Hi! This is version 2 of the reworked parallel bringup series. Version 1 can be found here: https://lore.kernel.org/lkml/20230414225551.858160935@xxxxxxxxxxxxx Background ---------- The reason why people are interested in parallel bringup is to shorten the (kexec) reboot time of cloud servers to reduce the downtime of the VM tenants. The current fully serialized bringup does the following per AP: 1) Prepare callbacks (allocate, intialize, create threads) 2) Kick the AP alive (e.g. INIT/SIPI on x86) 3) Wait for the AP to report alive state 4) Let the AP continue through the atomic bringup 5) Let the AP run the threaded bringup to full online state There are two significant delays: #3 The time for an AP to report alive state in start_secondary() on x86 has been measured in the range between 350us and 3.5ms depending on vendor and CPU type, BIOS microcode size etc. #4 The atomic bringup does the microcode update. This has been measured to take up to ~8ms on the primary threads depending on the microcode patch size to apply. On a two socket SKL server with 56 cores (112 threads) the boot CPU spends on current mainline about 800ms busy waiting for the APs to come up and apply microcode. That's more than 80% of the actual onlining procedure. By splitting the actual bringup mechanism into two parts this can be reduced to waiting for the first AP to report alive or if the system is large enough the first AP is already waiting when the boot CPU finished the wake-up of the last AP. That reduces the AP bringup time on that SKL from ~800ms to ~80ms. The actual gain varies wildly depending on the system, CPU, microcode patch size and other factors. The V1 cover letter has more details and a deep analysis. Changes vs. V1: 1) Switch APIC ID retrieval from CPUID to reading the APIC itself. This is required because CPUID based APIC ID retrieval can only provide the initial APIC ID, which might have been overruled by the firmware. Some AMD APUs come up with APIC ID = initial APIC ID + 0x10, so the APIC ID to CPU number lookup would fail miserably if based on CPUID. The only requirement is that the actual APIC IDs are consistent with the APCI/MADT table. 2) As a consequence of #1 parallel bootup support for SEV guest has been dropped. Reading the APIC ID in a SEV guest is done via RDMSR. That RDMSR is intercepted and raises #VC which cannot be handled at that point as there is no stack and no IDT. There is no GHCB protocol for RDMSR like there is for CPUID. Left as an exercise for SEV wizards. 3) Address review comments from Brian and the fallout reported by the kernel robot 4) Unbreak i386 which exploded when bringing up the secondary CPUs due to the unconditinal load_ucode_ap() invocation in start_secondary(). That happens because on 32-bit load_ucode_ap() is invoked on the secondary CPUs from assembly code before paging is initialized and therefore uses physical addresses which are obviously invalid after paging is enabled. 5) Small enhancements and comment updates. 6) Rebased on Linux tree (1a5304fecee5) The series applies on Linus tree and is also available from git: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git hotplug Thanks, tglx --- Documentation/admin-guide/kernel-parameters.txt | 20 Documentation/core-api/cpu_hotplug.rst | 13 arch/Kconfig | 23 + arch/arm/Kconfig | 1 arch/arm/include/asm/smp.h | 2 arch/arm/kernel/smp.c | 18 arch/arm64/Kconfig | 1 arch/arm64/include/asm/smp.h | 2 arch/arm64/kernel/smp.c | 14 arch/csky/Kconfig | 1 arch/csky/include/asm/smp.h | 2 arch/csky/kernel/smp.c | 8 arch/mips/Kconfig | 1 arch/mips/cavium-octeon/smp.c | 1 arch/mips/include/asm/smp-ops.h | 1 arch/mips/kernel/smp-bmips.c | 1 arch/mips/kernel/smp-cps.c | 14 arch/mips/kernel/smp.c | 8 arch/mips/loongson64/smp.c | 1 arch/parisc/Kconfig | 1 arch/parisc/kernel/process.c | 4 arch/parisc/kernel/smp.c | 7 arch/riscv/Kconfig | 1 arch/riscv/include/asm/smp.h | 2 arch/riscv/kernel/cpu-hotplug.c | 14 arch/x86/Kconfig | 45 -- arch/x86/include/asm/apic.h | 5 arch/x86/include/asm/apicdef.h | 5 arch/x86/include/asm/cpu.h | 5 arch/x86/include/asm/cpumask.h | 5 arch/x86/include/asm/processor.h | 1 arch/x86/include/asm/realmode.h | 3 arch/x86/include/asm/smp.h | 24 - arch/x86/include/asm/topology.h | 23 - arch/x86/include/asm/tsc.h | 2 arch/x86/kernel/acpi/sleep.c | 9 arch/x86/kernel/apic/apic.c | 26 - arch/x86/kernel/callthunks.c | 4 arch/x86/kernel/cpu/amd.c | 2 arch/x86/kernel/cpu/cacheinfo.c | 21 arch/x86/kernel/cpu/common.c | 50 -- arch/x86/kernel/cpu/topology.c | 3 arch/x86/kernel/head_32.S | 14 arch/x86/kernel/head_64.S | 87 +++ arch/x86/kernel/sev.c | 2 arch/x86/kernel/smp.c | 3 arch/x86/kernel/smpboot.c | 526 ++++++++---------------- arch/x86/kernel/topology.c | 98 ---- arch/x86/kernel/tsc.c | 20 arch/x86/kernel/tsc_sync.c | 36 - arch/x86/power/cpu.c | 37 - arch/x86/realmode/init.c | 3 arch/x86/realmode/rm/trampoline_64.S | 27 + arch/x86/xen/enlighten_hvm.c | 11 arch/x86/xen/smp_hvm.c | 16 arch/x86/xen/smp_pv.c | 56 +- drivers/acpi/processor_idle.c | 4 include/linux/cpu.h | 4 include/linux/cpuhotplug.h | 17 kernel/cpu.c | 396 +++++++++++++++++- kernel/smp.c | 2 kernel/smpboot.c | 163 ------- 62 files changed, 934 insertions(+), 982 deletions(-)
