Skip to contentSkip to conversation

fsl/usb: fix phy_type checking

PATCH New Nikhil Badola 2 participants 4 messages Feb 17, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fsl/usb: fix phy_type checking

Strcmp should not be used to check the argument of phy_type which maybe parsed by hwconfig_subarg. Hwconfig_subarg returns part of hwconfig starting from the argument (if it has the argument) till the end of the string. Since phy_type could be either 'utmi' or 'ulpi', strncmp should be used along with length limited to 4

---
 drivers/usb/host/ehci-fsl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/host/ehci-fsl.c +1 -1
index 45e5d6a..1ca7cf5 100644
@@ -86,7 +86,7 @@ int ehci_hcd_init(int index, enum usb_init_type init,
 #endif
 	}
 
-	if (!strcmp(phy_type, "utmi")) {
+	if (!strncmp(phy_type, "utmi", 4)) {
 #if defined(CONFIG_SYS_FSL_USB_INTERNAL_UTMI_PHY)
 		setbits_be32(&ehci->control, PHY_CLK_SEL_UTMI);
 		setbits_be32(&ehci->control, UTMI_PHY_EN);
Signature · 1 line
1.7.11.7
#1 lore raw

PATCH 1/1 fsl/usb: fix phy_type checking

Hi Nikhil,

Quoted 4 lines · Nikhil Badola

Strcmp should not be used to check the argument of phy_type which maybe parsed by hwconfig_subarg. Hwconfig_subarg returns part of hwconfig starting from the argument (if it has the argument) till the end of the string. Since phy_type could be either 'utmi' or 'ulpi', strncmp should be used along with length limited to 4

Not sure I understand what exact problem you are considering here. If you know that phy_type is either "utmi" or "ulpi", then there is no benefit in switching from strcmp() to strncmp() since there is no risk that strcmp() overruns any of its arguments.

Quoted 11 lines
---
 drivers/usb/host/ehci-fsl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/host/ehci-fsl.c +0 -0
index 45e5d6a..1ca7cf5 100644

drivers/usb/host/ehci-fsl.c int ehci_hcd_init(int index, enum usb_init_type init, · line 92

Quoted 8 lines
drivers/usb/host/ehci-fsl.c +1 -1
@@ -86,7 +86,7 @@ int ehci_hcd_init(int index, enum usb_init_type init,
 #endif
 	}
 
-	if (!strcmp(phy_type, "utmi")) {
+	if (!strncmp(phy_type, "utmi", 4)) {
 #if defined(CONFIG_SYS_FSL_USB_INTERNAL_UTMI_PHY)
 		setbits_be32(&ehci->control, PHY_CLK_SEL_UTMI);
 		setbits_be32(&ehci->control, UTMI_PHY_EN);

Amicalement,

Signature · 1 line
Albert.
#2 lore raw

PATCH 1/1 fsl/usb: fix phy_type checking

Hi Albert,

-----Original Message-----
From: Albert ARIBAUD [mailto:albert.u.boot at aribaud.net]
Sent: Monday, February 17, 2014 3:05 PM
To: Badola Nikhil-B46172
Cc: u-boot at lists.denx.de; Xie Shaohui-B21989
Subject: Re: [U-Boot] [PATCH] fsl/usb: fix phy_type checking

Hi Nikhil,

Quoted 5 lines · Nikhil Badola

Strcmp should not be used to check the argument of phy_type which maybe parsed by hwconfig_subarg. Hwconfig_subarg returns part of hwconfig starting from the argument (if it has the argument) till the end of the string. Since phy_type could be either 'utmi' or 'ulpi', strncmp should be used along with length limited to 4

Not sure I understand what exact problem you are considering here. If you know that phy_type is either "utmi" or "ulpi", then there is no benefit in switching from strcmp() to strncmp() since there is no risk that strcmp() overruns any of its arguments.

[Nikhil Badola] :
Strcmp() won't overrun any of its arguments but it does check NULL character at the end of both strings. Let me explain this with an example:
Consider the hwconfig string to be defined as hwconfig=usb1:dr_mode=host,phy_type=utmi;fsl_ddr:bank_intlv=auto

When hwconfig string is parsed for "phy_type" sub-argument, it returns string "utmi;fsl_ddr:bank_intlv=auto" which gets stored in phy_type pointer. When strcmp() is used, then '\0' character of "utmi" string is compared with ';' of phy_type hence returning non-zero value.

Quoted 12 lines
---
 drivers/usb/host/ehci-fsl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/host/ehci-fsl.c +0 -0
index 45e5d6a..1ca7cf5 100644
@@ -86,7 +86,7 @@ int ehci_hcd_init(int index, enum usb_init_type 

init, #endif

drivers/usb/host/ehci-fsl.c int ehci_hcd_init(int index, enum usb_init_type init, · line 92

Quoted 7 lines
drivers/usb/host/ehci-fsl.c +1 -1
 	}
 
-	if (!strcmp(phy_type, "utmi")) {
+	if (!strncmp(phy_type, "utmi", 4)) {
 #if defined(CONFIG_SYS_FSL_USB_INTERNAL_UTMI_PHY)
 		setbits_be32(&ehci->control, PHY_CLK_SEL_UTMI);
 		setbits_be32(&ehci->control, UTMI_PHY_EN);

Amicalement,
--
Albert.

#3 lore raw

PATCH 1/1 fsl/usb: fix phy_type checking

Hi nikhil.badola at freescale.com,

Quoted 13 lines · nikhil.badola at freescale.com

Hi Albert,

-----Original Message-----
From: Albert ARIBAUD [mailto:albert.u.boot at aribaud.net]
Sent: Monday, February 17, 2014 3:05 PM
To: Badola Nikhil-B46172
Cc: u-boot at lists.denx.de; Xie Shaohui-B21989
Subject: Re: [U-Boot] [PATCH] fsl/usb: fix phy_type checking

Hi Nikhil,

On Mon, 17 Feb 2014 14:39:47 +0530, Nikhil Badola <[email protected]> wrote:

Quoted 7 lines
Quoted 5 lines

Strcmp should not be used to check the argument of phy_type which maybe parsed by hwconfig_subarg. Hwconfig_subarg returns part of hwconfig starting from the argument (if it has the argument) till the end of the string. Since phy_type could be either 'utmi' or 'ulpi', strncmp should be used along with length limited to 4

Not sure I understand what exact problem you are considering here. If you know that phy_type is either "utmi" or "ulpi", then there is no benefit in switching from strcmp() to strncmp() since there is no risk that strcmp() overruns any of its arguments.

(seems like your mailer does not quote properly. Can you fix this?)

Quoted 7 lines

[Nikhil Badola] :
Strcmp() won't overrun any of its arguments but it does check NULL character at the end of both strings. Let me explain this with an example:
Consider the hwconfig string to be defined as hwconfig=usb1:dr_mode=host,phy_type=utmi;fsl_ddr:bank_intlv=auto

When hwconfig string is parsed for "phy_type" sub-argument, it returns string "utmi;fsl_ddr:bank_intlv=auto" which gets stored in phy_type pointer. When strcmp() is used, then '\0' character of "utmi" string is compared with ';' of phy_type hence returning non-zero value.

That is much clearer. Can you post a V2 with an amended commit message, something like "limit phy_type comparison to the 4 first characters, so that a comparison of "utmi;fsl_ddr:bank_intlv=auto" with "utmi" will succeed".

Quoted 3 lines

Amicalement,
--
Albert.

Amicalement,

Signature · 1 line
Albert.