[U-Boot] [PATCH] net, phy, cpsw: fix unaligned access if no phy found

if phy_connect() did not find a phy, phydev is not initialized and following code in cpsw_phy_init() maybe crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
--- Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ...
resetting ... --- drivers/net/cpsw.c | 3 +++ 1 Datei geändert, 3 Zeilen hinzugefügt(+)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..b18d528 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -947,6 +947,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
+ if (!phydev) + return -1; + phydev->supported &= supported; phydev->advertising = phydev->supported;

On Wed, Sep 04, 2013 at 02:16:01PM +0200, Heiko Schocher wrote:
if phy_connect() did not find a phy, phydev is not initialized and following code in cpsw_phy_init() maybe crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ...
resetting ...
drivers/net/cpsw.c | 3 +++ 1 Datei ge??ndert, 3 Zeilen hinzugef??gt(+)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..b18d528 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -947,6 +947,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
- if (!phydev)
return -1;
- phydev->supported &= supported; phydev->advertising = phydev->supported;
This isn't really an unaligned access problem it's a NULL pointer dereference, so I'll re-word the commit message when I grab this for u-boot-ti soon.

Hello Tom,
Am 04.09.2013 15:14, schrieb Tom Rini:
On Wed, Sep 04, 2013 at 02:16:01PM +0200, Heiko Schocher wrote:
if phy_connect() did not find a phy, phydev is not initialized and following code in cpsw_phy_init() maybe crashes. Fix this.
Signed-off-by: Heiko Schocherhs@denx.de Cc: Joe Hershbergerjoe.hershberger@gmail.com Cc: Mugunthan V Nmugunthanvnm@ti.com Cc: Tom Rinitrini@ti.com
Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ...
resetting ...
drivers/net/cpsw.c | 3 +++ 1 Datei ge??ndert, 3 Zeilen hinzugef??gt(+)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..b18d528 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -947,6 +947,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
- if (!phydev)
return -1;
- phydev->supported&= supported; phydev->advertising = phydev->supported;
This isn't really an unaligned access problem it's a NULL pointer dereference, so I'll re-word the commit message when I grab this for u-boot-ti soon.
Yes, thanks ... Hmm.. there are also problems if starting tftp on such a hardware ... I found more issues in this driver ... I repost a v2 soon ...
bye, Heiko

On Wed, Sep 4, 2013 at 7:16 AM, Heiko Schocher hs@denx.de wrote:
if phy_connect() did not find a phy, phydev is not initialized and following code in cpsw_phy_init() maybe crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
Acked-by: Joe Hershberger joe.hershberger@ni.com

if phy_connect() did not find a phy, phydev is NULL and following code in cpsw_phy_init() crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
--- - changes for v2: - change commit message as it is a NULL pointer deference error not an unaligned access problem, as Tom Rini mentioned - fix more places in the driver with NULL pointer problem
Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ... --- drivers/net/cpsw.c | 26 +++++++++++++++++++++++++- 1 Datei geändert, 25 Zeilen hinzugefügt(+), 1 Zeile entfernt(-)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..186665b 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -561,6 +561,9 @@ static inline void setbit_and_wait_for_clear32(void *addr) static void cpsw_set_slave_mac(struct cpsw_slave *slave, struct cpsw_priv *priv) { + if (!priv) + return; + __raw_writel(mac_hi(priv->dev->enetaddr), &slave->regs->sa_hi); __raw_writel(mac_lo(priv->dev->enetaddr), &slave->regs->sa_lo); } @@ -568,9 +571,17 @@ static void cpsw_set_slave_mac(struct cpsw_slave *slave, static void cpsw_slave_update_link(struct cpsw_slave *slave, struct cpsw_priv *priv, int *link) { - struct phy_device *phy = priv->phydev; + struct phy_device *phy; u32 mac_control = 0;
+ if (!priv) + return; + + phy = priv->phydev; + + if (!phy) + return; + phy_startup(phy); *link = phy->link;
@@ -604,8 +615,12 @@ static int cpsw_update_link(struct cpsw_priv *priv) int link = 0; struct cpsw_slave *slave;
+ if (!priv) + return -1; + for_each_slave(slave, priv) cpsw_slave_update_link(slave, priv, &link); + priv->mdio_link = readl(&mdio_regs->link); return link; } @@ -614,6 +629,9 @@ static int cpsw_check_link(struct cpsw_priv *priv) { u32 link = 0;
+ if (!priv) + return -1; + link = __raw_readl(&mdio_regs->link) & priv->phy_mask; if ((link) && (link == priv->mdio_link)) return 1; @@ -623,6 +641,9 @@ static int cpsw_check_link(struct cpsw_priv *priv)
static inline u32 cpsw_get_slave_port(struct cpsw_priv *priv, u32 slave_num) { + if (!priv) + return -1; + if (priv->host_port == 0) return slave_num + 1; else @@ -947,6 +968,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
+ if (!phydev) + return -1; + phydev->supported &= supported; phydev->advertising = phydev->supported;

