Clocks are properly reference counted and do not need to be inside the
lock range.
Right now this triggers a false-positive lockdep warning on MT8192 based
Chromebooks, through a combination of mtk-scp that has a cros-ec-rpmsg
sub-device, the (actual) cros-ec I2C adapter registration, I2C client
(not on cros-ec) probe doing i2c transfers and enabling clocks.
This is a false positive because the cros-ec-rpmsg under mtk-scp does
not have an I2C adapter, and also each I2C adapter and cros-ec instance
have their own mutex.
Move the clk operations outside of the send_lock range.
Fixes: ("63c13d61eafe remoteproc/mediatek: add SCP support for mt8183")
Signed-off-by: Chen-Yu Tsai <[email protected]>
---
drivers/remoteproc/mtk_scp_ipi.c | 11 +++++------
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/remoteproc/mtk_scp_ipi.c b/drivers/remoteproc/mtk_scp_ipi.c
index 00f041ebcde6..4c0d121c2f54 100644
--- a/drivers/remoteproc/mtk_scp_ipi.c
+++ b/drivers/remoteproc/mtk_scp_ipi.c
@@ -164,21 +164,21 @@ int scp_ipi_send(struct mtk_scp *scp, u32 id, void *buf, unsigned int len,
WARN_ON(len > sizeof(send_obj->share_buf)) || WARN_ON(!buf))
return -EINVAL;
- mutex_lock(&scp->send_lock);
-
ret = clk_prepare_enable(scp->clk);
if (ret) {
dev_err(scp->dev, "failed to enable clock\n");
- goto unlock_mutex;
+ return ret;
}
+ mutex_lock(&scp->send_lock);
+
/* Wait until SCP receives the last command */
timeout = jiffies + msecs_to_jiffies(2000);
do {
if (time_after(jiffies, timeout)) {
dev_err(scp->dev, "%s: IPI timeout!\n", __func__);
ret = -ETIMEDOUT;
- goto clock_disable;
+ goto unlock_mutex;
}
} while (readl(scp->reg_base + scp->data->host_to_scp_reg));
@@ -205,10 +205,9 @@ int scp_ipi_send(struct mtk_scp *scp, u32 id, void *buf, unsigned int len,
ret = 0;
}
-clock_disable:
- clk_disable_unprepare(scp->clk);
unlock_mutex:
mutex_unlock(&scp->send_lock);
+ clk_disable_unprepare(scp->clk);
return ret;
}
--
2.39.0.314.g84b9a713c41-goog
Il 04/01/23 09:31, Chen-Yu Tsai ha scritto:
> Clocks are properly reference counted and do not need to be inside the
> lock range.
>
> Right now this triggers a false-positive lockdep warning on MT8192 based
> Chromebooks, through a combination of mtk-scp that has a cros-ec-rpmsg
> sub-device, the (actual) cros-ec I2C adapter registration, I2C client
> (not on cros-ec) probe doing i2c transfers and enabling clocks.
>
> This is a false positive because the cros-ec-rpmsg under mtk-scp does
> not have an I2C adapter, and also each I2C adapter and cros-ec instance
> have their own mutex.
>
> Move the clk operations outside of the send_lock range.
>
> Fixes: ("63c13d61eafe remoteproc/mediatek: add SCP support for mt8183")
> Signed-off-by: Chen-Yu Tsai <[email protected]>
Nice catch.
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
On Wed, Jan 04, 2023 at 04:31:10PM +0800, Chen-Yu Tsai wrote:
> Clocks are properly reference counted and do not need to be inside the
> lock range.
>
> Right now this triggers a false-positive lockdep warning on MT8192 based
> Chromebooks, through a combination of mtk-scp that has a cros-ec-rpmsg
> sub-device, the (actual) cros-ec I2C adapter registration, I2C client
> (not on cros-ec) probe doing i2c transfers and enabling clocks.
>
> This is a false positive because the cros-ec-rpmsg under mtk-scp does
> not have an I2C adapter, and also each I2C adapter and cros-ec instance
> have their own mutex.
>
> Move the clk operations outside of the send_lock range.
Thanks for providing such a clear explanation - it makes my job a lot easier.
>
> Fixes: ("63c13d61eafe remoteproc/mediatek: add SCP support for mt8183")
This is the wrong format for a "Fixes:" tag. Please see
Documentation/process/submitting-patches.rst for details.
