[PATCH v7] drm/xe: remove GuC reload in D3Hot path
Riana Tauro
riana.tauro at intel.com
Thu Oct 3 05:31:06 UTC 2024
Hi Matthew, Rodrigo
Sorry for the delay in response
On 8/15/2024 9:35 PM, Matthew Brost wrote:
> On Thu, Aug 15, 2024 at 11:32:50AM -0400, Rodrigo Vivi wrote:
>> On Thu, Aug 15, 2024 at 03:39:47AM +0000, Matthew Brost wrote:
>>> On Wed, Aug 14, 2024 at 04:29:19PM +0530, Riana Tauro wrote:
>>>> Currently GuC is reloaded for both runtime resume and system resume.
>>>> For D3hot <-> D0 transitions no power is lost during suspend so GuC reload
>>>> is not necessary.
>>>>
>>>> Remove GuC reload from D3Hot path and only enable/disable CTB
>>>> communication.
>>>>
>>>> v2: rebase
>>>>
>>>> v3: fix commit message
>>>> add kernel-doc for gt suspend and resume methods
>>>> fix comment
>>>> do not split register and enable calls of CT (Michal)
>>>>
>>>> v4: fix commit message
>>>> fix comment (Karthik)
>>>> split patches
>>>> correct kernel-doc (Rodrigo)
>>>>
>>>> v5: do not expose internal function of CT layer (Michal)
>>>> remove wait for outstanding g2h as it will be always zero,
>>>> use assert instead (Matthew Brost)
>>>> use runtime suspend and runtime resume pair for CT layer
>>>> (Michal / Matthew Brost)
>>>>
>>>> v6: use xe_gt_WARN_ON instead of xe_gt_assert (Michal)
>>>> assert and queue handler if g2h head and tail are
>>>> not equal (Matthew Brost)
>>>>
>>>> v7: split functions into runtime suspend and resume (Lucas, Rodrigo)
>>>> move enable irq code to ct runtime resume (Michal)
>>>>
>>>
>>> I know this has been a lot of revs, moving target, a bit conflicting
>>> feedback but IMO looks really good and ready to merge. The CI failures
>>> are a bit concerning though, can you hit retest button or send a rev8?
>>> Hard to tell if noise or genuine issues.
>>
>> Looking to all the test results for all the reviews I'm honestly
>> concerned with this change. It looks to break exactly in tests
>> that exercise suspend and resume with some strange kernel warnings
>> right after it resumed...
>>
>
> Concerned about all of CI too. Display is all foreign is all to me, so
> no idea what is happening their but from my PoV TLB timeouts in CI are
> also scary too. If the device goes into suspend prematurely TLB timeouts
> could occur. I thought I changed the PM [1] [2] for this to be
> impossible but might have screwed that up. So agree all of these failures
> to need to be investigated.
>
We tried to reproduce the issue locally but couldn't, will send out a
new rev to check if tlb invalidation failures are seen.
Thanks,
Riana Tauro
> Matt
>
> [1] https://patchwork.freedesktop.org/series/136256/
> [2] https://patchwork.freedesktop.org/series/136226/
>
>>>
>>> Matt
>>>
>>>> Signed-off-by: Riana Tauro <riana.tauro at intel.com>
>>>> ---
>>>> drivers/gpu/drm/xe/xe_gt.c | 76 ++++++++++++++++++++++++++++++++++
>>>> drivers/gpu/drm/xe/xe_gt.h | 2 +
>>>> drivers/gpu/drm/xe/xe_guc.c | 45 ++++++++++++++++++++
>>>> drivers/gpu/drm/xe/xe_guc.h | 2 +
>>>> drivers/gpu/drm/xe/xe_guc_ct.c | 36 ++++++++++++++++
>>>> drivers/gpu/drm/xe/xe_guc_ct.h | 3 ++
>>>> drivers/gpu/drm/xe/xe_pm.c | 9 ++--
>>>> drivers/gpu/drm/xe/xe_uc.c | 30 ++++++++++++++
>>>> drivers/gpu/drm/xe/xe_uc.h | 2 +
>>>> 9 files changed, 202 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
>>>> index 58895ed22f6e..466908c91f44 100644
>>>> --- a/drivers/gpu/drm/xe/xe_gt.c
>>>> +++ b/drivers/gpu/drm/xe/xe_gt.c
>>>> @@ -831,6 +831,12 @@ void xe_gt_suspend_prepare(struct xe_gt *gt)
>>>> XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
>>>> }
>>>>
>>>> +/**
>>>> + * xe_gt_suspend - GT suspend helper
