[PATCH v34 06/14] arm64: kdump: protect crash dump kernel memory

[takahiro.akashi@linaro.org (AKASHI Takahiro)]

Ard,

On Tue, Mar 28, 2017 at 11:07:05AM +0100, Ard Biesheuvel wrote:
> On 28 March 2017 at 07:51, AKASHI Takahiro <takahiro.akashi@linaro.org> wrote:
> > arch_kexec_protect_crashkres() and arch_kexec_unprotect_crashkres()
> > are meant to be called by kexec_load() in order to protect the memory
> > allocated for crash dump kernel once the image is loaded.
> >
> > The protection is implemented by unmapping the relevant segments in crash
> > dump kernel memory, rather than making it read-only as other archs do,
> > to prevent any corruption due to potential cache alias (with different
> > attributes) problem.
> >
> 
> I think it would be more accurate to replace 'corruption' with
> 'coherency issues', given that this patch does not solve the issue of
> writable aliases that may be used to modify the contents of the
> region, but it does prevent issues related to mismatched attributes
> (which are arguably a bigger concern)

OK

> > Page-level mappings are consistently used here so that we can change
> > the attributes of segments in page granularity as well as shrink the region
> > also in page granularity through /sys/kernel/kexec_crash_size, putting
> > the freed memory back to buddy system.
> >
> > Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
> 
> As a head's up, this patch is going to conflict heavily with patches
> that are queued up in arm64/for-next/core atm.

I'll look into it later, but

> Some questions below.
> 
> > ---
> >  arch/arm64/kernel/machine_kexec.c | 32 +++++++++++---
> >  arch/arm64/mm/mmu.c               | 90 ++++++++++++++++++++-------------------
> >  2 files changed, 72 insertions(+), 50 deletions(-)
> >
> > diff --git a/arch/arm64/kernel/machine_kexec.c b/arch/arm64/kernel/machine_kexec.c
> > index bc96c8a7fc79..b63baa749609 100644
> > --- a/arch/arm64/kernel/machine_kexec.c
> > +++ b/arch/arm64/kernel/machine_kexec.c
> > @@ -14,7 +14,9 @@
> >
> >  #include <asm/cacheflush.h>
> >  #include <asm/cpu_ops.h>
> > +#include <asm/mmu.h>
> >  #include <asm/mmu_context.h>
> > +#include <asm/page.h>
> >
> >  #include "cpu-reset.h"
> >
> > @@ -22,8 +24,6 @@
> >  extern const unsigned char arm64_relocate_new_kernel[];
> >  extern const unsigned long arm64_relocate_new_kernel_size;
> >
> > -static unsigned long kimage_start;
> > -
> >  /**
> >   * kexec_image_info - For debugging output.
> >   */
> > @@ -64,8 +64,6 @@ void machine_kexec_cleanup(struct kimage *kimage)
> >   */
> >  int machine_kexec_prepare(struct kimage *kimage)
> >  {
> > -       kimage_start = kimage->start;
> > -
> >         kexec_image_info(kimage);
> >
> >         if (kimage->type != KEXEC_TYPE_CRASH && cpus_are_stuck_in_kernel()) {
> > @@ -183,7 +181,7 @@ void machine_kexec(struct kimage *kimage)
> >         kexec_list_flush(kimage);
> >
> >         /* Flush the new image if already in place. */
> > -       if (kimage->head & IND_DONE)
> > +       if ((kimage != kexec_crash_image) && (kimage->head & IND_DONE))
> >                 kexec_segment_flush(kimage);
> >
> >         pr_info("Bye!\n");
> > @@ -201,7 +199,7 @@ void machine_kexec(struct kimage *kimage)
> >          */
> >
> >         cpu_soft_restart(1, reboot_code_buffer_phys, kimage->head,
> > -               kimage_start, 0);
> > +               kimage->start, 0);
> >
> >         BUG(); /* Should never get here. */
> >  }
> > @@ -210,3 +208,25 @@ void machine_crash_shutdown(struct pt_regs *regs)
> >  {
> >         /* Empty routine needed to avoid build errors. */
> >  }
> > +
> > +void arch_kexec_protect_crashkres(void)
> > +{
> > +       int i;
> > +
> > +       kexec_segment_flush(kexec_crash_image);
> > +
> > +       for (i = 0; i < kexec_crash_image->nr_segments; i++)
> > +               set_memory_valid(
> > +                       __phys_to_virt(kexec_crash_image->segment[i].mem),
> > +                       kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 0);
> > +}
> > +
> > +void arch_kexec_unprotect_crashkres(void)
> > +{
> > +       int i;
> > +
> > +       for (i = 0; i < kexec_crash_image->nr_segments; i++)
> > +               set_memory_valid(
> > +                       __phys_to_virt(kexec_crash_image->segment[i].mem),
> > +                       kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 1);
> > +}
> > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> > index d28dbcf596b6..f6a3c0e9d37f 100644
> > --- a/arch/arm64/mm/mmu.c
> > +++ b/arch/arm64/mm/mmu.c
> > @@ -22,6 +22,8 @@
> >  #include <linux/kernel.h>
> >  #include <linux/errno.h>
> >  #include <linux/init.h>
> > +#include <linux/ioport.h>
> > +#include <linux/kexec.h>
> >  #include <linux/libfdt.h>
> >  #include <linux/mman.h>
> >  #include <linux/nodemask.h>
> > @@ -332,56 +334,31 @@ static void create_mapping_late(phys_addr_t phys, unsigned long virt,
> >                              NULL, debug_pagealloc_enabled());
> >  }
> >
> > -static void __init __map_memblock(pgd_t *pgd, phys_addr_t start, phys_addr_t end)
> > +static void __init __map_memblock(pgd_t *pgd, phys_addr_t start,
> > +                                 phys_addr_t end, pgprot_t prot,
> > +                                 bool page_mappings_only)
> > +{
> > +       __create_pgd_mapping(pgd, start, __phys_to_virt(start), end - start,
> > +                            prot, early_pgtable_alloc,
> > +                            page_mappings_only);
> > +}
> > +
> > +static void __init map_mem(pgd_t *pgd)
> >  {
> >         phys_addr_t kernel_start = __pa_symbol(_text);
> >         phys_addr_t kernel_end = __pa_symbol(__init_begin);
> > +       struct memblock_region *reg;
> >
> >         /*
> > -        * Take care not to create a writable alias for the
> > -        * read-only text and rodata sections of the kernel image.
> > +        * Temporarily marked as NOMAP to skip mapping in the next for-loop
> >          */
> > +       memblock_mark_nomap(kernel_start, kernel_end - kernel_start);
> >
> 
> OK, so the trick is to mark a memblock region NOMAP temporarily, so
> that we can iterate over the regions more easily?
> Is that the sole reason for using NOMAP in this series?

