Account NvMap allocated memory into both RSS and CG tracking to make
efficient OOM kill decisions during memory pressure.
NvMap allocates memory via kernel APIs like alloc_pages, the kernel
memory is not accounted on behalf of process who requests the
allocation. Hence in case OOM, the OOM killer never kills the process
who has allocated memory via NvMap even though this process might be
holding most of the memory.
Solve this issue using following approach:
- Use __GFP_ACCOUNT and __GFP_NORETRY flag
- __GFP_NORETRY will not let the current allocation flow to go into OOM
path, so that it will never trigger OOM.
- __GFP_ACCOUNT causes the allocation to be accounted to kmemcg. So any
allocation done by NvMap will be definitely accounted to kmemcg and
cgroups can be used to define memory limits.
- Add RSS counting for the process which allocates by NvMap, so that OOM
score for that process will get updated and OOM killer can pick this
process based upon the OOM score.
- Every process that has a reference to NvMap Handle would have the
memory size accounted into its RSS. On releasing the reference to
handle, the RSS would be reduced.
Bug 5222690
Change-Id: I3fa9b76ec9fc8d7f805111cb96e11e2ab1db42ce
Signed-off-by: Ketan Patil <ketanp@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3427871
Reviewed-by: Krishna Reddy <vdumpa@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
In Linux v5.18, commits 20c238dfb26c ("media: v4l2-mediabus: Drop legacy
V4L2_MBUS_CSI2_*_LANE flags"), 5a6ac3f4b46f ("media: v4l2-mediabus: Drop
legacy V4L2_MBUS_CSI2_CHANNEL_* flags") and b9f7caa7753a ("media:
v4l2-mediabus: Drop V4L2_MBUS_CSI2_CONTINUOUS_CLOCK flag") removed
various CSI2 flags. It turns out that out of these flags we only need to
actually set the number of lanes and the other flags are redundant. The
configuration was moved by commit 94d964e58ad6 ("media: v4l2-fwnode:
Move bus config structure to v4l2_mediabus.h") to the structure
v4l2_mbus_config_mipi_csi2 which itself is under the structure
v4l2_fwnode_endpoint. Use conftest to detect if the structure
v4l2_mbus_config_mipi_csi2 is present and avoid using kernel version
checks.
Bug 4425688
Change-Id: Icd597492e48831326c203a6556176ac3767472a3
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3354475
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
Reviewed-by: Paritosh Dixit <paritoshd@nvidia.com>
In Linux v6.12, commit 641bb4394f40 ("fs: move FMODE_UNSIGNED_OFFSET to
fop_flags") moves the setting of the FMODE_UNSIGNED_OFFSET into the
various drivers and renames this flag. This change is causing various
tests to fail because opening the DRM device is failing.
Use the presence of the new FOP_UNSIGNED_OFFSET flag in the kernel to
determine if we need to set this flag in the Tegra DRM driver.
Bug 4876974
Change-Id: Ie72a9340187b1c206f0cda6d7e3dab44d603a41c
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3269063
(cherry picked from commit a680b4881fb8e6ace46f409c33307f3bc10366a4)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3338872
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
Tested-by: Ilies Chergui <ichergui@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: Ilies Chergui <ichergui@nvidia.com>
After programming the TX Register, it takes 3-4 clock cycles for TX
data synchronization to the CEC core domain. The TX_EMPTY interrupt
should not be cleared until this data synchronization is complete.
Enable CEC HW feature to automatically clears the TX_EMPTY interrupt
after TX data synchronization is complete. It takes time for HW to
clear the interrupt and TX_EMPTY can still appear high, hence SW
needs to poll WR_LOCK until it goes to 0. This will avoid SW to
attempt the next TX block during the same TX_EMPTY interrupt.
Also read RX_REGISTER based on the buffer occupancy which is
indicated by the CEC_RX_BUFFER_STAT_0 register.
Bug 4954851
Change-Id: I3ec3792c9ae3b8a00c800c921cf4e4d09369e6b9
Signed-off-by: Ken Chang <kenc@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3322519
(cherry picked from commit 94cf6106402dc833dbaa4305c3e1a8f85fd80d0e)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3336992
Reviewed-by: Bitan Biswas <bbiswas@nvidia.com>
Reviewed-by: Prafull Suryawanshi <prafulls@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
The RealTek drivers do not build against all Linux distributions because
some Linux distributions have back-ported upstream changes and so the
KERNEL_VERSION checks in the driver do not work for these distributions.
Add compilation flags for these drivers so that we can skip the building
of these drivers for certain Linux distributions.
Bug 5180190
Change-Id: I48101b5b9e199c6887a2cbde916d7df5ae1c53a7
Signed-off-by: Paritosh Dixit <paritoshd@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3325283
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
The tegra-virt-storage driver is currently not used for the L4T in the
rel-36 codeline. Add an option to disable building the driver. It
provides flexibility to disable the driver in case there are build
issues with 3rd-party distro kernels.
Bug 5180190
Change-Id: If58d8e782c564aa7b5229dc96093050c19882011
Signed-off-by: Paritosh Dixit <paritoshd@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3323464
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: svcacv <svcacv@nvidia.com>
It may happen that the variable req->iv may have stale values or
zero sized buffer by default and may end up getting used during
encryption/decryption. This in turn may corrupt the results or break
the operation. Ensure not use IV for algorithms like AES ECB where IV
is not expected.
