Re: [patch V4 00/37] cpu/hotplug, x86: Reworked parallel CPU bringup

[Oleksandr Natalenko <oleksandr@natalenko.name>]

Hello.

On pátek 12. května 2023 23:06:56 CEST Thomas Gleixner wrote:
> Hi!
> 
> This is version 4 of the reworked parallel bringup series. Version 3 can be
> found here:
> 
>    https://lore.kernel.org/lkml/20230508181633.089804905@linutronix.de
> 
> This is just a reiteration to address the following details:
> 
>   1) Address review feedback (Peter Zijlstra)
> 
>   2) Fix a MIPS related build problem (0day)
> 
> Other than that there are no changes and the other details are all the same
> as in V3 and V2.
> 
> It's also available from git:
> 
>     git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git hotplug
> 
> Diff to V3 below.
> 
> Thanks,
> 
> 	tglx

With this patchset:

```

[    0.137719] smpboot: Allowing 32 CPUs, 0 hotplug CPUs
[    0.777312] smpboot: CPU0: AMD Ryzen 9 5950X 16-Core Processor (family: 0x19, model: 0x21, stepping: 0x2)
[    0.777896] smpboot: Parallel CPU startup enabled: 0x80000000
```

Seems to survive suspend/resume cycle too.

Hence:

Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>

Thanks.

