Using kobject_get_path() means a dynamic memory allocation gets done, which
doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
spinlock.

To fix the problem, let's convert into using the simpler dev_name(). This
means the information about the path doesn't get presented in debugfs, but
hopefully this shouldn't be an issue.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
---
Changes in v2:
	- New patch.
---
 drivers/pmdomain/core.c | 23 +++--------------------
 1 file changed, 3 insertions(+), 20 deletions(-)

diff --git a/drivers/pmdomain/core.c b/drivers/pmdomain/core.c
index 454fccc38df1..90a65bd09d52 100644
--- a/drivers/pmdomain/core.c
+++ b/drivers/pmdomain/core.c
@@ -3182,7 +3182,6 @@ static int genpd_summary_one(struct seq_file *s,
 		[GENPD_STATE_OFF] = "off"
 	};
 	struct pm_domain_data *pm_data;
-	const char *kobj_path;
 	struct gpd_link *link;
 	char state[16];
 	int ret;
@@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
 	}
 
 	list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
-		kobj_path = kobject_get_path(&pm_data->dev->kobj,
-				genpd_is_irq_safe(genpd) ?
-				GFP_ATOMIC : GFP_KERNEL);
-		if (kobj_path == NULL)
-			continue;
-
-		seq_printf(s, "\n    %-50s  ", kobj_path);
+		seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));
 		rtpm_status_str(s, pm_data->dev);
 		perf_status_str(s, pm_data->dev);
-		kfree(kobj_path);
 	}
 
 	seq_puts(s, "\n");
@@ -3393,23 +3385,14 @@ static int devices_show(struct seq_file *s, void *data)
 {
 	struct generic_pm_domain *genpd = s->private;
 	struct pm_domain_data *pm_data;
-	const char *kobj_path;
 	int ret = 0;
 
 	ret = genpd_lock_interruptible(genpd);
 	if (ret)
 		return -ERESTARTSYS;
 
-	list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
-		kobj_path = kobject_get_path(&pm_data->dev->kobj,
-				genpd_is_irq_safe(genpd) ?
-				GFP_ATOMIC : GFP_KERNEL);
-		if (kobj_path == NULL)
-			continue;
-
-		seq_printf(s, "%s\n", kobj_path);
-		kfree(kobj_path);
-	}
+	list_for_each_entry(pm_data, &genpd->dev_list, list_node)
+		seq_printf(s, "%s\n", dev_name(pm_data->dev));
 
 	genpd_unlock(genpd);
 	return ret;
-- 
2.34.1

Hi Ulf,

On Mon, May 27, 2024 at 4:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> Using kobject_get_path() means a dynamic memory allocation gets done, which
> doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
> spinlock.
>
> To fix the problem, let's convert into using the simpler dev_name(). This
> means the information about the path doesn't get presented in debugfs, but
> hopefully this shouldn't be an issue.
>
> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> ---
> Changes in v2:
>         - New patch.

Thanks for your patch, which is now commit 9094e53ff5c86ebe ("pmdomain:
core: Use dev_name() instead of kobject_get_path() in debugfs")
in pmdomain/next.

> --- a/drivers/pmdomain/core.c
> +++ b/drivers/pmdomain/core.c
> @@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
>         }
>
>         list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
> -               kobj_path = kobject_get_path(&pm_data->dev->kobj,
> -                               genpd_is_irq_safe(genpd) ?
> -                               GFP_ATOMIC : GFP_KERNEL);
> -               if (kobj_path == NULL)
> -                       continue;
> -
> -               seq_printf(s, "\n    %-50s  ", kobj_path);
> +               seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));

While some of the old names didn't even fit in 50 characters, the new
names need much less space, so perhaps this is a good opportunity to
decrease the table width?

>                 rtpm_status_str(s, pm_data->dev);
>                 perf_status_str(s, pm_data->dev);
> -               kfree(kobj_path);
>         }
>
>         seq_puts(s, "\n");

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

On Tue, 20 Aug 2024 at 10:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>
> Hi Ulf,
>
> On Mon, May 27, 2024 at 4:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > Using kobject_get_path() means a dynamic memory allocation gets done, which
> > doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
> > spinlock.
> >
> > To fix the problem, let's convert into using the simpler dev_name(). This
> > means the information about the path doesn't get presented in debugfs, but
> > hopefully this shouldn't be an issue.
> >
> > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > ---
> > Changes in v2:
> >         - New patch.
>
> Thanks for your patch, which is now commit 9094e53ff5c86ebe ("pmdomain:
> core: Use dev_name() instead of kobject_get_path() in debugfs")
> in pmdomain/next.
>
> > --- a/drivers/pmdomain/core.c
> > +++ b/drivers/pmdomain/core.c
> > @@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
> >         }
> >
> >         list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
> > -               kobj_path = kobject_get_path(&pm_data->dev->kobj,
> > -                               genpd_is_irq_safe(genpd) ?
> > -                               GFP_ATOMIC : GFP_KERNEL);
> > -               if (kobj_path == NULL)
> > -                       continue;
> > -
> > -               seq_printf(s, "\n    %-50s  ", kobj_path);
> > +               seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));
>
> While some of the old names didn't even fit in 50 characters, the new
> names need much less space, so perhaps this is a good opportunity to
> decrease the table width?

