Porting Linux to a new processor architectureKernel Recipes 2015

Joël Porquet

September 30th, 2015

Background

Research engineer at LIP6 from 2013 to 2015

TSAREU-funded projectMassively parallel architectureShared and hardware-maintained coherent memory

cluster(n,n)

cluster(0,0)

MIPS32

L1 cache+ MMU

MemoryCache

MIPS32

L1 cache+ MMU

MIPS32

L1 cache+ MMU

MIPS32

L1 cache+ MMU

Local Crossbar

ExternalRAM

INT/timercontroller

DMAcontroller

Joël Porquet Kernel Recipes 2015 September 30th, 2015 2 / 14

Is a new port necessary?*

New board with already-supported processorNew processor within an existing, already supported processor familyNew processor architecture

$ mkdir arch/tsar

*https://lwn.net/Articles/597351/Joël Porquet Kernel Recipes 2015 September 30th, 2015 3 / 14

How to start?

Two-step process1 Minimal set of files that define a minimal set of symbols2 Gradual implementation of the boot functions

$ ls -l arch/tsar/configs/

tsar_defconfiginclude/

asm/uapi/asm/

kernel/vmlinux.lds.S

lib/mm/KconfigMakefile

Joël Porquet Kernel Recipes 2015 September 30th, 2015 4 / 14

The boot sequence

kernel_entry*start_kernel

setup_arch*trap_init*mm_init

mem_init*init_IRQ*time_init*rest_init

kernel_thread(kernel_init)kernel_thread(kthreadd)cpu_startup_entry

Joël Porquet Kernel Recipes 2015 September 30th, 2015 5 / 14

Early assembly boot code

kernel_entry()resets the processor to a default stateclears the bss segmentsaves the bootloader argument(s) (e.g. device tree)initializes the first page table

maps the kernel image

enables the virtual memory and jumps into the virtual address spacesets up the stack register (and optionally the current thread info register)jumps to start_kernel()

Joël Porquet Kernel Recipes 2015 September 30th, 2015 6 / 14

setup_arch()

Scans the flattened device tree, discovers the physical memory banks andregisters them into the memblock layerParses the early arguments (e.g. early_printk)Configures memblock and maps the low memoryMemory zones (ZONE_DMA, ZONE_NORMAL, ZONE_HIGHMEM)

PhysicalMemory

VirtualMemory

user space kernel space

DirectMapping vmalloc

0GiB 3GiB 4GiB

lowmem highmem

pkmap

Joël Porquet Kernel Recipes 2015 September 30th, 2015 7 / 14

trap_init()

Exception vectorThe exception vector acts as a dispatcher:

mfc0 k1, CP0_CAUSEandi k1, k1, 0x7clw k0, exception_handlers(k1)jr k0

Configures the processor to use this exception vectorInitializes exception_handlers with the sub-handlers (handle_int,handle_bp, etc.)

Joël Porquet Kernel Recipes 2015 September 30th, 2015 8 / 14

Trap infrastructure

ENTRY(handle_int)SAVE_ALLCLImove a0, spla ra, ret_from_intrj do_IRQ

ENDPROC(handle_int)

ENTRY(handle_bp)SAVE_ALLSTImove a0, spla ra, ret_from_exceptionj do_bp

ENDPROC(handle_bp)

/* CLI: switch to pure kernel mode and disable interruptions *//* STI: switch to pure kernel mode and enable interruptions */

do_* are C functions:

void do_bp(struct pt_regs *regs){

die_if_kernel("do_bp in kernel",regs);

force_sig(SIGTRAP, current);}

Joël Porquet Kernel Recipes 2015 September 30th, 2015 9 / 14

mem_init()

Releases the free memory from memblock to the buddy allocator (aka thepage allocator)

Memory: 257916k/262144k available (1412k kernel code, 4228k reserved,267k data, 84k bss, 169k init, 0k highmem)Virtual kernel memory layout:

vmalloc : 0xd0800000 - 0xfffff000 ( 759 MB)lowmem : 0xc0000000 - 0xd0000000 ( 256 MB)

.init : 0xc01a5000 - 0xc01ba000 ( 84 kB)

.data : 0xc01621f8 - 0xc01a4fe0 ( 267 kB)

.text : 0xc00010c0 - 0xc01621f8 (1412 kB)

Joël Porquet Kernel Recipes 2015 September 30th, 2015 10 / 14

init_IRQ()

Scans device tree and finds all the nodes identified as interrupt controllers.icu: xicu {

compatible = "soclib,vci_xicu";interrupt-controller;#interrupt-cells = <1>;reg = <0x0 0xf0000000 0x1000>;

};

→ First device driver!

Joël Porquet Kernel Recipes 2015 September 30th, 2015 11 / 14

time_init()

Parses clock provider nodesclocks {

freq: frequency@25MHz {#clock-cells = <0>;compatible = "fixed-clock";clock-frequency = <25000000>;

};};

Parses clocksource nodesClock-source device (monotonic counter)Clock-event device (counts periods of time and raises interrupts)

→ Second device driver!

Joël Porquet Kernel Recipes 2015 September 30th, 2015 12 / 14

To init

Process managementSetting up the stack for new threadsSwitching between threads (switch_to())

Page fault handlerCatching memory faults

System callsList of system callsEnhancement of the interrupt and exception handler

Signal managementExecution of the signal handlers

User-space memory accessSetting up the exception table

Joël Porquet Kernel Recipes 2015 September 30th, 2015 13 / 14

Conclusion

After the initial portPush it upstream if possibleKeep it up-to-dateEnhance it: SMP, NUMA, drivers, etc.

Bedtime readingSeries of articles on LWN.net

The basics: https://lwn.net/Articles/654783/The early code: https://lwn.net/Articles/656286/To the finish line: https://lwn.net/Articles/657939/

Joël Porquet Kernel Recipes 2015 September 30th, 2015 14 / 14