Re: [PATCH] usb: dwc3: gadget: don't rely on jiffies while holding spinlock

[Felipe Balbi]

Hi,

Nicolas Saenz Julienne  writes:
> __dwc3_gadget_wakeup() is called while holding a spinlock, then depends on
> jiffies in order to timeout while polling the USB core for a link state
> update. In the case the wakeup failed, the timeout will never happen and
> will also cause the cpu to stall until rcu_preempt kicks in.
>
> This switches to a "decrement variable and wait" timeout scheme.
>
> Signed-off-by: Nicolas Saenz Julienne 
> ---
>  drivers/usb/dwc3/gadget.c | 8 +---
>  1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
> index 8f8c215..d0c711f 100644
> --- a/drivers/usb/dwc3/gadget.c
> +++ b/drivers/usb/dwc3/gadget.c
> @@ -1433,7 +1433,7 @@ static int dwc3_gadget_get_frame(struct usb_gadget *g)
>  
>  static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
>  {
> - unsigned long   timeout;
> + int retries;
>  
>   int ret;
>   u32 reg;
> @@ -1484,14 +1484,16 @@ static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
>   }
>  
>   /* poll until Link State changes to ON */
> - timeout = jiffies + msecs_to_jiffies(100);
> + retries = 2;
>  
> - while (!time_after(jiffies, timeout)) {
> + while (retries--) {
>   reg = dwc3_readl(dwc->regs, DWC3_DSTS);
>  
>   /* in HS, means ON */
>   if (DWC3_DSTS_USBLNKST(reg) == DWC3_LINK_STATE_U0)
>   break;
> +
> + udelay(5);

let's just drop this udelay() here. It's very likely this will switch
the link state much quicker than 100ms anyway. Other than that, nice
catch :-)

-- 
balbi


signature.asc
Description: PGP signature

Re: [PATCH] usb: dwc3: gadget: don't rely on jiffies while holding spinlock

[Felipe Balbi]

Hi,

Nicolas Saenz Julienne  writes:
> __dwc3_gadget_wakeup() is called while holding a spinlock, then depends on
> jiffies in order to timeout while polling the USB core for a link state
> update. In the case the wakeup failed, the timeout will never happen and
> will also cause the cpu to stall until rcu_preempt kicks in.
>
> This switches to a "decrement variable and wait" timeout scheme.
>
> Signed-off-by: Nicolas Saenz Julienne 
> ---
>  drivers/usb/dwc3/gadget.c | 8 +---
>  1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
> index 8f8c215..d0c711f 100644
> --- a/drivers/usb/dwc3/gadget.c
> +++ b/drivers/usb/dwc3/gadget.c
> @@ -1433,7 +1433,7 @@ static int dwc3_gadget_get_frame(struct usb_gadget *g)
>  
>  static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
>  {
> - unsigned long   timeout;
> + int retries;
>  
>   int ret;
>   u32 reg;
> @@ -1484,14 +1484,16 @@ static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
>   }
>  
>   /* poll until Link State changes to ON */
> - timeout = jiffies + msecs_to_jiffies(100);
> + retries = 2;
>  
> - while (!time_after(jiffies, timeout)) {
> + while (retries--) {
>   reg = dwc3_readl(dwc->regs, DWC3_DSTS);
>  
>   /* in HS, means ON */
>   if (DWC3_DSTS_USBLNKST(reg) == DWC3_LINK_STATE_U0)
>   break;
> +
> + udelay(5);

let's just drop this udelay() here. It's very likely this will switch
the link state much quicker than 100ms anyway. Other than that, nice
catch :-)

-- 
balbi


signature.asc
Description: PGP signature

[PATCH] usb: dwc3: gadget: don't rely on jiffies while holding spinlock

[Nicolas Saenz Julienne]

__dwc3_gadget_wakeup() is called while holding a spinlock, then depends on
jiffies in order to timeout while polling the USB core for a link state
update. In the case the wakeup failed, the timeout will never happen and
will also cause the cpu to stall until rcu_preempt kicks in.

This switches to a "decrement variable and wait" timeout scheme.

Signed-off-by: Nicolas Saenz Julienne 
---
 drivers/usb/dwc3/gadget.c | 8 +---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 8f8c215..d0c711f 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1433,7 +1433,7 @@ static int dwc3_gadget_get_frame(struct usb_gadget *g)
 
 static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
 {
-   unsigned long   timeout;
+   int retries;
 
int ret;
u32 reg;
@@ -1484,14 +1484,16 @@ static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
}
 
/* poll until Link State changes to ON */
-   timeout = jiffies + msecs_to_jiffies(100);
+   retries = 2;
 
-   while (!time_after(jiffies, timeout)) {
+   while (retries--) {
reg = dwc3_readl(dwc->regs, DWC3_DSTS);
 
/* in HS, means ON */
if (DWC3_DSTS_USBLNKST(reg) == DWC3_LINK_STATE_U0)
break;
+
+   udelay(5);
}
 
if (DWC3_DSTS_USBLNKST(reg) != DWC3_LINK_STATE_U0) {
-- 
2.7.4

[PATCH] usb: dwc3: gadget: don't rely on jiffies while holding spinlock

[Nicolas Saenz Julienne]

__dwc3_gadget_wakeup() is called while holding a spinlock, then depends on
jiffies in order to timeout while polling the USB core for a link state
update. In the case the wakeup failed, the timeout will never happen and
will also cause the cpu to stall until rcu_preempt kicks in.

This switches to a "decrement variable and wait" timeout scheme.

Signed-off-by: Nicolas Saenz Julienne 
---
 drivers/usb/dwc3/gadget.c | 8 +---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 8f8c215..d0c711f 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1433,7 +1433,7 @@ static int dwc3_gadget_get_frame(struct usb_gadget *g)
 
 static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
 {
-   unsigned long   timeout;
+   int retries;
 
int ret;
u32 reg;
@@ -1484,14 +1484,16 @@ static int __dwc3_gadget_wakeup(struct dwc3 *dwc)
}
 
/* poll until Link State changes to ON */
-   timeout = jiffies + msecs_to_jiffies(100);
+   retries = 2;
 
-   while (!time_after(jiffies, timeout)) {
+   while (retries--) {
reg = dwc3_readl(dwc->regs, DWC3_DSTS);
 
/* in HS, means ON */
if (DWC3_DSTS_USBLNKST(reg) == DWC3_LINK_STATE_U0)
break;
+
+   udelay(5);
}
 
if (DWC3_DSTS_USBLNKST(reg) != DWC3_LINK_STATE_U0) {
-- 
2.7.4