515.57

Author: aritger

CHANGELOG.md

@@ -2,6 +2,12 @@
 
 ## Release 515 Entries
 
+### [515.57] 2022-06-28
+
+#### Fixed
+
+- Backtick is deprecated, [#273](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/273) by @arch-user-france1
+
 ### [515.48.07] 2022-05-31
 
 #### Added
@@ -10,20 +16,20 @@
 
 #### Fixed
 
-- Fix various README capitalizations, [#8 by @lx-is](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/8)
-- Automatically tag bug report issues, [#15 by @thebeanogamer](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/15)
-- Improve conftest.sh Script, [#37 by @Nitepone](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/37)
-- Update HTTP link to HTTPS, [#101 by @alcaparra](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/101)
-- moved array sanity check to before the array access, [#117 by @RealAstolfo](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/117)
-- Fixed some typos, [#122 by @FEDOyt](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/122)
-- Fixed capitalization, [#123 by @keroeslux](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/123)
-- Fix typos in NVDEC Engine Descriptor, [#126 from @TrickyDmitriy](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/126)
-- Extranous apostrohpes in a makefile script [sic], [#14 by @kiroma](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/14)
-- HDMI no audio @4K above 60Hz, [#75 by @adolfotregosa](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/75)
-- dp_configcaps.cpp:405: array index sanity check in wrong place?, [#110 by @dcb314](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/110)
-- NVRM kgspInitRm_IMPL: missing NVDEC0 engine, cannot initialize GSP-RM, [#116 by @kfazz](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/116)
-- ERROR: modpost: "backlight_device_register" [...nvidia-modeset.ko] undefined, [#135 by @sndirsch](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/135)
-- aarch64 build fails, [#151 by @frezbo](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/151)
+- Fix various README capitalizations, [#8](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/8) by @27lx 
+- Automatically tag bug report issues, [#15](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/15) by @thebeanogamer
+- Improve conftest.sh Script, [#37](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/37) by @Nitepone
+- Update HTTP link to HTTPS, [#101](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/101) by @alcaparra
+- moved array sanity check to before the array access, [#117](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/117) by @RealAstolfo
+- Fixed some typos, [#122](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/122) by @FEDOyt
+- Fixed capitalization, [#123](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/123) by @keroeslux
+- Fix typos in NVDEC Engine Descriptor, [#126](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/126) from @TrickyDmitriy
+- Extranous apostrohpes in a makefile script [sic], [#14](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/14) by @kiroma
+- HDMI no audio @ 4K above 60Hz, [#75](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/75) by @adolfotregosa
+- dp_configcaps.cpp:405: array index sanity check in wrong place?, [#110](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/110) by @dcb314
+- NVRM kgspInitRm_IMPL: missing NVDEC0 engine, cannot initialize GSP-RM, [#116](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/116) by @kfazz
+- ERROR: modpost: "backlight_device_register" [...nvidia-modeset.ko] undefined, [#135](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/135) by @sndirsch
+- aarch64 build fails, [#151](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/151) by @frezbo
 
 ### [515.43.04] 2022-05-11
 

README.md

@@ -1,23 +1,23 @@
 # NVIDIA Linux Open GPU Kernel Module Source
 
 This is the source release of the NVIDIA Linux open GPU kernel modules,
-version 515.48.07.
+version 515.57.
 
 
 ## How to Build
 
 To build:
 
-    make modules -j`nproc`
+    make modules -j$(nproc)
 
 To install, first uninstall any existing NVIDIA kernel modules.  Then,
 as root:
 
-    make modules_install -j`nproc`
+    make modules_install -j$(nproc)
 
 Note that the kernel modules built here must be used with gsp.bin
 firmware and user-space NVIDIA GPU driver components from a corresponding
-515.48.07 driver release.  This can be achieved by installing
+515.57 driver release.  This can be achieved by installing
 the NVIDIA GPU driver from the .run file using the `--no-kernel-modules`
 option.  E.g.,
 
@@ -39,7 +39,7 @@ If cross-compiling, set these variables on the make command line:
 E.g.,
 