> Signed-off-by: Chen-Yu Tsai <[email protected]>
I have fixed the above and applied this patch.
Thanks,
Mathieu
> ---
> drivers/remoteproc/mtk_scp_ipi.c | 11 +++++------
> 1 file changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/remoteproc/mtk_scp_ipi.c b/drivers/remoteproc/mtk_scp_ipi.c
> index 00f041ebcde6..4c0d121c2f54 100644
> --- a/drivers/remoteproc/mtk_scp_ipi.c
> +++ b/drivers/remoteproc/mtk_scp_ipi.c
> @@ -164,21 +164,21 @@ int scp_ipi_send(struct mtk_scp *scp, u32 id, void *buf, unsigned int len,
> WARN_ON(len > sizeof(send_obj->share_buf)) || WARN_ON(!buf))
> return -EINVAL;
>
> - mutex_lock(&scp->send_lock);
> -
> ret = clk_prepare_enable(scp->clk);
> if (ret) {
> dev_err(scp->dev, "failed to enable clock\n");
> - goto unlock_mutex;
> + return ret;
> }
>
> + mutex_lock(&scp->send_lock);
> +
> /* Wait until SCP receives the last command */
> timeout = jiffies + msecs_to_jiffies(2000);
> do {
> if (time_after(jiffies, timeout)) {
> dev_err(scp->dev, "%s: IPI timeout!\n", __func__);
> ret = -ETIMEDOUT;
> - goto clock_disable;
> + goto unlock_mutex;
> }
> } while (readl(scp->reg_base + scp->data->host_to_scp_reg));
>
> @@ -205,10 +205,9 @@ int scp_ipi_send(struct mtk_scp *scp, u32 id, void *buf, unsigned int len,
> ret = 0;
> }
>
> -clock_disable:
> - clk_disable_unprepare(scp->clk);
> unlock_mutex:
> mutex_unlock(&scp->send_lock);
> + clk_disable_unprepare(scp->clk);
>
> return ret;
> }
> --
> 2.39.0.314.g84b9a713c41-goog
>
On Thu, Jan 5, 2023 at 6:22 AM Mathieu Poirier
<[email protected]> wrote:
>
> On Wed, Jan 04, 2023 at 04:31:10PM +0800, Chen-Yu Tsai wrote:
> > Clocks are properly reference counted and do not need to be inside the
> > lock range.
> >
> > Right now this triggers a false-positive lockdep warning on MT8192 based
> > Chromebooks, through a combination of mtk-scp that has a cros-ec-rpmsg
> > sub-device, the (actual) cros-ec I2C adapter registration, I2C client
> > (not on cros-ec) probe doing i2c transfers and enabling clocks.
> >
> > This is a false positive because the cros-ec-rpmsg under mtk-scp does
> > not have an I2C adapter, and also each I2C adapter and cros-ec instance
> > have their own mutex.
> >
> > Move the clk operations outside of the send_lock range.
>
> Thanks for providing such a clear explanation - it makes my job a lot easier.
>
> >
> > Fixes: ("63c13d61eafe remoteproc/mediatek: add SCP support for mt8183")
>
> This is the wrong format for a "Fixes:" tag. Please see
> Documentation/process/submitting-patches.rst for details.
>
> > Signed-off-by: Chen-Yu Tsai <[email protected]>
>
> I have fixed the above and applied this patch.
Thanks, and sorry about the brain fart on my end.
ChenYu
On Wed, Jan 4, 2023 at 4:31 PM Chen-Yu Tsai <[email protected]> wrote:
>
> Clocks are properly reference counted and do not need to be inside the
> lock range.
>
> Right now this triggers a false-positive lockdep warning on MT8192 based
> Chromebooks, through a combination of mtk-scp that has a cros-ec-rpmsg
> sub-device, the (actual) cros-ec I2C adapter registration, I2C client
> (not on cros-ec) probe doing i2c transfers and enabling clocks.
>
> This is a false positive because the cros-ec-rpmsg under mtk-scp does
> not have an I2C adapter, and also each I2C adapter and cros-ec instance
> have their own mutex.
>
> Move the clk operations outside of the send_lock range.
Well this managed to untangle the lockdep warning a bit, but it still
appears. I'll keep looking into it.