Bug 4883011
Change-Id: I309ad26f99fe54005ff71b270b3b115dc62ac168
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3319062
Reviewed-by: svcacv <svcacv@nvidia.com>
Tested-by: Brad Griffis <bgriffis@nvidia.com>
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
The HW supports only storing 15 keys at a time. This limits the number
of tfms that can work without failutes. Reserve keyslots to solve this
and use the reserved ones during the encryption/decryption operation.
This allow users to have the capability of hardware protected keys
and faster operations if there are limited number of tfms while not
halting the operation if there are more tfms.
Bug 4883011
Change-Id: I220f1e8205dde1f078be6ed4cb09b699b6d5dfa2
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3263283
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
In Linux v6.13, commit 94a20fb9af16 ("sysfs: treewide: constify
attribute callback of bin_attribute::mmap()") changed the type of the
'struct bin_attribute' argument of the bin_attribute:mmap function
pointer to const. Use conftest to detect if this argument is const or
not in the kernel the driver is being compiled against to fix the build.
Bug 4991705
Change-Id: I791d6c9dde50f1444d0339ebd8a18434045f9026
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3261699
(cherry picked from commit fae042552e211199a160e5d6a82a67edd7b7e03c)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3321698
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
During SC7 tsec driver is unable to find the desc file for the
tsec fw and thus tsec boot fails. This might result into system
hang. Hence this change adds the tsec fw desc file path as the
module firmware so that the firmware is available to it during
SC7.
Bug 5053273
Change-Id: I4f2c9bb5288dac6973f574159fdebc48009edf2a
Signed-off-by: spatki <spatki@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3308515
Reviewed-by: Nikesh Oswal <noswal@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
This merge imports tegra-cec driver from linux-nvidia repo to nvidia-oot repo
with maintaining git history of tegra-cec from start.
bug 4954851
Change-Id: Icb8d066df5eeaa083a70a7f6657caf1c1593ec17
Signed-off-by: Prafull Suryawanshi <prafulls@nvidia.com>
This change includes below -
1. For T23x, CEC engine is moved to AON partition. As AON partition
does not powergate, skip all powergate/unpowergate calls for t23x.
2. The clock on which CEC engine operates, 32k/rtc_clk, it is always
running clock in AON partition. So skip clock enable/disable APIs.
3. This change also skips calls to TEGRA DC engine which is not available
from T23x.
4. RX Buffer size is increased to 64 depth from 1 depth. So new register is
provided to program depth so that interrupt gets triggered when certain
depth is filled. By default it is 64 depth so program it to 1 depth
to match old behavior.
5. When tegradc is absent, read physical address from SPARE register which
will be populated by UEFI.
6. This change also adds new register definitions in header file which
needs to program when adding support for CEC SC7 wakeup.
7. Timing registers updated for T23x as per HW guidelines.
8. Some initialization code is changed for T23x.
9. Add devicetree binding documentation.
bug 200727002
Change-Id: I71c2d323632c61e4c4b82bcdbca9e72179761224
Signed-off-by: Prafull Suryawanshi <prafulls@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2523595
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: svc_kernel_abi <svc_kernel_abi@nvidia.com>
Reviewed-by: Shu Zhong <shuz@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Add genpd/pm_runtime support to handle power domains
in Tegra Display driver. Also updated CEC and HDA_DC
drivers to add pm_runtime calls.
Added new APIs in CEC and SOR drivers to abstract
calls to pm_runtime and tegra_powergate APIs based
on specific Kernel version. Guarded powergate_id
with CONFIG_TEGRA_POWERGATE since its not needed
when pm_runtime APIs are used, and the powergate IDs
are defined in tegra_powergate.h which is not used
in K5.9.
Removed pm_runtime autosuspend setting since display
driver doesn't assign PM suspend/resume hooks at
present. Suspend/resume calls are linked to platform
device callbacks.
Removed pm_runtime_enable() call in dc_common probe
since there's logically no need of power-domains
in dc_common and also there is no power-domains node
in display_hub node in device-tree. Same with DSI.
All the power-domains handled by each DC are listed
in "power-domains" and "power-domain-names" properties
under respective dc nodes in device tree.
bug 200601926
Change-Id: I0fc9d21ff56185a69de6e0663b5932a93cf40e2b
Signed-off-by: Naveen Kumar S <nkumars@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2358953
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Shu Zhong <shuz@nvidia.com>
Reviewed-by: Ujwal Patel <ujwalp@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
This change fixes below follow-up issues with the earlier fix
to access_ok() API done as part of
https://git-master.nvidia.com/r/c/linux-nvidia/+/2361042:
1. Definition of VERIFY_READ and VERIFY_WRITE have been removed
after K4.14. Hence, used them only for kernel versions < K5.4.
2. Changed access_type parameter of tegra_cec_access_ok() API
to bool. WRITE or READ will be chosen based on the bool value.
3. In the last CL, VERIFY_WRITE was changed to VERIFY_READ in
two access_ok() calls by mistake. Corrected them in this CL.
4. Defined access_type variable with attribute __maybe_unused
since gcc doesn't seem to consider a variable as used if only
passed as function parameter.
bug 200601926
Change-Id: I9f3540cd028280ca679b4b69d5b4a72aa943b2ae
Signed-off-by: Naveen Kumar S <nkumars@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvidia/+/2394585
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Shu Zhong <shuz@nvidia.com>
Reviewed-by: Ujwal Patel <ujwalp@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
