Re: [PATCH v4 02/17] clk: at91: pmc: execute suspend/resume only for backup mode

[Claudiu.Beznea]

On 08.10.2021 06:51, Stephen Boyd wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> Quoting Claudiu Beznea (2021-09-23 06:20:31)
>> Before going to backup mode architecture specific PM code sets the first
>> word in securam (file arch/arm/mach-at91/pm.c, function at91_pm_begin()).
>> Thus take this into account when suspending/resuming clocks. This will
>> avoid executing unnecessary instructions when suspending to non backup
>> modes. Also this commit changed the postcore_initcall() with
>> subsys_initcall() to be able to execute of_find_compatible_node() since
>> this was not available at the moment of postcore_initcall(). This should
>> not alter the tcb_clksrc since the changes are related to clocks
>> suspend/resume procedure that will be executed at the user space request,
>> thus long ago after subsys_initcall().
> 
> Is the comment still relevant though?

For architecture PM code yes, the securam is set in [1].

Related to replacing postcore_init() with subsys_initcall() to be able to
have the proper result of of_find_compatible_node() I have to re-check
(don't know if something has been changed in this area since January). If
you know something please let me know.

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm/mach-at91/pm.c#n290

> 
>>
>> Signed-off-by: Claudiu Beznea <claudiu.beznea@xxxxxxxxxxxxx>
>> ---
>> diff --git a/drivers/clk/at91/pmc.c b/drivers/clk/at91/pmc.c
>> index b2806946a77a..58e9c088cb22 100644
>> --- a/drivers/clk/at91/pmc.c
>> +++ b/drivers/clk/at91/pmc.c
>> @@ -110,13 +112,35 @@ struct pmc_data *pmc_data_allocate(unsigned int ncore, unsigned int nsystem,
>>  }
>>
>>  #ifdef CONFIG_PM
>> +
>> +/* Address in SECURAM that say if we suspend to backup mode. */
>> +static void __iomem *at91_pmc_backup_suspend;
>> +
>>  static int at91_pmc_suspend(void)
>>  {
>> +       unsigned int backup;
>> +
>> +       if (!at91_pmc_backup_suspend)
>> +               return 0;
>> +
>> +       backup = *(unsigned int *)at91_pmc_backup_suspend;
> 
> This will fail sparse. Why are we reading iomem without using iomem
> reading wrapper?

By mistake. I'll switch to iomem reading wrapper.

Is it OK to send soon a new version with these adjustments or do you have
other patches in this series to review?

Thank you,
Claudiu Beznea

> 
>> +       if (!backup)