nvidia-tegra-modules.git/drivers/iommu, branch gpu-paging NVIDIA's kernel modules to support tegra chips (used in Jetson boards) tegra-smmu: Fix grep in 12000000.iommu debugfs. 2021-10-06T15:56:17+00:00 Achal Verma achalv@nvidia.com 2021-09-20T08:28:34+00:00 916bb90add1c9a589b70d80d6c409e8caddce673 grep in /sys/kernel/debug/12000000.iommu/ failing because inside masters/ some devices didn't have pointers set for cfg, so added check in debugfs show for NULL pointer. And other issue is accessing registers inside gnsr dir. These registers are not ioremapped or access to them is not allowed in case of hypervisor. So removed gnsr node creation in hypervisor case. Bug 200772838 Change-Id: I2dc0cbe9f71db90246a60de5db27158b555737a4 Signed-off-by: Achal Verma <achalv@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2597304 Reviewed-by: Dmitry Pervushin <dpervushin@nvidia.com> Reviewed-by: Amrita Deshmukh <amritad@nvidia.com> GVS: Gerrit_Virtual_Submit Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2606024 Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> Reviewed-by: Phoenix Jung <pjung@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> 
grep in /sys/kernel/debug/12000000.iommu/ failing
because inside masters/ some devices didn't have pointers
set for cfg, so added check in debugfs show for NULL pointer.

And other issue is accessing registers inside gnsr dir.
These registers are not ioremapped or access to them is
not allowed in case of hypervisor. So removed gnsr node
creation in hypervisor case.

Bug 200772838

Change-Id: I2dc0cbe9f71db90246a60de5db27158b555737a4
Signed-off-by: Achal Verma <achalv@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2597304
Reviewed-by: Dmitry Pervushin <dpervushin@nvidia.com>
Reviewed-by: Amrita Deshmukh <amritad@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2606024
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: Phoenix Jung <pjung@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: arm-smmu-t19x: fix linearmap 2020-02-27T15:41:07+00:00 Pritesh Raithatha praithatha@nvidia.com 2020-02-26T07:16:39+00:00 ee27a7ee99182192485923d20fc554e3e43c0d5e linearmap should be done before SMMU translation is enabled. Update code accordingly. Bug 200572719 Change-Id: I14fa18edf7b7d20619757ace8da48b978f087faa Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2302997 Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> Reviewed-by: Automatic_Commit_Validation_User Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> 
linearmap should be done before SMMU translation is enabled.
Update code accordingly.

Bug 200572719

Change-Id: I14fa18edf7b7d20619757ace8da48b978f087faa
Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2302997
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: arm-smmu: t19x: free pgd using free_pages() 2020-01-17T00:01:53+00:00 Stephen Warren swarren@nvidia.com 2020-01-14T22:50:37+00:00 4fa874bc385332d4d768d8faaf2ecab3208b9d5a The pgd value was allocated using __get_free_pages() and so must be freed using a page-oriented API, not a kmalloc-oriented API. Bug 2792255 Change-Id: If48da608fa92d788909efcc10a0a950a949dd146 Fixes: 9ff237010c8e ("iommu: arm-smmu: copy t18x version driver") Signed-off-by: Stephen Warren <swarren@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2279146 GVS: Gerrit_Virtual_Submit Reviewed-by: Puneet Saxena <puneets@nvidia.com> Reviewed-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
The pgd value was allocated using __get_free_pages() and so must be freed
using a page-oriented API, not a kmalloc-oriented API.

Bug 2792255

Change-Id: If48da608fa92d788909efcc10a0a950a949dd146
Fixes: 9ff237010c8e ("iommu: arm-smmu: copy t18x version driver")
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2279146
GVS: Gerrit_Virtual_Submit
Reviewed-by: Puneet Saxena <puneets@nvidia.com>
Reviewed-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
kernel: nvidia: smmu: Add check for SMR reg init 2019-04-25T14:12:37+00:00 Suresh Venkatachalam skathirampat@nvidia.com 2019-04-24T02:41:52+00:00 ce9d0873bcb1b141a47257076d3a6bd6b565327f Add check to initialize SMR registers only when IDR0 indicates SMR mode This saves ~1.6ms during boot in L+EBP config Bug 2441521 Change-Id: I55d6b59270c7bf618ab79c14ccb557e7680bca7c Signed-off-by: Suresh Venkatachalam <skathirampat@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/2103851 GVS: Gerrit_Virtual_Submit Reviewed-by: Ashok Kumar Sekar <akumars@nvidia.com> Reviewed-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-by: Sachin Nikam <snikam@nvidia.com> Tested-by: Prakhar Gupta <prakharg@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Add check to initialize SMR registers only when IDR0 indicates
SMR mode
This saves ~1.6ms during boot in L+EBP config