     # compile on x86_64 for aarch64
-    make modules -j`nproc`          \
+    make modules -j$(nproc)         \
         TARGET_ARCH=aarch64         \
         CC=aarch64-linux-gnu-gcc    \
         LD=aarch64-linux-gnu-ld     \
@@ -59,7 +59,7 @@ DEBUG - Set this to "1" to build the kernel modules as debug.  By default, the
 
 These variables can be set on the make command line.  E.g.,
 
-    make modules -j`nproc` NV_VERBOSE=1
+    make modules -j$(nproc) NV_VERBOSE=1
 
 
 ## Supported Toolchains
@@ -167,17 +167,17 @@ for the target kernel.
 ## Compatible GPUs
 
 The open-gpu-kernel-modules can be used on any Turing or later GPU
-(see the table below). However, in the 515.48.07 release,
+(see the table below). However, in the 515.57 release,
 GeForce and Workstation support is still considered alpha-quality.
 
 To enable use of the open kernel modules on GeForce and Workstation GPUs,
 set the "NVreg_OpenRmEnableUnsupportedGpus" nvidia.ko kernel module
 parameter to 1. For more details, see the NVIDIA GPU driver end user
 README here:
 
-https://us.download.nvidia.com/XFree86/Linux-x86_64/515.48.07/README/kernel_open.html
+https://us.download.nvidia.com/XFree86/Linux-x86_64/515.57/README/kernel_open.html
 
-In the below table, if three IDs are listed, the first is the PCI Device
+In the below table, if three IDs are listed, the first is the PCI Device 
 ID, the second is the PCI Subsystem Vendor ID, and the third is the PCI
 Subsystem Device ID.
 
@@ -501,6 +501,7 @@ Subsystem Device ID.
 | Matrox D-Series D2450                           | 1F76 102B 2800 |
 | Matrox D-Series D2480                           | 1F76 102B 2900 |
 | NVIDIA GeForce GTX 1650                         | 1F82           |
+| NVIDIA GeForce GTX 1630                         | 1F83           |
 | NVIDIA GeForce GTX 1650                         | 1F91           |
 | NVIDIA GeForce GTX 1650 with Max-Q Design       | 1F91 103C 863E |
 | NVIDIA GeForce GTX 1650 with Max-Q Design       | 1F91 103C 86E7 |
@@ -703,8 +704,6 @@ Subsystem Device ID.
 | NVIDIA RTX A5500                                | 2233 10DE 165A |
 | NVIDIA RTX A5500                                | 2233 17AA 165A |
 | NVIDIA A40                                      | 2235 10DE 145A |
-| NVIDIA A10                                      | 2236 10DE 1482 |
-| NVIDIA A10G                                     | 2237 10DE 152F |
 | NVIDIA A10M                                     | 2238 10DE 1677 |
 | NVIDIA GeForce RTX 3060 Ti                      | 2414           |
 | NVIDIA GeForce RTX 3080 Ti Laptop GPU           | 2420           |

kernel-open/Kbuild

@@ -72,7 +72,7 @@ EXTRA_CFLAGS += -I$(src)/common/inc
 EXTRA_CFLAGS += -I$(src)
 EXTRA_CFLAGS += -Wall -MD $(DEFINES) $(INCLUDES) -Wno-cast-qual -Wno-error -Wno-format-extra-args
 EXTRA_CFLAGS += -D__KERNEL__ -DMODULE -DNVRM
-EXTRA_CFLAGS += -DNV_VERSION_STRING=\"515.48.07\"
+EXTRA_CFLAGS += -DNV_VERSION_STRING=\"515.57\"
 
 EXTRA_CFLAGS += -Wno-unused-function
 

kernel-open/nvidia-uvm/uvm_gpu_replayable_faults.c

@@ -1942,45 +1942,32 @@ static NV_STATUS cancel_faults_precise_va(uvm_gpu_t *gpu,
 
 // Function called when the system has found a global error and needs to
 // trigger RC in RM.
-// We cancel one entry per uTLB
 static void cancel_fault_batch_tlb(uvm_gpu_t *gpu,
                                    uvm_fault_service_batch_context_t *batch_context,
                                    UvmEventFatalReason reason)
 {
     NvU32 i;
 
