| Message ID | a078eb2e46d00ec59c8a91ea0afa5190730c9e58.1651212016.git.mchehab@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Let userspace know when snd-hda-intel needs i915 | expand |
On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > Sometimes, device drivers are bound using indirect references, > which is not visible when looking at /proc/modules or lsmod. > > Add a function to allow setting up module references for such > cases. > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> This sounds like duct tape at the wrong level. We should have a device_link connecting these devices, and maybe device_link internally needs to make sure the respective driver modules stay around for long enough too. But open-coding this all over the place into every driver that has some kind of cross-driver dependency sounds terrible. Or maybe the bug is that the snd driver keeps accessing the hw/component side when that is just plain gone. Iirc there's still fundamental issues there on the sound side of things, which have been attempted to paper over by timeouts and stuff like that in the past instead of enforcing a hard link between the snd and i915 side. Adding Greg for device model questions like this. -Daniel > --- > > See [PATCH 0/2] at: https://lore.kernel.org/all/cover.1651212016.git.mchehab@kernel.org/ > > include/linux/module.h | 7 +++++++ > kernel/module/main.c | 31 +++++++++++++++++++++++++++++++ > 2 files changed, 38 insertions(+) > > diff --git a/include/linux/module.h b/include/linux/module.h > index 46d4d5f2516e..be74f807e41d 100644 > --- a/include/linux/module.h > +++ b/include/linux/module.h > @@ -596,6 +596,8 @@ static inline bool within_module(unsigned long addr, const struct module *mod) > /* Search for module by name: must be in a RCU-sched critical section. */ > struct module *find_module(const char *name); > > +int module_add_named_dependency(const char *name, struct module *this); > + > /* Returns 0 and fills in value, defined and namebuf, or -ERANGE if > symnum out of range. */ > int module_get_kallsym(unsigned int symnum, unsigned long *value, char *type, > @@ -772,6 +774,11 @@ static inline int lookup_module_symbol_attrs(unsigned long addr, unsigned long * > return -ERANGE; > } > > +static inline int module_add_named_dependency(const char *name, struct module *this) > +{ > + return 0; > +} > + > static inline int module_get_kallsym(unsigned int symnum, unsigned long *value, > char *type, char *name, > char *module_name, int *exported) > diff --git a/kernel/module/main.c b/kernel/module/main.c > index 05a42d8fcd7a..dbd577ccc38c 100644 > --- a/kernel/module/main.c > +++ b/kernel/module/main.c > @@ -631,6 +631,37 @@ static int ref_module(struct module *a, struct module *b) > return 0; > } > > +int module_add_named_dependency(const char *name, struct module *this) > +{ > + struct module *mod; > + int ret; > + > + if (!name || !this || !this->name) { > + return -EINVAL; > + } > + > + mutex_lock(&module_mutex); > + mod = find_module(name); > + if (!mod) { > + ret = -EINVAL; > + goto ret; > + } > + > + ret = ref_module(this, mod); > + if (ret) > + goto ret; > + > +#ifdef CONFIG_MODULE_UNLOAD > + ret = sysfs_create_link(mod->holders_dir, > + &this->mkobj.kobj, this->name); > +#endif > + > +ret: > + mutex_unlock(&module_mutex); > + return ret; > +} > +EXPORT_SYMBOL_GPL(module_add_named_dependency); > + > /* Clear the unload stuff of the module. */ > static void module_unload_free(struct module *mod) > { > -- > 2.35.1 >
Hi Daniel, Em Fri, 29 Apr 2022 09:54:10 +0200 Daniel Vetter <daniel@ffwll.ch> escreveu: > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > Sometimes, device drivers are bound using indirect references, > > which is not visible when looking at /proc/modules or lsmod. > > > > Add a function to allow setting up module references for such > > cases. > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > This sounds like duct tape at the wrong level. We should have a > device_link connecting these devices, and maybe device_link internally > needs to make sure the respective driver modules stay around for long > enough too. But open-coding this all over the place into every driver that > has some kind of cross-driver dependency sounds terrible. > > Or maybe the bug is that the snd driver keeps accessing the hw/component > side when that is just plain gone. Iirc there's still fundamental issues > there on the sound side of things, which have been attempted to