scripts/gdb: Fix lx-device-list-bus and lx-device-list-class

  After the conversion to bus_to_subsys() and class_to_subsys(), the gdb
scripts listing the system buses and classes respectively was broken,
fix those by returning the subsys_priv pointer and have the various
caller de-reference either the 'bus' or 'class' structure members
accordingly.

Fixes: 83b9148df2c9 ("driver core: bus: bus iterator cleanups")
Fixes: 7b884b7f24b4 ("driver core: class.c: convert to only use class_to_subsys")
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
---
 scripts/gdb/linux/device.py | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
  

On Wed, 29 Nov 2023 20:33:16 -0800 Florian Fainelli <florian.fainelli@broadcom.com> wrote:

> After the conversion to bus_to_subsys() and class_to_subsys(), the gdb
> scripts listing the system buses and classes respectively was broken,
> fix those by returning the subsys_priv pointer and have the various
> caller de-reference either the 'bus' or 'class' structure members
> accordingly.
> 
> Fixes: 83b9148df2c9 ("driver core: bus: bus iterator cleanups")
> Fixes: 7b884b7f24b4 ("driver core: class.c: convert to only use class_to_subsys")

I don't know how -stable maintainers handle more than one Fixes:.  One
of the above was released in 6.3, the other in 6.4.  What are they to
do?

My opinion: developers should use Fixes: as a request to -stable
maintainers to target a particular kernel version (and later).

If the developers indeed want to fix both kernel versions then wouldn't
two patches be best?
  

On 11/30/23 13:51, Andrew Morton wrote:
> On Wed, 29 Nov 2023 20:33:16 -0800 Florian Fainelli <florian.fainelli@broadcom.com> wrote:
> 
>> After the conversion to bus_to_subsys() and class_to_subsys(), the gdb
>> scripts listing the system buses and classes respectively was broken,
>> fix those by returning the subsys_priv pointer and have the various
>> caller de-reference either the 'bus' or 'class' structure members
>> accordingly.
>>
>> Fixes: 83b9148df2c9 ("driver core: bus: bus iterator cleanups")
>> Fixes: 7b884b7f24b4 ("driver core: class.c: convert to only use class_to_subsys")
> 
> I don't know how -stable maintainers handle more than one Fixes:.  One
> of the above was released in 6.3, the other in 6.4.  What are they to
> do?

Neither are stable kernels fortunately, so any kernel >= 6.4 where both 
issues are present should be able to cleanly cherry-pick the change.

I suppose this should have called for me to issue two patches, one 
fixing lx-device-list-bus and the other one fixing lx-device-list-class, 
though it seems like this is such a niche usage that it may not matter 
that much.

> 
> My opinion: developers should use Fixes: as a request to -stable
> maintainers to target a particular kernel version (and later).
> 
> If the developers indeed want to fix both kernel versions then wouldn't
> two patches be best?
> 

They probably would, or I could submit a targeted fix for 6.3 for instance.
  

On Thu, 30 Nov 2023 14:00:32 -0800 Florian Fainelli <florian.fainelli@broadcom.com> wrote:

> >> Fixes: 83b9148df2c9 ("driver core: bus: bus iterator cleanups")
> >> Fixes: 7b884b7f24b4 ("driver core: class.c: convert to only use class_to_subsys")
> > 
> > I don't know how -stable maintainers handle more than one Fixes:.  One
> > of the above was released in 6.3, the other in 6.4.  What are they to
> > do?
> 
> Neither are stable kernels fortunately, so any kernel >= 6.4 where both 
> issues are present should be able to cleanly cherry-pick the change.

OK, I'll change it to just 6.4's 7b884b7f24b4 to simplify life.

> I suppose this should have called for me to issue two patches, one 
> fixing lx-device-list-bus and the other one fixing lx-device-list-class, 
> though it seems like this is such a niche usage that it may not matter 
> that much.

