- 02 Jun, 2023 2 commits
-
-
Nathan Chancellor authored
When building with clang's -Wincompatible-function-pointer-types-strict, the following warnings occur: drivers/gpu/drm/i915/gt/intel_ggtt_gmch.c:102:23: error: incompatible function pointer types assigning to 'void (*)(struct i915_address_space *, dma_addr_t, u64, unsigned int, u32)' (aka 'void (*)(struct i915_address_space *, unsigned int, unsigned long long, unsigned int, unsigned int)') from 'void (struct i915_address_space *, dma_addr_t, u64, enum i915_cache_level, u32)' (aka 'void (struct i915_address_space *, unsigned int, unsigned long long, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.insert_page = gmch_ggtt_insert_page; ^ ~~~~~~~~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gt/intel_ggtt_gmch.c:103:26: error: incompatible function pointer types assigning to 'void (*)(struct i915_address_space *, struct i915_vma_resource *, unsigned int, u32)' (aka 'void (*)(struct i915_address_space *, struct i915_vma_resource *, unsigned int, unsigned int)') from 'void (struct i915_address_space *, struct i915_vma_resource *, enum i915_cache_level, u32)' (aka 'void (struct i915_address_space *, struct i915_vma_resource *, enum i915_cache_level, unsigned int)') [-Werror, -Wincompatible-function-pointer-types-strict] ggtt->vm.insert_entries = gmch_ggtt_insert_entries; ^ ~~~~~~~~~~~~~~~~~~~~~~~~ 2 errors generated. The warning is pointing out that while 'enum i915_cache_level' and 'unsigned int' are ABI compatible, these indirect calls will fail clang's kernel Control Flow Integrity (kCFI) checks, as the callback's signature does not exactly match the prototype's signature. To fix this, replace the cache_level parameter with pat_index, as was done in other places within i915 where there is no difference between cache_level and pat_index on certain generations. Fixes: 9275277d ("drm/i915: use pat_index instead of cache_level") Signed-off-by:Nathan Chancellor <nathan@kernel.org> Reviewed-by:
Andi Shyti <andi.shyti@linux.intel.com> Reviewed-by:
Fei Yang <fei.yang@intel.com> Signed-off-by:
-
Nathan Chancellor authored
When booting a kernel compiled with CONFIG_CFI_CLANG (kCFI), there is a CFI failure in ggtt_probe_common() when trying to call hsw_pte_encode() via an indirect call: [ 5.030027] CFI failure at ggtt_probe_common+0xd1/0x130 [i915] (target: hsw_pte_encode+0x0/0x30 [i915]; expected type: 0xf5c1d0fc) With kCFI, indirect calls are validated against their expected type versus actual type and failures occur when the two types do not match. clang's -Wincompatible-function-pointer-types-strict can catch this at compile time but it is not enabled for the kernel yet: drivers/gpu/drm/i915/gt/intel_ggtt.c:1155:23: error: incompatible function pointer types assigning to 'u64 (*)(dma_addr_t, unsigned int, u32)' (aka 'unsigned long long (*)(unsigned int, unsigned int, unsigned int)') from 'u64 (dma_addr_t, enum i915_cache_level, u32)' (aka 'unsigned long long (unsigned int, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.pte_encode = iris_pte_encode; ^ ~~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gt/intel_ggtt.c:1157:23: error: incompatible function pointer types assigning to 'u64 (*)(dma_addr_t, unsigned int, u32)' (aka 'unsigned long long (*)(unsigned int, unsigned int, unsigned int)') from 'u64 (dma_addr_t, enum i915_cache_level, u32)' (aka 'unsigned long long (unsigned int, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.pte_encode = hsw_pte_encode; ^ ~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gt/intel_ggtt.c:1159:23: error: incompatible function pointer types assigning to 'u64 (*)(dma_addr_t, unsigned int, u32)' (aka 'unsigned long long (*)(unsigned int, unsigned int, unsigned int)') from 'u64 (dma_addr_t, enum i915_cache_level, u32)' (aka 'unsigned long long (unsigned int, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.pte_encode = byt_pte_encode; ^ ~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gt/intel_ggtt.c:1161:23: error: incompatible function pointer types assigning to 'u64 (*)(dma_addr_t, unsigned int, u32)' (aka 'unsigned long long (*)(unsigned int, unsigned int, unsigned int)') from 'u64 (dma_addr_t, enum i915_cache_level, u32)' (aka 'unsigned long long (unsigned int, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.pte_encode = ivb_pte_encode; ^ ~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gt/intel_ggtt.c:1163:23: error: incompatible function pointer types assigning