On Fri, Nov 01, 2024 at 12:50:26PM +0100, Simon Glass wrote:

Rather than doing autoprobe within the driver model code, move it out to the board-init code. This makes it clear that it is a separate step from binding devices.

For now this is always done twice, before and after relocation, but we should discuss whether it might be possible to drop the post-relocation probe.

For boards with SPL, the autoprobe is still done there as well.

Note that with this change, autoprobe happens after the EVT_DM_POST_INIT_R/F events are sent, rather than before.

Update the docs a little, for this feature.

Signed-off-by: Simon Glass sjg@chromium.org

This commit is where we need: Link: https://lore.kernel.org/u-boot/20240626235717.272219-1-marex@denx.de/

So that it's easy to go from git to fuller history.

diff --git a/doc/develop/driver-model/design.rst b/doc/develop/driver-model/design.rst index 8c2c81d7ac9..1dcac1408ff 100644 --- a/doc/develop/driver-model/design.rst +++ b/doc/develop/driver-model/design.rst @@ -842,6 +842,21 @@ steps (see device_probe()): cause the uclass to do some housekeeping to record the device as activated and 'known' by the uclass.

+For some platforms, certain devices must be probed to get the platform into +a working state. To help with this, drivers marked with +``DM_FLAG_PROBE_AFTER_BIND`` will be probed immediately after all devices are +bound. For now, this happens in each xPL build as well as in U-Boot proper, +both before relocation and after relocation. See the call to ``dm_autoprobe()`` +for where this is done.

+Note that autoprobe happens after the ``EVT_DM_POST_INIT_R`` and +``EVT_DM_POST_INIT_F`` events are sent, since these events relate to +device-binding.

+See here for discussion of this feature:

+:Link: https://patchwork.ozlabs.org/project/uboot/patch/20240626235717.272219-1-mar...

There's 40+ messages in the thread (but I know not all of them were about design), so is the above really a sufficient summary?

Hi Quentin,

On Wed, 6 Nov 2024 at 08:31, Quentin Schulz quentin.schulz@cherry.de wrote:

Hi Simon,

Nitpicks below.

On 11/1/24 12:50 PM, Simon Glass wrote:

...

Rather than doing autoprobe within the driver model code, move it out to the board-init code. This makes it clear that it is a separate step from binding devices.

For now this is always done twice, before and after relocation, but we should discuss whether it might be possible to drop the post-relocation probe.

For boards with SPL, the autoprobe is still done there as well.

Note that with this change, autoprobe happens after the EVT_DM_POST_INIT_R/F events are sent, rather than before.

Update the docs a little, for this feature.

Signed-off-by: Simon Glass sjg@chromium.org

Changes in v2:

• Add autoprobe to SPL also

• Leave the function name the same

• Fix 'Prove' typo

• Update cover letter since SPL is now covered also

• Update the documentation too, with a discussion link

  common/board_f.c | 4 ++++ common/board_r.c | 4 ++++ common/spl/spl.c | 4 ++++ doc/develop/driver-model/design.rst | 15 +++++++++++++++ drivers/core/root.c | 13 ++++++++++++- include/dm/root.h | 10 ++++++++++ 6 files changed, 49 insertions(+), 1 deletion(-)

diff --git a/common/board_f.c b/common/board_f.c index 1e1aa08530a..974eb4f39e9 100644 --- a/common/board_f.c +++ b/common/board_f.c @@ -822,6 +822,10 @@ static int initf_dm(void) if (ret) return ret;

• ret = dm_autoprobe();
  

• if (ret)
  

•         return ret;
  

• if (IS_ENABLED(CONFIG_TIMER_EARLY)) {
          ret = dm_timer_init();
          if (ret)
  

diff --git a/common/board_r.c b/common/board_r.c index 62228a723e1..d1507b06f8d 100644 --- a/common/board_r.c +++ b/common/board_r.c @@ -241,6 +241,10 @@ static int initr_dm(void) if (ret) return ret;

• ret = dm_autoprobe();
  

• if (ret)
  

•         return ret;
  

• return 0;
  

return dm_autoprobe();

OK

?