paper over > by timeouts and stuff like that in the past instead of enforcing a hard > link between the snd and i915 side. I agree with you that the device link between snd-hda and the DRM driver should properly handle unbinding on both directions. This is something that require further discussions with ALSA and DRM people, and we should keep working on it. Yet, the binding between those drivers do exist, but, despite other similar inter-driver bindings being properly reported by lsmod, this one is invisible for userspace. What this series does is to make such binding visible. As simple as that. > > Adding Greg for device model questions like this. > -Daniel > > > --- > > > > See [PATCH 0/2] at: https://lore.kernel.org/all/cover.1651212016.git.mchehab@kernel.org/ > > > > include/linux/module.h | 7 +++++++ > > kernel/module/main.c | 31 +++++++++++++++++++++++++++++++ > > 2 files changed, 38 insertions(+) > > > > diff --git a/include/linux/module.h b/include/linux/module.h > > index 46d4d5f2516e..be74f807e41d 100644 > > --- a/include/linux/module.h > > +++ b/include/linux/module.h > > @@ -596,6 +596,8 @@ static inline bool within_module(unsigned long addr, const struct module *mod) > > /* Search for module by name: must be in a RCU-sched critical section. */ > > struct module *find_module(const char *name); > > > > +int module_add_named_dependency(const char *name, struct module *this); > > + > > /* Returns 0 and fills in value, defined and namebuf, or -ERANGE if > > symnum out of range. */ > > int module_get_kallsym(unsigned int symnum, unsigned long *value, char *type, > > @@ -772,6 +774,11 @@ static inline int lookup_module_symbol_attrs(unsigned long addr, unsigned long * > > return -ERANGE; > > } > > > > +static inline int module_add_named_dependency(const char *name, struct module *this) > > +{ > > + return 0; > > +} > > + > > static inline int module_get_kallsym(unsigned int symnum, unsigned long *value, > > char *type, char *name, > > char *module_name, int *exported) > > diff --git a/kernel/module/main.c b/kernel/module/main.c > > index 05a42d8fcd7a..dbd577ccc38c 100644 > > --- a/kernel/module/main.c > > +++ b/kernel/module/main.c > > @@ -631,6 +631,37 @@ static int ref_module(struct module *a, struct module *b) > > return 0; > > } > > > > +int module_add_named_dependency(const char *name, struct module *this) > > +{ > > + struct module *mod; > > + int ret; > > + > > + if (!name || !this || !this->name) { > > + return -EINVAL; > > + } > > + > > + mutex_lock(&module_mutex); > > + mod = find_module(name); > > + if (!mod) { > > + ret = -EINVAL; > > + goto ret; > > + } > > + > > + ret = ref_module(this, mod); > > + if (ret) > > + goto ret; > > + > > +#ifdef CONFIG_MODULE_UNLOAD > > + ret = sysfs_create_link(mod->holders_dir, > > + &this->mkobj.kobj, this->name); > > +#endif > > + > > +ret: > > + mutex_unlock(&module_mutex); > > + return ret; > > +} > > +EXPORT_SYMBOL_GPL(module_add_named_dependency); > > + > > /* Clear the unload stuff of the module. */ > > static void module_unload_free(struct module *mod) > > { > > -- > > 2.35.1 > > >
On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > Hi Daniel, > > Em Fri, 29 Apr 2022 09:54:10 +0200 > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > Sometimes, device drivers are bound using indirect references, > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > Add a function to allow setting up module references for such > > > cases. > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > This sounds like duct tape at the wrong level. We should have a > > device_link connecting these devices, and maybe device_link internally > > needs to make sure the respective driver modules stay around for long > > enough too. But open-coding this all over the place into every driver that > > has some kind of cross-driver dependency sounds terrible. > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > side when that is just plain gone. Iirc there's still fundamental issues > > there on the sound side of things, which have been attempted to paper over > > by timeouts and stuff like that in the past instead of enforcing a hard > > link between the snd and i915 side. > > I agree with you that the device link between snd-hda and the DRM driver > should properly handle unbinding on both directions. This is something > that require further discussions with ALSA and DRM people, and we should > keep working on it. > > Yet, the binding between those drivers do exist, but, despite other > similar inter-driver bindings being properly reported by lsmod, this one > is invisible for userspace. > > What this series does is to make such binding visible. As simple as that. It also increases the reference count, and creates a user/kernel api with the symlinks, right? Will the reference count increase prevent the modules from now being unloadable? This feels like a very "weak" link between modules that should not be needed if reference counting is implemented properly (so that things are cleaned up in the correct order.) Please don't paper over the real issue with this hack. thanks, greg k-h