to 'u64 (*)(dma_addr_t, unsigned int, u32)' (aka 'unsigned long long (*)(unsigned int, unsigned int, unsigned int)') from 'u64 (dma_addr_t, enum i915_cache_level, u32)' (aka 'unsigned long long (unsigned int, enum i915_cache_level, unsigned int)') [-Werror,-Wincompatible-function-pointer-types-strict] ggtt->vm.pte_encode = snb_pte_encode; ^ ~~~~~~~~~~~~~~ 5 errors generated. In this case, the pre-gen8 pte_encode functions have a second parameter type of 'enum i915_cache_level' whereas the function pointer prototype in 'struct i915_address_space' expects a second parameter type of 'unsigned int'. Update the second parameter of the callbacks and the comment above them noting that these statements are still valid, which matches other functions and files, to clear up the kCFI failures at run time. Fixes: 9275277d ("drm/i915: use pat_index instead of cache_level") Signed-off-by:
-
- 01 Jun, 2023 1 commit
-
-
Andi Shyti authored
kernel_context() returns an error pointer. Use pointer-error conversion functions to evaluate its return value, rather than checking for a '0' return. Fixes: eb5c10cb ("drm/i915: Remove I915_USER_PRIORITY_SHIFT") Reported-by:
Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by:
Andrzej Hajda <andrzej.hajda@intel.com> Acked-by:
Tejas Upadhyay <tejas.upadhyay@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230526124138.2006110-1-andi.shyti@linux.intel.com
-
- 31 May, 2023 4 commits
-
-
Ashutosh Dixit authored
No functional changes but we can remove some unsightly index computation and read/write functions if we convert the PMU sample array from a one-dimensional to a two-dimensional array. v2: Retain read/store helpers (Tvrtko) Suggested-by:
Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Reviewed-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Signed-off-by:
Ashutosh Dixit <ashutosh.dixit@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230524215629.97920-3-ashutosh.dixit@intel.com
-
Ashutosh Dixit authored
pmu_needs_timer() keeps the timer running even when GT is parked, ostensibly to sample requested/actual frequencies. However frequency_sample() has the following: /* Report 0/0 (actual/requested) frequency while parked. */ if (!intel_gt_pm_get_if_awake(gt)) return; The above code prevents frequencies to be sampled while the GT is parked. So we might as well turn off the sampling timer itself in this case and save CPU cycles/power. v2: Instead of turning freq bits off, return false, since no counters will run after this change when GT is parked (Tvrtko) v3: Remove gpu_active argument of pmu_needs_timer (Andrzej) Signed-off-by: -
Ashutosh Dixit authored
Clearing out report id and timestamp as means to detect unlanded reports only works if report size is power of 2. That is, only when report size is a sub-multiple of the OA buffer size can we be certain that reports will land at the same place each time in the OA buffer (after rewind). If report size is not a power of 2, we need to zero out the entire report to be able to detect unlanded reports reliably. v2: Add Fixes tag (Umesh) Fixes: 1cc064dc ("drm/i915/perf: Add support for OA media units") Reviewed-by:
Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> Reviewed-by:
Lionel Landwerlin <lionel.g.landwerlin@intel.com> Signed-off-by:
-
Michal Wajdeczko authored
We've already switched to new HXG definitions some time ago, drop legacy CTB definitions to avoid mistakes. Signed-off-by:
Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Piotr Piórkowski <piotr.piorkowski@intel.com> Cc: John Harrison <John.C.Harrison@Intel.com> Reviewed-by:
Piotr Piórkowski <piotr.piorkowski@intel.com> Signed-off-by:
John Harrison <John.C.Harrison@Intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230509201103.538-1-michal.wajdeczko@intel.com
-
- 30 May, 2023 4 commits
-
-
Michal Wajdeczko authored
For easier debug of any unexpected error responses from GuC that might be related to non-blocking fast requests, track action code (and stack if under DEBUG_GUC config) for every H2G request. Signed-off-by:
-
Michal Wajdeczko authored
Instead of printing message fence twice, include HXG header of the unexpected message and its len. Signed-off-by:
-
Michal Wajdeczko authored
