Re: [PATCH v2 1/3] LoongArch: Add kexec support

[Youling Tang]

Hi, Huacai

On 09/09/2022 04:16 PM, Huacai Chen wrote:
> Hi, Youling,
>
> On Fri, Sep 9, 2022 at 11:20 AM Youling Tang <tangyouling@xxxxxxxxxxx> wrote:
> > Add three new files, kexec.h, machine_kexec.c and relocate_kernel.S to the
> > LoongArch architecture that add support for the kexec re-boot mechanis
> > (CONFIG_KEXEC) on LoongArch platforms.
> >
> > Supports loading vmlinux (vmlinux.elf) in ELF format and vmlinux.efi in
> > PE format.
> >
> > I tested this on  LoongArch 3A5000 machine and works as expected,
> >
> >  $ sudo kexec -l /boot/vmlinux.efi --reuse-cmdline
> >  $ sudo kexec -e
> >
> > Signed-off-by: Youling Tang <tangyouling@xxxxxxxxxxx>
> > ---
> >  arch/loongarch/Kconfig                  |  11 ++
> >  arch/loongarch/include/asm/kexec.h      |  58 ++++++++
> >  arch/loongarch/kernel/Makefile          |   2 +
> >  arch/loongarch/kernel/head.S            |   7 +-
> >  arch/loongarch/kernel/machine_kexec.c   | 188 ++++++++++++++++++++++++
> >  arch/loongarch/kernel/relocate_kernel.S | 106 +++++++++++++
> >  6 files changed, 371 insertions(+), 1 deletion(-)
> >  create mode 100644 arch/loongarch/include/asm/kexec.h
> >  create mode 100644 arch/loongarch/kernel/machine_kexec.c
> >  create mode 100644 arch/loongarch/kernel/relocate_kernel.S
> >
> > diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
> > index 9b1f2ab878e9..08e063aaf847 100644
> > --- a/arch/loongarch/Kconfig
> > +++ b/arch/loongarch/Kconfig
> > @@ -422,6 +422,17 @@ config ARCH_IOREMAP
> >           protection support. However, you can enable LoongArch DMW-based
> >           ioremap() for better performance.
> >
> > +config KEXEC
> > +       bool "Kexec system call"
> > +       select KEXEC_CORE
> > +       help
> > +         kexec is a system call that implements the ability to shutdown your
> > +         current kernel, and to start another kernel.  It is like a reboot
> > +         but it is independent of the system firmware.   And like a reboot
> > +         you can start any kernel with it, not just Linux.
> > +
> > +         The name comes from the similarity to the exec system call.
> > +
> >  config SECCOMP
> >         bool "Enable seccomp to safely compute untrusted bytecode"
> >         depends on PROC_FS
> > diff --git a/arch/loongarch/include/asm/kexec.h b/arch/loongarch/include/asm/kexec.h
> > new file mode 100644
> > index 000000000000..f23506725e00
> > --- /dev/null
> > +++ b/arch/loongarch/include/asm/kexec.h
> > @@ -0,0 +1,58 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +/*
> > + * kexec.h for kexec
> > + *
> > + * Copyright (C) 2022 Loongson Technology Corporation Limited
> > + */
> > +
> > +#ifndef _ASM_KEXEC_H
> > +#define _ASM_KEXEC_H
> > +
> > +#include <asm/stacktrace.h>
> > +#include <asm/page.h>
> > +
> > +/* Maximum physical address we can use pages from */
> > +#define KEXEC_SOURCE_MEMORY_LIMIT (-1UL)
> > +/* Maximum address we can reach in physical address mode */
> > +#define KEXEC_DESTINATION_MEMORY_LIMIT (-1UL)
> > + /* Maximum address we can use for the control code buffer */
> > +#define KEXEC_CONTROL_MEMORY_LIMIT (-1UL)
> > +
> > +/* Reserve a page for the control code buffer */
> > +#define KEXEC_CONTROL_PAGE_SIZE PAGE_SIZE
> > +
> > +/* The native architecture */
> > +#define KEXEC_ARCH KEXEC_ARCH_LOONGARCH
> > +
> > +static inline void crash_setup_regs(struct pt_regs *newregs,
> > +                                   struct pt_regs *oldregs)
> > +{
> > +       if (oldregs)
> > +               memcpy(newregs, oldregs, sizeof(*newregs));
> > +       else
> > +               prepare_frametrace(newregs);
> > +}
> > +
> > +#define ARCH_HAS_KIMAGE_ARCH
> > +
> > +struct kimage_arch {
> > +       unsigned long boot_flag;
> > +       unsigned long fdt_addr;
> > +};
> I prefer to change boot_flag to efi_boot, the latter is better to
> correspond the current usage, and keeps consistency with efistub.

