ERROR: modpost: "boot_cpu_hartid" [arch/riscv/kvm/kvm.ko] undefined!

[Anup Patel]

+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.

@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);
    --