On Thursday 05 September 2013 11:35 AM, Heiko Schocher wrote:
if phy_connect() did not find a phy, phydev is NULL and following code in cpsw_phy_init() crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
- changes for v2:
- change commit message as it is a NULL pointer deference error not an unaligned access problem, as Tom Rini mentioned
- fix more places in the driver with NULL pointer problem
Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ...
drivers/net/cpsw.c | 26 +++++++++++++++++++++++++- 1 Datei geändert, 25 Zeilen hinzugefügt(+), 1 Zeile entfernt(-)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..186665b 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -561,6 +561,9 @@ static inline void setbit_and_wait_for_clear32(void *addr) static void cpsw_set_slave_mac(struct cpsw_slave *slave, struct cpsw_priv *priv) {
- if (!priv)
return;
- __raw_writel(mac_hi(priv->dev->enetaddr), &slave->regs->sa_hi); __raw_writel(mac_lo(priv->dev->enetaddr), &slave->regs->sa_lo);
} @@ -568,9 +571,17 @@ static void cpsw_set_slave_mac(struct cpsw_slave *slave, static void cpsw_slave_update_link(struct cpsw_slave *slave, struct cpsw_priv *priv, int *link) {
- struct phy_device *phy = priv->phydev;
struct phy_device *phy; u32 mac_control = 0;
if (!priv)
return;
phy = priv->phydev;
if (!phy)
return;
phy_startup(phy); *link = phy->link;
@@ -604,8 +615,12 @@ static int cpsw_update_link(struct cpsw_priv *priv) int link = 0; struct cpsw_slave *slave;
- if (!priv)
return -1;
- for_each_slave(slave, priv) cpsw_slave_update_link(slave, priv, &link);
- priv->mdio_link = readl(&mdio_regs->link); return link;
} @@ -614,6 +629,9 @@ static int cpsw_check_link(struct cpsw_priv *priv) { u32 link = 0;
All the above *priv* check can be remove as priv is already validated in this function and all the above functions are called only from this function.
- if (!priv)
return -1;
- link = __raw_readl(&mdio_regs->link) & priv->phy_mask; if ((link) && (link == priv->mdio_link)) return 1;
@@ -623,6 +641,9 @@ static int cpsw_check_link(struct cpsw_priv *priv)
static inline u32 cpsw_get_slave_port(struct cpsw_priv *priv, u32 slave_num) {
- if (!priv)
return -1;
- if (priv->host_port == 0) return slave_num + 1; else
@@ -947,6 +968,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
- if (!phydev)
return -1;
- phydev->supported &= supported; phydev->advertising = phydev->supported;
I don't think you need to validate *priv* as it is already validated in driver register (cpsw_register()). In any case priv is NULL, then there is some corruption in dev structure as without priv, dev should not exist and stack is not supposed to call any cpsw apis.
Regards Mugunthan V N