> ---
> diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c
> index f5e0f4235746..90c71d800b59 100644
> --- a/arch/mips/kernel/smp.c
> +++ b/arch/mips/kernel/smp.c
> @@ -690,7 +690,7 @@ void flush_tlb_one(unsigned long vaddr)
>  EXPORT_SYMBOL(flush_tlb_page);
>  EXPORT_SYMBOL(flush_tlb_one);
>  
> -#ifdef CONFIG_HOTPLUG_CPU
> +#ifdef CONFIG_HOTPLUG_CORE_SYNC_DEAD
>  void arch_cpuhp_cleanup_dead_cpu(unsigned int cpu)
>  {
>  	if (mp_ops->cleanup_dead_cpu)
> diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
> index 0438802031c3..9cd77d319555 100644
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -290,8 +290,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>  
>  	/*  APIC ID not found in the table. Drop the trampoline lock and bail. */
>  	movq	trampoline_lock(%rip), %rax
> -	lock
> -	btrl	$0, (%rax)
> +	movl	$0, (%rax)
>  
>  1:	cli
>  	hlt
> @@ -320,8 +319,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>  	movq	trampoline_lock(%rip), %rax
>  	testq	%rax, %rax
>  	jz	.Lsetup_gdt
> -	lock
> -	btrl	$0, (%rax)
> +	movl	$0, (%rax)
>  
>  .Lsetup_gdt:
>  	/*
> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
> index 5caf4897b507..660709e94823 100644
> --- a/arch/x86/kernel/smpboot.c
> +++ b/arch/x86/kernel/smpboot.c
> @@ -161,31 +161,28 @@ static inline void smpboot_restore_warm_reset_vector(void)
>  
>  }
>  
> -/*
> - * Report back to the Boot Processor during boot time or to the caller processor
> - * during CPU online.
> - */
> -static void smp_callin(void)
> +/* Run the next set of setup steps for the upcoming CPU */
> +static void ap_starting(void)
>  {
>  	int cpuid = smp_processor_id();
>  
>  	/*
> -	 * If waken up by an INIT in an 82489DX configuration the alive
> -	 * synchronization guarantees we don't get here before an
> -	 * INIT_deassert IPI reaches our local APIC, so it is now safe to
> -	 * touch our local APIC.
> +	 * If woken up by an INIT in an 82489DX configuration the alive
> +	 * synchronization guarantees that the CPU does not reach this
> +	 * point before an INIT_deassert IPI reaches the local APIC, so it
> +	 * is now safe to touch the local APIC.
>  	 *
>  	 * Set up this CPU, first the APIC, which is probably redundant on
>  	 * most boards.
>  	 */
>  	apic_ap_setup();
>  
> -	/* Save our processor parameters. */
> +	/* Save the processor parameters. */
>  	smp_store_cpu_info(cpuid);
>  
>  	/*
>  	 * The topology information must be up to date before
> -	 * calibrate_delay() and notify_cpu_starting().
> +	 * notify_cpu_starting().
>  	 */
>  	set_cpu_sibling_map(cpuid);
>  
> @@ -197,7 +194,7 @@ static void smp_callin(void)
>  
>  	/*
>  	 * This runs the AP through all the cpuhp states to its target
> -	 * state (CPUHP_ONLINE in the case of serial bringup).
> +	 * state CPUHP_ONLINE.
>  	 */
>  	notify_cpu_starting(cpuid);
>  }
> @@ -274,10 +271,7 @@ static void notrace start_secondary(void *unused)
>  	rcu_cpu_starting(raw_smp_processor_id());
>  	x86_cpuinit.early_percpu_clock_init();
>  
> -	smp_callin();
> -
> -	/* Otherwise gcc will move up smp_processor_id() before cpu_init() */
> -	barrier();
> +	ap_starting();
>  
>  	/* Check TSC synchronization with the control CPU. */
>  	check_tsc_sync_target();
> diff --git a/arch/x86/realmode/rm/trampoline_64.S b/arch/x86/realmode/rm/trampoline_64.S
> index 2dfb1c400167..c6de4deec746 100644
> --- a/arch/x86/realmode/rm/trampoline_64.S
> +++ b/arch/x86/realmode/rm/trampoline_64.S
> @@ -40,17 +40,13 @@
>  .macro LOAD_REALMODE_ESP
>  	/*
>  	 * Make sure only one CPU fiddles with the realmode stack
> -	 */
> +	*/
>  .Llock_rm\@:
> -	btl	$0, tr_lock
> -	jnc	2f
> -	pause
> -	jmp	.Llock_rm\@
> +        lock btsl       $0, tr_lock
> +        jnc             2f
> +        pause
> +        jmp             .Llock_rm\@
>  2:
> -	lock
> -	btsl	$0, tr_lock
> -	jc	.Llock_rm\@
> -
>  	# Setup stack
>  	movl	$rm_stack_end, %esp
>  .endm
> diff --git a/kernel/cpu.c b/kernel/cpu.c
> index 60b4093fae9e..005f863a3d2b 100644
> --- a/kernel/cpu.c
> +++ b/kernel/cpu.c
> @@ -294,14 +294,14 @@ enum cpuhp_sync_state {
>   * cpuhp_ap_update_sync_state - Update synchronization state during bringup/teardown
>   * @state:	The synchronization state to set
>   *
> - * No synchronization point. Just update of the synchronization state.
> + * No synchronization point. Just update of the synchronization state, but implies
> + * a full barrier so that the AP changes are visible before the control CPU proceeds.
>   */
>  static inline void cpuhp_ap_update_sync_state(enum cpuhp_sync_state state)
>  {
>  	atomic_t *st = this_cpu_ptr(&cpuhp_state.ap_sync_state);
> -	int sync = atomic_read(st);
>  
> -	while (!atomic_try_cmpxchg(st, &sync, state));
> +	(void)atomic_xchg(st, state);
>  }
>  
>  void __weak arch_cpuhp_sync_state_poll(void) { cpu_relax(); }
> @@ -829,7 +829,11 @@ static int bringup_cpu(unsigned int cpu)
>  	/*
>  	 * Some architectures have to walk the irq descriptors to
>  	 * setup the vector space for the cpu which comes online.
> -	 * Prevent irq alloc/free across the bringup.
> +	 *
> +	 * Prevent irq alloc/free across the bringup by acquiring the
> +	 * sparse irq lock. Hold it until the upcoming CPU completes the
> +	 * startup in cpuhp_online_idle() which allows to avoid
> +	 * intermediate synchronization points in the architecture code.
>  	 */
>  	irq_lock_sparse();
>  
> 
> 
> 


-- 
Oleksandr Natalenko (post-factum)