linux.git/arch/x86/kernel/devicetree.c, branch v6.8

x86/of: Move the x86_flattree_get_config() call out of x86_dtb_init()

2023-10-02T19:30:09+00:00

Fetching the device tree configuration before initmem_init() is necessary
to allow the parsing of NUMA node information. However moving the entire
x86_dtb_init() call before initmem_init() is not correct as APIC/IO-APIC enumeration
has to be after initmem_init().

Thus, move the x86_flattree_get_config() call out of x86_dtb_init(),
into setup_arch(), to call it before initmem_init(), and
leave the ACPI/IOAPIC registration sequence as-is.

[ mingo: Updated the changelog for clarity. ]

Signed-off-by: Saurabh Sengar 
Signed-off-by: Ingo Molnar 
Cc: Sebastian Andrzej Siewior 
Link: https://lore.kernel.org/r/1692949657-16446-1-git-send-email-ssengar@linux.microsoft.com

x86/apic: Make some APIC init functions bool

2023-08-09T18:58:20+00:00

Quite some APIC init functions are pure boolean, but use the success = 0,
fail < 0 model. That's confusing as hell when reading through the code.

Convert them to boolean.

Signed-off-by: Thomas Gleixner 
Signed-off-by: Dave Hansen 
Acked-by: Peter Zijlstra (Intel) 
Tested-by: Michael Kelley 
Tested-by: Sohil Mehta 
Tested-by: Juergen Gross  # Xen PV (dom0 and unpriv. guest)

x86/of: Fix the APIC address registration

2023-08-09T18:58:19+00:00

The device tree APIC parser tries to force-enable the local APIC when it is
not set in CPUID. apic_force_enable() registers the boot CPU apic on
success.

If that succeeds then dtb_lapic_setup() registers the local APIC again
eventually with a different address.

Rewrite the code so that it only registers it once.

Signed-off-by: Thomas Gleixner 
Signed-off-by: Dave Hansen 
Acked-by: Peter Zijlstra (Intel) 
Tested-by: Michael Kelley 
Tested-by: Sohil Mehta 
Tested-by: Juergen Gross  # Xen PV (dom0 and unpriv. guest)

x86/apic: Remove the pointless APIC version check

2023-08-09T18:58:19+00:00

This historical leftover is really uninteresting today. Whatever MPTABLE or
MADT delivers we only trust the hardware anyway.

Signed-off-by: Thomas Gleixner 
Signed-off-by: Dave Hansen 
Acked-by: Peter Zijlstra (Intel) 
Tested-by: Michael Kelley 
Tested-by: Sohil Mehta 
Tested-by: Juergen Gross  # Xen PV (dom0 and unpriv. guest)

x86/of: Add support for boot time interrupt delivery mode configuration

2022-12-02T13:57:14+00:00

Presently, init/boot time interrupt delivery mode is enumerated only for
ACPI enabled systems by parsing MADT table or for older systems by parsing
MP table. But for OF based x86 systems, it is assumed & hardcoded to be
legacy PIC mode. This causes a boot time crash for platforms which do not
provide a 8259 compliant legacy PIC.

Add support for configuration of init time interrupt delivery mode for x86
OF based systems by introducing a new optional boolean property
'intel,virtual-wire-mode' for the local APIC interrupt-controller
node. This property emulates IMCRP Bit 7 of MP feature info byte 2 of MP
floating pointer structure.

Defaults to legacy PIC mode if absent. Configures it to virtual wire
compatibility mode if present.

Signed-off-by: Rahul Tanwar 
Signed-off-by: Thomas Gleixner 
Reviewed-by: Andy Shevchenko 
Link: https://lore.kernel.org/r/20221124084143.21841-5-rtanwar@maxlinear.com

x86/of: Replace printk(KERN_LVL) with pr_lvl()

2022-12-02T13:57:14+00:00

Use pr_lvl() instead of the deprecated printk(KERN_LVL).

Just a upgrade of print utilities usage. no functional changes.

Suggested-by: Andy Shevchenko 
Signed-off-by: Rahul Tanwar 
Signed-off-by: Thomas Gleixner 
Reviewed-by: Andy Shevchenko 
Link: https://lore.kernel.org/r/20221124084143.21841-4-rtanwar@maxlinear.com

x86/of: Remove unused early_init_dt_add_memory_arch()

2022-12-02T13:57:13+00:00

Recently objtool started complaining about dead code in the object files,
in particular

vmlinux.o: warning: objtool: early_init_dt_scan_memory+0x191: unreachable instruction

when CONFIG_OF=y.