>>>> + * @gt: GT object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> int xe_gt_suspend(struct xe_gt *gt)
>>>> {
>>>> int err;
>>>> @@ -881,6 +887,12 @@ int xe_gt_sanitize_freq(struct xe_gt *gt)
>>>> return ret;
>>>> }
>>>>
>>>> +/**
>>>> + * xe_gt_resume - GT resume helper
>>>> + * @gt: GT object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> int xe_gt_resume(struct xe_gt *gt)
>>>> {
>>>> int err;
>>>> @@ -909,6 +921,70 @@ int xe_gt_resume(struct xe_gt *gt)
>>>> return err;
>>>> }
>>>>
>>>> +/**
>>>> + * xe_gt_runtime_suspend - GT runtime suspend helper
>>>> + * @gt: GT object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> +int xe_gt_runtime_suspend(struct xe_gt *gt)
>>>> +{
>>>> + struct xe_device *xe = gt_to_xe(gt);
>>>> + int ret = 0;
>>>> +
>>>> + if (xe->d3cold.allowed)
>>>> + return xe_gt_suspend(gt);
>>>> +
>>>> + xe_gt_dbg(gt, "suspending\n");
>>>> + xe_gt_sanitize(gt);
>>>> +
>>>> + ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
>>>> + if (ret)
>>>> + goto err_force_wake;
>>>> +
>>>> + ret = xe_uc_runtime_suspend(>->uc);
>>>> +
>>>> +err_force_wake:
>>>> + XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
>>>> + if (ret)
>>>> + xe_gt_err(gt, "runtime suspend failed (%pe)\n", ERR_PTR(ret));
>>>> + else
>>>> + xe_gt_dbg(gt, "suspended\n");
>>>> +
>>>> + return ret;
>>>> +}
>>>> +
>>>> +/**
>>>> + * xe_gt_runtime_resume - GT runtime resume helper
>>>> + * @gt: GT object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> +int xe_gt_runtime_resume(struct xe_gt *gt)
>>>> +{
>>>> + struct xe_device *xe = gt_to_xe(gt);
>>>> + int ret = 0;
>>>> +
>>>> + if (xe->d3cold.allowed)
>>>> + return xe_gt_resume(gt);
>>>> +
>>>> + xe_gt_dbg(gt, "resuming\n");
>>>> + ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
>>>> + if (ret)
>>>> + goto err_force_wake;
>>>> +
>>>> + ret = xe_uc_runtime_resume(>->uc);
>>>> +
>>>> +err_force_wake:
>>>> + XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
>>>> + if (ret)
>>>> + xe_gt_err(gt, "runtime resume failed (%pe)\n", ERR_PTR(ret));
>>>> + else
>>>> + xe_gt_dbg(gt, "resumed\n");
>>>> +
>>>> + return ret;
>>>> +}
>>>> +
>>>> struct xe_hw_engine *xe_gt_hw_engine(struct xe_gt *gt,
>>>> enum xe_engine_class class,
>>>> u16 instance, bool logical)
>>>> diff --git a/drivers/gpu/drm/xe/xe_gt.h b/drivers/gpu/drm/xe/xe_gt.h
>>>> index 8b1a5027dcf2..402322ab8096 100644
>>>> --- a/drivers/gpu/drm/xe/xe_gt.h
>>>> +++ b/drivers/gpu/drm/xe/xe_gt.h
>>>> @@ -55,6 +55,8 @@ void xe_gt_record_user_engines(struct xe_gt *gt);
>>>> void xe_gt_suspend_prepare(struct xe_gt *gt);
>>>> int xe_gt_suspend(struct xe_gt *gt);
>>>> int xe_gt_resume(struct xe_gt *gt);
>>>> +int xe_gt_runtime_suspend(struct xe_gt *gt);
>>>> +int xe_gt_runtime_resume(struct xe_gt *gt);
>>>> void xe_gt_reset_async(struct xe_gt *gt);
>>>> void xe_gt_sanitize(struct xe_gt *gt);
>>>> int xe_gt_sanitize_freq(struct xe_gt *gt);
>>>> diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
>>>> index de0fe9e65746..bf365c6cea4d 100644
>>>> --- a/drivers/gpu/drm/xe/xe_guc.c
>>>> +++ b/drivers/gpu/drm/xe/xe_guc.c
>>>> @@ -1101,6 +1101,51 @@ void xe_guc_sanitize(struct xe_guc *guc)
>>>> guc->submission_state.enabled = false;
>>>> }
>>>>
>>>> +/**
>>>> + * xe_guc_runtime_suspend - GuC runtime suspend
>>>> + * @guc: GuC object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> +int xe_guc_runtime_suspend(struct xe_guc *guc)
>>>> +{
>>>> + return xe_guc_ct_runtime_suspend(&guc->ct);
>>>> +}
>>>> +
>>>> +/**
>>>> + * xe_guc_runtime_resume - GuC runtime resume
>>>> + * @guc: GuC object
>>>> + *
>>>> + * This function enables interrupts and CTB communication
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise.