In addition to the already defined REQUEST HXG message format, which is used when sender expects some confirmation or data, HXG protocol includes definition of the FAST REQUEST message, that may be used when sender does not expect any useful data to be returned. Using this instead of GUC_HXG_TYPE_EVENT for non-blocking CTB requests will allow GuC to send back GUC_HXG_TYPE_RESPONSE_FAILURE in case of errors. Note that it is not possible to return such errors to the caller, since this is for non-blocking calls and the related fence is not stored. Instead such messages are treated as unexpected, which will give an indication of potential GuC misprogramming that warrants extra debugging effort. Signed-off-by:
-
Chris Wilson authored
Allow compute contexts to submit the maximal amount of work without blocking userspace. The original size for user LRC ring's (SZ_16K) was chosen to minimise memory consumption, without being so small as to frequently stall in the middle of workloads. With the main consumers being GL / media pipelines of 2 or 3 batches per frame, we want to support ~10 requests in flight to allow for the application to control throttling without stalling within a frame. v2: - cover with else part Signed-off-by:
Chris Wilson <chris.p.wilson@intel.com> Signed-off-by:
-
- 26 May, 2023 1 commit
-
-
Jani Nikula authored
Use <> instead of "" for including headers from include/. Fixes: 8a9bf295 ("drm/i915/gsc: add initial support for GSC proxy") Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Alan Previn <alan.previn.teres.alexis@intel.com> Signed-off-by:
Jani Nikula <jani.nikula@intel.com> Reviewed-by:
Luca Coelho <luciano.coelho@intel.com> Reviewed-by:
Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230525094942.941123-1-jani.nikula@intel.com
-
- 23 May, 2023 1 commit
-
-
Fei Yang authored
The design is to keep Buffer Object's caching policy immutable through out its life cycle. This patch ends the support for set caching ioctl from MTL onward. While doing that we also set BO's to be 1-way coherent at creation time because GPU is no longer automatically snooping CPU cache. For userspace components needing to fine tune the caching policy for BO's, a follow up patch will extend the GEM_CREATE uAPI to allow them specify caching mode at BO creation time. Signed-off-by:
-
- 22 May, 2023 7 commits
-
-
Tvrtko Ursulin authored
Start exporting frequency and RC6 counters from all tiles. Existing counters keep their names and config values and new one use the namespace added in the previous patch, with the "-gtN" added to their names. Interrupts counter is an odd one off. Because it is the global device counters (not only GT) we choose not to add per tile versions for now. Signed-off-by:
Aravind Iddamsetty <aravind.iddamsetty@intel.com> Reviewed-by:
-
Tvrtko Ursulin authored
Reserve some bits in the counter config namespace which will carry the tile id and prepare the code to handle this. No per tile counters have been added yet. v2: - Fix checkpatch issues - Use 4 bits for gt id in non-engine counters. Drop FIXME. - Set MAX GTs to 4. Drop FIXME. v3: (Ashutosh, Tvrtko) - Drop BUG_ON that would never fire - Make enable u64 - Pull in some code from next patch v4: Set I915_PMU_MAX_GTS to 2 (Tvrtko) v5: s/u64/u32 where needed (Ashutosh) Signed-off-by:
-
Tvrtko Ursulin authored
We do not want to have timers per tile and waste CPU cycles and energy via multiple wake-up sources, for a relatively un-important task of PMU sampling, so keeping a single timer works well. But we also do not want the first GT which goes idle to turn off the timer. Add some reference counting, via a mask of unparked GTs, to solve this. v2: Drop the check for unparked in i915_sample (Ashutosh) v3: Revert v2 (Tvrtko) Signed-off-by:
-
Tvrtko Ursulin authored
Trivial prep work for full multi-tile enablement later. Signed-off-by:
Vinay Belgaumkar <vinay.belgaumkar@intel.com> Reviewed-by:
-
Tvrtko Ursulin authored
As we have more and more engines do not waste time sampling the ones no- one is monitoring. Signed-off-by:
-
Tvrtko Ursulin authored
Given how the metrics are already exported, we also need to run sampling over engines from all GTs. Problem of GT frequencies is left for later. Signed-off-by:
-
Tvrtko Ursulin authored
Having it as u64 was a confusing (but harmless) mistake. Also add some asserts to make sure the internal field does not overflow in the future. v2: Fix WARN_ON firing for INTERRUPT event (Umesh) Signed-off-by:
-
- 19 May, 2023 1 commit
-
-
Cong Liu authored
Be sure to properly free the allocated memory before exiting the live_nop_switch function. Signed-off-by:
Cong Liu <liucong2@kylinos.cn> Suggested-by:
Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by:
-
- 18 May, 2023 3 commits
-
-
Radhakrishna Sripada authored
Like DG2, MTL a-step hardware is subject to Wa_16014892111 which requires that any changes made to the DRAW_WATERMARK register be done via an INDIRECT_CTX batch buffer rather than through a regular context workaround. The bspec gives the same non-default recommended tuning value for DRAW_WATERMARK as DG2, so we can re-use the INDIRECT_CTX code to apply that tuning setting on A-step hardware. Application of the tuning setting on B-step and later does not need INDIRECT_CTX handling and is already done in mtl_ctx_workarounds_init() as usual. v2: Limit the WA for A-step v3: Update the commit message. v4: Reorder platform checks and update commit message. Bspec: 68331 Cc: Haridhar Kalvala <haridhar.kalvala@intel.com> Cc: Gustavo Sousa <gustavo.sousa@intel.com> Reviewed-by:
Matt Roper <matthew.d.roper@intel.com> Signed-off-by:
Radhakrishna Sripada <radhakrishna.sripada@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230517233111.297542-2-radhakrishna.sripada@intel.com
-
Radhakrishna Sripada authored
MTL reuses the tuning parameters for DG2. Extend the dg2 performance tuning parameters to MTL. v2: Add DRAW_WATERMARK tuning parameter. v3: Limit DRAW_WATERMARK tuning to non A0 step. v4: Reorder platform checks. Restrict Blend fill caching optimization to Render GT. v5: Move mtl tuning params to its own function Bspec: 68331 Cc: Haridhar Kalvala <haridhar.kalvala@intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Reviewed-by:Gustavo Sousa <gustavo.sousa@intel.com> Signed-off-by:
-
Andrzej Hajda authored
Multiple CI tests fails with forcewake ack timeouts if render power gating is enabled. BSpec 52698 states it should be 0 for MTL, but apparently this info is outdated. Anyway since the patch makes MTL pass basic tests added FIXME tag informing this is temporary workaround. v2: added FIXME tag Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4983Signed-off-by:
Nirmoy Das <nirmoy.das@intel.com> Reviewed-by:
-
- 17 May, 2023 1 commit
-
-
Vinay Belgaumkar authored
rps_boost debugfs shows host turbo related info. This is not valid when SLPC is enabled. guc_slpc_info already shows the number of boosts. Add num_waiters there as well and disable rps_boost when SLPC is enabled. v2: Replace Bug with Link to resolve checkpatch warning Link: https://gitlab.freedesktop.org/drm/intel/-/issues/7632Reviewed-by:
-
- 16 May, 2023 2 commits
-
-
John Harrison authored
In the past, There have been sporadic CTB failures which proved hard to reproduce manually. The most effective solution was to dump the GuC log at the point of failure and let the CI system do the repro. It is preferable not to dump the GuC log via dmesg for all issues as it is not always necessary and is not helpful for end users. But rather than trying to re-invent the code to do this each time it is wanted, commit the code but for DEBUG_GUC builds only. v2: Use IS_ENABLED for testing config options. Signed-off-by:
-
John Harrison authored
This is useful for getting debug information out in certain situations, such as failing kernel selftests and CI runs that don't log error captures. It is especially useful for things like retrieving GuC logs as GuC operation can't be tracked by adding printk or ftrace entries. v2: Add CONFIG_DRM_I915_DEBUG_GEM wrapper (review feedback by Rodrigo). Signed-off-by:
-
- 15 May, 2023 1 commit
-
-
Ashutosh Dixit authored
Loading i915 on UBSAN enabled kernels (CONFIG_UBSAN/CONFIG_UBSAN_BOOL) causes the following warning: UBSAN: invalid-load in drivers/gpu/drm/i915/gt/uc/intel_uc.c:558:2 load of value 255 is not a valid value for type '_Bool' Call Trace: dump_stack_lvl+0x57/0x7d ubsan_epilogue+0x5/0x40 __ubsan_handle_load_invalid_value.cold+0x43/0x48 __uc_init_hw+0x76a/0x903 [i915] ... i915_driver_probe+0xfb1/0x1eb0 [i915] i915_pci_probe+0xbe/0x2d0 [i915] The warning happens because during probe i915_hwmon is still not available which results in the output boolean variable *old remaining uninitialized. Silence the warning by initializing the variable to an arbitrary value. v2: Move variable initialization to the declaration (Andi) Signed-off-by:
-
- 12 May, 2023 10 commits
-
-
John Harrison authored
The GuC has a completely separate engine class enum when referring to register capture lists, which combines render and compute. The driver was using the 'normal' GuC specific engine class enum instead. That meant that it thought it was defining a capture list for compute engines, the list was actually being applied to the GSC engine. And if a platform didn't have a render engine, then it would get no compute register captures at all. Fix that. Signed-off-by:
Alan Previn <alan.previn.teres.alexis@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230512013544.3367606-1-John.C.Harrison@Intel.com
-
John Harrison authored
A recent change bumped a 'notice' message up to 'error' level for debug builds to help trap incorrect configurations in CI systems. Unfortunately, the error condition in question is triggered by the error injection probe test. So change the message again to be 'probe error' level instead. Signed-off-by:
-
Alan Previn authored
Enable PXP with MTL-GSC-CS: add the has_pxp into device info and increase the debugfs teardown timeouts to align with new GSC-CS + firmware specs. Now that we have 3 places that are selecting pxp timeouts based on tee vs gsccs back-end, let's add a helper. Signed-off-by:
-
Alan Previn authored
On legacy platforms, KCR HW enabling is done at the time the mei component interface is bound. It's also disabled during unbind. However, for MTL onwards, we don't depend on a tee component to start sending GSC-CS firmware messages. Thus, immediately enable (or disable) KCR HW on PXP's init, fini and resume. Signed-off-by:
-
Alan Previn authored
Because of the additional firmware, component-driver and initialization depedencies required on MTL platform before a PXP context can be created, UMD calling for PXP creation as a way to get-caps can take a long time. An actual real world customer stack has seen this happen in the 4-to-8 second range after the kernel starts (which sees MESA's init appear in the middle of this range as the compositor comes up). To avoid unncessary delays experienced by the UMD for get-caps purposes, add a GET_PARAM for I915_PARAM_PXP_SUPPORT. However, some failures can still occur after all the depedencies are met (such as firmware init flow failure, bios configurations or SOC fusing not allowing PXP enablement). Those scenarios will only be known to user space when it attempts creating a PXP context and is documented in the GEM UAPI headers. While making this change, create a helper that is common to both GET_PARAM caller and intel_pxp_start since the latter does similar checks. Signed-off-by:
Jordan Justen <jordan.l.justen@intel.com> Signed-off-by:
-
Alan Previn authored
Add MTL's function for ARB session creation using PXP firmware version 4.3 ABI structure format. While relooking at the ARB session creation flow in intel_pxp_start, let's address missing UAPI documentation. Without actually changing backward compatible behavior, update i915's drm-uapi comments that describe the possible error values when creating a context with I915_CONTEXT_PARAM_PROTECTED_CONTENT: Since the first merge of PXP support on ADL, i915 returns -ENXIO if a dependency such as firmware or component driver was yet to be loaded or returns -EIO if the creation attempt failed when requested by the PXP firmware (specific firmware error responses are reported in dmesg). Add MTL's function for ARB session invalidation but this reuses PXP firmware version 4.2 ABI structure format. For both cases, in the back-end gsccs functions for sending messages to the firmware inspect the GSC-CS-Mem-Header's pending-bit which means the GSC firmware is busy and we should retry. Given the last hw requirement, lets also update functions in front-end layer that wait for session creation or teardown completion to use new worst case timeout periods. Signed-off-by:
-
Alan Previn authored
Add GSC engine based method for sending PXP firmware packets to the GSC firmware for MTL (and future) products. Use the newly added helpers to populate the GSC-CS memory header and send the message packet to the FW by dispatching the GSC_HECI_CMD_PKT instruction on the GSC engine. We use non-priveleged batches for submission to GSC engine which require two buffers for the request: - a buffer for the HECI packet that contains PXP FW commands - a batch-buffer that contains the engine instruction for sending the HECI packet to the GSC firmware. Thus, add the allocation and freeing of these buffers in gsccs init and fini. The GSC-fw may reply to commands with a SUCCESS but with an additional pending-bit set in the reply packet. This bit means the GSC-FW is currently busy and the caller needs to try again with the gsc_message_handle the fw returned. Thus, add a wrapper to continuously retry send_message while replaying the gsc_message_handle. Retries need to follow the arch-spec count and delay until GSC-FW replies with the real SUCCESS or timeout after that spec'd delay. The GSC-fw requires a non-zero host_session_handle provided by the caller to enable gsc_message_handle tracking. Thus, allocate the host_session_handle at init and destroy it at fini (the latter requiring an FYI to the gsc-firmware). Signed-off-by: -
