Re: [PATCH v16 16/16] arm64: kexec_file: add kaslr support

[AKASHI Takahiro]

On Tue, Dec 11, 2018 at 02:50:02PM +0900, AKASHI Takahiro wrote:
> On Fri, Nov 30, 2018 at 01:19:44PM +0000, Will Deacon wrote:
> > On Thu, Nov 15, 2018 at 02:52:55PM +0900, AKASHI Takahiro wrote:
> > > Adding "kaslr-seed" to dtb enables triggering kaslr, or kernel virtual
> > > address randomization, at secondary kernel boot. We always do this as
> > > it will have no harm on kaslr-incapable kernel.
> > > 
> > > We don't have any "switch" to turn off this feature directly, but still
> > > can suppress it by passing "nokaslr" as a kernel boot argument.
> > > 
> > > Signed-off-by: AKASHI Takahiro <takahiro.akashi@xxxxxxxxxx>
> > > Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
> > > Cc: Will Deacon <will.deacon@xxxxxxx>
> > > ---
> > >  arch/arm64/kernel/machine_kexec_file.c | 46 +++++++++++++++++++++++++-
> > >  1 file changed, 45 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/arch/arm64/kernel/machine_kexec_file.c b/arch/arm64/kernel/machine_kexec_file.c
> > > index ab296b98d633..a0a730bd9be6 100644
> > > --- a/arch/arm64/kernel/machine_kexec_file.c
> > > +++ b/arch/arm64/kernel/machine_kexec_file.c
> > > @@ -16,6 +16,7 @@
> > >  #include <linux/libfdt.h>
> > >  #include <linux/memblock.h>
> > >  #include <linux/of_fdt.h>
> > > +#include <linux/random.h>
> > >  #include <linux/slab.h>
> > >  #include <linux/string.h>
> > >  #include <linux/types.h>
> > > @@ -28,6 +29,7 @@
> > >  #define FDT_PSTR_INITRD_STA	"linux,initrd-start"
> > >  #define FDT_PSTR_INITRD_END	"linux,initrd-end"
> > >  #define FDT_PSTR_BOOTARGS	"bootargs"
> > > +#define FDT_PSTR_KASLR_SEED	"kaslr-seed"
> > >  
> > >  const struct kexec_file_ops * const kexec_file_loaders[] = {
> > >  	&kexec_image_ops,
> > > @@ -46,11 +48,38 @@ int arch_kimage_file_post_load_cleanup(struct kimage *image)
> > >  	return kexec_image_post_load_cleanup_default(image);
> > >  }
> > >  
> > > +/* crng needs to have been initialized for providing kaslr-seed */
> > > +static int random_ready;
> > > +
> > > +static void random_ready_notified(struct random_ready_callback *unused)
> > > +{
> > > +	random_ready = 1;
> > > +}
> > > +
> > > +static struct random_ready_callback random_ready_cb = {
> > > +	.func = random_ready_notified,
> > > +};
> > > +
> > > +static __init int init_random_ready_cb(void)
> > > +{
> > > +	int ret;
> > > +
> > > +	ret = add_random_ready_callback(&random_ready_cb);
> > > +	if (ret == -EALREADY)
> > > +		random_ready = 1;
> > > +	else if (ret)
> > > +		pr_warn("failed to add a callback for random_ready\n");
> > > +
> > > +	return 0;
> > > +}
> > > +late_initcall(init_random_ready_cb)
> > 
> > Why can't we just call crng_ready()?
> 
> because crng_ready() is locally defined in drivers/char/random.c.
> Instead, I'd like to use
>         wait_for_random_bytes();
>         value = get_random_u64();

Correction:
After several tests, I now don't think that calling wait_for_random_bytes()
is a good idea since it can make kexec_file_load() syscall stalled.
So, I would go for
        if (rng_is_initialized())
                value = get_random_u64();

-Takahiro Akashi

> > > +
> > >  static int setup_dtb(struct kimage *image,
> > >  		     unsigned long initrd_load_addr, unsigned long initrd_len,
> > >  		     char *cmdline, void *dtb)
> > >  {
> > >  	int nodeoffset;
> > > +	u64 value;
> > >  	int ret;
> > >  
> > >  	nodeoffset = fdt_path_offset(dtb, "/chosen");
> > > @@ -106,12 +135,27 @@ static int setup_dtb(struct kimage *image,
> > >  			return -EINVAL;
> > >  	}
> > >  
> > > +	/* add kaslr-seed */
> > > +	ret = fdt_delprop(dtb, nodeoffset, FDT_PSTR_KASLR_SEED);
> > > +	if (ret && (ret != -FDT_ERR_NOTFOUND))
> > > +		return -EINVAL;
> > > +
> > > +	if (random_ready) {
> > > +		get_random_bytes(&value, sizeof(value));
> > 
> > get_random_u64() ?
> 
> OK.
> 
> > > +		ret = fdt_setprop_u64(dtb, nodeoffset, FDT_PSTR_KASLR_SEED,
> > > +							value);
> > > +		if (ret)
> > > +			return (ret == -FDT_ERR_NOSPACE ? -ENOMEM : -EINVAL);
> > > +	} else {
> > 
> > Wouldn't we be better off preserving the previous seed here, if it was
> > present?
> 
> While there's no guarantee that dtb won't be (partially) broken
> on failure, I will let this function return successfully
> by deleting succeeding fdt_delprop().
> 
> 
> > > +		pr_notice("kaslr-seed won't be fed\n");
> > 
> > "fed" is probably not the right word here.
> 
> => won't be *provided* on kexec?
> 
> -Takahiro Akashi
> 
> > Will