summaryrefslogtreecommitdiff
path: root/drivers/gpu
AgeCommit message (Collapse)Author
2023-08-31drm/amdkfd: use mask to get v9 interrupt sq data bits correctlyAlex Sierra
Interrupt sq data bits were not taken properly from contextid0 and contextid1. Use macro KFD_CONTEXT_ID_GET_SQ_INT_DATA instead. Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Allocate coredump memory in a nonblocking wayAndré Almeida
During a GPU reset, a normal memory reclaim could block to reclaim memory. Giving that coredump is a best effort mechanism, it shouldn't disturb the reset path. Change its memory allocation flag to a nonblocking one. Signed-off-by: André Almeida <andrealmeid@igalia.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Support query ecc cap for aqua_vanjaramHawking Zhang
Driver queries umc_info v4_0 to identify ecc cap for aqua_vanjaram Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Candice Li <candice.li@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Add umc_info v4_0 structureHawking Zhang
To be used by aqua_vanjaram Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Candice Li <candice.li@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: always switch off ODM before committing more streamsWenjing Liu
ODM power optimization is only supported with single stream. When ODM power optimization is enabled, we might not have enough free pipes for enabling other stream. So when we are committing more than 1 stream we should first switch off ODM power optimization to make room for new stream and then allocating pipe resource for the new stream. Cc: stable@vger.kernel.org Fixes: 59de751e3845 ("drm/amd/display: add ODM case when looking for first split pipe") Reviewed-by: Dillon Varone <dillon.varone@amd.com> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Wenjing Liu <wenjing.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: Remove wait while lockedGabe Teeger
[Why] We wait for mpc idle while in a locked state, leading to potential deadlock. [What] Move the wait_for_idle call to outside of HW lock. This and a call to wait_drr_doublebuffer_pending_clear are moved added to a new static helper function called wait_for_outstanding_hw_updates, to make the interface clearer. Cc: stable@vger.kernel.org Fixes: 8f0d304d21b3 ("drm/amd/display: Do not commit pipe when updating DRR") Reviewed-by: Jun Lei <jun.lei@amd.com> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Gabe Teeger <gabe.teeger@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: update blank state on ODM changesWenjing Liu
When we are dynamically adding new ODM slices, we didn't update blank state, if the pipe used by new ODM slice is previously blanked, we will continue outputting blank pixel data on that slice causing right half of the screen showing blank image. The previous fix was a temporary hack to directly update current state when committing new state. This could potentially cause hw and sw state synchronization issues and it is not permitted by dc commit design. Cc: stable@vger.kernel.org Fixes: 7fbf451e7639 ("drm/amd/display: Reinit DPG when exiting dynamic ODM") Reviewed-by: Dillon Varone <dillon.varone@amd.com> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Wenjing Liu <wenjing.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: Add smu write msg id fail retry processFudong Wang
A benchmark stress test (12-40 machines x 48hours) found that DCN315 has cases where DC writes to an indirect register to set the smu clock msg id, but when we go to read the same indirect register the returned msg id doesn't match with what we just set it to. So, to fix this retry the write until the register's value matches with the requested value. Cc: stable@vger.kernel.org # 6.1+ Fixes: f94903996140 ("drm/amd/display: Add DCN315 CLK_MGR") Reviewed-by: Charlene Liu <charlene.liu@amd.com> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Fudong Wang <fudong.wang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Add SMU v13.0.6 default reset methodsLijo Lazar
For APUs with SMU v13.0.6, mode-2 reset is kept as default and for others mode-1 is the default reset method. Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Tested-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31Partially revert "drm/amd/display: update add plane to context logic with a ↵Wenjing Liu
