| Message ID | 20231223233523.4411-3-maxtram95@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | r8152: Fix a regression with usbguard | expand |
On Sat, Dec 23, 2023 at 3:36 PM Maxim Mikityanskiy <maxtram95@gmail.com> wrote: > > With the introduction of r8152-cfgselector, the following regression > appeared on machines that use usbguard: the netdev appears only when the > USB device is inserted the first time (before the module is loaded), but > on the second and next insertions no netdev is registered. > > It happens because the device is probed as unauthorized, and usbguard > gives it an authorization a moment later. If the module is not loaded, > it's normally loaded slower than the authorization is given, and > everything works. If the module is already loaded, the cfgselector's > probe function runs first, but then usb_authorize_device kicks in and > changes the configuration to something chosen by the standard > usb_choose_configuration. rtl8152_probe refuses to probe non-vendor > configurations, and the user ends up without a netdev. > > The previous commit added possibility to override > usb_choose_configuration. Use it to fix the bug and pick the right > configuration on both probe and authorization. > > Fixes: ec51fbd1b8a2 ("r8152: add USB device driver for config selection") > Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com> Maxim, Does this solve the same problem that Doug Anderson posted patches for a few weeks ago? https://lore.kernel.org/all/20231201183113.343256-1-dianders@chromium.org/ Those went through the USB tree, so that probably explains why you didn't see them in the net.git tree. cheers, grant > --- > drivers/net/usb/r8152.c | 18 +++++++++--------- > 1 file changed, 9 insertions(+), 9 deletions(-) > > diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c > index 9bf2140fd0a1..f0ac31a94f3c 100644 > --- a/drivers/net/usb/r8152.c > +++ b/drivers/net/usb/r8152.c > @@ -10070,6 +10070,11 @@ static struct usb_driver rtl8152_driver = { > }; > > static int rtl8152_cfgselector_probe(struct usb_device *udev) > +{ > + return 0; > +} > + > +static int rtl8152_cfgselector_choose_configuration(struct usb_device *udev) > { > struct usb_host_config *c; > int i, num_configs; > @@ -10078,7 +10083,7 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > * driver supports it. > */ > if (__rtl_get_hw_ver(udev) == RTL_VER_UNKNOWN) > - return 0; > + return -EOPNOTSUPP; > > /* The vendor mode is not always config #1, so to find it out. */ > c = udev->config; > @@ -10094,20 +10099,15 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > } > > if (i == num_configs) > - return -ENODEV; > - > - if (usb_set_configuration(udev, c->desc.bConfigurationValue)) { > - dev_err(&udev->dev, "Failed to set configuration %d\n", > - c->desc.bConfigurationValue); > - return -ENODEV; > - } > + return -EOPNOTSUPP; > > - return 0; > + return c->desc.bConfigurationValue; > } > > static struct usb_device_driver rtl8152_cfgselector_driver = { > .name = MODULENAME "-cfgselector", > .probe = rtl8152_cfgselector_probe, > + .choose_configuration = rtl8152_cfgselector_choose_configuration, > .id_table = rtl8152_table, > .generic_subclass = 1, > .supports_autosuspend = 1, > -- > 2.43.0 >
On Mon, Dec 25, 2023 at 11:18 PM Grant Grundler <grundler@chromium.org> wrote: > > On Sat, Dec 23, 2023 at 3:36 PM Maxim Mikityanskiy <maxtram95@gmail.com> wrote: > > > > With the introduction of r8152-cfgselector, the following regression > > appeared on machines that use usbguard: the netdev appears only when the > > USB device is inserted the first time (before the module is loaded), but > > on the second and next insertions no netdev is registered. > > > > It happens because the device is probed as unauthorized, and usbguard > > gives it an authorization a moment later. If the module is not loaded, > > it's normally loaded slower than the authorization is given, and > > everything works. If the module is already loaded, the cfgselector's > > probe function runs first, but then usb_authorize_device kicks in and > > changes the configuration to something chosen by the standard > > usb_choose_configuration. rtl8152_probe refuses to probe non-vendor > > configurations, and the user ends up without a netdev. > > > > The previous commit added possibility to override > > usb_choose_configuration. Use it to fix the bug and pick the right > > configuration on both probe and authorization. > > > > Fixes: ec51fbd1b8a2 ("r8152: add USB device driver for config selection") > > Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com> > > Maxim, > Does this solve the same problem that Doug Anderson posted patches for > a few weeks