linux-kernel - Re: [PATCH] hv: use substraction to update ring buffer index

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170105113955.GD13756@mwanda>
Date:   Thu, 5 Jan 2017 14:39:55 +0300
From:   Dan Carpenter <dan.carpenter@...cle.com>
To:     Long Li <longli@...hange.microsoft.com>
Cc:     "K. Y. Srinivasan" <kys@...rosoft.com>,
        Haiyang Zhang <haiyangz@...rosoft.com>,
        devel@...uxdriverproject.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] hv: use substraction to update ring buffer index

On Wed, Jan 04, 2017 at 08:08:22PM -0800, Long Li wrote:
> From: Long Li <longli@...rosoft.com>
> 
> The ring buffer code uses %= to calculate index. For x86/64, %= compiles to
> div, more than 10 times slower than sub.
> 
> Replace div with sub for this data heavy code path.
> 
> Signed-off-by: Long Li <longli@...rosoft.com>
> ---
>  drivers/hv/ring_buffer.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
> index cd49cb1..f8eee6e 100644
> --- a/drivers/hv/ring_buffer.c
> +++ b/drivers/hv/ring_buffer.c
> @@ -135,7 +135,8 @@ hv_get_next_readlocation_withoffset(struct hv_ring_buffer_info *ring_info,
>  	u32 next = ring_info->ring_buffer->read_index;
>  
>  	next += offset;
> -	next %= ring_info->ring_datasize;
> +	if (next >= ring_info->ring_datasize)
> +		next -= ring_info->ring_datasize;

I take it that we trust that offset is roughly correct and not more than
2x ring_info->ring_datasize?  I guess there is only one caller so it's
probably true...

>  
>  	return next;
>  }
> @@ -179,7 +180,8 @@ static u32 hv_copyfrom_ringbuffer(
>  	memcpy(dest, ring_buffer + start_read_offset, destlen);
>  
>  	start_read_offset += destlen;
> -	start_read_offset %= ring_buffer_size;
> +	if (start_read_offset >= ring_buffer_size)
> +		start_read_offset -= ring_buffer_size;

I totally don't understand the original code here.  We do the memset
and then we verify that we are not copying beyond the end of the ring
buffer?  If feels like we should verify that offset + destlen aren't
more than ring_buffer_size before we do the memcpy().

regards,
dan carpenter