HI Greg, Em Fri, 29 Apr 2022 10:30:33 +0200 Greg KH <gregkh@linuxfoundation.org> escreveu: > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > > Hi Daniel, > > > > Em Fri, 29 Apr 2022 09:54:10 +0200 > > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > > Sometimes, device drivers are bound using indirect references, > > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > > > Add a function to allow setting up module references for such > > > > cases. > > > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > > > This sounds like duct tape at the wrong level. We should have a > > > device_link connecting these devices, and maybe device_link internally > > > needs to make sure the respective driver modules stay around for long > > > enough too. But open-coding this all over the place into every driver that > > > has some kind of cross-driver dependency sounds terrible. > > > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > > side when that is just plain gone. Iirc there's still fundamental issues > > > there on the sound side of things, which have been attempted to paper over > > > by timeouts and stuff like that in the past instead of enforcing a hard > > > link between the snd and i915 side. > > > > I agree with you that the device link between snd-hda and the DRM driver > > should properly handle unbinding on both directions. This is something > > that require further discussions with ALSA and DRM people, and we should > > keep working on it. > > > > Yet, the binding between those drivers do exist, but, despite other > > similar inter-driver bindings being properly reported by lsmod, this one > > is invisible for userspace. > > > > What this series does is to make such binding visible. As simple as that. > > It also increases the reference count, and creates a user/kernel api > with the symlinks, right? Will the reference count increase prevent the > modules from now being unloadable? > > This feels like a very "weak" link between modules that should not be > needed if reference counting is implemented properly (so that things are > cleaned up in the correct order.) The refcount increment exists even without this patch, as hda_component_master_bind() at sound/hda/hdac_component.c uses try_module_get() when it creates the device link. This series won't change anything with that regards. The only difference is that it will now properly report userspace that snd-hda will be using something inside the DRM driver (basically, it uses the DRM driver to power-control the HDA hardware on modern CPU/GPUs). - Btw, this is not the only case where userspace invisible bindings between two driver happen within the Kernel. On media, DVB drivers attach the frontend/tuner drivers using I2C bus on a way where lsmod doesn't current report any dependencies. See, for instance, PCTV 290e driver (partial) dependency chain: $ lsmod Module Size Used by rc_pinnacle_pctv_hd 16384 0 em28xx_rc 20480 0 tda18271 53248 1 cxd2820r 45056 1 em28xx_dvb 36864 0 dvb_core 155648 2 cxd2820r,em28xx_dvb em28xx 106496 2 em28xx_rc,em28xx_dvb tveeprom 28672 1 em28xx videobuf2_vmalloc 20480 1 uvcvideo videobuf2_memops 20480 1 videobuf2_vmalloc videobuf2_common 69632 4 videobuf2_vmalloc,videobuf2_v4l2,uvcvideo,videobuf2_memops videodev 266240 4 videobuf2_v4l2,uvcvideo,videobuf2_common,em28xx mc 65536 6 videodev,videobuf2_v4l2,uvcvideo,dvb_core,videobuf2_common,em28xx In the above example, tda18271 is an I2C tuner driver which talks to the hardware via the I2C bus registered by the em28xx driver. It is loaded during em28xx probing time. Again, lsmod doesn't show such dependencies. One can't remove the tuner driver without first removing the em28xx driver, which is the one that increments its refcount. - Back to the snd-hda issue, the problem is not with refcount. It is, instead, to provide a way for userspace to know what's the correct order to remove/unbind modules. Regards, Mauro
