[PATCH v5 4/5] powerpc/vdso: Wire up getrandom() vDSO implementation on VDSO32

[Christophe Leroy]

Hi Jason,

Le 06/09/2024 à 05:24, Jason A. Donenfeld a écrit :
> On Fri, Sep 06, 2024 at 04:48:28AM +0200, Jason A. Donenfeld wrote:
>> On Thu, Sep 05, 2024 at 10:41:40PM +0200, Jason A. Donenfeld wrote:
>>> On Thu, Sep 05, 2024 at 06:13:29PM +0200, Jason A. Donenfeld wrote:
>>>>> +/*
>>>>> + * The macro sets two stack frames, one for the caller and one for the callee
>>>>> + * because there are no requirement for the caller to set a stack frame when
>>>>> + * calling VDSO so it may have omitted to set one, especially on PPC64
>>>>> + */
>>>>> +
>>>>> +.macro cvdso_call funct
>>>>> +  .cfi_startproc
>>>>> +	PPC_STLU	r1, -PPC_MIN_STKFRM(r1)
>>>>> +  .cfi_adjust_cfa_offset PPC_MIN_STKFRM
>>>>> +	mflr		r0
>>>>> +	PPC_STLU	r1, -PPC_MIN_STKFRM(r1)
>>>>> +  .cfi_adjust_cfa_offset PPC_MIN_STKFRM
>>>>> +	PPC_STL		r0, PPC_MIN_STKFRM + PPC_LR_STKOFF(r1)
>>>>> +  .cfi_rel_offset lr, PPC_MIN_STKFRM + PPC_LR_STKOFF
>>>>> +	get_datapage	r8
>>>>> +	addi		r8, r8, VDSO_RNG_DATA_OFFSET
>>>>> +	bl		CFUNC(DOTSYM(\funct))
>>>>> +	PPC_LL		r0, PPC_MIN_STKFRM + PPC_LR_STKOFF(r1)
>>>>> +	cmpwi		r3, 0
>>>>> +	mtlr		r0
>>>>> +	addi		r1, r1, 2 * PPC_MIN_STKFRM
>>>>> +  .cfi_restore lr
>>>>> +  .cfi_def_cfa_offset 0
>>>>> +	crclr		so
>>>>> +	bgelr+
>>>>> +	crset		so
>>>>> +	neg		r3, r3
>>>>> +	blr
>>>>> +  .cfi_endproc
>>>>> +.endm
>>>>
>>>> Can you figure out what's going on and send a fix, which I'll squash
>>>> into this commit?
>>>
>>> This doesn't work, but I wonder if something like it is what we want. I
>>> need to head out for the day, but here's what I've got. It's all wrong
>>> but might be of interest.
>>
>> Oh, I just got one small detail wrong before. The below actually works,
>> and uses the same strategy as on arm64.
>>
>> Let me know if you'd like me to fix up this commit with the below patch,
>> or if you have another way you'd like to go about it.
> 
> And here's the much shorter version in assembly, which maybe you prefer.
> Also works, and is a bit less invasive than the other thing.
> 
> diff --git a/arch/powerpc/kernel/vdso/getrandom.S b/arch/powerpc/kernel/vdso/getrandom.S
> index a957cd2b2b03..070daba2d547 100644
> --- a/arch/powerpc/kernel/vdso/getrandom.S
> +++ b/arch/powerpc/kernel/vdso/getrandom.S
> @@ -32,6 +32,14 @@
>     .cfi_rel_offset r2, PPC_MIN_STKFRM + STK_GOT
>   #endif
>   	get_datapage	r8
> +#ifdef CONFIG_TIME_NS
> +	lis		r10, 0x7fff
> +	ori		r10, r10, 0xffff
> +	lwz		r9, VDSO_DATA_OFFSET + 4(r8)
> +	cmpw		r9, r10
> +	bne		+8
> +	addi		r8, r8, (1 << CONFIG_PAGE_SHIFT)
> +#endif
>   	addi		r8, r8, VDSO_RNG_DATA_OFFSET
>   	bl		CFUNC(DOTSYM(\funct))
>   	PPC_LL		r0, PPC_MIN_STKFRM + PPC_LR_STKOFF(r1)
> 

Thanks for looking.

I came to more or less the same solutions thnt you with the following 
that seems to work:

diff --git a/arch/powerpc/kernel/vdso/vgetrandom.c 
b/arch/powerpc/kernel/vdso/vgetrandom.c
index 5f855d45fb7b..9705344d39d0 100644
--- a/arch/powerpc/kernel/vdso/vgetrandom.c
+++ b/arch/powerpc/kernel/vdso/vgetrandom.c
@@ -4,11 +4,19 @@
   *
   * Copyright (C) 2024 Christophe Leroy <christophe.leroy at csgroup.eu>, 
CS GROUP France
   */
+#include <linux/container_of.h>
  #include <linux/time.h>
  #include <linux/types.h>

+#include <asm/vdso_datapage.h>
+
  ssize_t __c_kernel_getrandom(void *buffer, size_t len, unsigned int 
flags, void *opaque_state,
  			     size_t opaque_len, const struct vdso_rng_data *vd)
  {
+	struct vdso_arch_data *arch_data = container_of(vd, struct 
vdso_arch_data, rng_data);
+
+	if (IS_ENABLED(CONFIG_TIME_NS) && arch_data->data[0].clock_mode == 
VDSO_CLOCKMODE_TIMENS)
+		vd = (void *)vd + (1UL << CONFIG_PAGE_SHIFT);
+
  	return __cvdso_getrandom_data(vd, buffer, len, flags, opaque_state, 
opaque_len);
  }


However, if we have this problem with __kernel_getrandom, don't we also 
have it with: ?
		__kernel_get_syscall_map;
		__kernel_get_tbfreq;
		__kernel_sync_dicache;

If they are also affected, then get_page macro is the place to fix.

I will check all of this now and keep you updated before noon (Paris Time).

Christophe