AMD General


[cid:8cf4391d-80d1-47f4-90f5-119ec89cd9e8]<https://outlook.office.com/bookwithme/user/[email protected]?anonymous&ep=owaSlotsEmailSignature>
Book time to meet with 
me<https://outlook.office.com/bookwithme/user/[email protected]?anonymous&ep=owaSlotsEmailSignature>

________________________________
From: amd-gfx <[email protected]> on behalf of 
[email protected] <[email protected]>
Sent: Monday, November 16, 2026 2:12 PM
To: [email protected] <[email protected]>
Subject: amd-gfx Digest, Vol 126, Issue 488

Send amd-gfx mailing list submissions to
        [email protected]

To subscribe or unsubscribe via the World Wide Web, visit
        https://lists.freedesktop.org/mailman/listinfo/amd-gfx
or, via email, send a message with subject or body 'help' to
        [email protected]

You can reach the person managing the list at
        [email protected]

When replying, please edit your Subject line so it is more specific
than "Re: Contents of amd-gfx digest..."


Today's Topics:

   1. Re: [PATCH 1/5] Revert "drm/amdgpu: re-enable SPM updates in
      the the IB path" ([email protected])
   2. Re: [PATCH 4/5] drm/amdgpu/gfx9.4.3: program the RLC SPM in
      RLC resume ([email protected])
   3. Re: [PATCH 5/5] drm/amdgpu: bring back the ring parameter for
      update_spm_vmid() ([email protected])
   4. [PATCH 4/5 v2] drm/amdgpu/gfx9.4.3: program the RLC SPM in
      RLC resume (Alex Deucher)
   5. [PATCH] drm/amdgpu: Disable XNACK retry on GFX12.1 (Kent Russell)


----------------------------------------------------------------------

Message: 1
Date: Mon, 16 Nov 2026 17:08:47 +0000
From: [email protected]
To: "Alex Deucher" <[email protected]>,
        [email protected]
Subject: Re: [PATCH 1/5] Revert "drm/amdgpu: re-enable SPM updates in
        the the IB path"
Message-ID: <[email protected]>
Content-Type: text/plain; charset=utf-8

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] drm/amdgpu: missing Fixes trailer for reverted commit
  d44428868e5b
--

commit 36fa0c29e9f1b1b0ebcda0c9f28cb357bd14196e
Author: Alex Deucher <[email protected]>

> Revert "drm/amdgpu: re-enable SPM updates in the the IB path"
>
> This reverts commit d44428868e5b9f97c6216eee1b4f82aec7a6e8d2.
>
> This causes flakiness in gfx11 CI tests.  Revert.
>
> Signed-off-by: Alex Deucher <[email protected]>

[Severity: Low]
Since commit d44428868e5b ("drm/amdgpu: re-enable SPM updates in the the
IB path") carries its own Fixes: tag, should this revert include:

  Fixes: d44428868e5b ("drm/amdgpu: re-enable SPM updates in the the IB path")

so stable and backport trees do not pick up the original commit without
this revert?

--
Sashiko AI review ? 
https://sashiko.dev/#/patchset/[email protected]?part=1


------------------------------

Message: 2
Date: Mon, 16 Nov 2026 17:20:11 +0000
From: [email protected]
To: "Alex Deucher" <[email protected]>,
        [email protected]
Subject: Re: [PATCH 4/5] drm/amdgpu/gfx9.4.3: program the RLC SPM in
        RLC resume
Message-ID: <[email protected]>
Content-Type: text/plain; charset=utf-8

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [High] drm/amdgpu: gfx_v9_4_3_xcc_rlc_resume() passes 0 instead of
  xcc_id
- [Medium] drm/amdgpu: moving update_spm_vmid() to
  gfx_v9_4_3_xcc_rlc_resume() skips VFs
--

commit 754d99a96da547c9e401d40cfbeb309d0d3aa36f
Author: Alex Deucher <[email protected]>

drm/amdgpu/gfx9.4.3: program the RLC SPM in RLC resume

This commit moves the RLC SPM VMID programming from gfx_v9_4_3_rlc_init()
to gfx_v9_4_3_xcc_rlc_resume() so that the register is reprogrammed after
suspend/resume and GPU reset.

> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c 
> b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> index da031caa4e9e6..151336e6094f7 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> @@ -1395,10 +1395,6 @@ static void 
> gfx_v9_4_3_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
>
>  static int gfx_v9_4_3_rlc_init(struct amdgpu_device *adev)
>  {
> -     /* init spm vmid with 0xf */
> -     if (adev->gfx.rlc.funcs->update_spm_vmid)
> -             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, 
> adev->vm_manager.spm_vmid);
> -
>        return 0;
>  }
>
> @@ -1588,6 +1584,10 @@ static int gfx_v9_4_3_xcc_rlc_resume(struct 
> amdgpu_device *adev, int xcc_id)
>        gfx_v9_4_3_xcc_init_pg(adev, xcc_id);
>        amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
>
> +     /* init spm vmid with 0xf */
> +     if (adev->gfx.rlc.funcs->update_spm_vmid)
> +             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, 
> adev->vm_manager.spm_vmid);
> +
>        return 0;
>  }

[Severity: High]
Should update_spm_vmid() pass xcc_id instead of 0 here?

gfx_v9_4_3_rlc_resume() and gfx_v9_4_3_xcp_resume() call
gfx_v9_4_3_xcc_rlc_resume() for each XCC instance, and
gfx_v9_4_3_update_spm_vmid() uses the instance argument with
GET_INST(GC, inst) to select which XCC's regRLC_SPM_MC_CNTL register to
write.  Passing 0 programs XCC 0 on every iteration and leaves
regRLC_SPM_MC_CNTL on all other XCCs at its hardware reset default of
VMID 0.

[Severity: Medium]
Does moving update_spm_vmid() into gfx_v9_4_3_xcc_rlc_resume() skip SPM
VMID initialization on SR-IOV VFs?

Both callers of gfx_v9_4_3_xcc_rlc_resume() (gfx_v9_4_3_rlc_resume() and
gfx_v9_4_3_xcp_resume()) skip calling it when amdgpu_sriov_vf(adev) is
true:

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c:gfx_v9_4_3_rlc_resume() {
        if (amdgpu_sriov_vf(adev))
                return 0;

        num_xcc = NUM_XCC(adev->gfx.xcc_mask);
        for (i = 0; i < num_xcc; i++) {
                r = gfx_v9_4_3_xcc_rlc_resume(adev, i);
    ...
}

Unlike gfx_v9_4_3_rlc_init(), which ran during sw_init on VFs as well,
gfx_v9_4_3_update_spm_vmid() is no longer reached on SR-IOV VFs, leaving
regRLC_SPM_MC_CNTL unprogrammed in SR-IOV one-VF mode despite the
amdgpu_sriov_is_pp_one_vf(adev) handling in gfx_v9_4_3_update_spm_vmid().

--
Sashiko AI review ? 
https://sashiko.dev/#/patchset/[email protected]?part=4


------------------------------

Message: 3
Date: Mon, 16 Nov 2026 17:26:33 +0000
From: [email protected]
To: "Alex Deucher" <[email protected]>,
        [email protected]
Subject: Re: [PATCH 5/5] drm/amdgpu: bring back the ring parameter for
        update_spm_vmid()
Message-ID: <[email protected]>
Content-Type: text/plain; charset=utf-8

Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [High] drm/amdgpu: hardcoded XCC 0 in gfx_v9_4_3_xcc_rlc_resume()
  skips XCCs 1..N-1
- [Medium] drm/amdgpu: gfx9 compute emit_frame_size exceeds max_dw in
  amdgpu_ring_init()
- [Medium] drm/amdgpu: update_spm_vmid() skipped on SR-IOV VFs in
  gfx_v8_0 and gfx_v9_4_3
--

commit 08b45306962b59cde3c232193cc26534f308aaa2
Author: Alex Deucher <[email protected]>

drm/amdgpu: bring back the ring parameter for update_spm_vmid()

This commit restores the ring parameter to update_spm_vmid() so that GFX9,
GFX10, GFX11, and GFX12.0 emit the SPM VMID update on the ring during IB
scheduling while skipping MMIO updates when a ring is provided.