Bug 2441521

Change-Id: I55d6b59270c7bf618ab79c14ccb557e7680bca7c
Signed-off-by: Suresh Venkatachalam <skathirampat@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2103851
GVS: Gerrit_Virtual_Submit
Reviewed-by: Ashok Kumar Sekar <akumars@nvidia.com>
Reviewed-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
Tested-by: Prakhar Gupta <prakharg@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: arm-smmu: add master debugfs remove support 2018-09-20T10:31:24+00:00 Pritesh Raithatha praithatha@nvidia.com 2018-09-10T07:41:43+00:00 0337d318793b077f3c4acc8529b3379e241bf14d Current code is not removing master debugfs when device is removed. It caused crash when smmu_test is trying to access the debugfs. Add support to remove debugfs when device is removed. Bug 200438310 Change-Id: I19aedf79e9982ce1b77192def271a9741e696e62 Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1817976 Reviewed-by: Sachin Nikam <snikam@nvidia.com> GVS: Gerrit_Virtual_Submit Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Current code is not removing master debugfs when device is removed.
It caused crash when smmu_test is trying to access the debugfs.
Add support to remove debugfs when device is removed.

Bug 200438310

Change-Id: I19aedf79e9982ce1b77192def271a9741e696e62
Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1817976
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: of_tegra-smmu: Include file header 2018-09-06T04:23:43+00:00 David Gilhooley dgilhooley@nvidia.com 2018-09-04T21:36:16+00:00 5ec70d925888780c6ee12233ade4b6dde89808c2 Include the of-tegra-smmu header which defines the functions that are in this file Bug 200434802 Change-Id: I9d3354afe3acd1b5d28cf4a42bcea54a3e742ba1 Signed-off-by: David Gilhooley <dgilhooley@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1812743 Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Include the of-tegra-smmu header which
defines the functions that are in this file

Bug 200434802

Change-Id: I9d3354afe3acd1b5d28cf4a42bcea54a3e742ba1
Signed-off-by: David Gilhooley <dgilhooley@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1812743
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: arm-smmu-suspend: use double pointer for single SMMU 2018-08-31T18:23:55+00:00 Pritesh Raithatha praithatha@nvidia.com 2018-08-31T06:17:17+00:00 8f6d16c9fde5362ddb5a8e9d369bf6e243e6074b Current code uses smmu base pointer based on number of SMMUs. For single SMMU base addr is used as single pointer and for multiple SMMUs base addr is used as double pointer to extract base address of all SMMUs. Update code to use double pointer base addr for all. Bug 2196203 Change-Id: I197c3d89acf6d3d365b664e521753257f4b5bae8 Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1810491 GVS: Gerrit_Virtual_Submit Reviewed-by: Sachin Nikam <snikam@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Current code uses smmu base pointer based on number of SMMUs.
For single SMMU base addr is used as single pointer and for
multiple SMMUs base addr is used as double pointer to extract
base address of all SMMUs.

Update code to use double pointer base addr for all.

Bug 2196203

Change-Id: I197c3d89acf6d3d365b664e521753257f4b5bae8
Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1810491
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
iommu: of_tegra-smmu: Use non-device memory 2018-08-30T05:40:05+00:00 David Gilhooley dgilhooley@nvidia.com 2018-08-27T17:15:40+00:00 0f8378d7eff2425f5ef358e53acb6c21fff9fc04 Use normal memory allocation instead of device memory for tegra_iommu_groups. We attach these structs to a global linked list, so we shouldn't be freeing the memory when a device exits. Add a global free function instead. Bug 2343965 Change-Id: I2bc466d65f6024c6c731503b92cec7075d22c532 Signed-off-by: David Gilhooley <dgilhooley@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1807583 Tested-by: Bitan Biswas <bbiswas@nvidia.com> GVS: Gerrit_Virtual_Submit Reviewed-by: Krishna Reddy <vdumpa@nvidia.com> Reviewed-by: Puneet Saxena <puneets@nvidia.com> Reviewed-by: Ashish Mhetre <amhetre@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Use normal memory allocation instead of device
memory for tegra_iommu_groups. We attach
these structs to a global linked list, so
we shouldn't be freeing the memory when
a device exits.
Add a global free function instead.