...

} #endif diff --git a/common/spl/spl.c b/common/spl/spl.c index 1ceb63daf31..c4256f237ee 100644 --- a/common/spl/spl.c +++ b/common/spl/spl.c @@ -500,6 +500,10 @@ static int spl_common_init(bool setup_malloc) debug("dm_init_and_scan() returned error %d\n", ret); return ret; }

•         ret = dm_autoprobe();
  

•         if (ret)
  

•                 return ret;
  }
  
  return 0;
  

diff --git a/doc/develop/driver-model/design.rst b/doc/develop/driver-model/design.rst index 8c2c81d7ac9..1dcac1408ff 100644 --- a/doc/develop/driver-model/design.rst +++ b/doc/develop/driver-model/design.rst @@ -842,6 +842,21 @@ steps (see device_probe()): cause the uclass to do some housekeeping to record the device as activated and 'known' by the uclass.

+For some platforms, certain devices must be probed to get the platform into +a working state. To help with this, drivers marked with +``DM_FLAG_PROBE_AFTER_BIND`` will be probed immediately after all devices are +bound. For now, this happens in each xPL build as well as in U-Boot proper, +both before relocation and after relocation. See the call to ``dm_autoprobe()`` +for where this is done.

+Note that autoprobe happens after the ``EVT_DM_POST_INIT_R`` and +``EVT_DM_POST_INIT_F`` events are sent, since these events relate to +device-binding.

+See here for discussion of this feature:

+:Link: https://patchwork.ozlabs.org/project/uboot/patch/20240626235717.272219-1-mar...

• Running stage ^^^^^^^^^^^^^

diff --git a/drivers/core/root.c b/drivers/core/root.c index 2d4f078f97f..371a37fe752 100644 --- a/drivers/core/root.c +++ b/drivers/core/root.c @@ -299,6 +299,17 @@ static int dm_probe_devices(struct udevice *dev) return 0; }

+int dm_autoprobe(void) +{

• int ret;
  

• ret = dm_probe_devices(gd->dm_root);
  

• if (ret)
  

•         return log_msg_ret("pro", ret);
  

• return 0;
  

1. Should be log_msg_retz since log_msg_ret will only print the message

on < 0 and here we test for ret

Well we have this all over the place. I'll add a comment to dm_probe_devices(), but log_msg_retz() is really only for cases where the function can return a positive value (e.g. a size, or node offset). Most functions just return 0 on success and -ve error, which is what log_msg_ret() is for.

2. Simplify:

return log_msg_retz("pro", dm_probe_devices(gd->dm_root));

the message won't be printed if the return value is 0.

I don't like combining things like this, so tend to write it out in full. It is easier to add to the code later. Also the semantics of putting a required call inside a debugging call are not that clear. Perhaps we could document this in log.h ? But I always like to have a simple 'return 0' at the end of a function, after all the error cases have been handled.

...

+}

• /**
  • dm_scan() - Scan tables to bind devices

@@ -331,7 +342,7 @@ static int dm_scan(bool pre_reloc_only) if (ret) return ret;

• return dm_probe_devices(gd->dm_root);
  

• return 0;
  

Simplify with:

return dm_scan_other(pre_reloc_only);

See above :-)

...

}

int dm_init_and_scan(bool pre_reloc_only) diff --git a/include/dm/root.h b/include/dm/root.h index b2f30a842f5..cd240c6a716 100644 --- a/include/dm/root.h +++ b/include/dm/root.h @@ -136,6 +136,16 @@ int dm_scan_other(bool pre_reloc_only); */ int dm_init_and_scan(bool pre_reloc_only);

+/**

• • dm_autoprobe() - Probe devices which are marked for probe-after-bind
• • This probes all devices with a DM_FLAG_PROBE_AFTER_BIND flag. It checks the
• • entire tree, so parent nodes need not have the flag set.

Can suggest: """ It recursively probes parent nodes, so they do not need to have the flag set themselves. """ which is what I believe you wanted to say? "Checking the entire tree" induces the opposite of "parent nodes need not have the flag set".

Thanks, added and I will expand a little, too.

Regards, Simon