[ ... ]

> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c 
> b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> index df935827b002a..e818280139791 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> @@ -4084,7 +4084,8 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device 
> *adev)
>
>        /* init spm vmid with 0xf */
>        if (adev->gfx.rlc.funcs->update_spm_vmid)
> -             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, 
> adev->vm_manager.spm_vmid);
> +             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, NULL, 
> adev->vm_manager.spm_vmid);
> +
>        return 0;
>  }
>
> @@ -5483,10 +5484,13 @@ static void gfx_v8_0_unset_safe_mode(struct 
> amdgpu_device *adev, int xcc_id)
>  }
>
>  static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, int xcc_id,
> -             unsigned int vmid)
> +             struct amdgpu_ring *ring, unsigned vmid)
>  {
>        u32 data;
>
> +     if (ring)
> +             return;
> +
>        amdgpu_gfx_off_ctrl(adev, false);
>
>        if (amdgpu_sriov_is_pp_one_vf(adev))

[Severity: Medium]
Does returning early when ring is non-NULL leave mmRLC_SPM_VMID unprogrammed
on SR-IOV VFs in one-VF mode?

gfx_v8_0_rlc_resume() returns early when amdgpu_sriov_vf(adev) is true before
reaching the update_spm_vmid(adev, 0, NULL, ...) call:

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c:gfx_v8_0_rlc_resume() {
        if (amdgpu_sriov_vf(adev)) {
                gfx_v8_0_init_csb(adev);
                return 0;
        }
    ...
}

With amdgpu_ib_schedule() passing a non-NULL ring, gfx_v8_0_update_spm_vmid()
now returns before reaching the amdgpu_sriov_is_pp_one_vf(adev) branch, so
mmRLC_SPM_VMID is never programmed on SR-IOV VFs.  The same issue affects
gfx_v9_4_3_update_spm_vmid(), because both gfx_v9_4_3_rlc_resume() and
gfx_v9_4_3_xcp_resume() skip calling gfx_v9_4_3_xcc_rlc_resume() when
amdgpu_sriov_vf(adev) is true.

> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c 
> b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> index 2e2673a6dbbf5..be45c10b6f579 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

[ ... ]

> @@ -7663,6 +7675,7 @@ static const struct amdgpu_ring_funcs 
> gfx_v9_0_ring_funcs_compute = {
>        .get_wptr = gfx_v9_0_ring_get_wptr_compute,
>        .set_wptr = gfx_v9_0_ring_set_wptr_compute,
>        .emit_frame_size =
> +             5 + /* update_spm_vmid */
>                20 + /* gfx_v9_0_ring_emit_gds_switch */
>                7 + /* gfx_v9_0_ring_emit_hdp_flush */
>                5 + /* hdp invalidate */

[Severity: Medium]
Does adding 5 dwords here trigger WARN_ON(max_ibs_dw > max_dw) in
amdgpu_ring_init() for every GFX9 compute ring?

This increases gfx_v9_0_ring_funcs_compute.emit_frame_size from 145 to 150.
In amdgpu_ring_init(), max_ibs_dw is 150 + 125 * 7 = 1025, which rounds up to
1280 with align_mask (0xff), exceeding the max_dw of 1024 passed by
gfx_v9_0_compute_ring_init().

> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c 
> b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> index 151336e6094f7..f28ffe8a5ac55 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
> @@ -1586,7 +1586,7 @@ static int gfx_v9_4_3_xcc_rlc_resume(struct 
> amdgpu_device *adev, int xcc_id)
>
>        /* init spm vmid with 0xf */
>        if (adev->gfx.rlc.funcs->update_spm_vmid)
> -             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, 
> adev->vm_manager.spm_vmid);
> +             adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, NULL, 
> adev->vm_manager.spm_vmid);
>
>        return 0;
>  }
> @@ -1609,10 +1609,13 @@ static int gfx_v9_4_3_rlc_resume(struct amdgpu_device 
> *adev)
>  }
>
>  static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev,
> -                                           int inst, unsigned int vmid)
> +                                           int inst, struct amdgpu_ring 
> *ring, unsigned int vmid)
>  {
>        u32 reg, pre_data, data;
>
> +     if (ring)
> +             return;
> +
>        reg = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regRLC_SPM_MC_CNTL);

[Severity: High]
Should gfx_v9_4_3_xcc_rlc_resume() pass xcc_id instead of 0 to
update_spm_vmid()?

