Hi Greg,
On Mon, Sep 20, 2021 at 10:56:55AM +0200, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
>
> The patch below does not apply to the 5.14-stable tree.
> If someone wants it applied there, or to any other stable or longterm
> tree, then please email the backport, including the original git commit
> id to <stable@xxxxxxxxxxxxxxx>.
There is one prerequisite patch for this patch. I can backport both to
all relevant stable branches, but I would like to know how you would
like to receive backport patches. Several git bundles to stable@?
Several two-patch series to stable@? I can also give you several tags /
branches to pull from if you like.
Thanks,
Wei.
>
> thanks,
>
> greg k-h
>
> ------------------ original commit in Linus's tree ------------------
>
> From dfb5c1e12c28e35e4d4e5bc8022b0e9d585b89a7 Mon Sep 17 00:00:00 2001
> From: Wei Liu <wei.liu@xxxxxxxxxx>
> Date: Fri, 10 Sep 2021 18:57:14 +0000
> Subject: [PATCH] x86/hyperv: remove on-stack cpumask from
> hv_send_ipi_mask_allbutself
>
> It is not a good practice to allocate a cpumask on stack, given it may
> consume up to 1 kilobytes of stack space if the kernel is configured to
> have 8192 cpus.
>
> The internal helper functions __send_ipi_mask{,_ex} need to loop over
> the provided mask anyway, so it is not too difficult to skip `self'
> there. We can thus do away with the on-stack cpumask in
> hv_send_ipi_mask_allbutself.
>
> Adjust call sites of __send_ipi_mask as needed.
>
> Reported-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> Suggested-by: Michael Kelley <mikelley@xxxxxxxxxxxxx>
> Suggested-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> Fixes: 68bb7bfb7985d ("X86/Hyper-V: Enable IPI enlightenments")
> Signed-off-by: Wei Liu <wei.liu@xxxxxxxxxx>
> Reviewed-by: Michael Kelley <mikelley@xxxxxxxxxxxxx>
> Link: https://lore.kernel.org/r/20210910185714.299411-3-wei.liu@xxxxxxxxxx
>
> diff --git a/arch/x86/hyperv/hv_apic.c b/arch/x86/hyperv/hv_apic.c
> index 90e682a92820..32a1ad356c18 100644
> --- a/arch/x86/hyperv/hv_apic.c
> +++ b/arch/x86/hyperv/hv_apic.c
> @@ -99,7 +99,8 @@ static void hv_apic_eoi_write(u32 reg, u32 val)
> /*
> * IPI implementation on Hyper-V.
> */
> -static bool __send_ipi_mask_ex(const struct cpumask *mask, int vector)
> +static bool __send_ipi_mask_ex(const struct cpumask *mask, int vector,
> + bool exclude_self)
> {
> struct hv_send_ipi_ex **arg;
> struct hv_send_ipi_ex *ipi_arg;
> @@ -123,7 +124,10 @@ static bool __send_ipi_mask_ex(const struct cpumask *mask, int vector)
>
> if (!cpumask_equal(mask, cpu_present_mask)) {
> ipi_arg->vp_set.format = HV_GENERIC_SET_SPARSE_4K;
> - nr_bank = cpumask_to_vpset(&(ipi_arg->vp_set), mask);
> + if (exclude_self)
> + nr_bank = cpumask_to_vpset_noself(&(ipi_arg->vp_set), mask);
> + else
> + nr_bank = cpumask_to_vpset(&(ipi_arg->vp_set), mask);
> }
> if (nr_bank < 0)
> goto ipi_mask_ex_done;
> @@ -138,15 +142,25 @@ static bool __send_ipi_mask_ex(const struct cpumask *mask, int vector)
> return hv_result_success(status);
> }
>
> -static bool __send_ipi_mask(const struct cpumask *mask, int vector)
> +static bool __send_ipi_mask(const struct cpumask *mask, int vector,
> + bool exclude_self)
> {
> - int cur_cpu, vcpu;
> + int cur_cpu, vcpu, this_cpu = smp_processor_id();
> struct hv_send_ipi ipi_arg;
> u64 status;
> + unsigned int weight;
>
> trace_hyperv_send_ipi_mask(mask, vector);
>
> - if (cpumask_empty(mask))
> + weight = cpumask_weight(mask);
> +
> + /*
> + * Do nothing if
> + * 1. the mask is empty
> + * 2. the mask only contains self when exclude_self is true
> + */
> + if (weight == 0 ||
> + (exclude_self && weight == 1 && cpumask_test_cpu(this_cpu, mask)))
> return true;
>
> if (!hv_hypercall_pg)
> @@ -172,6 +186,8 @@ static bool __send_ipi_mask(const struct cpumask *mask, int vector)
> ipi_arg.cpu_mask = 0;
>
> for_each_cpu(cur_cpu, mask) {
> + if (exclude_self && cur_cpu == this_cpu)
> + continue;
> vcpu = hv_cpu_number_to_vp_number(cur_cpu);
> if (vcpu == VP_INVAL)
> return false;
> @@ -191,7 +207,7 @@ static bool __send_ipi_mask(const struct cpumask *mask, int vector)
> return hv_result_success(status);
>
> do_ex_hypercall:
> - return __send_ipi_mask_ex(mask, vector);
> + return __send_ipi_mask_ex(mask, vector, exclude_self);
> }
>
> static bool __send_ipi_one(int cpu, int vector)
> @@ -208,7 +224,7 @@ static bool __send_ipi_one(int cpu, int vector)
> return false;
>
> if (vp >= 64)
> - return __send_ipi_mask_ex(cpumask_of(cpu), vector);
> + return __send_ipi_mask_ex(cpumask_of(cpu), vector, false);
>
> status = hv_do_fast_hypercall16(HVCALL_SEND_IPI, vector, BIT_ULL(vp));
> return hv_result_success(status);
> @@ -222,20 +238,13 @@ static void hv_send_ipi(int cpu, int vector)
>
> static void hv_send_ipi_mask(const struct cpumask *mask, int vector)
> {
> - if (!__send_ipi_mask(mask, vector))
> + if (!__send_ipi_mask(mask, vector, false))
> orig_apic.send_IPI_mask(mask, vector);
> }
>
> static void hv_send_ipi_mask_allbutself(const struct cpumask *mask, int vector)
> {
> - unsigned int this_cpu = smp_processor_id();
> - struct cpumask new_mask;
> - const struct cpumask *local_mask;
> -
> - cpumask_copy(&new_mask, mask);
> - cpumask_clear_cpu(this_cpu, &new_mask);
> - local_mask = &new_mask;
> - if (!__send_ipi_mask(local_mask, vector))
> + if (!__send_ipi_mask(mask, vector, true))
> orig_apic.send_IPI_mask_allbutself(mask, vector);
> }
>
> @@ -246,7 +255,7 @@ static void hv_send_ipi_allbutself(int vector)
>
> static void hv_send_ipi_all(int vector)
> {
> - if (!__send_ipi_mask(cpu_online_mask, vector))
> + if (!__send_ipi_mask(cpu_online_mask, vector, false))
> orig_apic.send_IPI_all(vector);
> }
>
>
