Re: [PATCH v6 4/6] Drivers: hv: Mark shared memory as decrypted for CCA Realms

From: Mukesh R

Date: Tue Nov 17 2026 - 20:29:47 EST


On 11/17/26 11:57, Mukesh R wrote:
On 11/16/26 11:53, Kameron Carr wrote:
On 11/13/2026 6:59 PM, Mukesh R wrote:
On 11/12/26 16:46, Mukesh R wrote:
On 11/12/26 09:49, Kameron Carr wrote:
On 11/11/2026 6:43 PM, Mukesh R wrote:
On 11/11/26 10:01, Michael Kelley wrote:
From: Kameron Carr <kameroncarr@xxxxxxxxxxxxxxxxxxx> Sent: Friday, October 30, 2026 5:14 PM

In hv_common_cpu_init(), the per-CPU hypercall input/output pages need
to be marked as decrypted (shared) for confidential VM isolation types.
This is already done for SNP and TDX isolation; extend the same handling
to Arm CCA Realm guests so that the host hypervisor can access the
shared hypercall buffers.

is_realm_world() is declared in include/linux/arm-rsi-cmds.h which only
builds on arm64 (it includes <asm/memory.h>), so using it directly in
the arch-neutral drivers/hv/hv_common.c would break the x86 build.
Introduce a Hyper-V-specific helper following the established
hv_isolation_type_snp() / hv_isolation_type_tdx() pattern.

On architectures other than arm64 the weak default keeps the existing
behaviour.

Reviewed-by: Michael Kelley <mhklinux@xxxxxxxxxxx>
Signed-off-by: Kameron Carr <kameroncarr@xxxxxxxxxxxxxxxxxxx>
---
    arch/arm64/hyperv/mshyperv.c   |  5 +++++
    drivers/hv/hv_common.c         | 17 +++++++++++++----
    include/asm-generic/mshyperv.h |  1 +
    3 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/arch/arm64/hyperv/mshyperv.c b/arch/arm64/hyperv/mshyperv.c
index 46ab2da..da6c3aa0 100644
--- a/arch/arm64/hyperv/mshyperv.c
+++ b/arch/arm64/hyperv/mshyperv.c
@@ -182,3 +182,8 @@ bool hv_is_hyperv_initialized(void)
        return hyperv_initialized;
    }
    EXPORT_SYMBOL_GPL(hv_is_hyperv_initialized);
+
+bool hv_isolation_type_cca(void)
+{
+    return is_realm_world();
+}
diff --git a/drivers/hv/hv_common.c b/drivers/hv/hv_common.c
index 31256cb..a60c92c 100644
--- a/drivers/hv/hv_common.c
+++ b/drivers/hv/hv_common.c
@@ -476,6 +476,7 @@ int hv_common_cpu_init(unsigned int cpu)
        u64 msr_vp_index;
        gfp_t flags;
        const int pgcount = hv_output_page_exists() ? 2 : 1;
+    const size_t alloc_size = ALIGN((size_t)pgcount * HV_HYP_PAGE_SIZE, PAGE_SIZE);
        void *mem;
        int ret = 0;

@@ -489,7 +490,7 @@ int hv_common_cpu_init(unsigned int cpu)
         * online and then taken offline
         */
        if (!*inputarg) {
-        mem = kmalloc_array(pgcount, HV_HYP_PAGE_SIZE, flags);
+        mem = kmalloc(alloc_size, flags);
            if (!mem)
                return -ENOMEM;

@@ -499,14 +500,16 @@ int hv_common_cpu_init(unsigned int cpu)
            }

            if (!ms_hyperv.paravisor_present &&
-            (hv_isolation_type_snp() || hv_isolation_type_tdx())) {
-            ret = set_memory_decrypted((unsigned long)mem, pgcount);
+            (hv_isolation_type_snp() || hv_isolation_type_tdx() ||
+             hv_isolation_type_cca())) {

Could this big "if" statement use "hv_is_isolation_supported()" instead
of explicitly testing for each CoCo VM type? I ask partly because
Mukesh Rathor has just submitted a patch [1] that introduces a new
function that does something almost exactly the same as
hv_is_isolation_supported() and uses it here.

+            (hv_isolation_type_snp() || hv_isolation_type_tdx() ||
+             hv_isolation_type_cca())) {

With my change, this will become:

          (hv_isolation_hw_cvm() || hv_isolation_type_cca())) {

Any reason you wouldn't change hv_isolation_hw_cvm() to

    return hv_isolation_type_snp() || hv_isolation_type_tdx() ||
      hv_isolation_type_cca()

That would seem more logical to me.


We could, I just don't know if on cca, the repeat hypercall has same
behavior. I was leaving that for the arm team. Anyways, I sent question
to the hyp team, if they say yes, i can update my patch to include that.

Thanks,
-Mukesh


Confirming, arm has same behavior and needs the fix also. So, we can def
expand the check to include _cca.

I just sent V1, renaming hv_isolation_hw_cvm() to hv_isolation_type_hw().
If your patch goes after mine, you just need to add to your 0004*:

diff --git a/arch/arm64/hyperv/mshyperv.c b/arch/arm64/hyperv/mshyperv.c
index 4fdc26ade1d7..67a16b4d2235 100644
--- a/arch/arm64/hyperv/mshyperv.c
+++ b/arch/arm64/hyperv/mshyperv.c
@@ -134,3 +134,8 @@ bool hv_is_hyperv_initialized(void)
         return hyperv_initialized;
  }
  EXPORT_SYMBOL_GPL(hv_is_hyperv_initialized);
+
+bool hv_isolation_type_hw(void)
+{
+       return is_realm_world();
+}


OR, you could pick mine 0001-* and make it part of your series, and I
can drop it from mine and make it dependent on yours. But mine is small
enough that it has higher chances of going in first.....


This patch set has already been applied to hyperv-next, so maybe you can
rebase off that.


Ok, I'll rebase my patches from hyperv-fixes to hyperv-next, and
resubmit.

Thanks,
-Mukesh


On second thought, given these are overloaded functions, adding a new
function makes it complicated. Since it's only two places, it's just
not worth it. Punting on it for now...

Thanks,
-Mukesh