Hello Mugunthan,
Am 05.09.2013 10:27, schrieb Mugunthan V N:
On Thursday 05 September 2013 11:35 AM, Heiko Schocher wrote:
if phy_connect() did not find a phy, phydev is NULL and following code in cpsw_phy_init() crashes. Fix this.
Signed-off-by: Heiko Schocherhs@denx.de Cc: Joe Hershbergerjoe.hershberger@gmail.com Cc: Mugunthan V Nmugunthanvnm@ti.com Cc: Tom Rinitrini@ti.com
[...]
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..186665b 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -561,6 +561,9 @@ static inline void setbit_and_wait_for_clear32(void *addr) static void cpsw_set_slave_mac(struct cpsw_slave *slave, struct cpsw_priv *priv) {
- if (!priv)
return;
- __raw_writel(mac_hi(priv->dev->enetaddr),&slave->regs->sa_hi); __raw_writel(mac_lo(priv->dev->enetaddr),&slave->regs->sa_lo); }
@@ -568,9 +571,17 @@ static void cpsw_set_slave_mac(struct cpsw_slave *slave, static void cpsw_slave_update_link(struct cpsw_slave *slave, struct cpsw_priv *priv, int *link) {
- struct phy_device *phy = priv->phydev;
struct phy_device *phy; u32 mac_control = 0;
if (!priv)
return;
phy = priv->phydev;
if (!phy)
return;
phy_startup(phy); *link = phy->link;
@@ -604,8 +615,12 @@ static int cpsw_update_link(struct cpsw_priv *priv) int link = 0; struct cpsw_slave *slave;
- if (!priv)
return -1;
- for_each_slave(slave, priv) cpsw_slave_update_link(slave, priv,&link);
- priv->mdio_link = readl(&mdio_regs->link); return link; }
@@ -614,6 +629,9 @@ static int cpsw_check_link(struct cpsw_priv *priv) { u32 link = 0;
All the above *priv* check can be remove as priv is already validated in this function and all the above functions are called only from this function.
No, cpsw_update_link() is called for example directly from cpsw_init() cpsw_set_slave_mac() from cpsw_slave_init()
- if (!priv)
return -1;
- link = __raw_readl(&mdio_regs->link)& priv->phy_mask; if ((link)&& (link == priv->mdio_link)) return 1;
@@ -623,6 +641,9 @@ static int cpsw_check_link(struct cpsw_priv *priv)
static inline u32 cpsw_get_slave_port(struct cpsw_priv *priv, u32 slave_num) {
- if (!priv)
return -1;
- if (priv->host_port == 0) return slave_num + 1; else
@@ -947,6 +968,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
- if (!phydev)
return -1;
- phydev->supported&= supported; phydev->advertising = phydev->supported;
I don't think you need to validate *priv* as it is already validated in driver register (cpsw_register()). In any case priv is NULL, then there is some corruption in dev structure as without priv, dev should not exist and stack is not supposed to call any cpsw apis.
Yes, you are right. Checked it on the dxr2 board. We do not need the "priv check" ... Thanks!
bye, Heiko

if phy_connect() did not find a phy, phydev is NULL and following code in cpsw_phy_init() crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
--- - changes for v2: - change commit message as it is a NULL pointer deference error not a unaligned access problem, as Tom Rini mentioned - fix more places in the driver with NULL pointer problem - changes for v3: - add comment from Mugunthan V N: - priv NULL pointer checks not needed
Found on the dxr2 board with no phy connected to the board, U-Boot crashes with:
U-Boot 2013.07-12701-gea98378-dirty (Sep 04 2013 - 06:58:16)
I2C: ready DRAM: 128 MiB Enable d-cache FactorySet is not right in eeprom. NAND: 256 MiB MMC: OMAP SD/MMC: 0, OMAP SD/MMC: 1 8-bit BCH HW ECC selected Net: Could not get PHY for cpsw: addr 0 data abort
MAYBE you should read doc/README.arm-unaligned-accesses
pc : [<87f80574>] lr : [<87f80fcc>] sp : 86f5aee0 ip : 00000034 fp : 80100020 r10: 00000014 r9 : 07e5d000 r8 : 86f5af30 r7 : 86f5f750 r6 : 86f5f804 r5 : 86f5f708 r4 : 86f5f750 r3 : 00000000 r2 : 00000000 r1 : 87fa4d08 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 Resetting CPU ... --- drivers/net/cpsw.c | 10 +++++++++- 1 Datei geändert, 9 Zeilen hinzugefügt(+), 1 Zeile entfernt(-)
diff --git a/drivers/net/cpsw.c b/drivers/net/cpsw.c index 9bab71a..39240d9 100644 --- a/drivers/net/cpsw.c +++ b/drivers/net/cpsw.c @@ -568,9 +568,14 @@ static void cpsw_set_slave_mac(struct cpsw_slave *slave, static void cpsw_slave_update_link(struct cpsw_slave *slave, struct cpsw_priv *priv, int *link) { - struct phy_device *phy = priv->phydev; + struct phy_device *phy; u32 mac_control = 0;
+ phy = priv->phydev; + + if (!phy) + return; + phy_startup(phy); *link = phy->link;
@@ -947,6 +952,9 @@ static int cpsw_phy_init(struct eth_device *dev, struct cpsw_slave *slave) dev, slave->data->phy_if);
+ if (!phydev) + return -1; + phydev->supported &= supported; phydev->advertising = phydev->supported;

On Thursday 05 September 2013 03:20 PM, Heiko Schocher wrote:
if phy_connect() did not find a phy, phydev is NULL and following code in cpsw_phy_init() crashes. Fix this.
Signed-off-by: Heiko Schocher hs@denx.de Cc: Joe Hershberger joe.hershberger@gmail.com Cc: Mugunthan V N mugunthanvnm@ti.com Cc: Tom Rini trini@ti.com
Acked-by: Mugunthan V N mugunthanvnm@ti.com
Regards Mugunthan V N
participants (4)
-
Heiko Schocher
-
Joe Hershberger
-
Mugunthan V N
-
Tom Rini