On 2021-04-21 09:45, Oliver Neukum wrote: > suspend() does its poisoning conditionally, resume() does it > unconditionally. On a device with combined interfaces this > will balance, on a device with two interfaces the counter will > go negative and resubmission will fail. > > Both actions need to be done conditionally. FWIW, this relates to the issue reported on the kernel bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=212725 Thanks for addressing this! > > Fixes: 6069e3e927c8f ("USB: cdc-acm: untangle a circular dependency > between callback and softint") > Signed-off-by: Oliver Neukum <oneukum@xxxxxxxx> Tested-by: Stefan Agner <stefan@xxxxxxxx> -- Stefan > --- > drivers/usb/class/cdc-acm.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/usb/class/cdc-acm.c b/drivers/usb/class/cdc-acm.c > index 3fda1ec961d7..f5886c512fec 100644 > --- a/drivers/usb/class/cdc-acm.c > +++ b/drivers/usb/class/cdc-acm.c > @@ -1634,12 +1634,13 @@ static int acm_resume(struct usb_interface *intf) > struct urb *urb; > int rv = 0; > > - acm_unpoison_urbs(acm); > spin_lock_irq(&acm->write_lock); > > if (--acm->susp_count) > goto out; > > + acm_unpoison_urbs(acm); > + > if (tty_port_initialized(&acm->port)) { > rv = usb_submit_urb(acm->ctrlurb, GFP_ATOMIC);
