| Message ID | d72e872e7fdb5c04e890b978575f4d86bc61ad36.1630515789.git.ryder.lee@mediatek.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v5,1/2] mt76: mt7915: fix hwmon temp sensor mem use-after-free | expand |
On 9/1/21 10:49 AM, Ryder Lee wrote: > From: Ben Greear <greearb@candelatech.com> > > Without this change, garbage is seen in the hwmon name and sensors output > for mt7915 is garbled. It appears that the hwmon logic does not make a > copy of the incoming string, but instead just copies a char* and expects > it to never go away. > > Fixes: d6938251bb5b ("mt76: mt7915: add thermal sensor device support") > Signed-off-by: Ben Greear <greearb@candelatech.com> > Signed-off-by: Ryder Lee <ryder.lee@mediatek.com> > --- > v5: Use devm_kstrdup on the wiphy name as suggested. I don't care a great deal either way, but phyname can change (which was original way to reproduce this corruption bug), so with this v5 change, then the hwmon 'id' could confusingly be 'phy0' while user has renamed the phy0 to wiphy0 or whatever. It won't break my usage either way, so if you are happy with this, then good enough for me. But, see below, there is spurious change... > v4: Simplify flow. > v3: Add 'fixes' tag to aid backports. > --- > drivers/net/wireless/mediatek/mt76/mt7915/init.c | 8 ++++---- > drivers/net/wireless/mediatek/mt76/mt7915/mcu.c | 2 +- > 2 files changed, 5 insertions(+), 5 deletions(-) > > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > index acc83e9f409b..78b9abbe63f3 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > @@ -160,9 +160,10 @@ static int mt7915_thermal_init(struct mt7915_phy *phy) > struct wiphy *wiphy = phy->mt76->hw->wiphy; > struct thermal_cooling_device *cdev; > struct device *hwmon; > + const char *name; > > - cdev = thermal_cooling_device_register(wiphy_name(wiphy), phy, > - &mt7915_thermal_ops); > + name = devm_kstrdup(&wiphy->dev, wiphy_name(wiphy), GFP_KERNEL); > + cdev = thermal_cooling_device_register(name, phy, &mt7915_thermal_ops); > if (!IS_ERR(cdev)) { > if (sysfs_create_link(&wiphy->dev.kobj, &cdev->device.kobj, > "cooling_device") < 0) > @@ -174,8 +175,7 @@ static int mt7915_thermal_init(struct mt7915_phy *phy) > if (!IS_REACHABLE(CONFIG_HWMON)) > return 0; > > - hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, > - wiphy_name(wiphy), phy, > + hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, name, phy, > mt7915_hwmon_groups); > if (IS_ERR(hwmon)) > return PTR_ERR(hwmon); > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > index 932cf5a629db..219bb353b56d 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > @@ -1962,7 +1962,7 @@ mt7915_mcu_sta_bfer_tlv(struct mt7915_dev *dev, struct sk_buff *skb, > else > return; > > - bf->bf_cap = BIT(!ebf && dev->ibf); > + bf->bf_cap = ebf ? ebf : dev->ibf << 1; And this should not be in this patch. Thanks, Ben > bf->bw = sta->bandwidth; > bf->ibf_dbw = sta->bandwidth; > bf->ibf_nrow = tx_ant; >
On Wed, 2021-09-01 at 10:55 -0700, Ben Greear wrote: > On 9/1/21 10:49 AM, Ryder Lee wrote: > > From: Ben Greear <greearb@candelatech.com> > > > > Without this change, garbage is seen in the hwmon name and sensors > > output > > for mt7915 is garbled. It appears that the hwmon logic does not > > make a > > copy of the incoming string, but instead just copies a char* and > > expects > > it to never go away. > > > > Fixes: d6938251bb5b ("mt76: mt7915: add thermal sensor device > > support") > > Signed-off-by: Ben Greear <greearb@candelatech.com> > > Signed-off-by: Ryder Lee <ryder.lee@mediatek.com> > > --- > > v5: Use devm_kstrdup on the wiphy name as suggested. > > I don't care a great deal either way, but phyname can change (which > was original > way to reproduce this corruption bug), so with this v5 change, then > the hwmon 'id' could confusingly > be 'phy0' while user has renamed the phy0 to wiphy0 or whatever. > > It won't break my usage either way, so if you are happy with this, > then good > enough for me. I