OK.
> > +
> > +typedef void (*do_kexec_t)(unsigned long boot_flag,
> > +                          unsigned long fdt_addr,
> > +                          unsigned long first_ind_entry,
> > +                          unsigned long jump_addr);
> I prefer change the order of jump_addr and first_ind_entry here.

OK.
> > +
> > +struct kimage;
> > +extern const unsigned char relocate_new_kernel[];
> > +extern const size_t relocate_new_kernel_size;
> > +extern void kexec_reboot(void);
> > +
> > +#ifdef CONFIG_SMP
> > +extern atomic_t kexec_ready_to_reboot;
> > +extern const unsigned char kexec_smp_wait[];
> > +#endif
> > +
> > +#endif /* !_ASM_KEXEC_H */
> > diff --git a/arch/loongarch/kernel/Makefile b/arch/loongarch/kernel/Makefile
> > index 7225916dd378..17dc8ce6b5ce 100644
> > --- a/arch/loongarch/kernel/Makefile
> > +++ b/arch/loongarch/kernel/Makefile
> > @@ -17,6 +17,8 @@ obj-$(CONFIG_CPU_HAS_FPU)     += fpu.o
> >  obj-$(CONFIG_MODULES)          += module.o module-sections.o
> >  obj-$(CONFIG_STACKTRACE)       += stacktrace.o
> >
> > +obj-$(CONFIG_KEXEC)             += machine_kexec.o relocate_kernel.o
> > +
> >  obj-$(CONFIG_PROC_FS)          += proc.o
> >
> >  obj-$(CONFIG_SMP)              += smp.o
> > diff --git a/arch/loongarch/kernel/head.S b/arch/loongarch/kernel/head.S
> > index eb3f641d5915..0f786d670e66 100644
> > --- a/arch/loongarch/kernel/head.S
> > +++ b/arch/loongarch/kernel/head.S
> > @@ -20,7 +20,12 @@
> >
> >  _head:
> >         .word   MZ_MAGIC                /* "MZ", MS-DOS header */
> > -       .org    0x3c                    /* 0x04 ~ 0x3b reserved */
> > +       .org    0x8
> > +       .quad   0                       /* Image load offset from start of RAM */
> > +       .dword  _end - _text            /* Effective size of kernel image */
> > +       .quad   0
> > +       .dword  kernel_entry            /* Kernel entry point */
> > +       .org    0x3c                    /* 0x28 ~ 0x3b reserved */
> >         .long   pe_header - _head       /* Offset to the PE header */
> As discussed before, I prefer kernel entry point before kernel image size.

I will modify it in the next version, and will modify the kexec-tools
tool at the same time.