ago? > > https://lore.kernel.org/all/20231201183113.343256-1-dianders@chromium.org/ > > Those went through the USB tree, so that probably explains why you > didn't see them in the net.git tree. Specifically, usb.git usb-next branch: "r8152: Choose our USB config with choose_configuration() rather than probe()" https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git/commit/drivers/net/usb/r8152.c?h=usb-next&id=aa4f2b3e418e8673e55145de8b8016a7a9920306 Given this has a fixes tag, and two people have now tracked this down and posted fixes for this problem, is there any chance of this landing in 6.7? And it looks like Maxim's change could be rebased on top of Doug's patch and clean up a few more things. Do I see that correctly? cheers, grant > > cheers, > grant > > > --- > > drivers/net/usb/r8152.c | 18 +++++++++--------- > > 1 file changed, 9 insertions(+), 9 deletions(-) > > > > diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c > > index 9bf2140fd0a1..f0ac31a94f3c 100644 > > --- a/drivers/net/usb/r8152.c > > +++ b/drivers/net/usb/r8152.c > > @@ -10070,6 +10070,11 @@ static struct usb_driver rtl8152_driver = { > > }; > > > > static int rtl8152_cfgselector_probe(struct usb_device *udev) > > +{ > > + return 0; > > +} > > + > > +static int rtl8152_cfgselector_choose_configuration(struct usb_device *udev) > > { > > struct usb_host_config *c; > > int i, num_configs; > > @@ -10078,7 +10083,7 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > > * driver supports it. > > */ > > if (__rtl_get_hw_ver(udev) == RTL_VER_UNKNOWN) > > - return 0; > > + return -EOPNOTSUPP; > > > > /* The vendor mode is not always config #1, so to find it out. */ > > c = udev->config; > > @@ -10094,20 +10099,15 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > > } > > > > if (i == num_configs) > > - return -ENODEV; > > - > > - if (usb_set_configuration(udev, c->desc.bConfigurationValue)) { > > - dev_err(&udev->dev, "Failed to set configuration %d\n", > > - c->desc.bConfigurationValue); > > - return -ENODEV; > > - } > > + return -EOPNOTSUPP; > > > > - return 0; > > + return c->desc.bConfigurationValue; > > } > > > > static struct usb_device_driver rtl8152_cfgselector_driver = { > > .name = MODULENAME "-cfgselector", > > .probe = rtl8152_cfgselector_probe, > > + .choose_configuration = rtl8152_cfgselector_choose_configuration, > > .id_table = rtl8152_table, > > .generic_subclass = 1, > > .supports_autosuspend = 1, > > -- > > 2.43.0 > >
On Mon, 25 Dec 2023 at 23:42:08 -0800, Grant Grundler wrote: > On Mon, Dec 25, 2023 at 11:18 PM Grant Grundler <grundler@chromium.org> wrote: > > > > On Sat, Dec 23, 2023 at 3:36 PM Maxim Mikityanskiy <maxtram95@gmail.com> wrote: > > > > > > With the introduction of r8152-cfgselector, the following regression > > > appeared on machines that use usbguard: the netdev appears only when the > > > USB device is inserted the first time (before the module is loaded), but > > > on the second and next insertions no netdev is registered. > > > > > > It happens because the device is probed as unauthorized, and usbguard > > > gives it an authorization a moment later. If the module is not loaded, > > > it's normally loaded slower than the authorization is given, and > > > everything works. If the module is already loaded, the cfgselector's > > > probe function runs first, but then usb_authorize_device kicks in and > > > changes the configuration to something chosen by the standard > > > usb_choose_configuration. rtl8152_probe refuses to probe non-vendor > > > configurations, and the user ends up without a netdev. > > > > > > The previous commit added possibility to override > > > usb_choose_configuration. Use it to fix the bug and pick the right > > > configuration on both probe and authorization. > > > > > > Fixes: ec51fbd1b8a2 ("r8152: add USB device driver for config selection") > > > Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com> > > > > Maxim, > > Does this solve the same problem that Doug Anderson posted patches for > > a few weeks ago? > > > > https://lore.kernel.org/all/20231201183113.343256-1-dianders@chromium.org/ > > > > Those went through the USB tree, so that probably explains why you > > didn't see them in the net.git tree. Oh right, I didn't see these... Sorry for making noise then, Doug's patches should solve my problem. > Specifically, usb.git usb-next branch: > "r8152: Choose our USB config with choose_configuration() rather than probe()" > https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git/commit/drivers/net/usb/r8152.c?h=usb-next&id=aa4f2b3e418e8673e55145de8b8016a7a9920306 > > Given this has a fixes tag, and two people have now tracked this down > and posted fixes