On Fri, Apr 29, 2022 at 10:15:03AM +0100, Mauro Carvalho Chehab wrote: > HI Greg, > > Em Fri, 29 Apr 2022 10:30:33 +0200 > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > > > Hi Daniel, > > > > > > Em Fri, 29 Apr 2022 09:54:10 +0200 > > > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > > > Sometimes, device drivers are bound using indirect references, > > > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > > > > > Add a function to allow setting up module references for such > > > > > cases. > > > > > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > > > > > This sounds like duct tape at the wrong level. We should have a > > > > device_link connecting these devices, and maybe device_link internally > > > > needs to make sure the respective driver modules stay around for long > > > > enough too. But open-coding this all over the place into every driver that > > > > has some kind of cross-driver dependency sounds terrible. > > > > > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > > > side when that is just plain gone. Iirc there's still fundamental issues > > > > there on the sound side of things, which have been attempted to paper over > > > > by timeouts and stuff like that in the past instead of enforcing a hard > > > > link between the snd and i915 side. > > > > > > I agree with you that the device link between snd-hda and the DRM driver > > > should properly handle unbinding on both directions. This is something > > > that require further discussions with ALSA and DRM people, and we should > > > keep working on it. > > > > > > Yet, the binding between those drivers do exist, but, despite other > > > similar inter-driver bindings being properly reported by lsmod, this one > > > is invisible for userspace. > > > > > > What this series does is to make such binding visible. As simple as that. > > > > It also increases the reference count, and creates a user/kernel api > > with the symlinks, right? Will the reference count increase prevent the > > modules from now being unloadable? > > > > This feels like a very "weak" link between modules that should not be > > needed if reference counting is implemented properly (so that things are > > cleaned up in the correct order.) > > The refcount increment exists even without this patch, as > hda_component_master_bind() at sound/hda/hdac_component.c uses > try_module_get() when it creates the device link. Ok, then why shouldn't try_module_get() be creating this link instead of having to manually do it this way again? You don't want to have to go around and add this call to all users of that function, right? thanks, greg k-h
Em Fri, 29 Apr 2022 12:10:07 +0200 Greg KH <gregkh@linuxfoundation.org> escreveu: > On Fri, Apr 29, 2022 at 10:15:03AM +0100, Mauro Carvalho Chehab wrote: > > HI Greg, > > > > Em Fri, 29 Apr 2022 10:30:33 +0200 > > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > > > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > > > > Hi Daniel, > > > > > > > > Em Fri, 29 Apr 2022 09:54:10 +0200 > > > > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > > > > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > > > > Sometimes, device drivers are bound using indirect references, > > > > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > > > > > > > Add a function to allow setting up module references for such > > > > > > cases. > > > > > > > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > > > > > > > This sounds like duct tape at the wrong level. We should have a > > > > > device_link connecting these devices, and maybe device_link internally > > > > > needs to make sure the respective driver modules stay around for long > > > > > enough too. But open-coding this all over the place into every driver that > > > > > has some kind of cross-driver dependency sounds terrible. > > > > > > > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > > > > side when that is just plain gone. Iirc there's still fundamental issues > > > > > there on the sound side of things, which have been attempted to paper over > > > > > by timeouts and stuff like that in the past instead of enforcing a hard > > > > > link between the snd and i915 side. > > > > > > > > I agree with you that the device link between snd-hda and the DRM driver > > > > should properly handle unbinding on both directions. This is something > > > > that require further