It is indeed nichey :)
  

On Wed, 2023-11-29 at 20:33 -0800, Florian Fainelli wrote:
> After the conversion to bus_to_subsys() and class_to_subsys(), the
> gdb
> scripts listing the system buses and classes respectively was broken,
> fix those by returning the subsys_priv pointer and have the various
> caller de-reference either the 'bus' or 'class' structure members
> accordingly.
> 
> Fixes: 83b9148df2c9 ("driver core: bus: bus iterator cleanups")
> Fixes: 7b884b7f24b4 ("driver core: class.c: convert to only use
> class_to_subsys")
> Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>

Tested-by: Kuan-Ying Lee <Kuan-Ying.Lee@mediatek.com>

Thanks,
Kuan-Ying Lee

> ---
>  scripts/gdb/linux/device.py | 16 ++++++++--------
>  1 file changed, 8 insertions(+), 8 deletions(-)
> 
> diff --git a/scripts/gdb/linux/device.py
> b/scripts/gdb/linux/device.py
> index 16376c5cfec6..0eabc5f4f8ca 100644
> --- a/scripts/gdb/linux/device.py
> +++ b/scripts/gdb/linux/device.py
> @@ -36,26 +36,26 @@ def for_each_bus():
>      for kobj in
> kset_for_each_object(gdb.parse_and_eval('bus_kset')):
>          subsys = container_of(kobj, kset_type.get_type().pointer(),
> 'kobj')
>          subsys_priv = container_of(subsys,
> subsys_private_type.get_type().pointer(), 'subsys')
> -        yield subsys_priv['bus']
> +        yield subsys_priv
>  
>  
>  def for_each_class():
>      for kobj in
> kset_for_each_object(gdb.parse_and_eval('class_kset')):
>          subsys = container_of(kobj, kset_type.get_type().pointer(),
> 'kobj')
>          subsys_priv = container_of(subsys,
> subsys_private_type.get_type().pointer(), 'subsys')
> -        yield subsys_priv['class']
> +        yield subsys_priv
>  
>  
>  def get_bus_by_name(name):
>      for item in for_each_bus():
> -        if item['name'].string() == name:
> +        if item['bus']['name'].string() == name:
>              return item
>      raise gdb.GdbError("Can't find bus type {!r}".format(name))
>  
>  
>  def get_class_by_name(name):
>      for item in for_each_class():
> -        if item['name'].string() == name:
> +        if item['class']['name'].string() == name:
>              return item
>      raise gdb.GdbError("Can't find device class {!r}".format(name))
>  
> @@ -70,13 +70,13 @@ def klist_for_each(klist):
>  
>  
>  def bus_for_each_device(bus):
> -    for kn in klist_for_each(bus['p']['klist_devices']):
> +    for kn in klist_for_each(bus['klist_devices']):
>          dp = container_of(kn,
> device_private_type.get_type().pointer(), 'knode_bus')
>          yield dp['device']
>  
>  
>  def class_for_each_device(cls):
> -    for kn in klist_for_each(cls['p']['klist_devices']):
> +    for kn in klist_for_each(cls['klist_devices']):
>          dp = container_of(kn,
> device_private_type.get_type().pointer(), 'knode_class')
>          yield dp['device']
>  
> @@ -103,7 +103,7 @@ class LxDeviceListBus(gdb.Command):
>      def invoke(self, arg, from_tty):
>          if not arg:
>              for bus in for_each_bus():
> -                gdb.write('bus
> {}:\t{}\n'.format(bus['name'].string(), bus))
> +                gdb.write('bus
> {}:\t{}\n'.format(bus['bus']['name'].string(), bus))
>                  for dev in bus_for_each_device(bus):
>                      _show_device(dev, level=1)
>          else:
> @@ -123,7 +123,7 @@ class LxDeviceListClass(gdb.Command):
>      def invoke(self, arg, from_tty):
>          if not arg:
>              for cls in for_each_class():
> -                gdb.write("class
> {}:\t{}\n".format(cls['name'].string(), cls))
> +                gdb.write("class
> {}:\t{}\n".format(cls['class']['name'].string(), cls))
>                  for dev in class_for_each_device(cls):
>                      _show_device(dev, level=1)
>          else: