> > In addition, do some cleanup: add infor about reloction function to
>
> infor ? reloction?
Typo. I have fixed commit log. It meant to be "info about
arm64_relocate_new_kernel function"
>
>
> > kexec_image_info(), and remove extra messages from machine_kexec().
>
>
> > Make dtb_mem, always available, if CONFIG_KEXEC_FILE is not configured
> > dtb_mem is set to zero anyway.
>
> This is unrelated cleanup, please do it as an earlier patch to make it clearer what you
> are changing here.
Ok.
>
> (I'm not convinced you need to cache va<->pa)
>
>
> > diff --git a/arch/arm64/include/asm/kexec.h b/arch/arm64/include/asm/kexec.h
> > index 12a561a54128..d15ca1ca1e83 100644
> > --- a/arch/arm64/include/asm/kexec.h
> > +++ b/arch/arm64/include/asm/kexec.h
> > @@ -90,14 +90,15 @@ static inline void crash_prepare_suspend(void) {}
> > static inline void crash_post_resume(void) {}
> > #endif
> >
> > -#ifdef CONFIG_KEXEC_FILE
> > #define ARCH_HAS_KIMAGE_ARCH
> >
> > struct kimage_arch {
> > void *dtb;
> > unsigned long dtb_mem;
>
> > + unsigned long kern_reloc;
>
> This is cache-ing the physical address of an all-architectures value from struct kimage,
> in the arch specific part of struct kiamge. Why?
> (You must have the struct kimage on hand to access this thing at all!)
Because, currently only one physical page is used in order to do reboot:
control_code_page; this is where arm64_relocate_new_kernel is copied.
So, PA of control_code_page is used as a handler. But with MMU enabled
kexec reboot, a number of pages are allocated for reboot purposes,
they contain page table, arm64_relocate_new_kernel, and el2 vector
table. This is why we need handlers. control_code_page is simply one
of the pages that can contains any of the kexec reboot specific data.
> If its supposed to be a physical address, please use phys_addr_t.
Done that, also changed dtb_mem to phys_addr_t.
>
> > };
>
>
> > diff --git a/arch/arm64/kernel/machine_kexec.c b/arch/arm64/kernel/machine_kexec.c
> > index 0df8493624e0..9b41da50e6f7 100644
> > --- a/arch/arm64/kernel/machine_kexec.c
> > +++ b/arch/arm64/kernel/machine_kexec.c
> > @@ -42,6 +42,7 @@ static void _kexec_image_info(const char *func, int line,
> > pr_debug(" start: %lx\n", kimage->start);
> > pr_debug(" head: %lx\n", kimage->head);
> > pr_debug(" nr_segments: %lu\n", kimage->nr_segments);
> > + pr_debug(" kern_reloc: %pa\n", &kimage->arch.kern_reloc);
> >
> > for (i = 0; i < kimage->nr_segments; i++) {
> > pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
> > @@ -58,6 +59,19 @@ void machine_kexec_cleanup(struct kimage *kimage)
> > /* Empty routine needed to avoid build errors. */
> > }
> >
> > +int machine_kexec_post_load(struct kimage *kimage)
> > +{
> > + unsigned long kern_reloc;
> > +
> > + kern_reloc = page_to_phys(kimage->control_code_page);
>
> kern_reloc should be phys_addr_t.
Ok
>
>
> > + memcpy(__va(kern_reloc), arm64_relocate_new_kernel,
> > + arm64_relocate_new_kernel_size);
> > + kimage->arch.kern_reloc = kern_reloc;
>
>
> Please move the cache maintenance in here too. This will save us doing it late during
> kdump. This will also group the mmu-on changes together.
OK, but I think we should do it in a separate patch. I assume you mean
this code to be moved to load time:
177 /* Flush the reboot_code_buffer in preparation for its execution. */
178 __flush_dcache_area(reboot_code_buffer,
arm64_relocate_new_kernel_size);
179
180 /*
181 * Although we've killed off the secondary CPUs, we don't update
182 * the online mask if we're handling a crash kernel and consequently
183 * need to avoid flush_icache_range(), which will attempt to IPI
184 * the offline CPUs. Therefore, we must use the __* variant here.
185 */
186 __flush_icache_range((uintptr_t)reboot_code_buffer,
187 arm64_relocate_new_kernel_size);
Is it safe to do? We do not know what CPU is going to be executing
kexec reboot for us at the load time.
Pasha
