| Message ID | 164610295699.2682974.3646198829625502263.stgit@dwillia2-desk3.amr.corp.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | device-core: Generic device-lock lockdep validation | expand |
On Mon, 28 Feb 2022 18:49:17 -0800 Dan Williams <dan.j.williams@intel.com> wrote: > Update CONFIG_PROVE_CXL_LOCKING to use the common device-core helpers > for device_lock validation. > > When CONFIG_PROVE_LOCKING is enabled and device_set_lock_class() is > passed a non-zero lock class the core acquires the 'struct device' > @lockdep_mutex everywhere it acquires the device_lock. Where > lockdep_mutex does not skip lockdep validation like device_lock. > > cxl_set_lock_class() wraps device_set_lock_class() as to not collide > with other subsystems that may also support this lockdep validation > scheme. See the 'choice' for the various CONFIG_PROVE_$SUBSYS_LOCKING > options in lib/Kconfig.debug. > > Cc: Alison Schofield <alison.schofield@intel.com> > Cc: Vishal Verma <vishal.l.verma@intel.com> > Cc: Ira Weiny <ira.weiny@intel.com> > Cc: Ben Widawsky <ben.widawsky@intel.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> One query inline - otherwise looks good to me. > --- > diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c > index b1a4ba622739..f0a821de94cf 100644 > --- a/drivers/cxl/core/region.c > +++ b/drivers/cxl/core/region.c > @@ -476,6 +476,7 @@ int cxl_add_region(struct cxl_decoder *cxld, struct cxl_region *cxlr) > if (rc) > goto err; > > + cxl_set_lock_class(dev); I didn't see a cxl_lock_class for regions. Or is this meant to use the ANON_LOCK? > rc = device_add(dev); > if (rc) > goto err; > diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h > index 1357a245037d..f94eff659cce 100644 > --- a/drivers/cxl/cxl.h > +++ b/drivers/cxl/cxl.h > @@ -543,5 +543,14 @@ static inline int cxl_lock_class(struct device *dev) > else > return CXL_ANON_LOCK; > } > + > +static inline void cxl_set_lock_class(struct device *dev) > +{ > + device_set_lock_class(dev, cxl_lock_class(dev)); > +} > +#else > +static inline void cxl_set_lock_class(struct device *dev) > +{ > +} > #endif > #endif /* __CXL_H__ */ >
On Wed, Mar 9, 2022 at 10:33 AM Jonathan Cameron <Jonathan.Cameron@huawei.com> wrote: > > On Mon, 28 Feb 2022 18:49:17 -0800 > Dan Williams <dan.j.williams@intel.com> wrote: > > > Update CONFIG_PROVE_CXL_LOCKING to use the common device-core helpers > > for device_lock validation. > > > > When CONFIG_PROVE_LOCKING is enabled and device_set_lock_class() is > > passed a non-zero lock class the core acquires the 'struct device' > > @lockdep_mutex everywhere it acquires the device_lock. Where > > lockdep_mutex does not skip lockdep validation like device_lock. > > > > cxl_set_lock_class() wraps device_set_lock_class() as to not collide > > with other subsystems that may also support this lockdep validation > > scheme. See the 'choice' for the various CONFIG_PROVE_$SUBSYS_LOCKING > > options in lib/Kconfig.debug. > > > > Cc: Alison Schofield <alison.schofield@intel.com> > > Cc: Vishal Verma <vishal.l.verma@intel.com> > > Cc: Ira Weiny <ira.weiny@intel.com> > > Cc: Ben Widawsky <ben.widawsky@intel.com> > > Signed-off-by: Dan Williams <dan.j.williams@intel.com> > > One query inline - otherwise looks good to me. > > > --- > > > diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c > > index b1a4ba622739..f0a821de94cf 100644 > > --- a/drivers/cxl/core/region.c > > +++ b/drivers/cxl/core/region.c > > @@ -476,6 +476,7 @@ int cxl_add_region(struct cxl_decoder *cxld, struct cxl_region *cxlr) > > if (rc) > > goto err; > > > > + cxl_set_lock_class(dev); > > I didn't see a cxl_lock_class for regions. Or is this meant to use > the ANON_LOCK? Oh, yes, first I need to rebase this set before the region series which is going through a major revision. Second, I expect that the region lock_class may end up needing to nest inside the decoder lock class in order to facilitate decoders disconnecting themselves from regions if a memdev goes through ->remove() while an associated region is active. This series was motivated by wanting to validate the locking of the region creation sysfs-ABI versus device removal events.