> >  pe_header:
> > diff --git a/arch/loongarch/kernel/machine_kexec.c b/arch/loongarch/kernel/machine_kexec.c
> > new file mode 100644
> > index 000000000000..7b3fea506c6d
> > --- /dev/null
> > +++ b/arch/loongarch/kernel/machine_kexec.c
> > @@ -0,0 +1,188 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * machine_kexec.c for kexec
> > + *
> > + * Copyright (C) 2022 Loongson Technology Corporation Limited
> > + */
> > +#include <linux/compiler.h>
> > +#include <linux/cpu.h>
> > +#include <linux/kexec.h>
> > +#include <linux/mm.h>
> > +#include <linux/delay.h>
> > +#include <linux/libfdt.h>
> > +#include <linux/of_fdt.h>
> > +
> > +#include <asm/bootinfo.h>
> > +#include <asm/cacheflush.h>
> > +#include <asm/page.h>
> > +
> > +/* 0x100000 ~ 0x200000 is safe */
> > +#define KEXEC_CTRL_CODE        TO_CACHE(0x100000UL)
> > +#define KEXEC_BLOB_ADDR        TO_CACHE(0x108000UL)
> > +
> > +static unsigned long reboot_code_buffer;
> > +
> > +#ifdef CONFIG_SMP
> > +static void (*relocated_kexec_smp_wait)(void *);
> > +atomic_t kexec_ready_to_reboot = ATOMIC_INIT(0);
> > +#endif
> > +
> > +static unsigned long jump_addr;
> > +static unsigned long first_ind_entry;
> > +static unsigned long boot_flag;
> > +static unsigned long fdt_addr;
> Keep the same order as do_kexec() may be better.

OK.
> > +
> > +static void kexec_image_info(const struct kimage *kimage)
> > +{
> > +       unsigned long i;
> > +
> > +       pr_debug("kexec kimage info:\n");
> > +       pr_debug("\ttype:        %d\n", kimage->type);
> > +       pr_debug("\tstart:       %lx\n", kimage->start);
> > +       pr_debug("\thead:        %lx\n", kimage->head);
> > +       pr_debug("\tnr_segments: %lu\n", kimage->nr_segments);
> > +
> > +       for (i = 0; i < kimage->nr_segments; i++) {
> > +               pr_debug("\t    segment[%lu]: %016lx - %016lx", i,
> > +                       kimage->segment[i].mem,
> > +                       kimage->segment[i].mem + kimage->segment[i].memsz);
> > +               pr_debug("\t\t0x%lx bytes, %lu pages\n",
> > +                       (unsigned long)kimage->segment[i].memsz,
> > +                       (unsigned long)kimage->segment[i].memsz /  PAGE_SIZE);
> > +       }
> > +}
> > +
> > +int machine_kexec_prepare(struct kimage *kimage)
> > +{
> > +       int i;
> > +       void *dtb = (void *)KEXEC_BLOB_ADDR;
> > +
> > +       kexec_image_info(kimage);
> > +
> > +       /* Find the Flattened Device Tree */
> > +       for (i = 0; i < kimage->nr_segments; i++) {
> > +               if (!fdt_check_header(kimage->segment[i].buf)) {
> > +                       memcpy(dtb, kimage->segment[i].buf, SZ_64K);
> > +                       kimage->arch.boot_flag = fw_arg0;
> > +                       kimage->arch.fdt_addr = (unsigned long) dtb;
> > +                       break;
> > +               }
> > +               continue;
> > +       }
> > +
> > +       /* kexec need a safe page to save reboot_code_buffer */
> > +       kimage->control_code_page = virt_to_page((void *)KEXEC_CTRL_CODE);
> > +
> > +       reboot_code_buffer =
> > +         (unsigned long)page_address(kimage->control_code_page);
> > +       memcpy((void *)reboot_code_buffer, relocate_new_kernel,
> > +              relocate_new_kernel_size);
> > +
> > +#ifdef CONFIG_SMP
> > +       /* All secondary cpus now may jump to kexec_smp_wait cycle */
> > +       relocated_kexec_smp_wait = reboot_code_buffer +
> > +               (void *)(kexec_smp_wait - relocate_new_kernel);
> > +#endif
> > +
> > +       return 0;
> > +}
> > +
> > +void machine_kexec_cleanup(struct kimage *kimage)
> > +{
> > +}
> > +
> > +void kexec_reboot(void)
> > +{
> > +       do_kexec_t do_kexec = NULL;
> > +
> The old implementation has an "ibar" here, I suppose it is still needed?

The new implementation no longer has to turn off the non-boot cpus and
then turn it back on when doing multi-core kexec operations, so I don't
think there's any need to use ibar to make sure the remote CPUs are
online (as I didn't make it go offline).

