ERROR: modpost: "boot_cpu_hartid" [arch/riscv/kvm/kvm.ko] undefined!
Atish Patra
atishp at atishpatra.org
Tue Nov 23 07:07:10 PST 2021
On Mon, Nov 22, 2021 at 9:33 PM Anup Patel <Anup.Patel at wdc.com> wrote:
>
> +Atish, +Palmer
>
> On 23/11/21, 12:04 AM, "Sean Christopherson" <seanjc at google.com> wrote:
>
> On Mon, Nov 22, 2021, Adam Borowski wrote:
> > Hi!
> > With CONFIG_KVM=m and CONFIG_SMP=n, the build fails with:
> > ERROR: modpost: "boot_cpu_hartid" [arch/riscv/kvm/kvm.ko] undefined!
> >
> > An obvious fix would be to change riscv_cpuid_to_hartid_mask() from a
> > static inline function to a real exported symbol
>
> Amusingly, the !SMP version is buggy as it assumes the boot/only CPU is in the
> target mask, which will not always be true, e.g. if a VM has more vCPUs than there
> are pCPUs in the system, kvm_riscv_vcpu_sbi_ecall() can send in a mask without any
> CPUs set. Even if that's somewhat of a bogus use case, arguably a common helper
> shouldn't have different semantics for SMP vs !SMP.
>
> > but I wonder: is there a point in allowing such combinations that no one will
> > test?
>
> Eh, even if there is currently no known use case, unless the fix is really nasty,
> there's no reason to actively prevent a potentially legitimate use case. But it's
> somewhat of a moot point as the easy fix is to use the SMP version for both cases.
>
> I think this would be an appropriate fix?
>
>
> From f6510acd68feb90aa3ec393031cbcde725c52420 Mon Sep 17 00:00:00 2001
> From: Sean Christopherson <seanjc at google.com>
> Date: Mon, 22 Nov 2021 10:01:15 -0800
> Subject: [PATCH] RISC-V: Use common riscv_cpuid_to_hartid_mask() for both
> SMP=y and SMP=n
>
> Use what is currently the SMP=y version of riscv_cpuid_to_hartid_mask()
> for both SMP=y and SMP=n to fix a build failure with KVM=m and SMP=n due
> to boot_cpu_hartid not being exported. This also fixes a second bug
> where the SMP=n version assumes the sole CPU in the system is in the
> incoming mask, which may not hold true in kvm_riscv_vcpu_sbi_ecall() if
> the KVM guest VM has multiple vCPUs (on a SMP=n system).
>
> Reported-by: Adam Borowski <kilobyte at angband.pl>
> Fixes: 1ef46c231df4 ("RISC-V: Implement new SBI v0.2 extensions")
> Cc: stable at vger.kernel.org
> Signed-off-by: Sean Christopherson <seanjc at google.com>
>
> Yes, this fix looks good for now.
>
> Reviewed-by: Anup Patel <anup.patel at wdc.com>
>
> As a separate series, we have to totally remove riscv_cpuid_to_hartid_mask()
> because this function is going to break for sparse HART ids. In fact, we
> should avoid using "struct cpumask" (which is limited by NR_CPUS) for
> representing set of HART ids.
Yeah. I am working on it. We need to disassociate all the instances of hartid
with NR_CPUs to support sparse hartid schemes in the future.
>
> @Sean, can you re-send this patch to the Linux RISC-V mailing list ? It
> seems this email did not land correctly for Atish and other folks.
>
> Regards,
> Anup
>
> ---
> arch/riscv/include/asm/smp.h | 10 ++--------
> arch/riscv/kernel/setup.c | 10 ++++++++++
> arch/riscv/kernel/smp.c | 10 ----------
> 3 files changed, 12 insertions(+), 18 deletions(-)
>
> diff --git a/arch/riscv/include/asm/smp.h b/arch/riscv/include/asm/smp.h
> index a7d2811f3536..62d0e6e61da8 100644
> --- a/arch/riscv/include/asm/smp.h
> +++ b/arch/riscv/include/asm/smp.h
> @@ -43,7 +43,6 @@ void arch_send_call_function_ipi_mask(struct cpumask *mask);
> void arch_send_call_function_single_ipi(int cpu);
>
> int riscv_hartid_to_cpuid(int hartid);
> -void riscv_cpuid_to_hartid_mask(const struct cpumask *in, struct cpumask *out);
>
> /* Set custom IPI operations */
> void riscv_set_ipi_ops(const struct riscv_ipi_ops *ops);
> @@ -85,13 +84,6 @@ static inline unsigned long cpuid_to_hartid_map(int cpu)
> return boot_cpu_hartid;
> }
>
> -static inline void riscv_cpuid_to_hartid_mask(const struct cpumask *in,
> - struct cpumask *out)
> -{
> - cpumask_clear(out);
> - cpumask_set_cpu(boot_cpu_hartid, out);
> -}
> -
> static inline void riscv_set_ipi_ops(const struct riscv_ipi_ops *ops)
> {
> }
> @@ -102,6 +94,8 @@ static inline void riscv_clear_ipi(void)
>
> #endif /* CONFIG_SMP */
>
> +void riscv_cpuid_to_hartid_mask(const struct cpumask *in, struct cpumask *out);
> +
> #if defined(CONFIG_HOTPLUG_CPU) && (CONFIG_SMP)
> bool cpu_has_hotplug(unsigned int cpu);
> #else
> diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
> index b9620e5f00ba..6c5caf5eb906 100644
> --- a/arch/riscv/kernel/setup.c
> +++ b/arch/riscv/kernel/setup.c
> @@ -59,6 +59,16 @@ atomic_t hart_lottery __section(".sdata")
> unsigned long boot_cpu_hartid;
> static DEFINE_PER_CPU(struct cpu, cpu_devices);
>
> +void riscv_cpuid_to_hartid_mask(const struct cpumask *in, struct cpumask *out)
> +{
> + int cpu;
> +
> + cpumask_clear(out);
> + for_each_cpu(cpu, in)
> + cpumask_set_cpu(cpuid_to_hartid_map(cpu), out);
> +}
> +EXPORT_SYMBOL_GPL(riscv_cpuid_to_hartid_mask);
> +
> /*
> * Place kernel memory regions on the resource tree so that
> * kexec-tools can retrieve them from /proc/iomem. While there
> diff --git a/arch/riscv/kernel/smp.c b/arch/riscv/kernel/smp.c
> index 921d9d7df400..d0147294691d 100644
> --- a/arch/riscv/kernel/smp.c
> +++ b/arch/riscv/kernel/smp.c
> @@ -59,16 +59,6 @@ int riscv_hartid_to_cpuid(int hartid)
> return -ENOENT;
> }
>
> -void riscv_cpuid_to_hartid_mask(const struct cpumask *in, struct cpumask *out)
> -{
> - int cpu;
> -
> - cpumask_clear(out);
> - for_each_cpu(cpu, in)
> - cpumask_set_cpu(cpuid_to_hartid_map(cpu), out);
> -}
> -EXPORT_SYMBOL_GPL(riscv_cpuid_to_hartid_mask);
> -
> bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
> {
> return phys_id == cpuid_to_hartid_map(cpu);
> --
>
--
Regards,
Atish
More information about the kvm-riscv
mailing list