discussions with ALSA and DRM people, and we should > > > > keep working on it. > > > > > > > > Yet, the binding between those drivers do exist, but, despite other > > > > similar inter-driver bindings being properly reported by lsmod, this one > > > > is invisible for userspace. > > > > > > > > What this series does is to make such binding visible. As simple as that. > > > > > > It also increases the reference count, and creates a user/kernel api > > > with the symlinks, right? Will the reference count increase prevent the > > > modules from now being unloadable? > > > > > > This feels like a very "weak" link between modules that should not be > > > needed if reference counting is implemented properly (so that things are > > > cleaned up in the correct order.) > > > > The refcount increment exists even without this patch, as > > hda_component_master_bind() at sound/hda/hdac_component.c uses > > try_module_get() when it creates the device link. > > Ok, then why shouldn't try_module_get() be creating this link instead of > having to manually do it this way again? You don't want to have to go > around and add this call to all users of that function, right? Works for me, but this is not a too trivial change, as the new try_module_get() function will require two parameters, instead of one: - the module to be referenced; - the module which will reference it. On trivial cases, one will be THIS_MODULE, but, in the specific case of snd_hda, the binding is done via an ancillary routine under snd_hda_core, but the actual binding happens at snd_hda_intel. Ok, we could add a __try_module_get() (or whatever other name that would properly express what it does) with two parameters, and then define try_module_get() as: #define try_module_get(mod) __try_module_get(mod, THIS_MODULE) Would that work for you? Regards, Mauro
On Fri, Apr 29, 2022 at 11:23:51AM +0100, Mauro Carvalho Chehab wrote: > Em Fri, 29 Apr 2022 12:10:07 +0200 > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > On Fri, Apr 29, 2022 at 10:15:03AM +0100, Mauro Carvalho Chehab wrote: > > > HI Greg, > > > > > > Em Fri, 29 Apr 2022 10:30:33 +0200 > > > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > > > > > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > > > > > Hi Daniel, > > > > > > > > > > Em Fri, 29 Apr 2022 09:54:10 +0200 > > > > > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > > > > > > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > > > > > Sometimes, device drivers are bound using indirect references, > > > > > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > > > > > > > > > Add a function to allow setting up module references for such > > > > > > > cases. > > > > > > > > > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > > > > > > > > > This sounds like duct tape at the wrong level. We should have a > > > > > > device_link connecting these devices, and maybe device_link internally > > > > > > needs to make sure the respective driver modules stay around for long > > > > > > enough too. But open-coding this all over the place into every driver that > > > > > > has some kind of cross-driver dependency sounds terrible. > > > > > > > > > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > > > > > side when that is just plain gone. Iirc there's still fundamental issues > > > > > > there on the sound side of things, which have been attempted to paper over > > > > > > by timeouts and stuff like that in the past instead of enforcing a hard > > > > > > link between the snd and i915 side. > > > > > > > > > > I agree with you that the device link between snd-hda and the DRM driver > > > > > should properly handle unbinding on both directions. This is something > > > > > that require further discussions with ALSA and DRM people, and we should > > > > > keep working on it. > > > > > > > > > > Yet, the binding between those drivers do exist, but, despite other > > > > > similar inter-driver bindings being properly reported by lsmod, this one > > > > > is invisible for userspace. > > > > > > > > > > What this series does is to make such binding visible. As simple as that. > > > > > > > > It also increases the reference count, and creates a user/kernel api > > > > with the symlinks, right? Will the reference count increase prevent the > > > > modules from now being unloadable? > > > > > > > > This feels like a very "weak" link between modules that should not be > > > > needed if reference counting is implemented