| Message ID | 20240807062729.3159701-1-chaitanya.kumar.borah@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [core-for-CI] usb: typec: ucsi: Fix a potential deadlock in ucsi_send_command_common() | expand |
On Wed, 2024-08-07 at 11:57 +0530, Chaitanya Kumar Borah wrote: > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > The function returns with the ppm_lock held if there's an > error or the PPM reports BUSY condition. > > This is a core-for-ci patch for [1] > > [1] https://lore.kernel.org/linux-usb/20240806112029.2984319-1-heikki.krogerus@linux.intel.com/ > > Reported-by: Luciano Coelho <luciano.coelho@intel.com> > Fixes: 5e9c1662a89b ("usb: typec: ucsi: rework command execution functions") > References: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/11849 > Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> > --- > drivers/usb/typec/ucsi/ucsi.c | 11 ++++------- > 1 file changed, 4 insertions(+), 7 deletions(-) > > diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c > index dcd3765cc1f5..432a2d6266d7 100644 > --- a/drivers/usb/typec/ucsi/ucsi.c > +++ b/drivers/usb/typec/ucsi/ucsi.c > @@ -238,13 +238,10 @@ static int ucsi_send_command_common(struct ucsi *ucsi, u64 cmd, > mutex_lock(&ucsi->ppm_lock); > > ret = ucsi_run_command(ucsi, cmd, &cci, data, size, conn_ack); > - if (cci & UCSI_CCI_BUSY) { > - ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false); > - return ret ? ret : -EBUSY; > - } > - > - if (cci & UCSI_CCI_ERROR) > - return ucsi_read_error(ucsi, connector_num); > + if (cci & UCSI_CCI_BUSY) > + ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false) ?: -EBUSY; > + else if (cci & UCSI_CCI_ERROR) > + ret = ucsi_read_error(ucsi, connector_num); > > mutex_unlock(&ucsi->ppm_lock); > return ret; This is not the exact patch that was sent upstream, where Heikki changed my Reported-by to Reported-and-Tested-by (at least). I think the best is to have the patch that was actually sent upstream, because it's easier to match later on, when it gets merged in the mainline. In any case, I think this can be merged as is, so: Acked-by: Luca Coelho <luciano.coelho@intel.com> -- Cheers, Luca.
On Wed, 07 Aug 2024, "Coelho, Luciano" <luciano.coelho@intel.com> wrote: > On Wed, 2024-08-07 at 11:57 +0530, Chaitanya Kumar Borah wrote: >> From: Heikki Krogerus <heikki.krogerus@linux.intel.com> >> >> The function returns with the ppm_lock held if there's an >> error or the PPM reports BUSY condition. >> >> This is a core-for-ci patch for [1] >> >> [1] https://lore.kernel.org/linux-usb/20240806112029.2984319-1-heikki.krogerus@linux.intel.com/ >> >> Reported-by: Luciano Coelho <luciano.coelho@intel.com> >> Fixes: 5e9c1662a89b ("usb: typec: ucsi: rework command execution functions") >> References: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/11849 >> Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> >> --- >> drivers/usb/typec/ucsi/ucsi.c | 11 ++++------- >> 1 file changed, 4 insertions(+), 7 deletions(-) >> >> diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c >> index dcd3765cc1f5..432a2d6266d7 100644 >> --- a/drivers/usb/typec/ucsi/ucsi.c >> +++ b/drivers/usb/typec/ucsi/ucsi.c >> @@ -238,13 +238,10 @@ static int ucsi_send_command_common(struct ucsi *ucsi, u64 cmd, >> mutex_lock(&ucsi->ppm_lock); >> >> ret = ucsi_run_command(ucsi, cmd, &cci, data, size, conn_ack); >> - if (cci & UCSI_CCI_BUSY) { >> - ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false); >> - return ret ? ret : -EBUSY; >> - } >> - >> - if (cci & UCSI_CCI_ERROR) >> - return ucsi_read_error(ucsi, connector_num); >> + if (cci & UCSI_CCI_BUSY) >> + ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false) ?: -EBUSY; >> + else if (cci & UCSI_CCI_ERROR) >> + ret = ucsi_read_error(ucsi, connector_num); >> >> mutex_unlock(&ucsi->ppm_lock); >> return ret; > > This is not the exact patch that was sent upstream, where Heikki > changed my Reported-by to Reported-and-Tested-by (at least). I think > the best is to have the patch that was actually sent upstream, because > it's easier to match later on, when it gets merged in the mainline. > > In any case, I think this can be merged as is, so: > > Acked-by: Luca Coelho <luciano.coelho@intel.com> Thanks, pushed to topic/core-for-CI. BR, Jani. > > -- > Cheers, > Luca.
diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c index dcd3765cc1f5..432a2d6266d7 100644 --- a/drivers/usb/typec/ucsi/ucsi.c +++ b/drivers/usb/typec/ucsi/ucsi.c @@ -238,13 +238,10 @@ static int ucsi_send_command_common(struct ucsi *ucsi, u64 cmd, mutex_lock(&ucsi->ppm_lock); ret = ucsi_run_command(ucsi, cmd, &cci, data, size, conn_ack); - if (cci & UCSI_CCI_BUSY) { - ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false); - return ret ? ret : -EBUSY; - } - - if (cci & UCSI_CCI_ERROR) - return ucsi_read_error(ucsi, connector_num); + if (cci & UCSI_CCI_BUSY) + ret = ucsi_run_command(ucsi, UCSI_CANCEL, &cci, NULL, 0, false) ?: -EBUSY; + else if (cci & UCSI_CCI_ERROR) + ret = ucsi_read_error(ucsi, connector_num); mutex_unlock(&ucsi->ppm_lock); return ret;