new algorithm" This partially reverts commit 460ea8980511 ("drm/amd/display: update add plane to context logic with a new algorithm"). The new secondary pipe allocation logic triggers an issue with a specific hardware state transition and causes a frame of corruption when toggling between windowed MPO and ODM desktop only mode. Ideally hwss is supposed to handle this scenario. We are temporarily reverting the logic and investigate the root cause why this transition would cause corruptions. Fixes: 460ea8980511 ("drm/amd/display: update add plane to context logic with a new algorithm") Reviewed-by: Martin Leung <martin.leung@amd.com> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Wenjing Liu <wenjing.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: register edp_backlight_control() for DCN301Hamza Mahfooz
As made mention of in commit 099303e9a9bd ("drm/amd/display: eDP intermittent black screen during PnP"), we need to turn off the display's backlight before powering off an eDP display. Not doing so will result in undefined behaviour according to the eDP spec. So, set DCN301's edp_backlight_control() function pointer to dce110_edp_backlight_control(). Cc: stable@vger.kernel.org Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2765 Fixes: 9c75891feef0 ("drm/amd/display: rework recent update PHY state commit") Suggested-by: Swapnil Patel <swapnil.patel@amd.com> Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Add bootloader wait for PSP v13Lijo Lazar
Implement the wait for bootloader call back for PSP v13.0 ASICs. Only for ASICs with PSP v13.0.6, it needs an additional check for VBIOS mailbox status. Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Tested-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: register a dirty framebuffer callback for fbconHamza Mahfooz
fbcon requires that we implement &drm_framebuffer_funcs.dirty. Otherwise, the framebuffer might take a while to flush (which would manifest as noticeable lag). However, we can't enable this callback for non-fbcon cases since it may cause too many atomic commits to be made at once. So, implement amdgpu_dirtyfb() and only enable it for fbcon framebuffers (we can use the "struct drm_file file" parameter in the callback to check for this since it is only NULL when called by fbcon, at least in the mainline kernel) on devices that support atomic KMS. Cc: Aurabindo Pillai <aurabindo.pillai@amd.com> Cc: Mario Limonciello <mario.limonciello@amd.com> Cc: stable@vger.kernel.org # 6.1+ Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2519 Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Updated TCP/UTCL1 programmingMangesh Gadre
Update TCP/UTCL1 thrashing control settings v2: updated rev_id check Signed-off-by: Mangesh Gadre <Mangesh.Gadre@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/pm: fix debugfs pm_info outputAlex Deucher
Print both input and avg power. Fixes: 47f1724db4fe ("drm/amd: Introduce `AMDGPU_PP_SENSOR_GPU_INPUT_POWER`") Reviewed-by: Guchun Chen <guchun.chen@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Fix the return for gpu mode1_resetHawking Zhang
amdgpu_device_mode1_reset will return gpu mode1_reset succeed (ret = 0) as long as wait_for_bootloader call succeed, regardless of the status reported by smu or psp firmware. This results to driver continue executing recovery even smu or psp fail to perform mode1 reset. Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: fix static screen detection settingSungHuai Wang
[WHY] OTG_STATIC_SCREEN_EVENT_MASK is changed in DCN3, but we still follow DCN2 to apply setting for OTG_STATIC_SCREEN_EVENT_MASK. [How] Add new function to apply correct settings for DCN3 series. Reviewed-by: Anthony Koo <anthony.koo@amd.com> Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: SungHuai Wang <danny.wang@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: fix mode scaling (RMX_.*)Hamza Mahfooz
As made mention of in commit 4a2df0d1f28e ("drm/amd/display: Fixed non-native modes not lighting up"), we shouldn't call drm_mode_set_crtcinfo() once the crtc timings have been decided. Since, it can cause settings to be unintentionally overwritten. So, since dm_state is never NULL now, we can use old_stream to determine if we should call drm_mode_set_crtcinfo() because we only need to set the crtc timing parameters for entirely new streams. Cc: Harry Wentland <harry.wentland@amd.com> Cc: Rodrigo Siqueira <rodrigo.siqueira@amd.com> Fixes: bd49f19039c1 ("drm/amd/display: Always set crtcinfo from create_stream_for_sink") Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/pm: Fix critical temp unit of SMU v13.0.6Asad Kamal