> > +#ifdef CONFIG_SMP
> > +       /* All secondary cpus go to kexec_smp_wait */
> > +       if (smp_processor_id() > 0) {
> > +               relocated_kexec_smp_wait(NULL);
> > +               unreachable();
> > +       }
> > +#endif
> > +
> > +       do_kexec = (void *)reboot_code_buffer;
> > +       do_kexec(boot_flag, fdt_addr, first_ind_entry, jump_addr);
> > +
> > +       unreachable();
> > +}
> > +
> > +
> > +#ifdef CONFIG_SMP
> > +static void kexec_shutdown_secondary(void *)
> > +{
> > +       local_irq_disable();
> > +       while (!atomic_read(&kexec_ready_to_reboot))
> > +               cpu_relax();
> > +
> > +       kexec_reboot();
> The old implementation has an unreachable() after kexec_reboot(), but
> I don't know whether it is really needed.

I have marked as unreachable() in kexec_reboot, IMO, the place where
kexec_reboot is called no longer needs to be marked as unreachable().

> > +}
> > +#endif
> > +
> > +void machine_crash_shutdown(struct pt_regs *regs)
> > +{
> > +}
> > +
> > +void machine_shutdown(void)
> > +{
> > +#ifdef CONFIG_SMP
> > +       smp_call_function(kexec_shutdown_secondary, NULL, 0);
> > +#endif
> > +}
> > +
> > +void machine_kexec(struct kimage *image)
> > +{
> > +       unsigned long entry;
> > +       unsigned long *ptr;
> > +       struct kimage_arch *internal = &image->arch;
> > +
> > +       boot_flag = internal->boot_flag;
> > +       fdt_addr = internal->fdt_addr;
> > +
> > +       jump_addr = (unsigned long)phys_to_virt(image->start);
> > +
> > +       first_ind_entry = (unsigned long)phys_to_virt(image->head & PAGE_MASK);
> > +
> > +       /*
> > +        * The generic kexec code builds a page list with physical
> > +        * addresses. they are directly accessible through XKPRANGE
> > +        * hence the phys_to_virt() call.
> > +        */
> > +       for (ptr = &image->head; (entry = *ptr) && !(entry & IND_DONE);
> > +            ptr = (entry & IND_INDIRECTION) ?
> > +              phys_to_virt(entry & PAGE_MASK) : ptr + 1) {
> > +               if (*ptr & IND_SOURCE || *ptr & IND_INDIRECTION ||
> > +                   *ptr & IND_DESTINATION)
> > +                       *ptr = (unsigned long) phys_to_virt(*ptr);
> > +       }
> > +
> > +       /* Mark offline before disabling local irq. */
> > +       set_cpu_online(smp_processor_id(), false);
> > +
> > +       /* we do not want to be bothered. */
> > +       local_irq_disable();
> > +
> > +       pr_notice("Will call new kernel at %lx\n", jump_addr);
> > +       pr_notice("FDT image at %lx\n", fdt_addr);
> > +       pr_notice("Bye ...\n");
> > +
> > +       /* Make reboot code buffer available to the boot CPU. */
> > +       flush_cache_all();
> > +
> > +#ifdef CONFIG_SMP
> > +       atomic_set(&kexec_ready_to_reboot, 1);
> > +#endif
> > +
> > +       /*
> > +        * We know we were online, and there will be no incoming IPIs at
> > +        * this point.
> > +        */
> > +       set_cpu_online(smp_processor_id(), true);
> > +
> > +       /* Ensure remote CPUs observe that we're online before rebooting. */
> > +       smp_mb__after_atomic();
> > +
> > +       kexec_reboot();
> > +}
> > diff --git a/arch/loongarch/kernel/relocate_kernel.S b/arch/loongarch/kernel/relocate_kernel.S
> > new file mode 100644
> > index 000000000000..05a0c1ccfab6
> > --- /dev/null
> > +++ b/arch/loongarch/kernel/relocate_kernel.S
> > @@ -0,0 +1,106 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +/*
> > + * relocate_kernel.S for kexec
> > + *
> > + * Copyright (C) 2022 Loongson Technology Corporation Limited
> > + */
> > +
> > +#include <linux/kexec.h>
> > +
> > +#include <asm/asm.h>
> > +#include <asm/asmmacro.h>
> > +#include <asm/regdef.h>
> > +#include <asm/loongarch.h>
> > +#include <asm/stackframe.h>
> > +#include <asm/addrspace.h>
> > +
> > +SYM_CODE_START(relocate_new_kernel)
> > +       /*
> > +        * s0: Boot flag passed to the new kernel
> > +        * s1: Virt address of the FDT image
> > +        * s2: Pointer to the current entry
> "Pointer to the current entry" can be "Pointer to the current
> indirection page entry"? And I think we needn't backup Ax to Sx except
> this one in the assembly below.