for this problem, is there any chance of this landing > in 6.7? > > And it looks like Maxim's change could be rebased on top of Doug's > patch and clean up a few more things. Do I see that correctly? I only see some minor difference in error handling, not sure it's worth chasing for. > cheers, > grant > > > > > cheers, > > grant > > > > > --- > > > drivers/net/usb/r8152.c | 18 +++++++++--------- > > > 1 file changed, 9 insertions(+), 9 deletions(-) > > > > > > diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c > > > index 9bf2140fd0a1..f0ac31a94f3c 100644 > > > --- a/drivers/net/usb/r8152.c > > > +++ b/drivers/net/usb/r8152.c > > > @@ -10070,6 +10070,11 @@ static struct usb_driver rtl8152_driver = { > > > }; > > > > > > static int rtl8152_cfgselector_probe(struct usb_device *udev) > > > +{ > > > + return 0; > > > +} > > > + > > > +static int rtl8152_cfgselector_choose_configuration(struct usb_device *udev) > > > { > > > struct usb_host_config *c; > > > int i, num_configs; > > > @@ -10078,7 +10083,7 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > > > * driver supports it. > > > */ > > > if (__rtl_get_hw_ver(udev) == RTL_VER_UNKNOWN) > > > - return 0; > > > + return -EOPNOTSUPP; > > > > > > /* The vendor mode is not always config #1, so to find it out. */ > > > c = udev->config; > > > @@ -10094,20 +10099,15 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) > > > } > > > > > > if (i == num_configs) > > > - return -ENODEV; > > > - > > > - if (usb_set_configuration(udev, c->desc.bConfigurationValue)) { > > > - dev_err(&udev->dev, "Failed to set configuration %d\n", > > > - c->desc.bConfigurationValue); > > > - return -ENODEV; > > > - } > > > + return -EOPNOTSUPP; > > > > > > - return 0; > > > + return c->desc.bConfigurationValue; > > > } > > > > > > static struct usb_device_driver rtl8152_cfgselector_driver = { > > > .name = MODULENAME "-cfgselector", > > > .probe = rtl8152_cfgselector_probe, > > > + .choose_configuration = rtl8152_cfgselector_choose_configuration, > > > .id_table = rtl8152_table, > > > .generic_subclass = 1, > > > .supports_autosuspend = 1, > > > -- > > > 2.43.0 > > >
diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c index 9bf2140fd0a1..f0ac31a94f3c 100644 --- a/drivers/net/usb/r8152.c +++ b/drivers/net/usb/r8152.c @@ -10070,6 +10070,11 @@ static struct usb_driver rtl8152_driver = { }; static int rtl8152_cfgselector_probe(struct usb_device *udev) +{ + return 0; +} + +static int rtl8152_cfgselector_choose_configuration(struct usb_device *udev) { struct usb_host_config *c; int i, num_configs; @@ -10078,7 +10083,7 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) * driver supports it. */ if (__rtl_get_hw_ver(udev) == RTL_VER_UNKNOWN) - return 0; + return -EOPNOTSUPP; /* The vendor mode is not always config #1, so to find it out. */ c = udev->config; @@ -10094,20 +10099,15 @@ static int rtl8152_cfgselector_probe(struct usb_device *udev) } if (i == num_configs) - return -ENODEV; - - if (usb_set_configuration(udev, c->desc.bConfigurationValue)) { - dev_err(&udev->dev, "Failed to set configuration %d\n", - c->desc.bConfigurationValue); - return -ENODEV; - } + return -EOPNOTSUPP; - return 0; + return c->desc.bConfigurationValue; } static struct usb_device_driver rtl8152_cfgselector_driver = { .name = MODULENAME "-cfgselector", .probe = rtl8152_cfgselector_probe, + .choose_configuration = rtl8152_cfgselector_choose_configuration, .id_table = rtl8152_table, .generic_subclass = 1, .supports_autosuspend = 1,
With the introduction of r8152-cfgselector, the following regression appeared on machines that use usbguard: the netdev appears only when the USB device is inserted the first time (before the module is loaded), but on the second and next insertions no netdev is registered. It happens because the device is probed as unauthorized, and usbguard gives it an authorization a moment later. If the module is not loaded, it's normally loaded slower than the authorization is given, and everything works. If the module is already loaded, the cfgselector's probe function runs first, but then usb_authorize_device kicks in and changes the configuration to something chosen by the standard usb_choose_configuration. rtl8152_probe refuses to probe non-vendor configurations, and the user ends up without a netdev. The previous commit added possibility to override usb_choose_configuration. Use it to fix the bug and pick the right configuration on both probe and authorization. Fixes: ec51fbd1b8a2 ("r8152: add USB device driver for config selection") Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com> --- drivers/net/usb/r8152.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)