[PATCH v2] driver core: Make cycle dev_info() message dev_dbg()

[PATCH v2] driver core: Make cycle dev_info() message dev_dbg()

[Stephen Boyd]

This seems to mostly print debug information about device link stuff at
boot. It doesn't seem very useful outside of debugging so move it to
dev_dbg().

Cc: Saravana Kannan <saravanak@google.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
---

Changes from v1 (https://lore.kernel.org/r/20210603064849.1376107-1-swboyd@chromium.org)
 * Dropped second hunk and rebased to -rc4

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

diff --git a/drivers/base/core.c b/drivers/base/core.c
index 54ba506e5a89..378bb894869a 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -1736,7 +1736,7 @@ static int fw_devlink_create_devlink(struct device *con,
 		 */
 		if (!device_link_add(con, sup_dev, flags) &&
 		    !(flags & DL_FLAG_SYNC_STATE_ONLY)) {
-			dev_info(con, "Fixing up cyclic dependency with %s\n",
+			dev_dbg(con, "Fixing up cyclic dependency with %s\n",
 				 dev_name(sup_dev));
 			device_links_write_lock();
 			fw_devlink_relax_cycle(con, sup_dev);

base-commit: 8124c8a6b35386f73523d27eacb71b5364a68c4c
-- 
https://chromeos.dev

Re: [PATCH v2] driver core: Make cycle dev_info() message dev_dbg()

[Saravana Kannan]

On Thu, Jun 3, 2021 at 1:21 PM Stephen Boyd <swboyd@chromium.org> wrote:
>
> This seems to mostly print debug information about device link stuff at
> boot. It doesn't seem very useful outside of debugging so move it to
> dev_dbg().

As before, Nak at least for now.

This code path really shouldn't be exercised on most systems. I might
be open to this once fw_delink bug reports stop coming in for at least
1 full kernel minor version.

-Saravana


>
> Cc: Saravana Kannan <saravanak@google.com>
> Signed-off-by: Stephen Boyd <swboyd@chromium.org>
> ---
>
> Changes from v1 (https://lore.kernel.org/r/20210603064849.1376107-1-swboyd@chromium.org)
>  * Dropped second hunk and rebased to -rc4
>
>  drivers/base/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/base/core.c b/drivers/base/core.c
> index 54ba506e5a89..378bb894869a 100644
> --- a/drivers/base/core.c
> +++ b/drivers/base/core.c
> @@ -1736,7 +1736,7 @@ static int fw_devlink_create_devlink(struct device *con,
>                  */
>                 if (!device_link_add(con, sup_dev, flags) &&
>                     !(flags & DL_FLAG_SYNC_STATE_ONLY)) {
> -                       dev_info(con, "Fixing up cyclic dependency with %s\n",
> +                       dev_dbg(con, "Fixing up cyclic dependency with %s\n",
>                                  dev_name(sup_dev));
>                         device_links_write_lock();
>                         fw_devlink_relax_cycle(con, sup_dev);
>
> base-commit: 8124c8a6b35386f73523d27eacb71b5364a68c4c
> --
> https://chromeos.dev
>

Re: [PATCH v2] driver core: Make cycle dev_info() message dev_dbg()

[Greg Kroah-Hartman]

On Thu, Jun 03, 2021 at 01:21:16PM -0700, Stephen Boyd wrote:
> This seems to mostly print debug information about device link stuff at
> boot. It doesn't seem very useful outside of debugging so move it to
> dev_dbg().

What messages at boot time are you seeing with this that should not be
there?  Shouldn't we fix the root cause here and not paper over it by
just lowering the logging level?

What drivers are having problems to trigger this?

thanks,

greg k-h

Re: [PATCH v2] driver core: Make cycle dev_info() message dev_dbg()

[Stephen Boyd]

Quoting Greg Kroah-Hartman (2021-06-04 06:00:47)
> On Thu, Jun 03, 2021 at 01:21:16PM -0700, Stephen Boyd wrote:
> > This seems to mostly print debug information about device link stuff at
> > boot. It doesn't seem very useful outside of debugging so move it to
> > dev_dbg().
>
> What messages at boot time are you seeing with this that should not be
> there?  Shouldn't we fix the root cause here and not paper over it by
> just lowering the logging level?

Is there any problem? If they're a problem then shouldn't the printk be
at least warning level, pr_warn() if not pr_err()? I thought that
sometimes devices have cyclic links so we want to back off at that time
and stop trying.

>
> What drivers are having problems to trigger this?
>

It looks to be mostly the coresight devices that print this info message
for me. The other one is the display panel. The coresight devices are in
arch/arm64/boot/dts/qcom/sc7180.dtsi

[    0.545197] amba 6041000.funnel: Fixing up cyclic dependency with 6002000.stm
[    0.547173] amba 6045000.funnel: Fixing up cyclic dependency with
6042000.funnel
[    0.548032] amba 6045000.funnel: Fixing up cyclic dependency with
6041000.funnel
[    0.549899] amba 6048000.etr: Fixing up cyclic dependency with
6046000.replicator
[    0.551371] amba 6b04000.funnel: Fixing up cyclic dependency with
6045000.funnel
[    0.552781] amba 6b05000.etf: Fixing up cyclic dependency with 6b04000.funnel
[    0.554258] amba 6b06000.replicator: Fixing up cyclic dependency
with 6b05000.etf
[    0.555136] amba 6b06000.replicator: Fixing up cyclic dependency
with 6046000.replicator
[    0.560889] amba 7800000.funnel: Fixing up cyclic dependency with 7740000.etm
[    0.561726] amba 7800000.funnel: Fixing up cyclic dependency with 7640000.etm
[    0.562553] amba 7800000.funnel: Fixing up cyclic dependency with 7540000.etm
[    0.563388] amba 7800000.funnel: Fixing up cyclic dependency with 7440000.etm
[    0.564205] amba 7800000.funnel: Fixing up cyclic dependency with 7340000.etm
[    0.565034] amba 7800000.funnel: Fixing up cyclic dependency with 7240000.etm
[    0.565867] amba 7800000.funnel: Fixing up cyclic dependency with 7140000.etm
[    0.566693] amba 7800000.funnel: Fixing up cyclic dependency with 7040000.etm
[    0.568196] amba 7810000.funnel: Fixing up cyclic dependency with
7800000.funnel
[    0.569047] amba 7810000.funnel: Fixing up cyclic dependency with
6042000.funnel
[    0.623637] i2c 2-002d: Fixing up cyclic dependency with panel
[    1.108625] platform ae94000.dsi: Fixing up cyclic dependency with 2-002d
[    1.115646] platform ae94000.dsi: Fixing up cyclic dependency with
ae01000.mdp

I suppose with the graph binding making links we get these cyclic
dependency messages now.