Skip to content

Shivtej/GitHub action backport - #15

Open
shivtejp1 wants to merge 33 commits into
intel-gpu:backport/mainfrom
shivtejp1:shivtej/github-action-backport
Open

Shivtej/GitHub action backport#15
shivtejp1 wants to merge 33 commits into
intel-gpu:backport/mainfrom
shivtejp1:shivtej/github-action-backport

Conversation

@shivtejp1

Copy link
Copy Markdown

No description provided.

kolanupa and others added 30 commits June 29, 2026 12:05
…_queue placement event addition""

This reverts commit 383649c.
Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
List of eudebug tests to be executed by regular validation runs.

v2: Add missing tests
v3: Add missing tests

Signed-off-by: Katarzyna Piecielska <katarzyna.piecielska@intel.com>
Signed-off-by: Dominik Grzegorzek <dominik.grzegorzek@intel.com>
Signed-off-by: Christoph Manszewski <christoph.manszewski@intel.com>
Signed-off-by: Maciej Patelczyk <maciej.patelczyk@intel.com>
Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
Signed-off-by: Pawel Sikora <pawel.sikora@intel.com>
Signed-off-by: Karolina Stolarek <karolina.stolarek@intel.com>
Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Add tests to block list which are expected to be skipped as the
supported kernel changes are not present.

Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
There was a change in names of caching tests. Now instead of 2
we have 8 tests. Let's fix it.

Signed-off-by: Katarzyna Piecielska <katarzyna.piecielska@intel.com>

v2: Add missing tests
v3: Add missing tests
v4: Remove unnecessary tests
v5: Rename xe-fast-feedback to xe.fast-feedback

Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
Let's move ktests results at the end of fast-feedback and remove eudebug
ktest as this is expected to skip.

Signed-off-by: Katarzyna Piecielska <katarzyna.piecielska@intel.com>

v2: Rename xe-fast-feedback to xe.fast-feedback

Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
Remove eudebug tests from the core blocklist of BMG as we have support
for eudebug.

Suggested-by: Gandi Ramadevi <ramadevi.gandi@intel.com>
Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
Update test_config file to enable LNL/BMG eudebug runs on kobuk_6.17.

Signed-off-by: Muppidi, Venkateswara Rao <venkateswara.rao.muppidi@intel.com>
Update test_config file to enable BMG multigpu runs on kobuk_6.17.

Signed-off-by: Muppidi, Venkateswara Rao <venkateswara.rao.muppidi@intel.com>
Signed-off-by: Srinivasula Reddy Pondugula <p.srinivasula.reddy@intel.com>
Add the tests to the blocklist that are expected to fail due to the absence of backport support in XKB-6.17.
Tests:
	igt@xe_live_ktest@xe_dma_buf/xe_bo/xe_migrate
	igt@xe_pat@pat-index-all/pat-sanity
	igt@xe_configfs@gt-types-allowed
	igt@xe_configfs@ctx-restore*

Signed-off-by: Muppidi, Venkateswara Rao <venkateswara.rao.muppidi@intel.com>
Update 6.17 drm-uapi headers to align with latest v6.17 kernel,
including prelim eudebug macro updates for debug interface support.

Updated eudebug macros for v6.17 compatibility

Signed-off-by: Pravalika Gurram <pravalika.gurram@intel.com>
Update preliminary uapi headers with xe3p_lpg platform support and
eudebug interface enhancements.

- Added xe3p_lpg platform definitions and feature flags
- Updated eudebug macros for Xe3p-lpg compatibility
- Also allign xe_drm.h
- Add DRM_IOCTL_XE_EXEC_QUEUE_SET_PROPERTY in documentation

Signed-off-by: Pravalika Gurram <pravalika.gurram@intel.com>
In faultable VM mode (DRM_XE_VM_CREATE_FLAG_FAULT_MODE), GPU memory
accesses go through page fault handling. When the GPU writes the user
fence timeline value back to fence_data via a page fault, the kernel
implicitly creates a page table binding for that system memory page,
generating an additional VM_BIND_OP event.

