] == '\n')
> +	    buf[n - 1] = '\0';
> +	  else if (n == len)
> +	    res = ERANGE;
> +	  else
> +	    buf[n] = '\0';
> +	}
>        else
> -	buf[n] = '\0';
> +	res = errno;
> +
> +      __close_nocancel_nostatus (fd);
>      }
> +  else
> +    res = errno;

ENOENT for exited threads is inconsistent with ESRCH (or 0) in other
cases.

I wonder if we should cache the thread name in the TCB.  That would be
nice for debugging coredumps, too.

Thanks,
Florian