-    // Fault filtering is not allowed in the TLB-based fault cancel path
-    UVM_ASSERT(batch_context->num_cached_faults == batch_context->num_coalesced_faults);
-
-    for (i = 0; i < batch_context->num_cached_faults; ++i) {
-        NV_STATUS status;
+    for (i = 0; i < batch_context->num_coalesced_faults; ++i) {
+        NV_STATUS status = NV_OK;
         uvm_fault_buffer_entry_t *current_entry;
-        uvm_fault_utlb_info_t *utlb;
+        uvm_fault_buffer_entry_t *coalesced_entry;
 
-        current_entry = &batch_context->fault_cache[i];
-        utlb = &batch_context->utlbs[current_entry->fault_source.utlb_id];
+        current_entry = batch_context->ordered_fault_cache[i];
 
-        // If this uTLB has been already cancelled, skip it
-        if (utlb->cancelled)
-            continue;
+        // The list iteration below skips the entry used as 'head'.
+        // Report the 'head' entry explicitly.
+        uvm_va_space_down_read(current_entry->va_space);
+        uvm_tools_record_gpu_fatal_fault(gpu->parent->id, current_entry->va_space, current_entry, reason);
 
-        record_fatal_fault_helper(gpu, current_entry, reason);
-
-        // Although the global cancellation method can be used here instead of
-        // targeted, we still use the targeted method since this function is
-        // only invoked in GPUs without support for VA fault cancellation, for
-        // which the targeted version is already required in
-        // cancel_faults_precise_tlb(). To maintain consistency, we use the
-        // targeted variant in both cases.
-        status = push_cancel_on_gpu_targeted(gpu,
-                                             current_entry->instance_ptr,
-                                             current_entry->fault_source.gpc_id,
-                                             current_entry->fault_source.client_id,
-                                             &batch_context->tracker);
+        list_for_each_entry(coalesced_entry, &current_entry->merged_instances_list, merged_instances_list)
+            uvm_tools_record_gpu_fatal_fault(gpu->parent->id, current_entry->va_space, coalesced_entry, reason);
+        uvm_va_space_up_read(current_entry->va_space);
+
+        // We need to cancel each instance pointer to correctly handle faults from multiple contexts.
+        status = push_cancel_on_gpu_global(gpu, current_entry->instance_ptr, &batch_context->tracker);
         if (status != NV_OK)
             break;
-
-        utlb->cancelled = true;
     }
 }
 

kernel-open/nvidia-uvm/uvm_migrate_pageable.c

@@ -899,9 +899,14 @@ static NV_STATUS migrate_pageable(migrate_vma_state_t *state)
 
     // VMAs are validated and migrated one at a time, since migrate_vma works
     // on one vma at a time
-    for (; vma->vm_start <= prev_outer; vma = vma->vm_next) {
+    for (; vma->vm_start <= prev_outer; vma = find_vma_intersection(mm, prev_outer, outer)) {
         unsigned long next_addr = 0;
-        NV_STATUS status = migrate_pageable_vma(vma, start, outer, state, &next_addr);
+        NV_STATUS status;
+
+        // Callers have already validated the range so the vma should be valid.
+        UVM_ASSERT(vma);
+
+        status = migrate_pageable_vma(vma, start, outer, state, &next_addr);
         if (status == NV_WARN_NOTHING_TO_DO) {
             NV_STATUS populate_status = NV_OK;
             bool touch = uvm_migrate_args->touch;

kernel-open/nvidia-uvm/uvm_policy.c

@@ -44,7 +44,7 @@ bool uvm_is_valid_vma_range(struct mm_struct *mm, NvU64 start, NvU64 length)
         if (vma->vm_end >= end)
             return true;
         start = vma->vm_end;
-        vma = vma->vm_next;
+        vma = find_vma_intersection(mm, start, end);
     }
 
     return false;

kernel-open/nvidia-uvm/uvm_populate_pageable.c

@@ -158,7 +158,7 @@ NV_STATUS uvm_populate_pageable(struct mm_struct *mm,
     // VMAs are validated and populated one at a time, since they may have
     // different protection flags
     // Validation of VM_SPECIAL flags is delegated to get_user_pages
-    for (; vma->vm_start <= prev_end; vma = vma->vm_next) {
+    for (; vma && vma->vm_start <= prev_end; vma = find_vma_intersection(mm, prev_end, end)) {
         NV_STATUS status = uvm_populate_pageable_vma(vma, start, end - start, min_prot, touch, populate_permissions);
 
         if (status != NV_OK)

src/common/displayport/inc/dp_messages.h

@@ -116,7 +116,9 @@ namespace DisplayPort
         bool                isBeingDestroyed;
         bool                isPaused;
 