Previously, amdgpu_ib_schedule() called update_spm_vmid() with ring->xcc_id on
each IB submission, programming regRLC_SPM_MC_CNTL on all XCCs.  Now that
gfx_v9_4_3_update_spm_vmid() returns early when ring is non-NULL,
gfx_v9_4_3_xcc_rlc_resume() is the only caller that programs the register, and
passing hardcoded 0 leaves XCCs 1..num_xcc-1 at their hardware reset value of
VMID 0 (the kernel VMID).

--
Sashiko AI review ? 
https://sashiko.dev/#/patchset/[email protected]?part=5


------------------------------

Message: 4
Date: Mon, 16 Nov 2026 14:10:22 -0500
From: Alex Deucher <[email protected]>
To: <[email protected]>
Cc: Alex Deucher <[email protected]>
Subject: [PATCH 4/5 v2] drm/amdgpu/gfx9.4.3: program the RLC SPM in
        RLC resume
Message-ID: <[email protected]>
Content-Type: text/plain

The register needs to be reprogrammed after a suspend and resume
or a GPU reset so program it in rlc_resume rather than rlc_init.

v2: pass xcc_id when calling

Fixes: 7c0f7ee00c7d ("drm/amdgpu: add gc v9_4_3 rlc_funcs implementation")
Signed-off-by: Alex Deucher <[email protected]>

Acked-by: Vitaly Prosyak <[email protected]>

---
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c 
b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index da031caa4e9e6..7b8ff043c99ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -1395,10 +1395,6 @@ static void gfx_v9_4_3_init_rlcg_reg_access_ctrl(struct 
amdgpu_device *adev)

 static int gfx_v9_4_3_rlc_init(struct amdgpu_device *adev)
 {
-       /* init spm vmid with 0xf */
-       if (adev->gfx.rlc.funcs->update_spm_vmid)
-               adev->gfx.rlc.funcs->update_spm_vmid(adev, 0, 
adev->vm_manager.spm_vmid);
-
         return 0;
 }

@@ -1588,6 +1584,10 @@ static int gfx_v9_4_3_xcc_rlc_resume(struct 
amdgpu_device *adev, int xcc_id)
         gfx_v9_4_3_xcc_init_pg(adev, xcc_id);
         amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);

+       /* init spm vmid with 0xf */
+       if (adev->gfx.rlc.funcs->update_spm_vmid)
+               adev->gfx.rlc.funcs->update_spm_vmid(adev, xcc_id, 
adev->vm_manager.spm_vmid);
+
         return 0;
 }

--
2.56.1



------------------------------

Message: 5
Date: Mon, 16 Nov 2026 14:11:41 -0500
From: Kent Russell <[email protected]>
To: <[email protected]>
Cc: <[email protected]>, Kent Russell <[email protected]>
Subject: [PATCH] drm/amdgpu: Disable XNACK retry on GFX12.1
Message-ID: <[email protected]>
Content-Type: text/plain

This feature cannot be properly supported, so disable it both on the
amdgpu side and the amdkfd side.

Signed-off-by: Kent Russell <[email protected]>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c        |  7 +++++++
 drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c         |  5 ++---
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c       | 12 ++++++++++++
 .../amdkfd/kfd_device_queue_manager_v12_1.c    | 18 ++++--------------
 drivers/gpu/drm/amd/amdkfd/kfd_process.c       |  7 +++++++
 5 files changed, 32 insertions(+), 17 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 08d3ff61edcc..d0ede277aa85 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -1151,6 +1151,13 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
                 gmc->noretry = 1;
         else
                 gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : 
amdgpu_noretry;
+
+       /* XNACK cannot be enabled on GFX12.1, so keep noretry=1 */
+       if (gc_ver == IP_VERSION(12, 1, 0)) {
+               if (amdgpu_noretry == 0)
+                       dev_warn(adev->dev, "Unable to set noretry = 0 on 
GFX12.1. Keeping noretry = 1");
+               gmc->noretry = 1;
+       }
 }

 void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c 
b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c
index 1f57d0fa2c04..034cc3fd07f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c
@@ -1518,9 +1518,8 @@ static u32 gfx_v12_1_get_sh_mem_config(struct 
amdgpu_device *adev)
 {
         u32 sh_mem_config = DEFAULT_SH_MEM_CONFIG;

