Re: [PATCH 1/2] of: create of_root if no dtb provided

["Clément Léger" <clement.leger@bootlin.com>]

Le Fri, 24 Jun 2022 11:44:07 -0500,
Frank Rowand <frowand.list@gmail.com> a écrit :

> On 6/24/22 08:13, Clément Léger wrote:
> > Le Thu, 23 Jun 2022 22:43:26 -0500,
> > frowand.list@gmail.com a écrit :
> >   
> >>  
> >> +/*
> >> + * __dtb_empty_root_begin[] magically created by cmd_dt_S_dtb in
> >> + * scripts/Makefile.lib
> >> + */
> >> +extern void *__dtb_empty_root_begin;
> >> +
> >>  /*
> >>   * of_fdt_limit_memory - limit the number of regions in the /memory node
> >>   * @limit: maximum entries
> >> @@ -1332,8 +1338,13 @@ bool __init early_init_dt_scan(void *params)
> >>   */
> >>  void __init unflatten_device_tree(void)
> >>  {  
> > 
> > Hi Frank,
> > 
> > This function is only defined when CONFIG_OF_EARLY_FLATTREE is enabled.  
> 
> More precisely, only if CONFIG_OF_FLATTREE is enabled.  But that would
> most likely be seleved by CONFIG_OF_EARLY_FLATTREE, so in practice the
> issue you raise is valid.
> 
> > Which means that on platforms that do not select this, the default
> > empty device-tree creation will not be done.  
> 
> Yes, so platforms that need this functionality need to select this
> option.

Yes, but this seems a bit odd because this is not really a early
flattree that is provided by the firmware. This simply allows to have a
working support for overlays (As a "side effect" I agree).

> 
> > 
> > This configuration option is selected by the platform and not by the
> > user. On x86, only one config enables this (X86_INTEL_CE) which means
> > this won't work on all the other platforms even if CONFIG_OF is
> > selected. I would need this to work by only selected CONFIG_OF.  
> 
> Maybe this means that CONFIG_OF should be changed to select
> CONFIG_OF_FLATTREE.  Any opinions on this Rob?
> 
> > That's why I decided to add the of_root creation in of_core_init()
> > using a function (of_fdt_unflatten()) that is provided if CONFIG_OF is
> > defined.  
> 
> I mentioned this in response to the previous patch series, but will
> repeat here for those who might not have read that email thread.
> 
> I do not want the root live tree to be created buy different code in
> different places; I want one central place where this occurs.  When
> the tree can be created in multiple places by different code blocks,
> it becomes more difficult to understand the code and more likely that
> one of the tree creation code blocks is not updated when another is.

Understood, my point was more about the fact that I did not wanted to
select CONFIG_OF_FLATTREE to be able to use that support which does not
seems entirely tied to having a "early flattree".

Thanks,

Clément

> 
> >   
> >> -	__unflatten_device_tree(initial_boot_params, NULL, &of_root,
> >> +	if (!initial_boot_params) {
> >> +		initial_boot_params = (void *) __dtb_empty_root_begin;
> >> +		unflatten_and_copy_device_tree();
> >> +	} else {
> >> +		__unflatten_device_tree(initial_boot_params, NULL, &of_root,
> >>  				early_init_dt_alloc_memory_arch, false);
> >> +	}
> >>  
> >>  	/* Get pointer to "/chosen" and "/aliases" nodes for use everywhere */
> >>  	of_alias_scan(early_init_dt_alloc_memory_arch);
> >> @@ -1373,6 +1384,12 @@ void __init unflatten_and_copy_device_tree(void)
> >>  	unflatten_device_tree();
> >>  }
> >>  
> >> +void __init setup_of(void)
> >> +{
> >> +	if (!of_root)
> >> +		unflatten_device_tree();
> >> +}
> >> +
> >>  #ifdef CONFIG_SYSFS
> >>  static ssize_t of_fdt_raw_read(struct file *filp, struct kobject *kobj,
> >>  			       struct bin_attribute *bin_attr,
> >> diff --git a/include/linux/of_fdt.h b/include/linux/of_fdt.h
> >> index d69ad5bb1eb1..4566876db351 100644
> >> --- a/include/linux/of_fdt.h
> >> +++ b/include/linux/of_fdt.h
> >> @@ -81,6 +81,7 @@ extern const void *of_flat_dt_match_machine(const void *default_match,
> >>  /* Other Prototypes */
> >>  extern void unflatten_device_tree(void);
> >>  extern void unflatten_and_copy_device_tree(void);
> >> +extern void setup_of(void);
> >>  extern void early_init_devtree(void *);
> >>  extern void early_get_first_memblock_info(void *, phys_addr_t *);
> >>  #else /* CONFIG_OF_EARLY_FLATTREE */
> >> @@ -91,6 +92,7 @@ static inline void early_init_fdt_reserve_self(void) {}
> >>  static inline const char *of_flat_dt_get_machine_name(void) { return NULL; }
> >>  static inline void unflatten_device_tree(void) {}
> >>  static inline void unflatten_and_copy_device_tree(void) {}
> >> +static inline void of_setup(void) {}  
> >   
> 
> > Shouldn't this be setup_of(void) ?  
> 
> Yes, thanks!  Will fix.
> 
> One other thing I need to do is test this patch on a user mode linux
> kernel.
> 
> -Frank
> 



-- 
Clément Léger,
Embedded Linux and Kernel engineer at Bootlin
https://bootlin.com