| Message ID | 20210426065618.588144-1-tudor.ambarus@microchip.com |
|---|---|
| State | Accepted |
| Commit | bb4031b8af804244a7e4349d38f6624f68664bd6 |
| Headers | show |
| Series | [v2] clk: Skip clk provider registration when np is NULL | expand |
On Sun, Apr 25, 2021 at 11:56 PM Tudor Ambarus <tudor.ambarus@microchip.com> wrote: > > commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > revealed that clk/bcm/clk-raspberrypi.c driver calls > devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a > NULL pointer dereference in of_clk_add_hw_provider() when calling > fwnode_dev_initialized(). > > Returning 0 is reducing the if conditions in driver code and is being > consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF > is disabled. The downside is that drivers will maybe register clkdev lookups > when they don't need to and waste some memory. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") > Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> > Reviewed-by: Stephen Boyd <sboyd@kernel.org> > --- Reviewed-by: Saravana Kannan <saravanak@google.com> -Saravana > v2: > - s/return 0;/return; in void of_clk_del_provider() > - add second fixes tag and Stephen's R-b tag > The opinions on whether to return an error or zero were split. Returning 0 > and skipping the logic was considered safer as we don't know for sure if > other drivers are affected. See: > https://lore.kernel.org/lkml/d24bebc5-0f78-021f-293f-e58defa32531@samsung.com/ > https://lore.kernel.org/lkml/20210423171335.262316-1-tudor.ambarus@microchip.com/ > > drivers/clk/clk.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > index a3b30f7de2ef..b47460b40d14 100644 > --- a/drivers/clk/clk.c > +++ b/drivers/clk/clk.c > @@ -4552,6 +4552,9 @@ int of_clk_add_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4591,6 +4594,9 @@ int of_clk_add_hw_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4688,6 +4694,9 @@ void of_clk_del_provider(struct device_node *np) > { > struct of_clk_provider *cp; > > + if (!np) > + return; > + > mutex_lock(&of_clk_mutex); > list_for_each_entry(cp, &of_clk_providers, link) { > if (cp->node == np) { > -- > 2.25.1 >
On Mon, 2021-04-26 at 09:56 +0300, Tudor Ambarus wrote: > commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > revealed that clk/bcm/clk-raspberrypi.c driver calls > devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a > NULL pointer dereference in of_clk_add_hw_provider() when calling > fwnode_dev_initialized(). > > Returning 0 is reducing the if conditions in driver code and is being > consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF > is disabled. The downside is that drivers will maybe register clkdev lookups > when they don't need to and waste some memory. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") > Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> > Reviewed-by: Stephen Boyd <sboyd@kernel.org> > --- Reviewed-by: Nicolas Saenz Julienne <nsaenz@kernel.org> Regards, Nicolas
On Mon, Apr 26, 2021 at 09:56:18AM +0300, Tudor Ambarus wrote: > commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > revealed that clk/bcm/clk-raspberrypi.c driver calls > devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a > NULL pointer dereference in of_clk_add_hw_provider() when calling > fwnode_dev_initialized(). > > Returning 0 is reducing the if conditions in driver code and is being > consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF > is disabled. The downside is that drivers will maybe register clkdev lookups > when they don't need to and waste some memory. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") > Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> > Reviewed-by: Stephen Boyd <sboyd@kernel.org> Tested-by: Guenter Roeck <linux@roeck-us.net> > --- > v2: > - s/return 0;/return; in void of_clk_del_provider() > - add second fixes tag and Stephen's R-b tag > The opinions on whether to return an error or zero were split. Returning 0 > and skipping the logic was considered safer as we don't know for sure if > other drivers are affected. See: > https://lore.kernel.org/lkml/d24bebc5-0f78-021f-293f-e58defa32531@samsung.com/ > https://lore.kernel.org/lkml/20210423171335.262316-1-tudor.ambarus@microchip.com/ > > drivers/clk/clk.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > index a3b30f7de2ef..b47460b40d14 100644 > --- a/drivers/clk/clk.c > +++ b/drivers/clk/clk.c > @@ -4552,6 +4552,9 @@ int of_clk_add_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4591,6 +4594,9 @@ int of_clk_add_hw_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4688,6 +4694,9 @@ void of_clk_del_provider(struct device_node *np) > { > struct of_clk_provider *cp; > > + if (!np) > + return; > + > mutex_lock(&of_clk_mutex); > list_for_each_entry(cp, &of_clk_providers, link) { > if (cp->node == np) { > -- > 2.25.1 >