Critical Temperature needs to be reported in millidegree Celsius. Signed-off-by: Asad Kamal <asad.kamal@amd.com> Reviewed-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Remove SRAM clock gater override by driverMangesh Gadre
rlc firmware does required setting, driver need not do it. Signed-off-by: Mangesh Gadre <Mangesh.Gadre@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Add bootloader status checkLijo Lazar
Add a function to wait till bootloader has reached steady state. Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Tested-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdkfd: use correct method to get clock under SRIOVHorace Chen
[What] Current SRIOV still using adev->clock.default_XX which gets from atomfirmware. But these fields are abandoned in atomfirmware long ago. Which may cause function to return a 0 value. [How] We don't need to check whether SR-IOV. For SR-IOV one-vf-mode, pm is enabled and VF is able to read dpm clock from pmfw, so we can use dpm clock interface directly. For multi-VF mode, VF pm is disabled, so driver can just react as pm disabled. One-vf-mode is introduced from GFX9 so it shall not have any backward compatibility issue. Signed-off-by: Horace Chen <horace.chen@amd.com> Acked-by: Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Unset baco dummy mode on nbio v7.9Lijo Lazar
BACO dummy mode could be set under reset conditions and that affects framebuffer access. Check If baco dummy mode is set, unset it if so. Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Le Ma <le.ma@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Tested-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: set minimum of VBlank_nomChunTao Tso
[Why] If VBlank_nom is too small, it will cause VStartUP_Start smaller than VBackPorch + VSync width which is an invalid case for VStartUP_Start and where to send AS-SDP. [How] Setup a minimum value to VBlank_nom Reviewed-by: Reza Amini <reza.amini@amd.com> Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: ChunTao Tso <chuntao.tso@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: Correct unit conversion for vstartupReza Amini
[why] vstartup is calculated to be a large number. it works because it is within vertical blank, but it reduces region of blank that can be used for power gating. [how] Calculation needs to convert micro seconds to number of vertical lines. Reviewed-by: Kazlauskas Nicholas <Nicholas.Kazlauskas@amd.com> Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: Reza Amini <reza.amini@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: Roll back unit correctionOvidiu Bunea
[why] This Unit correction exposes a Replay corruption. [how] This reverts commit: commit dbd29029c7b5 ("drm/amd/display: Correct unit conversion for vstartup") Roll back unit conversion until Replay can fix their corruption. Fixes: dbd29029c7b5 ("drm/amd/display: Correct unit conversion for vstartup") Reviewed-by: Reza Amini <reza.amini@amd.com> Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: Ovidiu Bunea <ovidiu.bunea@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Enable ras for mp0 v13_0_6 sriovYiPeng Chai
Enable ras for mp0 v13_0_6 sriov Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com> Reviewed-by: Stanley.Yang <Stanley.Yang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdkfd: retry after EBUSY is returned from hmm_ranges_get_pagesAlex Sierra
if hmm_range_get_pages returns EBUSY error during svm_range_validate_and_map, within the context of a page fault interrupt. This should retry through svm_range_restore_pages callback. Therefore we treat this as EAGAIN error instead, and defer it to restore pages fallback. Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu/jpeg - skip change of power-gating state for sriovSamir Dhume
Powergating is handled in the host driver. Reviewed-by: Zhigang Luo <zhigang.luo@amd.com> Signed-off-by: Samir Dhume <samir.dhume@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/pm: Add critical temp for GC v9.4.3Asad Kamal
