Hi,
Johan Hovold johan@kernel.org writes:
On Tue, Jun 11, 2019 at 08:24:16PM +0300, Felipe Balbi wrote:
If we happen to have two XHCI controllers with DbC capability, then there's no hope this will ever work as the global pointer will be overwritten by the controller that probes last.
Avoid this problem by keeping the tty_driver struct pointer inside struct xhci_dbc.
How did you test this patch?
by running it on a machine that actually has two DbCs
@@ -279,52 +279,52 @@ static const struct tty_operations dbc_tty_ops = { .unthrottle = dbc_tty_unthrottle, }; -static struct tty_driver *dbc_tty_driver;
int xhci_dbc_tty_register_driver(struct xhci_hcd *xhci) { int status; struct xhci_dbc *dbc = xhci->dbc;
- dbc_tty_driver = tty_alloc_driver(1, TTY_DRIVER_REAL_RAW |
- dbc->tty_driver = tty_alloc_driver(1, TTY_DRIVER_REAL_RAW | TTY_DRIVER_DYNAMIC_DEV);
- if (IS_ERR(dbc_tty_driver)) {
status = PTR_ERR(dbc_tty_driver);dbc_tty_driver = NULL;
- if (IS_ERR(dbc->tty_driver)) {
status = PTR_ERR(dbc->tty_driver);dbc->tty_driver = NULL;
- dbc_tty_driver->driver_name = "dbc_serial";
- dbc_tty_driver->name = "ttyDBC";
- dbc->tty_driver->driver_name = "dbc_serial";
- dbc->tty_driver->name = "ttyDBC";
You're now registering multiple drivers for the same thing (and wasting a major number for each) and specifically using the same name, which should lead to name clashes when registering the second port.
No warnings were printed while running this, actually. Odd