Sure, it seems reasonable! Do you want to send a patch?

>
> >                 rtpm_status_str(s, pm_data->dev);
> >                 perf_status_str(s, pm_data->dev);
> > -               kfree(kobj_path);
> >         }
> >
> >         seq_puts(s, "\n");

Kind regards
Uffe

Hi Ulf,

On Tue, Aug 20, 2024 at 10:58 AM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> On Tue, 20 Aug 2024 at 10:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > On Mon, May 27, 2024 at 4:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > > Using kobject_get_path() means a dynamic memory allocation gets done, which
> > > doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
> > > spinlock.
> > >
> > > To fix the problem, let's convert into using the simpler dev_name(). This
> > > means the information about the path doesn't get presented in debugfs, but
> > > hopefully this shouldn't be an issue.
> > >
> > > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > ---
> > > Changes in v2:
> > >         - New patch.
> >
> > Thanks for your patch, which is now commit 9094e53ff5c86ebe ("pmdomain:
> > core: Use dev_name() instead of kobject_get_path() in debugfs")
> > in pmdomain/next.
> >
> > > --- a/drivers/pmdomain/core.c
> > > +++ b/drivers/pmdomain/core.c
> > > @@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
> > >         }
> > >
> > >         list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
> > > -               kobj_path = kobject_get_path(&pm_data->dev->kobj,
> > > -                               genpd_is_irq_safe(genpd) ?
> > > -                               GFP_ATOMIC : GFP_KERNEL);
> > > -               if (kobj_path == NULL)
> > > -                       continue;
> > > -
> > > -               seq_printf(s, "\n    %-50s  ", kobj_path);
> > > +               seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));
> >
> > While some of the old names didn't even fit in 50 characters, the new
> > names need much less space, so perhaps this is a good opportunity to
> > decrease the table width?
>
> Sure, it seems reasonable! Do you want to send a patch?

I started looking into it.  Then I noticed that on some systems
(e.g. TI am335x) the device names may have a longer format than
the typical <unit-address>.<nodename>. So I wanted to verify on
BeagleBone Black, but recent kernels crash during early boot.
Apparently that platform was broken between v6.8 and v6.9-rc1.
And during bisection, I encountered 3 different failure modes...

To be continued...

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Hi Ulf,

On Fri, Aug 30, 2024 at 5:50 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> On Tue, Aug 20, 2024 at 10:58 AM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > On Tue, 20 Aug 2024 at 10:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > > On Mon, May 27, 2024 at 4:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > > > Using kobject_get_path() means a dynamic memory allocation gets done, which
> > > > doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
> > > > spinlock.
> > > >
> > > > To fix the problem, let's convert into using the simpler dev_name(). This
> > > > means the information about the path doesn't get presented in debugfs, but
> > > > hopefully this shouldn't be an issue.
> > > >
> > > > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > > ---
> > > > Changes in v2:
> > > >         - New patch.
> > >
> > > Thanks for your patch, which is now commit 9094e53ff5c86ebe ("pmdomain:
> > > core: Use dev_name() instead of kobject_get_path() in debugfs")
> > > in pmdomain/next.
> > >
> > > > --- a/drivers/pmdomain/core.c
> > > > +++ b/drivers/pmdomain/core.c
> > > > @@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
> > > >         }
> > > >
> > > >         list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
> > > > -               kobj_path = kobject_get_path(&pm_data->dev->kobj,
> > > > -                               genpd_is_irq_safe(genpd) ?
> > > > -                               GFP_ATOMIC : GFP_KERNEL);
> > > > -               if (kobj_path == NULL)
> > > > -                       continue;
> > > > -
> > > > -               seq_printf(s, "\n    %-50s  ", kobj_path);
> > > > +               seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));
> > >
> > > While some of the old names didn't even fit in 50 characters, the new
> > > names need much less space, so perhaps this is a good opportunity to
> > > decrease the table width?
> >
> > Sure, it seems reasonable! Do you want to send a patch?
>
> I started looking into it.  Then I noticed that on some systems
> (e.g. TI am335x) the device names may have a longer format than
> the typical <unit-address>.<nodename>. So I wanted to verify on
> BeagleBone Black, but recent kernels crash during early boot.
> Apparently that platform was broken between v6.8 and v6.9-rc1.
> And during bisection, I encountered 3 different failure modes...
>
> To be continued...

