On 5 January 2015 at 09:21, Gal Hammer <gham...@redhat.com> wrote:
> The monitor's auto-completion feature stopped working when stdio is used
> as an input and qemu was resumed after it was suspended (using ctrl-z).
>
> Signed-off-by: Gal Hammer <gham...@redhat.com>
> ---
> qemu-char.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/qemu-char.c b/qemu-char.c
> index ef84b53..786df33 100644
> --- a/qemu-char.c
> +++ b/qemu-char.c
> @@ -1113,12 +1113,22 @@ static int old_fd0_flags;
> static bool stdio_in_use;
> static bool stdio_allow_signal;
>
> +static void qemu_chr_set_echo_stdio(CharDriverState *chr, bool echo);
> +
> static void term_exit(void)
> {
> tcsetattr (0, TCSANOW, &oldtty);
> fcntl(0, F_SETFL, old_fd0_flags);
> }
>
> +static void term_stdio_handler(int sig)
> +{
> + if (sig == SIGCONT) {
...why do we need this check? We don't register the function
for any other signals...
> + /* echo should be off after resume from suspend. */
> + qemu_chr_set_echo_stdio(NULL, false);
Should echo really be always off, even if the thing using the
char device had set it to on?
> + }
> +}
> +
> static void qemu_chr_set_echo_stdio(CharDriverState *chr, bool echo)
> {
> struct termios tty;
> @@ -1165,6 +1175,7 @@ static CharDriverState
> *qemu_chr_open_stdio(ChardevStdio *opts)
> tcgetattr(0, &oldtty);
> qemu_set_nonblock(0);
> atexit(term_exit);
> + signal(SIGCONT, term_stdio_handler);
This should probably be using sigaction() which is what we use
elsewhere for signal handler registration.
-- PMM