On Mon, Apr 26, 2021 at 09:56:18AM +0300, Tudor Ambarus wrote: > commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > revealed that clk/bcm/clk-raspberrypi.c driver calls > devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a > NULL pointer dereference in of_clk_add_hw_provider() when calling > fwnode_dev_initialized(). > > Returning 0 is reducing the if conditions in driver code and is being > consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF > is disabled. The downside is that drivers will maybe register clkdev lookups > when they don't need to and waste some memory. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") > Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> > Reviewed-by: Stephen Boyd <sboyd@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> > --- > v2: > - s/return 0;/return; in void of_clk_del_provider() > - add second fixes tag and Stephen's R-b tag > The opinions on whether to return an error or zero were split. Returning 0 > and skipping the logic was considered safer as we don't know for sure if > other drivers are affected. See: > https://lore.kernel.org/lkml/d24bebc5-0f78-021f-293f-e58defa32531@samsung.com/ > https://lore.kernel.org/lkml/20210423171335.262316-1-tudor.ambarus@microchip.com/ > > drivers/clk/clk.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > index a3b30f7de2ef..b47460b40d14 100644 > --- a/drivers/clk/clk.c > +++ b/drivers/clk/clk.c > @@ -4552,6 +4552,9 @@ int of_clk_add_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4591,6 +4594,9 @@ int of_clk_add_hw_provider(struct device_node *np, > struct of_clk_provider *cp; > int ret; > > + if (!np) > + return 0; > + > cp = kzalloc(sizeof(*cp), GFP_KERNEL); > if (!cp) > return -ENOMEM; > @@ -4688,6 +4694,9 @@ void of_clk_del_provider(struct device_node *np) > { > struct of_clk_provider *cp; > > + if (!np) > + return; > + > mutex_lock(&of_clk_mutex); > list_for_each_entry(cp, &of_clk_providers, link) { > if (cp->node == np) { > -- > 2.25.1 >
On Mon, May 10, 2021 at 12:36:45PM -0700, Guenter Roeck wrote: > On Mon, Apr 26, 2021 at 09:56:18AM +0300, Tudor Ambarus wrote: > > commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > > revealed that clk/bcm/clk-raspberrypi.c driver calls > > devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a > > NULL pointer dereference in of_clk_add_hw_provider() when calling > > fwnode_dev_initialized(). > > > > Returning 0 is reducing the if conditions in driver code and is being > > consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF > > is disabled. The downside is that drivers will maybe register clkdev lookups > > when they don't need to and waste some memory. > > > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > > Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") > > Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") > > Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> > > Reviewed-by: Stephen Boyd <sboyd@kernel.org> > > Tested-by: Guenter Roeck <linux@roeck-us.net> Now applied to my tree, thanks and sorry for the delay, I thought this was going through the clk tree. greg k-h
diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index a3b30f7de2ef..b47460b40d14 100644 --- a/drivers/clk/clk.c +++ b/drivers/clk/clk.c @@ -4552,6 +4552,9 @@ int of_clk_add_provider(struct device_node *np, struct of_clk_provider *cp; int ret; + if (!np) + return 0; + cp = kzalloc(sizeof(*cp), GFP_KERNEL); if (!cp) return -ENOMEM; @@ -4591,6 +4594,9 @@ int of_clk_add_hw_provider(struct device_node *np, struct of_clk_provider *cp; int ret; + if (!np) + return 0; + cp = kzalloc(sizeof(*cp), GFP_KERNEL); if (!cp) return -ENOMEM; @@ -4688,6 +4694,9 @@ void of_clk_del_provider(struct device_node *np) { struct of_clk_provider *cp; + if (!np) + return; + mutex_lock(&of_clk_mutex); list_for_each_entry(cp, &of_clk_providers, link) { if (cp->node == np) {
commit 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") revealed that clk/bcm/clk-raspberrypi.c driver calls devm_of_clk_add_hw_provider(), with a NULL dev->of_node, which resulted in a NULL pointer dereference in of_clk_add_hw_provider() when calling fwnode_dev_initialized(). Returning 0 is reducing the if conditions in driver code and is being consistent with the CONFIG_OF=n inline stub that returns 0 when CONFIG_OF is disabled. The downside is that drivers will maybe register clkdev lookups when they don't need to and waste some memory. Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> Fixes: 6579c8d97ad7 ("clk: Mark fwnodes when their clock provider is added") Fixes: 3c9ea42802a1 ("clk: Mark fwnodes when their clock provider is added/removed") Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com> Reviewed-by: Stephen Boyd <sboyd@kernel.org> --- v2: - s/return 0;/return; in void of_clk_del_provider() - add second fixes tag and Stephen's R-b tag The opinions on whether to return an error or zero were split. Returning 0 and skipping the logic was considered safer as we don't know for sure if other drivers are affected. See: https://lore.kernel.org/lkml/d24bebc5-0f78-021f-293f-e58defa32531@samsung.com/ https://lore.kernel.org/lkml/20210423171335.262316-1-tudor.ambarus@microchip.com/ drivers/clk/clk.c | 9 +++++++++ 1 file changed, 9 insertions(+)