RE: [PATCH] mshv: Do pre-commit before mapping a guest VM GPA
From: Michael Kelley
Date: Sat Nov 21 2026 - 23:25:08 EST
From: Mukesh R <mrathor@xxxxxxxxxxxxxxxxxxx> Sent: Friday, November 6, 2026 6:49 PM
>
> Pre commit reduces TLB flush overheads in the hypervisor. On systems
> with large number of CPUs, there is signficant advantage to doing
> this. Pre-commit allows the hypervisor to build all data structs needed
> in anticipation of the map gpa hypercall, thus reducing or eliminating
> deposit page returns from the map gpa hypercall, and as a result reducing
> TLB flushes. In some performance benchmarks, it made the maps of large
> ranges almost 30% faster.
>
> Signed-off-by: Mukesh R <mrathor@xxxxxxxxxxxxxxxxxxx>
> ---
> drivers/hv/mshv_root_hv_call.c | 52 ++++++++++++++++++++++++++++++++++
> include/asm-generic/mshyperv.h | 3 ++
> include/hyperv/hvgdk_mini.h | 1 +
> include/hyperv/hvhdk.h | 7 +++++
> 4 files changed, 63 insertions(+)
>
> diff --git a/drivers/hv/mshv_root_hv_call.c b/drivers/hv/mshv_root_hv_call.c
> index bf0f90515dad..5c44ba8f72b0 100644
> --- a/drivers/hv/mshv_root_hv_call.c
> +++ b/drivers/hv/mshv_root_hv_call.c
> @@ -181,6 +181,54 @@ int hv_call_delete_partition(u64 partition_id)
> return hv_result_to_errno(status);
> }
>
> +/*
> + * Pre commit to reduce TLB flush overheads in the hypervisor. On systems with
> + * large CPUs, there is signficant advantage to doing this. Pre-commit allows
> + * the hypervisor to build all data structs needed in anticipation of the map
> + * gpa hypercall, thus reducing or eliminating deposit page returns from the
> + * map gpa hypercall, and as a result reducing TLB flushes.
> + */
> +static int hv_pre_commit_gpa(u64 pt_id, u64 gfn, u64 num_pages)
> +{
> + ulong irq_flags;
> + struct hv_input_precommit_gpa_pages *input_page;
> + int rc = 0;
> + u64 status, done = 0;
> +
> + while (done < num_pages) {
> + u16 rep_count = min(num_pages - done, HV_REP_COUNT_MAX);
> +
> + local_irq_save(irq_flags);
> + input_page = *this_cpu_ptr(hyperv_pcpu_input_arg);
> + memset(input_page, 0, sizeof(*input_page));
> +
> + input_page->partition_id = pt_id;
> + input_page->target_gpa_base = gfn;
> +
> + status = hv_do_rep_hypercall(HVCALL_PRECOMMIT_GPA_PAGES,
> + rep_count, 0, input_page, NULL);
> + local_irq_restore(irq_flags);
> +
> + done += hv_repcomp(status);
> + gfn += hv_repcomp(status);
> +
> + if (hv_result_needs_memory(status)) {
> + rc = hv_deposit_memory(pt_id, status);
> + if (rc)
> + break;
> +
> + } else if (!hv_result_success(status)) {
> + pr_err("%s: failed to pre commit for gfn %#llx done %llu/%llu, status=%#llx (%s)\n",
> + __func__, gfn, done, num_pages, status,
> + hv_result_to_string(hv_result(status)));
> + rc = hv_result_to_errno(status);
> + break;
> + }
> + }
> +
> + return rc;
> +}
> +
> /* Ask the hypervisor to map guest ram pages or the guest mmio space */
> static int hv_do_map_gpa_hcall(u64 partition_id, u64 gfn, u64 page_struct_count,
> u32 flags, struct page **pages, u64 mmio_spa)
> @@ -205,6 +253,10 @@ static int hv_do_map_gpa_hcall(u64 partition_id, u64 gfn, u64 page_struct_count,
> page_count >>= large_shift;
> }
>
> + ret = hv_pre_commit_gpa(partition_id, gfn, page_struct_count);
> + if (ret)
> + return ret;
> +
The pre-commit call uses the original GFN and page struct count, so
evidently the pre-commit functionality in the hypervisor does not need
to know if the follow-on MAP_GPA_PAGES hypercall will be specifying
large pages. I don't know anything about the pre-commit functionality,
but my guess would have been that it is allocating page table pages, and
that the large page flag would let it know that only PMD level is needed.
Do you know those details? Just curious ....
> while (done < page_count) {
> ulong i, completed, remain = page_count - done;
> int rep_count = min(remain, HV_MAP_GPA_BATCH_SIZE);
> diff --git a/include/asm-generic/mshyperv.h b/include/asm-generic/mshyperv.h
> index 4c9f3353074a..a3d497d99b0b 100644
> --- a/include/asm-generic/mshyperv.h
> +++ b/include/asm-generic/mshyperv.h
> @@ -28,6 +28,9 @@
>
> #define VTPM_BASE_ADDRESS 0xfed40000
>
> +#define HV_REP_COUNT_MAX \
> + (HV_HYPERCALL_REP_COMP_MASK >> HV_HYPERCALL_REP_COMP_OFFSET)
> +
> enum hv_partition_type {
> HV_PARTITION_TYPE_GUEST,
> HV_PARTITION_TYPE_ROOT,
> diff --git a/include/hyperv/hvgdk_mini.h b/include/hyperv/hvgdk_mini.h
> index 966945f0504f..16c41ee5079a 100644
> --- a/include/hyperv/hvgdk_mini.h
> +++ b/include/hyperv/hvgdk_mini.h
> @@ -489,6 +489,7 @@ union hv_vp_assist_msr_contents { /*
> HV_REGISTER_VP_ASSIST_PAGE */
> #define HVCALL_FLUSH_GUEST_PHYSICAL_ADDRESS_LIST 0x00b0
> #define HVCALL_CREATE_DEVICE_DOMAIN 0x00b1
> #define HVCALL_ATTACH_DEVICE_DOMAIN 0x00b2
> +#define HVCALL_PRECOMMIT_GPA_PAGES 0x00be
> #define HVCALL_SIGNAL_EVENT_DIRECT 0x00c0
> #define HVCALL_POST_MESSAGE_DIRECT 0x00c1
> #define HVCALL_DISPATCH_VP 0x00c2
> diff --git a/include/hyperv/hvhdk.h b/include/hyperv/hvhdk.h
> index 0c89c62c9706..3944e51d0a2d 100644
> --- a/include/hyperv/hvhdk.h
> +++ b/include/hyperv/hvhdk.h
> @@ -953,4 +953,11 @@ struct hv_input_modify_sparse_spa_page_host_access {
> #define HV_MODIFY_SPA_PAGE_HOST_ACCESS_LARGE_PAGE 0x4
> #define HV_MODIFY_SPA_PAGE_HOST_ACCESS_HUGE_PAGE 0x8
>
> +struct hv_input_precommit_gpa_pages { /*
> HV_INPUT_PRECOMMIT_GPA_PAGES */
> + u64 partition_id;
> + u32 flags;
> + u32 reserved;
> + u64 target_gpa_base;
> +} __packed;
Does Windows put this struct in hvhdk.h? Its close relative from
a functional standpoint, struct hv_input_map_gpa_pages, is in
hvhdk_mini.h, so that would seem to be a more natural place for
this. But I never have understood the criteria for being in one vs.
the other.
Michael