Add critical temperature message support func for smu v13.0.6 and expose critical temperature as part of hw mon attributes for GC v9.4.3 v2: Added comment for pmfw version requirement & move the check to get_thermal_temperature_range function Signed-off-by: Asad Kamal <asad.kamal@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/pm: Update SMUv13.0.6 PMFW headersAsad Kamal
Update PMFW interface headers for updated metrics table and critical temperature message Signed-off-by: Asad Kamal <asad.kamal@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: update gc_info v2_1 from discoveryLe Ma
Several new fields are exposed in gc_info v2_1 Signed-off-by: Le Ma <le.ma@amd.com> Reviewed-by: Shiwu Zhang <shiwu.zhang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: update mall info v2 from discoveryLe Ma
Mall info v2 is introduced in ip discovery Signed-off-by: Le Ma <le.ma@amd.com> Reviewed-by: Shiwu Zhang <shiwu.zhang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu: Only support RAS EEPROM on dGPU platformCandice Li
RAS EEPROM device is only supported on dGPU platform for smu v13_0_6. Signed-off-by: Candice Li <candice.li@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amdgpu/pm: Add notification for no DC supportBokun Zhang
- There is a DPM issue where if DC is not present, FCLK will stay at low level. We need to send a SMU message to configure the DPM - Reuse smu_v13_0_notify_display_change() for this purpose Reviewed-by: Evan Quan <evan.quan@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Bokun Zhang <bokun.zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/amd/display: Enable Replay for static screen use casesBhawanpreet Lakha
- Setup replay config on device init. - Enable replay if feature is enabled (prioritize replay over PSR, since it can be enabled in more usecases) - Add debug masks to enable replay on supported ASICs Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com> Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2023-08-31drm/i915/gvt: Drop final dependencies on KVM internal detailsSean Christopherson
Open code gpa_to_gfn() in kvmgt_page_track_write() and drop KVMGT's dependency on kvm_host.h, i.e. include only kvm_page_track.h. KVMGT assumes "gfn == gpa >> PAGE_SHIFT" all over the place, including a few lines below in the same function with the same gpa, i.e. there's no reason to use KVM's helper for this one case. No functional change intended. Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-30-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31KVM: x86/mmu: Handle KVM bookkeeping in page-track APIs, not callersSean Christopherson
Get/put references to KVM when a page-track notifier is (un)registered instead of relying on the caller to do so. Forcing the caller to do the bookkeeping is unnecessary and adds one more thing for users to get wrong, e.g. see commit 9ed1fdee9ee3 ("drm/i915/gvt: Get reference to KVM iff attachment to VM is successful"). Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-29-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31KVM: x86/mmu: Drop @slot param from exported/external page-track APIsSean Christopherson
Refactor KVM's exported/external page-track, a.k.a. write-track, APIs to take only the gfn and do the required memslot lookup in KVM proper. Forcing users of the APIs to get the memslot unnecessarily bleeds KVM internals into KVMGT and complicates usage of the APIs. No functional change intended. Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-28-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31KVM: x86/mmu: Rename page-track APIs to reflect the new realitySean Christopherson
Rename the page-track APIs to capture that they're all about tracking writes, now that the facade of supporting multiple modes is gone. Opportunstically replace "slot" with "gfn" in anticipation of removing the @slot param from the external APIs. No functional change intended. Tested-by: Yongwei Ma <yongwei.ma@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-25-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31KVM: x86/mmu: Drop infrastructure for multiple page-track modesSean Christopherson