Indeed, early_init_dt_scan() is not used on x86 and making it compile (with
help of CONFIG_OF) will abrupt the code flow since in the middle of it
there is a BUG() instruction.

Remove the pointless function.

Signed-off-by: Andy Shevchenko 
Signed-off-by: Thomas Gleixner 
Link: https://lore.kernel.org/r/20221124184824.9548-1-andriy.shevchenko@linux.intel.com

Merge tag 'devicetree-for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux

2021-11-03T05:22:13+00:00

Pull devicetree updates from Rob Herring:

 - Convert /reserved-memory bindings to schemas

 - Convert a bunch of NFC bindings to schemas

 - Convert bindings to schema: Xilinx USB, Freescale DDR controller, Arm
   CCI-400, UBlox Neo-6M, 1-Wire GPIO, MSI controller, ASpeed LPC, OMAP
   and Inside-Secure HWRNG, register-bit-led, OV5640, Silead GSL1680,
   Elan ekth3000, Marvell bluetooth, TI wlcore, TI bluetooth, ESP
   ESP8089, tlm,trusted-foundations, Microchip cap11xx, Ralink SoCs and
   boards, and TI sysc

 - New binding schemas for: msi-ranges, Aspeed UART routing controller,
   palmbus, Xylon LogiCVC display controller, Mediatek's MT7621 SDRAM
   memory controller, and Apple M1 PCIe host

 - Run schema checks for %.dtb targets

 - Improve build time when using DT_SCHEMA_FILES

 - Improve error message when dtschema is not found

 - Various doc reference fixes in MAINTAINERS

 - Convert architectures to common CPU h/w ID parsing function
   of_get_cpu_hwid().

 - Allow for empty NUMA node IDs which may be hotplugged

 - Cleanup of __fdt_scan_reserved_mem()

 - Constify device_node parameters

 - Update dtc to upstream v1.6.1-19-g0a3a9d3449c8. Adds new checks
   'node_name_vs_property_name' and 'interrupt_map'.

 - Enable dtc 'unit_address_format' warning by default

 - Fix unittest EXPECT text for gpio hog errors

* tag 'devicetree-for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux: (97 commits)
  dt-bindings: net: ti,bluetooth: Document default max-speed
  dt-bindings: pci: rcar-pci-ep: Document r8a7795
  dt-bindings: net: qcom,ipa: IPA does support up to two iommus
  of/fdt: Remove of_scan_flat_dt() usage for __fdt_scan_reserved_mem()
  of: unittest: document intentional interrupt-map provider build warning
  of: unittest: fix EXPECT text for gpio hog errors
  of/unittest: Disable new dtc node_name_vs_property_name and interrupt_map warnings
  scripts/dtc: Update to upstream version v1.6.1-19-g0a3a9d3449c8
  dt-bindings: arm: firmware: tlm,trusted-foundations: Convert txt bindings to yaml
  dt-bindings: display: tilcd: Fix endpoint addressing in example
  dt-bindings: input: microchip,cap11xx: Convert txt bindings to yaml
  dt-bindings: ufs: exynos-ufs: add exynosautov9 compatible
  dt-bindings: ufs: exynos-ufs: add io-coherency property
  dt-bindings: mips: convert Ralink SoCs and boards to schema
  dt-bindings: display: xilinx: Fix example with psgtr
  dt-bindings: net: nfc: nxp,pn544: Convert txt bindings to yaml
  dt-bindings: Add a help message when dtschema tools are missing
  dt-bindings: bus: ti-sysc: Update to use yaml binding
  dt-bindings: sram: Allow numbers in sram region node name
  dt-bindings: display: Document the Xylon LogiCVC display controller
  ...

x86/of: Kill unused early_init_dt_scan_chosen_arch()

2021-10-25T08:56:37+00:00

There are no callers for early_init_dt_scan_chosen_arch(), so remove it.

Signed-off-by: Rob Herring 
Signed-off-by: Borislav Petkov 
Reviewed-by: Frank Rowand 
Link: https://lkml.kernel.org/r/20211022164642.2815706-1-robh@kernel.org

x86: dt: Use of_get_cpu_hwid()

2021-10-20T18:37:24+00:00

Replace open coded parsing of CPU nodes' 'reg' property with
of_get_cpu_hwid().

Cc: Thomas Gleixner 
Cc: Ingo Molnar 
Cc: Borislav Petkov 
Cc: x86@kernel.org
Cc: "H. Peter Anvin" 
Signed-off-by: Rob Herring 
Link: https://lore.kernel.org/r/20211006164332.1981454-11-robh@kernel.org