Identify and print the error source before rebooting the board due to an l3
application timeout error, by delaying the BUG_ON. This is helpful when
debugging, e.g. via serial.
Signed-off-by: Sicelo A. Mhlongo <[email protected]>
---
drivers/bus/omap_l3_smx.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/drivers/bus/omap_l3_smx.c b/drivers/bus/omap_l3_smx.c
index bb1606f5ce2d..0ccde12e33c0 100644
--- a/drivers/bus/omap_l3_smx.c
+++ b/drivers/bus/omap_l3_smx.c
@@ -170,11 +170,9 @@ static irqreturn_t omap3_l3_app_irq(int irq, void *_l3)
status = omap3_l3_readll(l3->rt, L3_SI_FLAG_STATUS_0);
/*
* if we have a timeout error, there's nothing we can
- * do besides rebooting the board. So let's BUG on any
- * of such errors and handle the others. timeout error
- * is severe and not expected to occur.
+ * do besides rebooting the board after identifying the
+ * error source.
*/
- BUG_ON(status & L3_STATUS_0_TIMEOUT_MASK);
} else {
status = omap3_l3_readll(l3->rt, L3_SI_FLAG_STATUS_1);
/* No timeout error for debug sources */
@@ -190,6 +188,12 @@ static irqreturn_t omap3_l3_app_irq(int irq, void *_l3)
ret |= omap3_l3_block_irq(l3, error, error_addr);
}
+ /*
+ * BUG on application timeout errors since they are severe and not
+ * expected to occur.
+ */
+ BUG_ON(!int_type && status & L3_STATUS_0_TIMEOUT_MASK);
+
/* Clear the status register */
clear = (L3_AGENT_STATUS_CLEAR_IA << int_type) |
L3_AGENT_STATUS_CLEAR_TA;
--
2.40.1
Hi,
On Mon, Jul 31, 2023 at 11:03:47PM +0200, Sicelo A. Mhlongo wrote:
> Identify and print the error source before rebooting the board due
> to an l3 application timeout error, by delaying the BUG_ON. This
> is helpful when debugging, e.g. via serial.
>
> Signed-off-by: Sicelo A. Mhlongo <[email protected]>
> ---
> drivers/bus/omap_l3_smx.c | 12 ++++++++----
> 1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/bus/omap_l3_smx.c b/drivers/bus/omap_l3_smx.c
> index bb1606f5ce2d..0ccde12e33c0 100644
> --- a/drivers/bus/omap_l3_smx.c
> +++ b/drivers/bus/omap_l3_smx.c
> @@ -170,11 +170,9 @@ static irqreturn_t omap3_l3_app_irq(int irq, void *_l3)
> status = omap3_l3_readll(l3->rt, L3_SI_FLAG_STATUS_0);
> /*
> * if we have a timeout error, there's nothing we can
> - * do besides rebooting the board. So let's BUG on any
> - * of such errors and handle the others. timeout error
> - * is severe and not expected to occur.
> + * do besides rebooting the board after identifying the
> + * error source.
> */
> - BUG_ON(status & L3_STATUS_0_TIMEOUT_MASK);
> } else {
> status = omap3_l3_readll(l3->rt, L3_SI_FLAG_STATUS_1);
> /* No timeout error for debug sources */
I think it's sensible to just remove the comments from both if/else
branches. They no longer make sense in this place. Afterwards you
should remove the curly brackets to follow kernel coding style.
> @@ -190,6 +188,12 @@ static irqreturn_t omap3_l3_app_irq(int irq, void *_l3)
> ret |= omap3_l3_block_irq(l3, error, error_addr);
> }
>
> + /*
> + * BUG on application timeout errors since they are severe and not
> + * expected to occur.
> + */
I suggest using the original text. It does mention "there's nothing
we can do besides rebooting the board".
> + BUG_ON(!int_type && status & L3_STATUS_0_TIMEOUT_MASK);
> +
> /* Clear the status register */
> clear = (L3_AGENT_STATUS_CLEAR_IA << int_type) |
> L3_AGENT_STATUS_CLEAR_TA;
-- Sebastian