Drop "support" for multiple page-track modes, as there is no evidence that array-based and refcounted metadata is the optimal solution for other modes, nor is there any evidence that other use cases, e.g. for access-tracking, will be a good fit for the page-track machinery in general. E.g. one potential use case of access-tracking would be to prevent guest access to poisoned memory (from the guest's perspective). In that case, the number of poisoned pages is likely to be a very small percentage of the guest memory, and there is no need to reference count the number of access-tracking users, i.e. expanding gfn_track[] for a new mode would be grossly inefficient. And for poisoned memory, host userspace would also likely want to trap accesses, e.g. to inject #MC into the guest, and that isn't currently supported by the page-track framework. A better alternative for that poisoned page use case is likely a variation of the proposed per-gfn attributes overlay (linked), which would allow efficiently tracking the sparse set of poisoned pages, and by default would exit to userspace on access. Link: https://lore.kernel.org/all/Y2WB48kD0J4VGynX@google.com Cc: Ben Gardon <bgardon@google.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-24-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: switch from ->track_flush_slot() to ->track_remove_region()Yan Zhao
Switch from the poorly named and flawed ->track_flush_slot() to the newly introduced ->track_remove_region(). From KVMGT's perspective, the two hooks are functionally equivalent, the only difference being that ->track_remove_region() is called only when KVM is 100% certain the memory region will be removed, i.e. is invoked slightly later in KVM's memslot modification flow. Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Suggested-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Yan Zhao <yan.y.zhao@intel.com> [sean: handle name change, massage changelog, rebase] Tested-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-20-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Don't bother removing write-protection on to-be-deleted slotSean Christopherson
When handling a slot "flush", don't call back into KVM to drop write protection for gfns in the slot. Now that KVM rejects attempts to move memory slots while KVMGT is attached, the only time a slot is "flushed" is when it's being removed, i.e. the memslot and all its write-tracking metadata is about to be deleted. Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-18-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31KVM: drm/i915/gvt: Drop @vcpu from KVM's ->track_write() hookSean Christopherson
Drop @vcpu from KVM's ->track_write() hook provided for external users of the page-track APIs now that KVM itself doesn't use the page-track mechanism. Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-16-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Protect gfn hash table with vgpu_lockSean Christopherson
Use vgpu_lock instead of KVM's mmu_lock to protect accesses to the hash table used to track which gfns are write-protected when shadowing the guest's GTT, and hoist the acquisition of vgpu_lock from intel_vgpu_page_track_handler() out to its sole caller, kvmgt_page_track_write(). This fixes a bug where kvmgt_page_track_write(), which doesn't hold kvm->mmu_lock, could race with intel_gvt_page_track_remove() and trigger a use-after-free. Fixing kvmgt_page_track_write() by taking kvm->mmu_lock is not an option as mmu_lock is a r/w spinlock, and intel_vgpu_page_track_handler() might sleep when acquiring vgpu->cache_lock deep down the callstack: intel_vgpu_page_track_handler() | |-> page_track->handler / ppgtt_write_protection_handler() | |-> ppgtt_handle_guest_write_page_table_bytes() | |-> ppgtt_handle_guest_write_page_table() | |-> ppgtt_handle_guest_entry_removal() | |-> ppgtt_invalidate_pte() | |-> intel_gvt_dma_unmap_guest_page() | |-> mutex_lock(&vgpu->cache_lock); Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-12-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Drop unused helper intel_vgpu_reset_gtt()Sean Christopherson
Drop intel_vgpu_reset_gtt() as it no longer has any callers. In addition to eliminating dead code, this eliminates the last possible scenario where __kvmgt_protect_table_find() can be reached without holding vgpu_lock. Requiring vgpu_lock to be held when calling __kvmgt_protect_table_find() will allow a protecting the gfn hash with vgpu_lock without too much fuss. No functional change intended. Fixes: ba25d977571e ("drm/i915/gvt: Do not destroy ppgtt_mm during vGPU D3->D0.") Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-11-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Use an "unsigned long" to iterate over memslot gfnsSean Christopherson