Bug 2343965

Change-Id: I2bc466d65f6024c6c731503b92cec7075d22c532
Signed-off-by: David Gilhooley <dgilhooley@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1807583
Tested-by: Bitan Biswas <bbiswas@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Krishna Reddy <vdumpa@nvidia.com>
Reviewed-by: Puneet Saxena <puneets@nvidia.com>
Reviewed-by: Ashish Mhetre <amhetre@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
arm: smmu: tegra: Fix ptdump address values 2018-08-27T16:13:23+00:00 Stephen Warren swarren@nvidia.com 2018-08-24T19:09:47+00:00 65089c2c315e044a24319d93f43b7851861c814f Fix smmu_ptdump_show() so that the body of each loop always increments addr by the appropriate amount, irrespective of whether any child tables or PTEs are present or not. This removes the need for bizarre special cases, and corrects the output in the debugfs file. For example, in the existing code, if 4M of entries at IOVA 0xff800000 were valid, the debugfs code would print 2M of entries at IOVA ff800000 and 2M of entries at ffc00000! Bug 2312071 Change-Id: I6d0df7d29909782bfbd950fea49c9d7a23ef0ccc Signed-off-by: Stephen Warren <swarren@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1806368 Reviewed-by: Puneet Saxena <puneets@nvidia.com> GVS: Gerrit_Virtual_Submit Reviewed-by: David Gilhooley <dgilhooley@nvidia.com> Reviewed-by: Krishna Reddy <vdumpa@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
Fix smmu_ptdump_show() so that the body of each loop always increments
addr by the appropriate amount, irrespective of whether any child tables
or PTEs are present or not. This removes the need for bizarre special
cases, and corrects the output in the debugfs file. For example, in the
existing code, if 4M of entries at IOVA 0xff800000 were valid, the debugfs
code would print 2M of entries at IOVA ff800000 and 2M of entries at
ffc00000!

Bug 2312071

Change-Id: I6d0df7d29909782bfbd950fea49c9d7a23ef0ccc
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1806368
Reviewed-by: Puneet Saxena <puneets@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: David Gilhooley <dgilhooley@nvidia.com>
Reviewed-by: Krishna Reddy <vdumpa@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Revert "iommu: arm-smmu: Remove Contiguous Hints" 2018-08-17T15:57:37+00:00 Pritesh Raithatha praithatha@nvidia.com 2018-08-16T18:06:32+00:00 6b7c821cdefacd90e6ef51118e91a58e1fcacb26 To improve perf, change 3a8db247e7f80e562b8e265adea637782a343aa8 and 81022ddbf62688c4f699da7c50eaf46e21ecf18f has removed contiguous page support and narrowed down domain spin lock from handle mapping to page table allocation. But the changes have caused intermediate issue where context faults are seen even though the pages are already mapped so reverting both changes. This reverts commit 3a8db247e7f80e562b8e265adea637782a343aa8. Bug 200433040 Bug 200434288 Bug 2327543 Change-Id: I4b73c04b5fb0d35ca93610d468172d3622d1deb4 Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1801270 GVS: Gerrit_Virtual_Submit Reviewed-by: Sachin Nikam <snikam@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> 
To improve perf, change 3a8db247e7f80e562b8e265adea637782a343aa8
and 81022ddbf62688c4f699da7c50eaf46e21ecf18f has removed contiguous
page support and narrowed down domain spin lock from handle mapping to
page table allocation. But the changes have caused intermediate issue
where context faults are seen even though the pages are already mapped
so reverting both changes.

This reverts commit 3a8db247e7f80e562b8e265adea637782a343aa8.

Bug 200433040
Bug 200434288
Bug 2327543

Change-Id: I4b73c04b5fb0d35ca93610d468172d3622d1deb4
Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1801270
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>