thought of this ... Considering the hwmon can't never know the phyname once it got changed. We can only specifiy a static "mt7915-xxx" to name it, or just use an initial phyname. Anyhow it's not the real phyname you change actually. > But, see below, there is spurious change... > > > > v4: Simplify flow. > > v3: Add 'fixes' tag to aid backports. > > --- > > drivers/net/wireless/mediatek/mt76/mt7915/init.c | 8 ++++---- > > drivers/net/wireless/mediatek/mt76/mt7915/mcu.c | 2 +- > > 2 files changed, 5 insertions(+), 5 deletions(-) > > > > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c > > b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > > index acc83e9f409b..78b9abbe63f3 100644 > > --- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c > > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > > @@ -160,9 +160,10 @@ static int mt7915_thermal_init(struct > > mt7915_phy *phy) > > struct wiphy *wiphy = phy->mt76->hw->wiphy; > > struct thermal_cooling_device *cdev; > > struct device *hwmon; > > + const char *name; > > > > - cdev = thermal_cooling_device_register(wiphy_name(wiphy), phy, > > - &mt7915_thermal_ops); > > + name = devm_kstrdup(&wiphy->dev, wiphy_name(wiphy), > > GFP_KERNEL); > > + cdev = thermal_cooling_device_register(name, phy, > > &mt7915_thermal_ops); > > if (!IS_ERR(cdev)) { > > if (sysfs_create_link(&wiphy->dev.kobj, &cdev- > > >device.kobj, > > "cooling_device") < 0) > > @@ -174,8 +175,7 @@ static int mt7915_thermal_init(struct > > mt7915_phy *phy) > > if (!IS_REACHABLE(CONFIG_HWMON)) > > return 0; > > > > - hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, > > - wiphy_name(wiphy > > ), phy, > > + hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, > > name, phy, > > mt7915_hwmon_gro > > ups); > > if (IS_ERR(hwmon)) > > return PTR_ERR(hwmon); > > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > > b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > > index 932cf5a629db..219bb353b56d 100644 > > --- a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c > > @@ -1962,7 +1962,7 @@ mt7915_mcu_sta_bfer_tlv(struct mt7915_dev > > *dev, struct sk_buff *skb, > > else > > return; > > > > - bf->bf_cap = BIT(!ebf && dev->ibf); > > + bf->bf_cap = ebf ? ebf : dev->ibf << 1; > > And this should not be in this patch. > Oops. I messed patch up while merging to other series. Will fix. Ryder >
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c b/drivers/net/wireless/mediatek/mt76/mt7915/init.c index acc83e9f409b..78b9abbe63f3 100644 --- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c +++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c @@ -160,9 +160,10 @@ static int mt7915_thermal_init(struct mt7915_phy *phy) struct wiphy *wiphy = phy->mt76->hw->wiphy; struct thermal_cooling_device *cdev; struct device *hwmon; + const char *name; - cdev = thermal_cooling_device_register(wiphy_name(wiphy), phy, - &mt7915_thermal_ops); + name = devm_kstrdup(&wiphy->dev, wiphy_name(wiphy), GFP_KERNEL); + cdev = thermal_cooling_device_register(name, phy, &mt7915_thermal_ops); if (!IS_ERR(cdev)) { if (sysfs_create_link(&wiphy->dev.kobj, &cdev->device.kobj, "cooling_device") < 0) @@ -174,8 +175,7 @@ static int mt7915_thermal_init(struct mt7915_phy *phy) if (!IS_REACHABLE(CONFIG_HWMON)) return 0; - hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, - wiphy_name(wiphy), phy, + hwmon = devm_hwmon_device_register_with_groups(&wiphy->dev, name, phy, mt7915_hwmon_groups); if (IS_ERR(hwmon)) return PTR_ERR(hwmon); diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c index 932cf5a629db..219bb353b56d 100644 --- a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c +++ b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.c @@ -1962,7 +1962,7 @@ mt7915_mcu_sta_bfer_tlv(struct mt7915_dev *dev, struct sk_buff *skb, else return; - bf->bf_cap = BIT(!ebf && dev->ibf); + bf->bf_cap = ebf ? ebf : dev->ibf << 1; bf->bw = sta->bandwidth; bf->ibf_dbw = sta->bandwidth; bf->ibf_nrow = tx_ant;