[6.18] Rebase ARM commits from 4.14 kernel

[chaitu236]

Rebase ARM commits from 4.14 kernel.

Majority of commits had merge conflicts or needed updates to be able to build on 6.18 kernel. I've recorded details in commit messages itself.

There are some commits I haven't cherry-picked either due to merge conflicts that required more detailed analysis or due to them failing build.

WI: AB#3966407, AB#3990497, AB#3990492, AB#4000899, AB#4000886

Testing

• Built ni-pkg target
• Boot tested resulting .itb on cRIO-9068, sbRIO-9651.
• Built scarthgap BSI, safemode with this kernel.
• Boot tested BSI on cRIO-9063, cRIO-9066, cRIO-9068, sbRIO-9607, sbRIO-9651, Elvis RIO CM, Elvis III, roboRIO, myRIO-1900, myRIO-1950, NI-9147, NI-9149, NI-7935.
• Hardware Manager/MAX can recognize devices and install software.
• Installed RSS on cRIO-9068 and could run a simple LV VI.
• Built for x64.
• Boot tested on x64 VM.

Issues

• Some warn splats related to ni16550 that needs looking into. Will create WI.

[chaitu236]

v2: Add /ni-install/ to gitignore.

[chaitu236]

v3: Updated some commit messages.

[chaitu236]

v4: savedefconfig (which I forgot to do previously). Enabled other configs required for CONFIG_RCU_NOCB_CPU_DEFAULT_ALL in nati_slsc_defconfig, nati_bluefin_defconfig.

Note: It's not yet clear if CONFIG_RCU_NOCB_CPU_DEFAULT_ALL is required in nati_slsc_defconfig, nati_bluefin_defconfig; will remove the changes later if not required.

[chaitu236]

v5: Cherry-pick few more commits that were left out as part of original rebase due to rebase/build failures. Expanded github-actions to add ARM checks.

[chaitu236]

Note: [TEST] Is rebase required failed due to fetch-depth 100 not being sufficient and not due to this branch not being based on latest nilrt/master/6.18. Not planning to increase fetch-depth as it only becomes a problem during rebases which happen infrequently.

[chaitu236]

v6: Disable CONFIG_LOCALVERSION_AUTO so some drivers can compile.

[gratian]

@chaitu236 I am about half way through the list of commits. Here is what I've found so far:

• The config options added by the commits below are not currently enabled in our defconfigs, ex. nati_zynq_defconfig. We either need to match the 4.14 configs or drop these commits if we decide they are not adding value (preferred option but would require running the kernel performance benchmarks with a network load):

  • "driver core: NI backwards compatibility for device_poll interface"
  • "net: add device_poll functionality common to all net devices"
  • "driver core: add device_poll interface"

• Please drop commit "watchdog: nic7018_wdt: Add support for timeout interrupt". It is already in tree. The new commit adds a duplicated devm_request_threaded_irq() call right below the one in-tree.

• Commit "Revert "irqchip/gic: Kill the xlate method"" is part of a set with other "irqchip: gic", currently not cherry-picked.

  • It is supposed to be used with the commits adding CONFIG_GIC_LEGACY_IRQDOMAIN which were not cherry-picked, see AB#3990488 and AB#4000886
  • Consider dropping or cherry picking as a set (if the workaround is needed).

• The whole set of pl353 set of commits needs a very close review.

  • We need to switch to the in-tree pl35x-nand-controller driver instead of cherry-picking an old out-of-tree Xilinx driver.
  • Bluefin already uses the in-tree pl353 NAND driver. We need to to the same for the rest of the Zynq-based devices. Also see: Rectify Bluefin NAND on-die ECC support (6.12) #229

• Commit "fpgaperipheral: Add driver to notify clients on FPGA downloads" has some leftover conflict resolution text in drivers/misc/Makefile (still has ">>>>>>> b00bd56 (fpgaperipheral: Add driver to notify clients on FPGA downloads)")

• From what I can tell the commits below are not needed (one is the logical revert of the other):

  • "nirtfeatures: BIOS writes HARD_BOOT_N, so driver doesn't need to"
  • "nirtfeatures: set HARD_BOOT_N if necessary when the driver loads"

• Nitpick, commit "shared: Adding mcopy syscall for Zynq and x64" has a white-space alignment issue.

[chaitu236]

@gratian

v7: Resolved review feedback

The config options added by the commits below are not currently enabled in our defconfigs, ex. nati_zynq_defconfig. We either need to match the 4.14 configs or drop these commits if we decide they are not adding value (preferred option but would require running the kernel performance benchmarks with a network load):

Dropped. Added them to AB#3990492 for further evaluation.

Please drop commit "watchdog: nic7018_wdt: Add support for timeout interrupt". It is already in tree. The new commit adds a duplicated devm_request_threaded_irq() call right below the one in-tree.

Done

Commit "Revert "irqchip/gic: Kill the xlate method"" is part of a set with other "irqchip: gic", currently not cherry-picked.

Dropped. Added to AB#3990488.

The whole set of pl353 set of commits needs a very close review.

It appears that we didn't have it's config enabled in 6.18 and things still seem to work anyway. Dropped all these commits.

mtd: nand: pl353: Use mtd_ooblayout_xxx() helpers
mtd: nand: pl353: Convert nand_ecclayout structs to mtd_ooblayout_ops
pl353_nand: Add dsb(st) (store-only data sync barrier) after reg writes
pl353_nand: Add module params to disable subpage read and write
nand: pl353: Fix private data's use of mtd_info
mtd: nand: xilinx: Rename the pl353 NAND driver to match bootloader
Xilinx: ARM: nand: Deassert CS when the chip is no longer in use
Xilinx: ARM: nand: Check read status when using on-die ECC
Xilinx: ARM: nand: Replace READ0 of the last read page with RNDOUT
Zynq NAND: enable subpage reads and writes
mtd: nand: pl353: add snapshot of driver from linux-xlnx tree
HACK: ARM: zynq: established early fixed mapping for PL353 NAND region

Commit "fpgaperipheral: Add driver to notify clients on FPGA downloads" has some leftover conflict resolution text in drivers/misc/Makefile (still has ">>>>>>> b00bd56 (fpgaperipheral: Add driver to notify clients on FPGA downloads)")

Fixed

From what I can tell the commits below are not needed (one is the logical revert of the other):

Dropped both

Nitpick, commit "shared: Adding mcopy syscall for Zynq and x64" has a white-space alignment issue.

Fixed

[chaitu236]

v8: Dropped mmc patches. i.e., following

mmc: sdhci: Do not disable interrupts in sdhci_set_clock
mmc: sdhci-pci: Use two delay quirks for Baytrail SD
mmc: sdhci: Add quirk work around double rescan
mmc: sdhci: Add quirk for delay between clock disable and param change
mmc: sdhci: Disable SD card clock before changing parameters
mmc: core: Wait for Vdd to settle on card power off

Build, installed BSI on NI-7935R and verified sdcard still works.