properly (so that things are > > > > cleaned up in the correct order.) > > > > > > The refcount increment exists even without this patch, as > > > hda_component_master_bind() at sound/hda/hdac_component.c uses > > > try_module_get() when it creates the device link. > > > > Ok, then why shouldn't try_module_get() be creating this link instead of > > having to manually do it this way again? You don't want to have to go > > around and add this call to all users of that function, right? > > Works for me, but this is not a too trivial change, as the new > try_module_get() function will require two parameters, instead of one: > > - the module to be referenced; > - the module which will reference it. > > On trivial cases, one will be THIS_MODULE, but, in the specific case > of snd_hda, the binding is done via an ancillary routine under > snd_hda_core, but the actual binding happens at snd_hda_intel. For calls that want to increment a module reference on behalf of a different code segment than is calling it, create a new function so we can audit-the-heck out of those code paths as odds are, they are unsafe. For the normal code path, just turn try_module_get() into a macro that includes THIS_MODULE as part of it like we do for the driver register functions (see usb_register_driver() in include/linux/usb.h as an example of how to do that.) > Ok, we could add a __try_module_get() (or whatever other name that > would properly express what it does) with two parameters, and then > define try_module_get() as: > > #define try_module_get(mod) __try_module_get(mod, THIS_MODULE) Yes, that's the way forward. thanks, greg k-h
On Fri, Apr 29, 2022 at 11:23:51AM +0100, Mauro Carvalho Chehab wrote: >Em Fri, 29 Apr 2022 12:10:07 +0200 >Greg KH <gregkh@linuxfoundation.org> escreveu: > >> On Fri, Apr 29, 2022 at 10:15:03AM +0100, Mauro Carvalho Chehab wrote: >> > HI Greg, >> > >> > Em Fri, 29 Apr 2022 10:30:33 +0200 >> > Greg KH <gregkh@linuxfoundation.org> escreveu: >> > >> > > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: >> > > > Hi Daniel, >> > > > >> > > > Em Fri, 29 Apr 2022 09:54:10 +0200 >> > > > Daniel Vetter <daniel@ffwll.ch> escreveu: >> > > > >> > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: >> > > > > > Sometimes, device drivers are bound using indirect references, >> > > > > > which is not visible when looking at /proc/modules or lsmod. >> > > > > > >> > > > > > Add a function to allow setting up module references for such >> > > > > > cases. >> > > > > > >> > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> >> > > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> >> > > > > >> > > > > This sounds like duct tape at the wrong level. We should have a >> > > > > device_link connecting these devices, and maybe device_link internally >> > > > > needs to make sure the respective driver modules stay around for long >> > > > > enough too. But open-coding this all over the place into every driver that >> > > > > has some kind of cross-driver dependency sounds terrible. >> > > > > >> > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component >> > > > > side when that is just plain gone. Iirc there's still fundamental issues >> > > > > there on the sound side of things, which have been attempted to paper over >> > > > > by timeouts and stuff like that in the past instead of enforcing a hard >> > > > > link between the snd and i915 side. >> > > > >> > > > I agree with you that the device link between snd-hda and the DRM driver >> > > > should properly handle unbinding on both directions. This is something >> > > > that require further discussions with ALSA and DRM people, and we should >> > > > keep working on it. >> > > > >> > > > Yet, the binding between those drivers do exist, but, despite other >> > > > similar inter-driver bindings being properly reported by lsmod, this one >> > > > is invisible for userspace. >> > > > >> > > > What this series does is to make such binding visible. As simple as that. >> > > >> > > It also increases the reference count, and creates a user/kernel api >> > > with the symlinks, right? Will the reference count increase prevent the >> > > modules from now being unloadable? >> > > >> > > This feels like a very "weak" link between modules that should not be >> > > needed if reference counting is implemented properly (so that things are >> > > cleaned up in the correct order.) >> > >> > The refcount increment exists even without this patch, as >> > hda_component_master_bind() at sound/hda/hdac_component.c uses >> > try_module_get() when it creates the device link. >> >> Ok, then why shouldn't try_module_get() be creating this link instead of >> having to manually do it this way again? You don't want to have to go >> around and add this call to all users of that function, right? > >Works for me, but this is not a too trivial change, as the new >try_module_get() function will require two parameters, instead of one: > > - the module to be referenced; > - the module which will reference it. > >On trivial cases, one will be THIS_MODULE, but, in the specific case >of snd_hda, the binding is done via an ancillary routine under >snd_hda_core, but the actual binding happens at snd_hda_intel. > >Ok, we could add a __try_module_get() (or whatever other name that >would properly express what it does) with two parameters, and then >define try_module_get() as: > > #define try_module_get(mod) __try_module_get(mod, THIS_MODULE) agree that this should be done at this level rather than open coding it at every driver. Main improvement being fixed here regardless of the snd-hda-intel issue is to properly annotate what is holding a module. Right now we have 1) symbol module dependencies; 2) kernel references; 3) userspace references. With (2) and (3) being unknown to the user from lsmod pov. Handling this any time try_module_get() is called would make (2) visible to lsmod. Paired with fixes to the (unreleased) kmod 30[1], this allows `modprobe -r --remove-holders <module>` to also try removing the holders before removing the module itself. thanks Lucas De Marchi [1] https://lore.kernel.org/linux-modules/20220329090912.geymr6xk7taq4rtq@ldmartin-desk2.jf.intel.com/T/#t > >Would that work for you? > >Regards, >Mauro
On Fri, Apr 29, 2022 at 12:36:01PM +0200, Greg KH wrote: > On Fri, Apr 29, 2022 at 11:23:51AM +0100, Mauro Carvalho Chehab wrote: > > Em Fri, 29 Apr 2022 12:10:07 +0200 > > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > > > On Fri, Apr 29, 2022 at 10:15:03AM +0100, Mauro Carvalho Chehab wrote: > > > > HI Greg, > > > > > > > > Em Fri, 29 Apr 2022 10:30:33 +0200 > > > > Greg KH <gregkh@linuxfoundation.org> escreveu: > > > > > > > > > On Fri, Apr 29, 2022 at 09:07:57AM +0100, Mauro Carvalho Chehab wrote: > > > > > > Hi Daniel, > > > > > > > > > > > > Em Fri, 29 Apr 2022 09:54:10 +0200 > > > > > > Daniel Vetter <daniel@ffwll.ch> escreveu: > > > > > > > > > > > > > On Fri, Apr 29, 2022 at 07:31:15AM +0100, Mauro Carvalho Chehab wrote: > > > > > > > > Sometimes, device drivers are bound using indirect references, > > > > > > > > which is not visible when looking at /proc/modules or lsmod. > > > > > > > > > > > > > > > > Add a function to allow setting up module references for such > > > > > > > > cases. > > > > > > > > > > > > > > > > Reviewed-by: Dan Williams <dan.j.williams@intel.com> > > > > > > > > Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> > > > > > > > > > > > > > > This sounds like duct tape at the wrong level. We should have a > > > > > > > device_link connecting these devices, and maybe device_link internally > > > > > > > needs to make sure the respective driver modules stay around for long > > > > > > > enough too. But open-coding this all over the place into every driver that > > > > > > > has some kind of cross-driver dependency sounds terrible. > > > > > > > > > > > > > > Or maybe the bug is that the snd driver keeps accessing the hw/component > > > > > > > side when that is just plain gone. Iirc there's still fundamental issues > > > > > > > there on the sound side of things, which have been attempted to paper over > > > > > > > by timeouts and stuff like that in the past instead of enforcing a hard > > > > > > > link between the snd and i915 side. > > > > > > > > > > > > I agree with you that the device link between snd-hda and the DRM driver > > > > > > should properly handle unbinding on both directions. This is something > > > > > > that require further discussions with ALSA and DRM people, and we should > > > > > > keep working on it. > > > > > > > > > > > > Yet, the binding between those drivers do exist, but, despite other > > > > > > similar inter-driver bindings being properly reported by lsmod, this one > > > > > > is invisible for userspace. > > > > > > > > > > > > What this series does is to make such binding visible. As simple as that. > > > > > > > > > > It also increases the reference count, and creates a user/kernel