[PATCH] lib: sbi: Add option to select boot hart

[Cheehong Ang]

> >> -----Original Message-----
> >> From: Jessica Clarke <jrtc27 at jrtc27.com>
> >> Sent: Wednesday, November 30, 2022 12:18 AM
> >> To: WeiLiang Lim <weiliang.lim at starfivetech.com>
> >> Cc: opensbi at lists.infradead.org; Cheehong Ang
> >> <cheehong.ang at starfivetech.com>; JunLiang Tan
> >> <junliang.tan at starfivetech.com>
> >> Subject: Re: [PATCH] lib: sbi: Add option to select boot hart
> >>
> >> On 29 Nov 2022, at 05:21, Wei Liang Lim
> >> <weiliang.lim at starfivetech.com>
> >> wrote:
> >>>
> >>> In certain scenario, user needs to select a particular hart ID as the boot
> hart.
> >> This option provide the flexibility for user to select the preferred boot
> hart.
> >>
> >> Why? What’s wrong with the others?
> >>
> >> Jess
> >>
> > We need this patch for 2 scenarios below:
> > 1) On our architecture, there are some harts supporting MMU sv39/sv48
> and some harts only supporting sv39. When those harts with sv48 support is
> picked as boot hart, the OS automatically assumes all the other non-boot
> harts support sv48 as well. Therefore, we need to select the harts with sv39
> supports only as the boot hart to avoid the incorrect page table issues in
> non-boot harts.
> 
> If you have a heterogenous environment then this isn’t the right fix, OSes
> need to be aware of heterogeneity otherwise other things can easily go
> wrong. And if they’re aware of that then you don’t need to control which OS
> boots. Nothing in the RISC-V world currently says that OSes should do their
> feature detection on the first hart that boots...
> 
> > 2) We have use case which requires the same boot hart to save OS context
> before going into 'hibernation' mode and to restore the OS context after
> resuming from 'hibernation' mode.
> 
> I can’t imagine current OpenSBI+Linux is at all prepared for hibernation so
> don’t think that’s a sufficient justification for it today?
> 

Appreciate the comments and responses.
We agree that there are other better/proper ways to fix the issues that we have encountered above.
Just like Bin Meng highlighted in other threads that instead of fixing the boot hart order, CONFIG options can be added to kernel code to enforce certain MMU type.
There is nothing wrong with lottery mechanism boot flow. We are not trying fix the boot hart selection mechanism in OpenSBI to resolve our problems. We just try to provide another option for user to specify a boot hart for certain use cases.
In some cases, we may also need to select certain hart as boot hart for debugging the SMP issues especially during hardware development stages.
My question here is do we think providing another option to select certain hart ID as boot hart is something unnecessary in RISC-V world at all ?

> Jess
> 
> >>> Signed-off-by: Wei Liang Lim <weiliang.lim at starfivetech.com>
> >>> Reviewed-by: Chee Hong Ang <cheehong.ang at starfivetech.com>
> >>> Reviewed-by: Jun Liang Tan <junliang.tan at starfivetech.com>
> >>> ---
> >>> firmware/objects.mk | 4 ++++
> >>> lib/sbi/sbi_init.c  | 9 ++++++++-
> >>> 2 files changed, 12 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/firmware/objects.mk b/firmware/objects.mk index
> >>> a1704c4..2b41718 100644
> >>> --- a/firmware/objects.mk
> >>> +++ b/firmware/objects.mk
> >>> @@ -28,6 +28,10 @@ ifdef FW_TEXT_START firmware-genflags-y +=
> >>> -DFW_TEXT_START=$(FW_TEXT_START) endif
> >>>
> >>> +ifdef FW_BOOT_HART_ID
> >>> +firmware-genflags-y += -DFW_BOOT_HART_ID=$(FW_BOOT_HART_ID)
> >>> +endif
> >>> +
> >>> ifdef FW_FDT_PATH
> >>> firmware-genflags-y += -DFW_FDT_PATH=\"$(FW_FDT_PATH)\"
> >>> ifdef FW_FDT_PADDING
> >>> diff --git a/lib/sbi/sbi_init.c b/lib/sbi/sbi_init.c index
> >>> d57efa7..5abfa79 100644
> >>> --- a/lib/sbi/sbi_init.c
> >>> +++ b/lib/sbi/sbi_init.c
> >>> @@ -443,7 +443,9 @@ static void __noreturn init_warmboot(struct
> >> sbi_scratch *scratch, u32 hartid)
> >>> 			     scratch->next_mode, FALSE);
> >>> }
> >>>
> >>> +#ifndef FW_BOOT_HART_ID
> >>> static atomic_t coldboot_lottery = ATOMIC_INITIALIZER(0);
> >>> +#endif
> >>>
> >>> /**
> >>> * Initialize OpenSBI library for current HART and jump to next @@
> >>> -494,7 +496,12 @@ void __noreturn sbi_init(struct sbi_scratch *scratch)
> >>> 	 * HARTs which satisfy above condition.
> >>> 	 */
> >>>
> >>> -	if (next_mode_supported && atomic_xchg(&coldboot_lottery, 1) ==
> 0)
> >>> +	if (next_mode_supported &&
> >>> +#ifdef FW_BOOT_HART_ID
> >>> +		hartid == FW_BOOT_HART_ID)
> >>> +#else
> >>> +		atomic_xchg(&coldboot_lottery, 1) == 0) #endif
> >>> 		coldboot = TRUE;
> >>>
> >>> 	/*
> >>> --
> >>> 2.25.1
> >>>
> >>>
> >>> --
> >>> opensbi mailing list
> >>> opensbi at lists.infradead.org
> >>> http://lists.infradead.org/mailman/listinfo/opensbi
> >