Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1573590 > unrolled thread

Re: [PATCH resend 4.9] hw_random: Don't use a stack buffer in add_early_randomness()

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH resend 4.9] hw_random: Don't use a stack buffer in  add_early_randomness() Yisheng Xie <xieyisheng1@huawei.com> - 2017-02-04 04:50 +0100
    Re: [PATCH resend 4.9] hw_random: Don't use a stack buffer in  add_early_randomness() Matt Mullins <mmullins@mmlx.us> - 2017-02-04 05:50 +0100
      Re: [PATCH resend 4.9] hw_random: Don't use a stack buffer in  add_early_randomness() Yisheng Xie <xieyisheng1@huawei.com> - 2017-02-04 11:40 +0100

#1573590 — Re: [PATCH resend 4.9] hw_random: Don't use a stack buffer in add_early_randomness()

Hi Andy,

On 2016/10/18 1:06, Andy Lutomirski wrote:
> hw_random carefully avoids using a stack buffer except in
> add_early_randomness().  This causes a crash in virtio_rng if
> CONFIG_VMAP_STACK=y.
I try to understand this patch, but I do not know why it will cause
a crash in virtio_rng with CONFIG_VMAP_STACK=y?
Could you please give me more info. about it.

Really thanks for that!
Yisheng Xie.

> 
> Reported-by: Matt Mullins <mmullins@mmlx.us>
> Tested-by: Matt Mullins <mmullins@mmlx.us>
> Fixes: d3cc7996473a ("hwrng: fetch randomness only after device init")
> Signed-off-by: Andy Lutomirski <luto@kernel.org>
> ---
> 
> This fixes a crash in 4.9-rc1.
> 
> resending because I typoed the git send-email command.  I stealthily added
> Matt's Tested-by, too.
> 
>  drivers/char/hw_random/core.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/char/hw_random/core.c b/drivers/char/hw_random/core.c
> index 9203f2d130c0..340f96e44642 100644
> --- a/drivers/char/hw_random/core.c
> +++ b/drivers/char/hw_random/core.c
> @@ -84,14 +84,14 @@ static size_t rng_buffer_size(void)
>  
>  static void add_early_randomness(struct hwrng *rng)
>  {
> -	unsigned char bytes[16];
>  	int bytes_read;
> +	size_t size = min_t(size_t, 16, rng_buffer_size());
>  
>  	mutex_lock(&reading_mutex);
> -	bytes_read = rng_get_data(rng, bytes, sizeof(bytes), 1);
> +	bytes_read = rng_get_data(rng, rng_buffer, size, 1);
>  	mutex_unlock(&reading_mutex);
>  	if (bytes_read > 0)
> -		add_device_randomness(bytes, bytes_read);
> +		add_device_randomness(rng_buffer, bytes_read);
>  }
>  
>  static inline void cleanup_rng(struct kref *kref)
> 

[toc] | [next] | [standalone]

#1573600

On Sat, Feb 04, 2017 at 11:47:38AM +0800, Yisheng Xie wrote:
> On 2016/10/18 1:06, Andy Lutomirski wrote:
> > hw_random carefully avoids using a stack buffer except in
> > add_early_randomness().  This causes a crash in virtio_rng if
> > CONFIG_VMAP_STACK=y.
> I try to understand this patch, but I do not know why it will cause
> a crash in virtio_rng with CONFIG_VMAP_STACK=y?
> Could you please give me more info. about it.

My original report was
https://lkml.kernel.org/r/20161016002151.GA18235@hydra.tuxags.com.

The virtio ring APIs use scatterlists to keep track of the buffers, and
scatterlist requires addresses to be in the kernel direct-mapped address range.
This is not the case for vmalloc()ed addresses, such as the original on-stack
"bytes" array when VMAP_STACK=y.

[toc] | [prev] | [next] | [standalone]

#1573644

hi, Matt,
Thanks for your reply.

On 2017/2/4 12:34, Matt Mullins wrote:
> On Sat, Feb 04, 2017 at 11:47:38AM +0800, Yisheng Xie wrote:
>> On 2016/10/18 1:06, Andy Lutomirski wrote:
>>> hw_random carefully avoids using a stack buffer except in
>>> add_early_randomness().  This causes a crash in virtio_rng if
>>> CONFIG_VMAP_STACK=y.
>> I try to understand this patch, but I do not know why it will cause
>> a crash in virtio_rng with CONFIG_VMAP_STACK=y?
>> Could you please give me more info. about it.
> 
> My original report was
> https://lkml.kernel.org/r/20161016002151.GA18235@hydra.tuxags.com.
> 
> The virtio ring APIs use scatterlists to keep track of the buffers, and
> scatterlist requires addresses to be in the kernel direct-mapped address range.
> This is not the case for vmalloc()ed addresses, such as the original on-stack
> "bytes" array when VMAP_STACK=y.
> 
I see, and will check the logic to get more detail about it.

Thanks.
Yisheng Xie

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web