Use an "unsigned long" instead of an "int" when iterating over the gfns in a memslot. The number of pages in the memslot is tracked as an "unsigned long", e.g. KVMGT could theoretically break if a KVM memslot larger than 16TiB were deleted (2^32 * 4KiB). Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-10-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Don't rely on KVM's gfn_to_pfn() to query possible 2M GTTSean Christopherson
Now that gvt_pin_guest_page() explicitly verifies the pinned PFN is a transparent hugepage page, don't use KVM's gfn_to_pfn() to pre-check if a 2MiB GTT entry is possible and instead just try to map the GFN with a 2MiB entry. Using KVM to query pfn that is ultimately managed through VFIO is odd, and KVM's gfn_to_pfn() is not intended for non-KVM consumption; it's exported only because of KVM vendor modules (x86 and PPC). Open code the check on 2MiB support instead of keeping is_2MB_gtt_possible() around for a single line of code. Move the call to intel_gvt_dma_map_guest_page() for a 4KiB entry into its case statement, i.e. fork the common path into the 4KiB and 2MiB "direct" shadow paths. Keeping the call in the "common" path is arguably more in the spirit of "one change per patch", but retaining the local "page_size" variable is silly, i.e. the call site will be changed either way, and jumping around the no-longer-common code is more subtle and rather odd, i.e. would just need to be immediately cleaned up. Drop the error message from gvt_pin_guest_page() when KVMGT attempts to shadow a 2MiB guest page that isn't backed by a compatible hugepage in the host. Dropping the pre-check on a THP makes it much more likely that the "error" will be encountered in normal operation. Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-9-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Error out on an attempt to shadowing an unknown GTT entry typeSean Christopherson
Bail from ppgtt_populate_shadow_entry() if an unexpected GTT entry type is encountered instead of subtly falling through to the common "direct shadow" path. Eliminating the default/error path's reliance on the common handling will allow hoisting intel_gvt_dma_map_guest_page() into the case statements so that the 2MiB case can try intel_gvt_dma_map_guest_page() and fallback to splitting the entry on failure. Reviewed-by: Zhi Wang <zhi.a.wang@intel.com> Tested-by: Yongwei Ma <yongwei.ma@intel.com> Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Link: https://lore.kernel.org/r/20230729013535.1070024-8-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-08-31drm/i915/gvt: Explicitly check that vGPU is attached before shadowingSean Christopherson
Move the check that a vGPU is attached from is_2MB_gtt_possible() all the way up to shadow_ppgtt_mm() to avoid unnecessary work, and to make it more obvious that a future cleanup of is_2MB_gtt_possible() isn't introducing a bug. is_2MB_gtt_possible() has only one caller, ppgtt_populate_shadow_entry(), and all paths in ppgtt_populate_shadow_entry() eventually check for attachment by way of intel_gvt_dma_map_guest_page(). And of the paths that lead to ppgtt_populate_shadow_entry(), shadow_ppgtt_mm() is the only one that doesn't already check for INTEL_VGPU_STATUS_ACTIVE or INTEL_VGPU_STATUS_ATTACHED. workload_thread() <= pick_next_workload() => INTEL_VGPU_STATUS_ACTIVE | -> dispatch_workload() | |-> prepare_workload() | -> intel_vgpu_sync_oos_pages() | | | |-> ppgtt_set_guest_page_sync() | | | |-> sync_oos_page() | | | |-> ppgtt_populate_shadow_entry() | |-> intel_vgpu_flush_post_shadow() | 1: |-> ppgtt_handle_guest_write_page_table() | |-> ppgtt_handle_guest_entry_add() | 2: | -> ppgtt_populate_spt_by_guest_entry() | | | |-> ppgtt_populate_spt() | | | |-> ppgtt_populate_shadow_entry() | | | |-> ppgtt_populate_spt_by_guest_entry() [see 2] | |-> ppgtt_populate_shadow_entry() kvmgt_page_track_write() <= KVM callback => INTEL_VGPU_STATUS_ATTACHED | |-> intel_vgpu_page_track_handler() | |-> ppgtt_write_protection_handler() | |-> ppgtt_handle_guest_write_page_table_bytes() | |-> ppgtt_handle_guest_write_page_table() [see 1] Reviewed-by: Yan Zhao <yan.y.zhao@intel.com> Tested-by: Yan Zhao <yan.y.zhao@intel.com> Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
generated by cgit (git 2.34.1) at 2024-07-01 03:06:28 +0000