diff --git a/drivers/cxl/core/memdev.c b/drivers/cxl/core/memdev.c index 1f76b28f9826..ad8c9f61c38a 100644 --- a/drivers/cxl/core/memdev.c +++ b/drivers/cxl/core/memdev.c @@ -343,6 +343,7 @@ struct cxl_memdev *devm_cxl_add_memdev(struct cxl_dev_state *cxlds) cxlmd->cxlds = cxlds; cdev = &cxlmd->cdev; + cxl_set_lock_class(dev); rc = cdev_device_add(cdev, dev); if (rc) goto err; diff --git a/drivers/cxl/core/pmem.c b/drivers/cxl/core/pmem.c index c3d7e6ce3fdf..e0bdda788b01 100644 --- a/drivers/cxl/core/pmem.c +++ b/drivers/cxl/core/pmem.c @@ -171,6 +171,7 @@ struct cxl_nvdimm_bridge *devm_cxl_add_nvdimm_bridge(struct device *host, if (rc) goto err; + cxl_set_lock_class(dev); rc = device_add(dev); if (rc) goto err; @@ -283,6 +284,7 @@ int devm_cxl_add_nvdimm(struct device *host, struct cxl_memdev *cxlmd) if (rc) goto err; + cxl_set_lock_class(dev); rc = device_add(dev); if (rc) goto err; diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c index 44006d8eb64d..5eee7de1c7f9 100644 --- a/drivers/cxl/core/port.c +++ b/drivers/cxl/core/port.c @@ -507,6 +507,7 @@ struct cxl_port *devm_cxl_add_port(struct device *host, struct device *uport, if (rc) goto err; + cxl_set_lock_class(dev); rc = device_add(dev); if (rc) goto err; @@ -1389,6 +1390,7 @@ int cxl_decoder_add_locked(struct cxl_decoder *cxld, int *target_map) if (is_root_decoder(dev)) cxld->platform_res.name = dev_name(dev); + cxl_set_lock_class(dev); return device_add(dev); } EXPORT_SYMBOL_NS_GPL(cxl_decoder_add_locked, CXL); diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c index b1a4ba622739..f0a821de94cf 100644 --- a/drivers/cxl/core/region.c +++ b/drivers/cxl/core/region.c @@ -476,6 +476,7 @@ int cxl_add_region(struct cxl_decoder *cxld, struct cxl_region *cxlr) if (rc) goto err; + cxl_set_lock_class(dev); rc = device_add(dev); if (rc) goto err; diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h index 1357a245037d..f94eff659cce 100644 --- a/drivers/cxl/cxl.h +++ b/drivers/cxl/cxl.h @@ -543,5 +543,14 @@ static inline int cxl_lock_class(struct device *dev) else return CXL_ANON_LOCK; } + +static inline void cxl_set_lock_class(struct device *dev) +{ + device_set_lock_class(dev, cxl_lock_class(dev)); +} +#else +static inline void cxl_set_lock_class(struct device *dev) +{ +} #endif #endif /* __CXL_H__ */
Update CONFIG_PROVE_CXL_LOCKING to use the common device-core helpers for device_lock validation. When CONFIG_PROVE_LOCKING is enabled and device_set_lock_class() is passed a non-zero lock class the core acquires the 'struct device' @lockdep_mutex everywhere it acquires the device_lock. Where lockdep_mutex does not skip lockdep validation like device_lock. cxl_set_lock_class() wraps device_set_lock_class() as to not collide with other subsystems that may also support this lockdep validation scheme. See the 'choice' for the various CONFIG_PROVE_$SUBSYS_LOCKING options in lib/Kconfig.debug. Cc: Alison Schofield <alison.schofield@intel.com> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Ben Widawsky <ben.widawsky@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- drivers/cxl/core/memdev.c | 1 + drivers/cxl/core/pmem.c | 2 ++ drivers/cxl/core/port.c | 2 ++ drivers/cxl/core/region.c | 1 + drivers/cxl/cxl.h | 9 +++++++++ 5 files changed, 15 insertions(+)