diff mbox

reset: fix shared reset triggered_count decrement on error

Message ID 1487182551-31517-1-git-send-email-jbrunet@baylibre.com (mailing list archive)
State Accepted
Headers show

Commit Message

Jerome Brunet Feb. 15, 2017, 6:15 p.m. UTC
For a shared reset, when the reset is successful, the triggered_count is
incremented when trying to call the reset callback, so that another device
sharing the same reset line won't trigger it again. If the reset has not
been triggered successfully, the trigger_count should be decremented.

The code does the opposite, and decrements the trigger_count on success.
As a consequence, another device sharing the reset will be able to trigger
it again.

Fixed be removing negation in from of the error code of the reset function.

Fixes: 7da33a37b48f ("reset: allow using reset_control_reset with shared reset")

Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
 Hi Philipp,

 I found this issue while testing your patch [0]
 It fixes a regression we have been having with usb. On meson-gxbb
 platforms, usb0 and usb1 share the same reset line. Martin had
 reports that usb0 recently got broken. In fact usb1 was able to
 trigger the reset again because the issue mentioned above.

[0]: http://lkml.kernel.org/r/20170130114116.22089-10-p.zabel@pengutronix.de

 drivers/reset/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Martin Blumenstingl Feb. 15, 2017, 8:09 p.m. UTC | #1
Hi Jerome,

ouch, thanks for spotting and fixing this!

On Wed, Feb 15, 2017 at 7:15 PM, Jerome Brunet <jbrunet@baylibre.com> wrote:
> For a shared reset, when the reset is successful, the triggered_count is
> incremented when trying to call the reset callback, so that another device
> sharing the same reset line won't trigger it again. If the reset has not
> been triggered successfully, the trigger_count should be decremented.
>
> The code does the opposite, and decrements the trigger_count on success.
> As a consequence, another device sharing the reset will be able to trigger
> it again.
>
> Fixed be removing negation in from of the error code of the reset function.
>
> Fixes: 7da33a37b48f ("reset: allow using reset_control_reset with shared reset")
>
> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Acked-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>

should this still go into 4.10?
I guess the meson8b-usb2 PHY driver is the only user of the shared
reset pulse, but if there are other users out there then these may
face strange issues as well!

> ---
>  Hi Philipp,
>
>  I found this issue while testing your patch [0]
>  It fixes a regression we have been having with usb. On meson-gxbb
>  platforms, usb0 and usb1 share the same reset line. Martin had
>  reports that usb0 recently got broken. In fact usb1 was able to
>  trigger the reset again because the issue mentioned above.
>
> [0]: http://lkml.kernel.org/r/20170130114116.22089-10-p.zabel@pengutronix.de
>
>  drivers/reset/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/reset/core.c b/drivers/reset/core.c
> index 71ccf281dce3..f1e5e65388bb 100644
> --- a/drivers/reset/core.c
> +++ b/drivers/reset/core.c
> @@ -169,7 +169,7 @@ int reset_control_reset(struct reset_control *rstc)
>         }
>
>         ret = rstc->rcdev->ops->reset(rstc->rcdev, rstc->id);
> -       if (rstc->shared && !ret)
> +       if (rstc->shared && ret)
>                 atomic_dec(&rstc->triggered_count);
>
>         return ret;
> --
> 2.7.4
>
Philipp Zabel Feb. 16, 2017, 9:52 a.m. UTC | #2
On Wed, 2017-02-15 at 19:15 +0100, Jerome Brunet wrote:
> For a shared reset, when the reset is successful, the triggered_count is
> incremented when trying to call the reset callback, so that another device
> sharing the same reset line won't trigger it again. If the reset has not
> been triggered successfully, the trigger_count should be decremented.
> 
> The code does the opposite, and decrements the trigger_count on success.
> As a consequence, another device sharing the reset will be able to trigger
> it again.
> 
> Fixed be removing negation in from of the error code of the reset function.
> 
> Fixes: 7da33a37b48f ("reset: allow using reset_control_reset with shared reset")
> 
> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
> ---
>  Hi Philipp,
> 
>  I found this issue while testing your patch [0]
>  It fixes a regression we have been having with usb. On meson-gxbb
>  platforms, usb0 and usb1 share the same reset line. Martin had
>  reports that usb0 recently got broken. In fact usb1 was able to
>  trigger the reset again because the issue mentioned above.

Thanks, applied with Martin's Acked-by. I'll send a pull request for
this later today.

regards
Philipp
diff mbox

Patch

diff --git a/drivers/reset/core.c b/drivers/reset/core.c
index 71ccf281dce3..f1e5e65388bb 100644
--- a/drivers/reset/core.c
+++ b/drivers/reset/core.c
@@ -169,7 +169,7 @@  int reset_control_reset(struct reset_control *rstc)
 	}
 
 	ret = rstc->rcdev->ops->reset(rstc->rcdev, rstc->id);
-	if (rstc->shared && !ret)
+	if (rstc->shared && ret)
 		atomic_dec(&rstc->triggered_count);
 
 	return ret;