The old vm_bind_clear_test_trigger fired on ALL VM_BIND_OP CREATE
events and performed a zero-content check (v == 0) on the bound
address. When triggered on the implicit fault-driven bind of the
fence_data page, it found the previously written timeline value (e.g.
0x2) instead of zero, causing a spurious assertion failure:

  CRITICAL: Failed assertion: v == 0
  CRITICAL: error: 0x2 != 0

Fix this by tagging only the clean_bo vm_bind with a metadata cookie
(VM_BIND_CLEAR_CLEAN_COOKIE) via prelim_drm_xe_vm_bind_op_ext_attach_debug.
Move the zero-content check into vm_bind_clear_ack_trigger, which
operates on VM_BIND_UFENCE events, and gate the check behind
vm_bind_clear_is_tagged_clean_bind(). Binds without the cookie
(batch_bo, implicit fault binds, etc.) are acked immediately without
the memory content check.
Ensure proper initialization of helper variables to avoid compiler
warnings.

Fixes: igt@xe_eudebug@vm-bind-clear-faultable

Signed-off-by: Krishnaiah Bommu <krishnaiah.bommu@intel.com>
Signed-off-by: Srinivasula Reddy Pondugula <p.srinivasula.reddy@intel.com>
Signed-off-by: Srinivasula Reddy Pondugula <p.srinivasula.reddy@intel.com>
…exclusion

The kernel returns -EBUSY (not -EPERM) when eudebug lockdown is
attempted while VFs are already enabled, as xe_sriov_pf_lockdown()
uses a guard that returns -EBUSY in that case.

Update the expected return value and add a comment explaining the
relationship to xe_sriov_pf_lockdown().

Signed-off-by: Bommu Krishnaiah <krishnaiah.bommu@intel.com>
…d remove duplicate entries

Signed-off-by: Muppidi, Venkateswara Rao <venkateswara.rao.muppidi@intel.com>
Adjust xe.fast-feedback.testlist regarding EUdebug tests to have proper
coverage of basic features.

Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
…ocklist

Unblock exhaustive eviction tests from xe.bmg.core.blocklist.txt for backport/main

Signed-off-by: Venkateswara Rao Muppidi <venkateswara.rao.muppidi@intel.com>
…blocklist

unblock xe_eudebug@multigpu-basic-client* tests from xe.multigpu.blocklist.txt

Signed-off-by: Venkateswara Rao Muppidi <venkateswara.rao.muppidi@intel.com>
…ocklist

Add the following tests to the xe.bmg.core blocklist

- page reclaim
- non-msix
- gpgpu with offset-4X4
- prefetch fault
- xa-app-transient-media
- xe_eudebug_online@breakpoint-many-sessions-tiles
- xe_compute@eu-busy-10s
- xe_pat@l2-flush-opt-svm-pat-restrict

Signed-off-by: Venkateswara Rao Muppidi <venkateswara.rao.muppidi@intel.com>
Add Blocklist for xe dkms kv5.15 and add it to meson.build

Signed-off-by: Aditya Chauhan <aditya.chauhan@intel.com>
Add entries to xe_test_config.json file to ensure that blocklists are
applied while generating testlists.

Signed-off-by: Aditya Chauhan <aditya.chauhan@intel.com>
Update dkms kv5.15 and bmg core blocklist files

Signed-off-by: Aditya Chauhan <aditya.chauhan@intel.com>
…ocklist

Block EUdebug tests on SR-IOV VF/VM runs
Block wedged-mode & pat-sw-hw tests on VF/VM runs

Signed-off-by: Venkateswara Rao Muppidi <venkateswara.rao.muppidi@intel.com>
…s custom timeout

The existing xe_eudebug_client_create() is creating the client with
default timeout. But there is no way to create the client with custom
timeout. Hence adding an API named xe_eudebug_client_create_timeout()
that takes the timeout as input and creates the client with provided
timeout value.

Suggested-by: Christoph Manszewski <christoph.manszewski@intel.com>
Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Reviewed-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
…ests

Use the new API to allow setting client timeout for caching tests.

Signed-off-by: Kolanupaka Naveena <kolanupaka.naveena@intel.com>
Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski@intel.com>
smithab-intc and others added 3 commits August 6, 2026 13:08
…low path

Increase XE_EUDEBUG_DEFAULT_CACHING_TIMEOUT_SEC, raising the
effective timeout from 60 seconds to 120 seconds.

The writes-caching-* tests can exceed the current timeout in
slow-path configurations, leading to failures. Increase the
timeout to provide additional execution headroom.

Suggested-by: Maciej Patelczyk <maciej.patelczyk@intel.com>
Signed-off-by: Sanjay Yadav <sanjay.kumar.yadav@intel.com>
Signed-off-by: Smitha Balasubramanyam <smitha.balasubramanyam@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.