Alan Previn authored
Add helper functions into a new file for heci-packet-submission. The helpers will handle generating the MTL GSC-CS Memory-Header and submission of the Heci-Cmd-Packet instructions to the engine. NOTE1: These common functions for heci-packet-submission will be used by different i915 callers: 1- GSC-SW-Proxy: This is pending upstream publication awaiting a few remaining opens 2- MTL-HDCP: An equivalent patch has also been published at: https://patchwork.freedesktop.org/series/111876/. (Patch 1) 3- PXP: This series. NOTE2: A difference in this patch vs what is appearing is in bullet 2 above is that HDCP (and SW-Proxy) will be using priveleged submission (GGTT and common gsc-uc-context) while PXP will be using non-priveleged PPGTT, context and batch buffer. Therefore this patch will only slightly overlap with the MTL-HDCP patches despite have very similar function names (emit_foo vs emit_nonpriv_foo). This is because HECI_CMD_PKT instructions require different flows and hw-specific code when done via PPGTT based submission (not different from other engines). MTL-HDCP contains the same intel_gsc_mtl_header_t structures as this but the helpers there are different. Both add the same new file names. NOTE3: Additional clarity about the heci-cmd-pkt layout and where the common helpers come in: - On MTL, when an i915 subsystem needs to send a command request to the security firmware, it will send that via the GSC- engine-command-streamer. - However those commands, (lets call them "gsc_specific_fw_api" calls), are not understood by the GSC command streamer hw. - The GSC CS only looks at the GSC_HECI_CMD_PKT instruction and passes it along to the GSC firmware. - The GSC FW on the other hand needs additional metadata to know which usage service is being called (PXP, HDCP, proxy, etc) along with session specific info. Thus an extra header called GSC-CS HECI Memory Header, (C) in below diagram is prepended before the FW specific API, (D). - Thus, the structural layout of the request submitted would need to look like the diagram below (for non-priv PXP). - In the diagram, the common helper for HDCP, (GSC-Sw-Proxy) and PXP (i.e. new function intel_gsc_uc_heci_cmd_emit_mtl_header) will populate blob (C) while additional helpers, different for PPGGTT (this patch) vs GGTT (HDCP series) will populate blobs (A) and (B) below. ___________________________________________________________ (A) | MI_BATCH_BUFFER_START (ppgtt, batchbuff-addr, ...) | | | | | _|________________________________________________ | | (B)| GSC_HECI_CMD_PKT (pkt-addr-in, pkt-size-in, | | | | pkt-addr-out, pkt-size-out) |-------- | | MI_BATCH_BUFFER_END | | | | |________________________________________________| | | | | | |_________________________________________________________| | | --------------------------------------------------------- | \|/ ______V___________________________________________ | _________________________________________ | |(C)| | | | | struct intel_gsc_mtl_header { | | | | validity marker | | | | heci_clent_id | | | | ... | | | | } | | | |_______________________________________| | |(D)| | | | | struct gsc_fw_specific_api_foobar { | | | | ... | | | | For an example, see | | | | 'struct pxp43_create_arb_in' at | | | | intel_pxp_cmd_interface_43.h | | | | | | | | } | | | | Struture depends on command type | | | | struct gsc_fw_specific_api_foobar { | | | |_______________________________________| | |________________________________________________| That said, this patch provides basic helpers but leaves the PXP subsystem (i.e. the caller) to handle (D) and everything else such as input/output size verification or handling the responses from security firmware (for example, requiring a retry). Signed-off-by: -
Alan Previn authored
Add MTL hw-plumbing enabling for KCR operation under PXP which includes: 1. Updating 'pick-gt' to get the media tile for KCR interrupt handling 2. Adding MTL's KCR registers for PXP operation (init, status-checking, etc.). While doing #2, lets create a separate registers header file for PXP to be consistent with other i915 global subsystems. Signed-off-by:
-
Alan Previn authored
For MTL, the PXP back-end transport uses the GSC engine to submit HECI packets through the HW to the GSC firmware for PXP arb session management. This submission uses a non-priveleged batch buffer, a buffer for the command packet and of course a context targeting the GSC-CS. Thus for MTL, we need to allocate and free a set of execution submission resources for the management of the arbitration session. Lets start with the context creation first since that object and its usage is very straight-forward. We'll add the buffer allocation and freeing later when we introduce the gsccs' send-message function. Do this one time allocation of gsccs specific resources in a new gsccs source file with intel_pxp_gsccs_init / fini functions and hook them up from the PXP front-end. Signed-off-by:
-
- 11 May, 2023 2 commits
-
-
Fei Yang authored
Currently the KMD is using enum i915_cache_level to set caching policy for buffer objects. This is flaky because the PAT index which really controls the caching behavior in PTE has far more levels than what's defined in the enum. In addition, the PAT index is platform dependent, having to translate between i915_cache_level and PAT index is not reliable, and makes the code more complicated. From UMD's perspective there is also a necessity to set caching policy for performance fine tuning. It's much easier for the UMD to directly use PAT index because the behavior of each PAT index is clearly defined in Bspec. Having the abstracted i915_cache_level sitting in between would only cause more ambiguity. PAT is expected to work much like MOCS already works today, and by design userspace is expected to select the index that exactly matches the desired behavior described in the hardware specification. For these reasons this patch replaces i915_cache_level with PAT index. Also note, the cache_level is not completely removed yet, because the KMD still has the need of creating buffer objects with simple cache settings such as cached, uncached, or writethrough. For kernel objects, cache_level is used for simplicity and backward compatibility. For Pre-gen12 platforms PAT can have 1:1 mapping to i915_cache_level, so these two are interchangeable. see the use of LEGACY_CACHELEVEL. One consequence of this change is that gen8_pte_encode is no longer working for gen12 platforms due to the fact that gen12 platforms has different PAT definitions. In the meantime the mtl_pte_encode introduced specfically for MTL becomes generic for all gen12 platforms. This patch renames the MTL PTE encode function into gen12_pte_encode and apply it to all gen12. Even though this change looks unrelated, but separating them would temporarily break gen12 PTE encoding, thus squash them in one patch. Special note: this patch changes the way caching behavior is controlled in the sense that some objects are left to be managed by userspace. For such objects we need to be careful not to change the userspace settings.There are kerneldoc and comments added around obj->cache_coherent, cache_dirty, and how to bypass the checkings by i915_gem_object_has_cache_level. For full understanding, these changes need to be looked at together with the two follow-up patches, one disables the {set|get}_caching ioctl's and the other adds set_pat extension to the GEM_CREATE uAPI. Bspec: 63019 Cc: Chris Wilson <chris.p.wilson@linux.intel.com> Signed-off-by: -
Fei Yang authored
This patch is a preparation for replacing enum i915_cache_level with PAT index. Caching policy for buffer objects is set through the PAT index in PTE, the old i915_cache_level is not sufficient to represent all caching modes supported by the hardware. Preparing the transition by adding some platform dependent data structures and helper functions to translate the cache_level to pat_index. cachelevel_to_pat: a platform dependent array mapping cache_level to pat_index. max_pat_index: the maximum PAT index recommended in hardware specification Needed for validating the PAT index passed in from user space. i915_gem_get_pat_index: function to convert cache_level to PAT index. obj_to_i915(obj): macro moved to header file for wider usage. I915_MAX_CACHE_LEVEL: upper bound of i915_cache_level for the convenience of coding. Cc: Chris Wilson <chris.p.wilson@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Signed-off-by:
-