api > > > > > with the symlinks, right? Will the reference count increase prevent the > > > > > modules from now being unloadable? > > > > > > > > > > This feels like a very "weak" link between modules that should not be > > > > > needed if reference counting is implemented properly (so that things are > > > > > cleaned up in the correct order.) > > > > > > > > The refcount increment exists even without this patch, as > > > > hda_component_master_bind() at sound/hda/hdac_component.c uses > > > > try_module_get() when it creates the device link. > > > > > > Ok, then why shouldn't try_module_get() be creating this link instead of > > > having to manually do it this way again? You don't want to have to go > > > around and add this call to all users of that function, right? > > > > Works for me, but this is not a too trivial change, as the new > > try_module_get() function will require two parameters, instead of one: > > > > - the module to be referenced; > > - the module which will reference it. > > > > On trivial cases, one will be THIS_MODULE, but, in the specific case > > of snd_hda, the binding is done via an ancillary routine under > > snd_hda_core, but the actual binding happens at snd_hda_intel. > > For calls that want to increment a module reference on behalf of a > different code segment than is calling it, create a new function so we > can audit-the-heck out of those code paths as odds are, they are unsafe. > > For the normal code path, just turn try_module_get() into a macro that > includes THIS_MODULE as part of it like we do for the driver register > functions (see usb_register_driver() in include/linux/usb.h as an > example of how to do that.) > > > Ok, we could add a __try_module_get() (or whatever other name that > > would properly express what it does) with two parameters, and then > > define try_module_get() as: > > > > #define try_module_get(mod) __try_module_get(mod, THIS_MODULE) > > Yes, that's the way forward. This might result in some complaints because it can create module refcount loops. Well, seemingly module refcount loops, you can break them by first unloading the driver, and then unloading the module. Currently a lot of folks rely on just unload the module to tear down the drivers (and all in right order through device links and component and all that). So maybe we want to make this some kind of weak reference, i.e. it shows up in lsmod, but the magic teardown still works and the module isn't actually pinnned. Either way I agree that something like this sounds like the right approach. -Daniel > > thanks, > > greg k-h
diff --git a/include/linux/module.h b/include/linux/module.h index 46d4d5f2516e..be74f807e41d 100644 --- a/include/linux/module.h +++ b/include/linux/module.h @@ -596,6 +596,8 @@ static inline bool within_module(unsigned long addr, const struct module *mod) /* Search for module by name: must be in a RCU-sched critical section. */ struct module *find_module(const char *name); +int module_add_named_dependency(const char *name, struct module *this); + /* Returns 0 and fills in value, defined and namebuf, or -ERANGE if symnum out of range. */ int module_get_kallsym(unsigned int symnum, unsigned long *value, char *type, @@ -772,6 +774,11 @@ static inline int lookup_module_symbol_attrs(unsigned long addr, unsigned long * return -ERANGE; } +static inline int module_add_named_dependency(const char *name, struct module *this) +{ + return 0; +} + static inline int module_get_kallsym(unsigned int symnum, unsigned long *value, char *type, char *name, char *module_name, int *exported) diff --git a/kernel/module/main.c b/kernel/module/main.c index 05a42d8fcd7a..dbd577ccc38c 100644 --- a/kernel/module/main.c +++ b/kernel/module/main.c @@ -631,6 +631,37 @@ static int ref_module(struct module *a, struct module *b) return 0; } +int module_add_named_dependency(const char *name, struct module *this) +{ + struct module *mod; + int ret; + + if (!name || !this || !this->name) { + return -EINVAL; + } + + mutex_lock(&module_mutex); + mod = find_module(name); + if (!mod) { + ret = -EINVAL; + goto ret; + } + + ret = ref_module(this, mod); + if (ret) + goto ret; + +#ifdef CONFIG_MODULE_UNLOAD + ret = sysfs_create_link(mod->holders_dir, + &this->mkobj.kobj, this->name); +#endif + +ret: + mutex_unlock(&module_mutex); + return ret; +} +EXPORT_SYMBOL_GPL(module_add_named_dependency); + /* Clear the unload stuff of the module. */ static void module_unload_free(struct module *mod) {