]> git.ipfire.org Git - thirdparty/linux.git/commit
Revert "drm/xe/display: use xe->display to decide whether to do anything"
authorVivek Kasireddy <vivek.kasireddy@intel.com>
Thu, 5 Jun 2025 05:42:47 +0000 (22:42 -0700)
committerJani Nikula <jani.nikula@intel.com>
Wed, 11 Jun 2025 10:31:33 +0000 (13:31 +0300)
commit5f105b2e49f71a98bf1d5b426ddc6f525a942b6b
treeeae65af719b96192fa7db0fba1b7dd067f65f653
parent188bdfb77615ab14da49bf1438ab3a1413da9898
Revert "drm/xe/display: use xe->display to decide whether to do anything"

This reverts commit 5a9f299f956ef9764f56044cfca7aafa23cea1d1.

The following crash/regression was seen with the reverted commit
on a specific BMG SKU with no display capabilities:

[  115.582833] BUG: kernel NULL pointer dereference, address: 00000000000005d0
[  115.589775] #PF: supervisor write access in kernel mode
[  115.594976] #PF: error_code(0x0002) - not-present page
[  115.600088] PGD 0 P4D 0
[  115.602617] Oops: Oops: 0002 [#1] SMP
[  115.606267] CPU: 14 UID: 0 PID: 1547 Comm: kworker/14:3 Tainted: G  U      E       6.15.0-local+ #62 PREEMPT(voluntary)
[  115.617332] Tainted: [U]=USER, [E]=UNSIGNED_MODULE
[  115.622100] Hardware name: Intel Corporation Meteor Lake Client Platform/MTL-P DDR5 SODIMM SBS RVP, BIOS MTLPEMI1.R00.3471.D49.2401260852 01/26/2024
[  115.635314] Workqueue: pm pm_runtime_work
[  115.639309] RIP: 0010:_raw_spin_lock+0x17/0x30
[  115.662382] RSP: 0018:ffffd13f82e7bc30 EFLAGS: 00010246
[  115.667581] RAX: 0000000000000000 RBX: ffff8be919076000 RCX: 0000000000000002
[  115.674675] RDX: 0000000000000001 RSI: 000000000000004b RDI: 00000000000005d0
[  115.681775] RBP: ffffd13f82e7bc60 R08: ffffd13f82e7bb00 R09: ffff8beb0c1b06c0
[  115.688869] R10: ffff8be7c034f4c0 R11: fefefefefefefeff R12: fffffffffffffff0
[  115.695965] R13: ffff8be9190762e8 R14: ffff8be919077798 R15: 00000000000005d0
[  115.703062] FS:  0000000000000000(0000) GS:ffff8beb552b6000(0000) knlGS:0000000000000000
[  115.711106] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  115.716826] CR2: 00000000000005d0 CR3: 000000024c68d002 CR4: 0000000000f72ef0
[  115.723921] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  115.731015] DR3: 0000000000000000 DR6: 00000000ffff07f0 DR7: 0000000000000400
[  115.738113] PKRU: 55555554
[  115.740816] Call Trace:
[  115.743258]  <TASK>
[  115.745363]  ? xe_display_flush_cleanup_work+0x92/0x120 [xe]
[  115.751102]  xe_display_pm_runtime_suspend+0x42/0x80 [xe]
[  115.756542]  xe_pm_runtime_suspend+0x11b/0x1b0 [xe]
[  115.761463]  xe_pci_runtime_suspend+0x23/0xd0 [xe]
[  115.766291]  pci_pm_runtime_suspend+0x6b/0x1a0
[  115.770717]  ? pci_pm_thaw_noirq+0xa0/0xa0
[  115.774797]  __rpm_callback+0x48/0x1e0
[  115.778531]  ? pci_pm_thaw_noirq+0xa0/0xa0
[  115.782614]  rpm_callback+0x66/0x70
[  115.786090]  ? pci_pm_thaw_noirq+0xa0/0xa0
[  115.790173]  rpm_suspend+0xe1/0x5e0
[  115.793647]  ? psi_task_switch+0xb8/0x200
[  115.797643]  ? finish_task_switch.isra.0+0x8d/0x270
[  115.802502]  pm_runtime_work+0xa6/0xc0
[  115.806238]  process_one_work+0x186/0x350
[  115.810234]  worker_thread+0x33a/0x480
[  115.813968]  ? process_one_work+0x350/0x350
[  115.818132]  kthread+0x10c/0x220
[  115.821350]  ? kthreads_online_cpu+0x120/0x120
[  115.825774]  ret_from_fork+0x3a/0x60
[  115.829339]  ? kthreads_online_cpu+0x120/0x120
[  115.833768]  ret_from_fork_asm+0x11/0x20
[  115.829339]  ? kthreads_online_cpu+0x120/0x120
[  115.833768]  ret_from_fork_asm+0x11/0x20
[  115.837680]  </TASK>
[  115.839907]  acpi_tad(E) drm(E)
[  115.931629] CR2: 00000000000005d0
[  115.934935] ---[ end trace 0000000000000000 ]---
[  115.939531] RIP: 0010:_raw_spin_lock+0x17/0x30

We cannot yet use xe->display to determine whether display hardware
has been successfully probed/initialized or not. This is because
xe->display would not be set to NULL even with GPUs with no display
capabilities (e.g, GMD_ID_DISPLAY = 0). However, this might change
in the future as Xe and i915 code is unified to deal with no display
cases.

Therefore, for now we have to continue to rely on xe->info.probe_display
(which would be set to false with display-less GPUs) to decide
whether to invoke any display related functions or not.

Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://lore.kernel.org/r/20250605054247.386633-1-vivek.kasireddy@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
drivers/gpu/drm/xe/display/xe_display.c