>>>> + */
>>>> +int xe_guc_runtime_resume(struct xe_guc *guc)
>>>> +{
>>>> + struct xe_device *xe = guc_to_xe(guc);
>>>> + int err;
>>>> +
>>>> + /*
>>>> + * Power is not lost when in D3Hot state, hence it is not necessary
>>>> + * to reload GuC everytime. Only enable interrupts and
>>>> + * CTB communication during resume
>>>> + */
>>>> + if (IS_SRIOV_VF(xe) && xe_device_has_memirq(xe)) {
>>>> + struct xe_gt *gt = guc_to_gt(guc);
>>>> + struct xe_tile *tile = gt_to_tile(gt);
>>>> +
>>>> + err = xe_memirq_init_guc(&tile->sriov.vf.memirq, guc);
>>>> + if (err)
>>>> + return err;
>>>> + } else {
>>>> + guc_enable_irq(guc);
>>>> + }
>>>> +
>>>> + xe_guc_ct_runtime_resume(&guc->ct);
>>>> +
>>>> + return 0;
>>>> +}
>>>> +
>>>> int xe_guc_reset_prepare(struct xe_guc *guc)
>>>> {
>>>> return xe_guc_submit_reset_prepare(guc);
>>>> diff --git a/drivers/gpu/drm/xe/xe_guc.h b/drivers/gpu/drm/xe/xe_guc.h
>>>> index c3e6b51f7a09..cb062f8ad868 100644
>>>> --- a/drivers/gpu/drm/xe/xe_guc.h
>>>> +++ b/drivers/gpu/drm/xe/xe_guc.h
>>>> @@ -32,6 +32,8 @@ int xe_guc_upload(struct xe_guc *guc);
>>>> int xe_guc_min_load_for_hwconfig(struct xe_guc *guc);
>>>> int xe_guc_enable_communication(struct xe_guc *guc);
>>>> int xe_guc_suspend(struct xe_guc *guc);
>>>> +int xe_guc_runtime_suspend(struct xe_guc *guc);
>>>> +int xe_guc_runtime_resume(struct xe_guc *guc);
>>>> void xe_guc_notify(struct xe_guc *guc);
>>>> int xe_guc_auth_huc(struct xe_guc *guc, u32 rsa_addr);
>>>> int xe_guc_mmio_send(struct xe_guc *guc, const u32 *request, u32 len);
>>>> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
>>>> index beeeb120d1fc..41bdb9437634 100644
>>>> --- a/drivers/gpu/drm/xe/xe_guc_ct.c
>>>> +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
>>>> @@ -419,6 +419,42 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct)
>>>> return err;
>>>> }
>>>>
>>>> +/**
>>>> + * xe_guc_ct_runtime_resume- GuC CT runtime resume
>>>> + * @ct: the &xe_guc_ct
>>>> + *
>>>> + * Mark GuC CT as enabled on runtime resume
>>>> + */
>>>> +void xe_guc_ct_runtime_resume(struct xe_guc_ct *ct)
>>>> +{
>>>> + struct guc_ctb *g2h = &ct->ctbs.g2h;
>>>> +
>>>> + xe_guc_ct_set_state(ct, XE_GUC_CT_STATE_ENABLED);
>>>> +
>>>> + /* Assert if g2h head and tail are unequal and queue g2h handler */
>>>> + if (xe_gt_WARN_ON(ct_to_gt(ct), desc_read(ct_to_xe(ct), g2h, tail) != g2h->info.head))
>>>> + queue_work(ct->g2h_wq, &ct->g2h_worker);
>>>> +}
>>>> +
>>>> +/**
>>>> + * xe_guc_ct_runtime_suspend- GuC CT runtime suspend
>>>> + * @ct: the &xe_guc_ct
>>>> + *
>>>> + * Mark GuC CT as disabled on runtime suspend
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise
>>>> + */
>>>> +int xe_guc_ct_runtime_suspend(struct xe_guc_ct *ct)
>>>> +{
>>>> + /* Assert if there are any outstanding g2h and abort suspend */
>>>> + if (xe_gt_WARN_ON(ct_to_gt(ct), ct->g2h_outstanding))
>>>> + return -EBUSY;
>>>> +
>>>> + xe_guc_ct_disable(ct);
>>>> +
>>>> + return 0;
>>>> +}
>>>> +
>>>> static void stop_g2h_handler(struct xe_guc_ct *ct)
>>>> {
>>>> cancel_work_sync(&ct->g2h_worker);
>>>> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.h b/drivers/gpu/drm/xe/xe_guc_ct.h
>>>> index 190202fce2d0..0cf9d77feb35 100644