The longest generic node names documented in the Devicetree
Specification are "air-pollution-sensor" and "interrupt-controller"
(both counting 20 characters), so a typical device name needs 8
(32-bit unit address) + 1 (dot) + 20 = 29 characters.
However, I assume some devices lie outside the 32-bit address space,
and thus need more space?

With the BeagleBone Black boot issue fixed:
"/devices/platform/ocp/5600fe00.target-module"
resp. "/devices/platform/ocp/44c00000.interconnect/44c00000.interconnect:segment@200000/44e3e074.target-module"
are now shortened to "5600fe00.target-module" resp. "44e3e074.target-module".
However, "/devices/platform/ocp/48000000.interconnect/48000000.interconnect:segment@200000/48000000.interconnect:segment@200000:target-module@0"
is shortened to "48000000.interconnect:segment@200000:target-module@0",
which is still longer than the old column width...

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

On Mon, 2 Sept 2024 at 16:43, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>
> Hi Ulf,
>
> On Fri, Aug 30, 2024 at 5:50 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > On Tue, Aug 20, 2024 at 10:58 AM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > > On Tue, 20 Aug 2024 at 10:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > > > On Mon, May 27, 2024 at 4:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > > > > Using kobject_get_path() means a dynamic memory allocation gets done, which
> > > > > doesn't work on a PREEMPT_RT based configuration while holding genpd's raw
> > > > > spinlock.
> > > > >
> > > > > To fix the problem, let's convert into using the simpler dev_name(). This
> > > > > means the information about the path doesn't get presented in debugfs, but
> > > > > hopefully this shouldn't be an issue.
> > > > >
> > > > > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > > > ---
> > > > > Changes in v2:
> > > > >         - New patch.
> > > >
> > > > Thanks for your patch, which is now commit 9094e53ff5c86ebe ("pmdomain:
> > > > core: Use dev_name() instead of kobject_get_path() in debugfs")
> > > > in pmdomain/next.
> > > >
> > > > > --- a/drivers/pmdomain/core.c
> > > > > +++ b/drivers/pmdomain/core.c
> > > > > @@ -3215,16 +3214,9 @@ static int genpd_summary_one(struct seq_file *s,
> > > > >         }
> > > > >
> > > > >         list_for_each_entry(pm_data, &genpd->dev_list, list_node) {
> > > > > -               kobj_path = kobject_get_path(&pm_data->dev->kobj,
> > > > > -                               genpd_is_irq_safe(genpd) ?
> > > > > -                               GFP_ATOMIC : GFP_KERNEL);
> > > > > -               if (kobj_path == NULL)
> > > > > -                       continue;
> > > > > -
> > > > > -               seq_printf(s, "\n    %-50s  ", kobj_path);
> > > > > +               seq_printf(s, "\n    %-50s  ", dev_name(pm_data->dev));
> > > >
> > > > While some of the old names didn't even fit in 50 characters, the new
> > > > names need much less space, so perhaps this is a good opportunity to
> > > > decrease the table width?
> > >
> > > Sure, it seems reasonable! Do you want to send a patch?
> >
> > I started looking into it.  Then I noticed that on some systems
> > (e.g. TI am335x) the device names may have a longer format than
> > the typical <unit-address>.<nodename>. So I wanted to verify on
> > BeagleBone Black, but recent kernels crash during early boot.
> > Apparently that platform was broken between v6.8 and v6.9-rc1.
> > And during bisection, I encountered 3 different failure modes...
> >
> > To be continued...
>
> The longest generic node names documented in the Devicetree
> Specification are "air-pollution-sensor" and "interrupt-controller"
> (both counting 20 characters), so a typical device name needs 8
> (32-bit unit address) + 1 (dot) + 20 = 29 characters.
> However, I assume some devices lie outside the 32-bit address space,
> and thus need more space?
>
> With the BeagleBone Black boot issue fixed:
> "/devices/platform/ocp/5600fe00.target-module"
> resp. "/devices/platform/ocp/44c00000.interconnect/44c00000.interconnect:segment@200000/44e3e074.target-module"
> are now shortened to "5600fe00.target-module" resp. "44e3e074.target-module".
> However, "/devices/platform/ocp/48000000.interconnect/48000000.interconnect:segment@200000/48000000.interconnect:segment@200000:target-module@0"
> is shortened to "48000000.interconnect:segment@200000:target-module@0",
> which is still longer than the old column width...

Should we really care about those silly long names? And are those
really a problem from genpd debugfs point of view?

That said, I don't have a suggestion for a new value of the table
width, but I am certainly open to adjusting it to whatever you
propose.

Kind regards
Uffe