Yes. (I followed Mark's suggestion.)

So I assume that my change here will be essentially orthogonal
with the chnages in for-next/core, at least, in its intent.

Thanks,
-Takahiro AKASHI

> > -       /* No overlap with the kernel text/rodata */
> > -       if (end < kernel_start || start >= kernel_end) {
> > -               __create_pgd_mapping(pgd, start, __phys_to_virt(start),
> > -                                    end - start, PAGE_KERNEL,
> > -                                    early_pgtable_alloc,
> > -                                    debug_pagealloc_enabled());
> > -               return;
> > -       }
> > -
> > -       /*
> > -        * This block overlaps the kernel text/rodata mappings.
> > -        * Map the portion(s) which don't overlap.
> > -        */
> > -       if (start < kernel_start)
> > -               __create_pgd_mapping(pgd, start,
> > -                                    __phys_to_virt(start),
> > -                                    kernel_start - start, PAGE_KERNEL,
> > -                                    early_pgtable_alloc,
> > -                                    debug_pagealloc_enabled());
> > -       if (kernel_end < end)
> > -               __create_pgd_mapping(pgd, kernel_end,
> > -                                    __phys_to_virt(kernel_end),
> > -                                    end - kernel_end, PAGE_KERNEL,
> > -                                    early_pgtable_alloc,
> > -                                    debug_pagealloc_enabled());
> > -
> > -       /*
> > -        * Map the linear alias of the [_text, __init_begin) interval as
> > -        * read-only/non-executable. This makes the contents of the
> > -        * region accessible to subsystems such as hibernate, but
> > -        * protects it from inadvertent modification or execution.
> > -        */
> > -       __create_pgd_mapping(pgd, kernel_start, __phys_to_virt(kernel_start),
> > -                            kernel_end - kernel_start, PAGE_KERNEL_RO,
> > -                            early_pgtable_alloc, debug_pagealloc_enabled());
> > -}
> > -
> > -static void __init map_mem(pgd_t *pgd)
> > -{
> > -       struct memblock_region *reg;
> > +#ifdef CONFIG_KEXEC_CORE
> > +       if (crashk_res.end)
> > +               memblock_mark_nomap(crashk_res.start,
> > +                                   resource_size(&crashk_res));
> > +#endif
> >
> >         /* map all the memory banks */
> >         for_each_memblock(memory, reg) {
> > @@ -393,8 +370,33 @@ static void __init map_mem(pgd_t *pgd)
> >                 if (memblock_is_nomap(reg))
> >                         continue;
> >
> > -               __map_memblock(pgd, start, end);
> > +               __map_memblock(pgd, start, end,
> > +                              PAGE_KERNEL, debug_pagealloc_enabled());
> > +       }
> > +
> > +       /*
> > +        * Map the linear alias of the [_text, __init_begin) interval as
> > +        * read-only/non-executable. This makes the contents of the
> > +        * region accessible to subsystems such as hibernate, but
> > +        * protects it from inadvertent modification or execution.
> > +        */
> > +       __map_memblock(pgd, kernel_start, kernel_end,
> > +                      PAGE_KERNEL_RO, debug_pagealloc_enabled());
> > +       memblock_clear_nomap(kernel_start, kernel_end - kernel_start);
> > +
> > +#ifdef CONFIG_KEXEC_CORE
> > +       /*
> > +        * User page-level mappings here so that we can shrink the region
> > +        * in page granularity and put back unused memory to buddy system
> > +        * through /sys/kernel/kexec_crash_size interface.
> > +        */
> > +       if (crashk_res.end) {
> > +               __map_memblock(pgd, crashk_res.start, crashk_res.end + 1,
> > +                              PAGE_KERNEL, true);
> > +               memblock_clear_nomap(crashk_res.start,
> > +                                    resource_size(&crashk_res));
> >         }
> > +#endif
> >  }
> >
> >  void mark_rodata_ro(void)
> > --
> > 2.11.1
> >