Since there are no function calls in between, I can remove the backup
of the a-series registers.

Thanks,
Youling.

> Huacai
>
> > +        * s3: Virt address to jump to after relocation
> > +        */
> > +       move            s0, a0
> > +       move            s1, a1
> > +       move            s2, a2
> > +       move            s3, a3
> > +
> > +process_entry:
> > +       PTR_L           s4, s2, 0
> > +       PTR_ADDI        s2, s2, SZREG
> > +
> > +       /* destination page */
> > +       andi            s5, s4, IND_DESTINATION
> > +       beqz            s5, 1f
> > +       li.w            t0, ~0x1
> > +       and             s6, s4, t0      /* store destination addr in s6 */
> > +       b               process_entry
> > +
> > +1:
> > +       /* indirection page, update s2  */
> > +       andi            s5, s4, IND_INDIRECTION
> > +       beqz            s5, 1f
> > +       li.w            t0, ~0x2
> > +       and             s2, s4, t0
> > +       b               process_entry
> > +
> > +1:
> > +       /* done page */
> > +       andi            s5, s4, IND_DONE
> > +       beqz            s5, 1f
> > +       b               done
> > +1:
> > +       /* source page */
> > +       andi            s5, s4, IND_SOURCE
> > +       beqz            s5, process_entry
> > +       li.w            t0, ~0x8
> > +       and             s4, s4, t0
> > +       li.w            s8, (1 << _PAGE_SHIFT) / SZREG
> > +
> > +copy_word:
> > +       /* copy page word by word */
> > +       REG_L           s7, s4, 0
> > +       REG_S           s7, s6, 0
> > +       PTR_ADDI        s6, s6, SZREG
> > +       PTR_ADDI        s4, s4, SZREG
> > +       LONG_ADDI       s8, s8, -1
> > +       beqz            s8, process_entry
> > +       b               copy_word
> > +       b               process_entry
> > +
> > +done:
> > +       ibar            0
> > +       dbar            0
> > +
> > +       move            a0, s0
> > +       move            a1, s1
> > +       /* jump to the new kernel */
> > +       jr              s3
> > +SYM_CODE_END(relocate_new_kernel)
> > +
> > +#ifdef CONFIG_SMP
> > +/*
> > + * Other CPUs should wait until code is relocated and
> > + * then start at entry (?) point.
> > + */
> > +SYM_CODE_START(kexec_smp_wait)
> > +1:     li.w            t0, 0x100                       /* wait for init loop */
> > +2:     addi.w          t0, t0, -1                      /* limit mailbox access */
> > +       bnez            t0, 2b
> > +       li.w            t1, LOONGARCH_IOCSR_MBUF0
> > +       iocsrrd.w       s1, t1                          /* check PC as an indicator */
> > +       beqz            s1, 1b
> > +       iocsrrd.d       s1, t1                          /* get PC via mailbox */
> > +
> > +       li.d            t0, CACHE_BASE
> > +       or              s1, s1, t0                      /* s1 = TO_CACHE(s1) */
> > +       jr              s1                              /* jump to initial PC */
> > +SYM_CODE_END(kexec_smp_wait)
> > +#endif
> > +
> > +relocate_new_kernel_end:
> > +
> > +SYM_DATA_START(relocate_new_kernel_size)
> > +       PTR             relocate_new_kernel_end - relocate_new_kernel
> > +SYM_DATA_END(relocate_new_kernel_size)
> > --
> > 2.36.0