>>>> --- a/drivers/gpu/drm/xe/xe_guc_ct.h
>>>> +++ b/drivers/gpu/drm/xe/xe_guc_ct.h
>>>> @@ -16,6 +16,9 @@ void xe_guc_ct_disable(struct xe_guc_ct *ct);
>>>> void xe_guc_ct_stop(struct xe_guc_ct *ct);
>>>> void xe_guc_ct_fast_path(struct xe_guc_ct *ct);
>>>>
>>>> +void xe_guc_ct_runtime_resume(struct xe_guc_ct *ct);
>>>> +int xe_guc_ct_runtime_suspend(struct xe_guc_ct *ct);
>>>> +
>>>> struct xe_guc_ct_snapshot *
>>>> xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic);
>>>> void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot,
>>>> diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
>>>> index 9f3c14fd9f33..0b8e984ffb5c 100644
>>>> --- a/drivers/gpu/drm/xe/xe_pm.c
>>>> +++ b/drivers/gpu/drm/xe/xe_pm.c
>>>> @@ -374,7 +374,7 @@ int xe_pm_runtime_suspend(struct xe_device *xe)
>>>> }
>>>>
>>>> for_each_gt(gt, xe, id) {
>>>> - err = xe_gt_suspend(gt);
>>>> + err = xe_gt_runtime_suspend(gt);
>>>> if (err)
>>>> goto out;
>>>> }
>>>> @@ -427,8 +427,11 @@ int xe_pm_runtime_resume(struct xe_device *xe)
>>>>
>>>> xe_irq_resume(xe);
>>>>
>>>> - for_each_gt(gt, xe, id)
>>>> - xe_gt_resume(gt);
>>>> + for_each_gt(gt, xe, id) {
>>>> + err = xe_gt_runtime_resume(gt);
>>>> + if (err)
>>>> + goto out;
>>>> + }
>>>>
>>>> if (xe->d3cold.allowed) {
>>>> xe_display_pm_resume(xe, true);
>>>> diff --git a/drivers/gpu/drm/xe/xe_uc.c b/drivers/gpu/drm/xe/xe_uc.c
>>>> index 0d073a9987c2..7bbc8b3c556f 100644
>>>> --- a/drivers/gpu/drm/xe/xe_uc.c
>>>> +++ b/drivers/gpu/drm/xe/xe_uc.c
>>>> @@ -288,6 +288,36 @@ int xe_uc_suspend(struct xe_uc *uc)
>>>> return xe_guc_suspend(&uc->guc);
>>>> }
>>>>
>>>> +/**
>>>> + * xe_uc_runtime_suspend - uC runtime suspend
>>>> + * @uc: uC object
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise
>>>> + */
>>>> +int xe_uc_runtime_suspend(struct xe_uc *uc)
>>>> +{
>>>> + if (!xe_device_uc_enabled(uc_to_xe(uc)))
>>>> + return 0;
>>>> +
>>>> + return xe_guc_runtime_suspend(&uc->guc);
>>>> +}
>>>> +
>>>> +/**
>>>> + * xe_uc_runtime_resume - uC runtime resume
>>>> + * @uc: uC object
>>>> + *
>>>> + * Called while resuming from D3Hot
>>>> + *
>>>> + * Return: 0 on success, negative error code otherwise
>>>> + */
>>>> +int xe_uc_runtime_resume(struct xe_uc *uc)
>>>> +{
>>>> + if (!xe_device_uc_enabled(uc_to_xe(uc)))
>>>> + return 0;
>>>> +
>>>> + return xe_guc_runtime_resume(&uc->guc);
>>>> +}
>>>> +
>>>> /**
>>>> * xe_uc_remove() - Clean up the UC structures before driver removal
>>>> * @uc: the UC object
>>>> diff --git a/drivers/gpu/drm/xe/xe_uc.h b/drivers/gpu/drm/xe/xe_uc.h
>>>> index 506517c11333..0beec5efde6b 100644
>>>> --- a/drivers/gpu/drm/xe/xe_uc.h
>>>> +++ b/drivers/gpu/drm/xe/xe_uc.h
>>>> @@ -15,6 +15,8 @@ int xe_uc_init_hw(struct xe_uc *uc);
>>>> int xe_uc_fini_hw(struct xe_uc *uc);
>>>> void xe_uc_gucrc_disable(struct xe_uc *uc);
>>>> int xe_uc_reset_prepare(struct xe_uc *uc);
>>>> +int xe_uc_runtime_suspend(struct xe_uc *uc);
>>>> +int xe_uc_runtime_resume(struct xe_uc *uc);
>>>> void xe_uc_stop_prepare(struct xe_uc *uc);
>>>> void xe_uc_stop(struct xe_uc *uc);
>>>> int xe_uc_start(struct xe_uc *uc);
>>>> --
>>>> 2.40.0
>>>>
More information about the Intel-xe
mailing list