+        // Properties from regkey
         bool                bNoReplyTimerForBusyWaiting;
+        bool                bDpcdProbingForBusyWaiting;
 
         List                messageReceivers;
         List                notYetSentDownRequest;    // Down Messages yet to be processed
@@ -159,7 +161,8 @@ namespace DisplayPort
         {
             DP_ASSERT(dpRegkeyDatabase.bInitialized &&
                       "All regkeys are invalid because dpRegkeyDatabase is not initialized!");
-            bNoReplyTimerForBusyWaiting = dpRegkeyDatabase.bNoReplyTimerForBusyWaiting;
+            bNoReplyTimerForBusyWaiting  = dpRegkeyDatabase.bNoReplyTimerForBusyWaiting;
+            bDpcdProbingForBusyWaiting   = dpRegkeyDatabase.bDpcdProbingForBusyWaiting;
         }
 
         MessageManager(DPCDHAL * hal, Timer * timer)

src/common/displayport/inc/dp_regkeydatabase.h

@@ -72,6 +72,9 @@
 #define NV_DP_DSC_OPTIMIZE_LT_BUG_3534707              "DP_DSC_OPTIMIZE_LT_BUG_3534707"
 
 #define NV_DP_REGKEY_NO_REPLY_TIMER_FOR_BUSY_WAITING   "NO_REPLY_TIMER_FOR_BUSY_WAITING"
+
+#define NV_DP_REGKEY_DPCD_PROBING_FOR_BUSY_WAITING     "DP_DPCD_PROBING_FOR_BUSY_WAITING"
+
 //
 // Data Base used to store all the regkey values.
 // The actual data base is declared statically in dp_evoadapter.cpp.
@@ -106,6 +109,7 @@ struct DP_REGKEY_DATABASE
     bool  bDscMstEnablePassThrough;
     bool  bDscOptimizeLTBug3534707;
     bool  bNoReplyTimerForBusyWaiting;
+    bool  bDpcdProbingForBusyWaiting;
 };
 
 #endif //INCLUDED_DP_REGKEYDATABASE_H

src/common/displayport/src/dp_evoadapter.cpp

@@ -96,7 +96,8 @@ const struct
     {NV_DP_DSC_MST_CAP_BUG_3143315,                 &dpRegkeyDatabase.bDscMstCapBug3143315,            DP_REG_VAL_BOOL},
     {NV_DP_DSC_MST_ENABLE_PASS_THROUGH,             &dpRegkeyDatabase.bDscMstEnablePassThrough,        DP_REG_VAL_BOOL},
     {NV_DP_DSC_OPTIMIZE_LT_BUG_3534707,             &dpRegkeyDatabase.bDscOptimizeLTBug3534707,        DP_REG_VAL_BOOL},
-    {NV_DP_REGKEY_NO_REPLY_TIMER_FOR_BUSY_WAITING,  &dpRegkeyDatabase.bNoReplyTimerForBusyWaiting,     DP_REG_VAL_BOOL}
+    {NV_DP_REGKEY_NO_REPLY_TIMER_FOR_BUSY_WAITING,  &dpRegkeyDatabase.bNoReplyTimerForBusyWaiting,     DP_REG_VAL_BOOL},
+    {NV_DP_REGKEY_DPCD_PROBING_FOR_BUSY_WAITING,    &dpRegkeyDatabase.bDpcdProbingForBusyWaiting,      DP_REG_VAL_BOOL}
 };
 
 EvoMainLink::EvoMainLink(EvoInterface * provider, Timer * timer) :