Re: [PATCH v1 1/4] xen/riscv: introduce preinit_xen_time()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.03.2025 16:29, Oleksii Kurochko wrote:
> 
> On 3/20/25 8:36 AM, Jan Beulich wrote:
>> On 19.03.2025 18:29, Oleksii Kurochko wrote:
>>> On 3/17/25 4:24 PM, Jan Beulich wrote:
>>>> On 11.03.2025 17:19, Oleksii Kurochko wrote:
>>>>> --- /dev/null
>>>>> +++ b/xen/arch/riscv/time.c
>>>>> @@ -0,0 +1,38 @@
>>>>> +#include <xen/device_tree.h>
>>>>> +#include <xen/init.h>
>>>>> +#include <xen/lib.h>
>>>>> +#include <xen/sections.h>
>>>>> +
>>>>> +unsigned long __ro_after_init cpu_khz; /* CPU clock frequency in kHz. */
>>>>> +unsigned long __read_mostly boot_count;
>>>> Why not also __ro_after_init? And what is this variable actually needed
>>>> for? Common code doesn't use it, so a better name (describing what it
>>>> really holds) might be desirable, even if this then means not being in
>>>> sync with Arm code.
>>> To calculate more accurate amount of time since boot.
>> Okay. But how does the name of the variable reflect that? I.e. what it
>> is that the count of is being stored? The only meaning I could associate
>> to a variable of this name is the number of boot cycles a system went
>> through. I.e. nothing that an OS (or hypervisor) would normally count.
> 
> But an OS (or hypervisor) doesn't count it, they initialize a variable
> once (in my case, it was named as boot_count) and then just subtract it from
> get_cycles() to get time relative to this variable (so since Xen boot) and not
> since power on as nothing guarantee (at least, I can't find that in the 
> RISC-V spec)
> that after power on the value in CSR_TIME will start from 0 what could lead to
> some issues, if my understanding is correct, such as if on SoC A timer starts 
> from
> let it be 1000 and on SoC B timer value starts from 5000 then all 
> measurements will be
> incorrect as Xen will think that for SoC B it was spent more time then for 
> SoC A.
> 
> What do you think if boot_count will be renamed to xen_start_time or 
> {initial_}boot_time?

Something like that, yes. Whether "time" in there is unambiguous enough I'm not
sure. "cycles" or "clock_cycles" or some such may help. I don't really want to
restrict you in what name you choose, just so long as the name reflects the
purpose in good enough a way.

Jan