-       /* GC 12.1.0 A0 requires RETRY_DISABLE in SH_MEM_CONFIG. */
-       if (SOC_V1_0_DIE_REV_XCD(adev->rev_id) == 0)
-               sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
+       /* XNACK cannot be enabled for GFX12.1 */
+       sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;

         return sh_mem_config;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 3d2136adb62b..3b62b9e5ecda 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1766,6 +1766,8 @@ static int kfd_ioctl_set_xnack_mode(struct file *filep,

         mutex_lock(&p->mutex);
         if (args->xnack_enabled >= 0) {
+               int i;
+
                 if (!list_empty(&p->pqm.queues)) {
                         pr_debug("Process has user queues running\n");
                         r = -EBUSY;
@@ -1780,6 +1782,16 @@ static int kfd_ioctl_set_xnack_mode(struct file *filep,
                         goto out_unlock;
                 }

+               /* XNACK cannot be enabled for gfx12.1 */
+               for (i = 0; i < p->n_pdds; i++) {
+                       struct kfd_node *dev = p->pdds[i]->dev;
+
+                       if (KFD_GC_VERSION(dev) == IP_VERSION(12, 1, 0)) {
+                               r = -EPERM;
+                               goto out_unlock;
+                       }
+               }
+
                 r = svm_range_switch_xnack_reserve_mem(p, args->xnack_enabled);
         } else {
                 args->xnack_enabled = p->xnack_enabled;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12_1.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12_1.c
index 525989371378..63ba13a60b98 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12_1.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12_1.c
@@ -54,7 +54,6 @@ static int update_qpd_v12_1(struct device_queue_manager *dqm,
         struct kfd_process_device *pdd;
         struct amdgpu_device *adev = dqm->dev->adev;
         struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
-       bool xnack_enabled;

         pdd = qpd_to_pdd(qpd);
         qpd->vm_cntx_cntl = hub->vm_cntx_cntl;
@@ -72,19 +71,10 @@ static int update_qpd_v12_1(struct device_queue_manager 
*dqm,
                 qpd->sh_mem_ape1_base = 0;
         }

-       xnack_enabled = KFD_SUPPORT_XNACK_PER_PROCESS(dqm->dev) ?
-                       pdd->process->xnack_enabled :
-                       !pdd->dev->kfd->noretry;
-
-       if (!xnack_enabled) {
-               qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
-               qpd->vm_cntx_cntl &=
-                       ~(1 << 
GCVM_CONTEXT0_CNTL__RETRY_PERMISSION_OR_INVALID_PAGE_FAULT__SHIFT);
-       } else {
-               qpd->sh_mem_config &= ~(1 << 
SH_MEM_CONFIG__RETRY_DISABLE__SHIFT);
-               qpd->vm_cntx_cntl |=
-                       (1 << 
GCVM_CONTEXT0_CNTL__RETRY_PERMISSION_OR_INVALID_PAGE_FAULT__SHIFT);
-       }
+       /* XNACK cannot be enabled for gfx12.1 */
+       qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
+       qpd->vm_cntx_cntl &=
+               ~(1 << 
GCVM_CONTEXT0_CNTL__RETRY_PERMISSION_OR_INVALID_PAGE_FAULT__SHIFT);

         qpd->sh_mem_bases = compute_sh_mem_bases_64bit(pdd);

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 546721b4d59a..059e22cd7f1a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1670,6 +1670,13 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool 
supported)
                     KFD_GC_VERSION(dev) < IP_VERSION(12, 1, 0))
                         return false;

+               /* XNACK cannot be enabled on GFX12.1 . While dev->kfd->noretry
+                * should propagate this requirement, adding it explicitly here
+                * will help cover any possible future reworks
+                */
+               if (KFD_GC_VERSION(dev) == IP_VERSION(12, 1, 0))
+                       return false;
+
                 if (dev->kfd->noretry)
                         return false;
         }
--
2.43.0



------------------------------

Subject: Digest Footer

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx


------------------------------

End of amd-gfx Digest, Vol 126, Issue 488
*****************************************

Reply via email to