sata, highbank: clear whole array in highbank_initialize_phys()
Subsystems (2)
PATCH v2 1/1 sata, highbank: clear whole array in highbank_initialize_phys()
The original code used the wrong parameter to clear tx_atten[]. It passed the number of elements instead of sizeof() the array to memset.
The other potential issue was that cphy_base[] wasn't cleared. I'm not sure if that was a real problem or not, but I have cleared it in my patch.
Instead of using memset(), this patch uses empty initializers as a cleanup.
---
v2: used {} initialization instead of memset. I also just went ahead
and cleared cphy_base[].
drivers/ata/sata_highbank.c +3 -5
@@ -343,13 +343,11 @@ static int highbank_initialize_phys(struct device *dev, void __iomem *addr) {struct device_node *sata_node = dev->of_node;int phy_count =0 , phy, port =0 , i; -void __iomem *cphy_base[CPHY_PHY_COUNT ]; -struct device_node *phy_nodes[CPHY_PHY_COUNT ]; -u32 tx_atten[CPHY_PORT_COUNT ]; +void __iomem *cphy_base[CPHY_PHY_COUNT ] = {}; +struct device_node *phy_nodes[CPHY_PHY_COUNT ] = {}; +u32 tx_atten[CPHY_PORT_COUNT ] = {};memset (port_data,0 ,sizeof (struct phy_lane_info ) *CPHY_PORT_COUNT ); -memset (phy_nodes,0 ,sizeof (struct device_node *) *CPHY_PHY_COUNT ); -memset (tx_atten,0xff ,CPHY_PORT_COUNT ); do { u32 tmp;
Quoted 11 lines · Dan Carpenter
The original code used the wrong parameter to clear tx_atten[]. It passed the number of elements instead of
sizeof()the array to memset.The other potential issue was that cphy_base[] wasn't cleared. I'm not sure if that was a real problem or not, but I have cleared it in my patch.
Instead of using
memset(), this patch uses empty initializers as a cleanup.Signed-off-by Dan Carpenter <[email protected]>
Quoted 26 lines
---
v2: used {} initialization instead of memset. I also just went aheadand cleared cphy_base[].drivers/ata/sata_highbank.c +3
-5@@ -343,13 +343,11 @@ static int highbank_initialize_phys(struct device *dev, void __iomem *addr) {struct device_node *sata_node = dev->of_node;int phy_count =0 , phy, port =0 , i; -void __iomem *cphy_base[CPHY_PHY_COUNT ]; -struct device_node *phy_nodes[CPHY_PHY_COUNT ]; -u32 tx_atten[CPHY_PORT_COUNT ]; +void __iomem *cphy_base[CPHY_PHY_COUNT ] = {}; +struct device_node *phy_nodes[CPHY_PHY_COUNT ] = {}; +u32 tx_atten[CPHY_PORT_COUNT ] = {};memset (port_data,0 ,sizeof (struct phy_lane_info ) *CPHY_PORT_COUNT ); -memset (phy_nodes,0 ,sizeof (struct device_node *) *CPHY_PHY_COUNT ); -memset (tx_atten,0xff ,CPHY_PORT_COUNT ); do { u32 tmp;
Quoted 11 lines · Dan Carpenter
The original code used the wrong parameter to clear tx_atten[]. It passed the number of elements instead of
sizeof()the array to memset.The other potential issue was that cphy_base[] wasn't cleared. I'm not sure if that was a real problem or not, but I have cleared it in my patch.
Instead of using
memset(), this patch uses empty initializers as a cleanup.Signed-off-by Dan Carpenter <[email protected]>
Applied to libata/for-3.13. I didn't cc stable as it doesn't look like it caused actual problems. Reading the code, I'm curious how port_data[] is supposed to work. It's a static array used throughout the driver which is zeroed on each probe(), which doesn't make whole lot of sense - if I connect a second sata_highbank controller in, the struct used by the first one will be cleared. Maybe the whole PHY thing is supposed to be completely static, but if so, it'd be great if the code is explicitly structured / explained that way.
Thanks.
Signature · 1 line
tejun