Re: [PATCH 2 of 4] xentrace: replace ASSERT with printk in __trace_var

[George Dunlap <george.dunlap@citrix.com>]

This seems likely to spam the console if there is a trace which violates
this; and this may happen in production environments if the loglevel is
increased.  I think putting in something to warn just once would be a
better idea.

 -George

On Fri, 2011-07-15 at 16:38 +0100, Olaf Hering wrote:
> # HG changeset patch
> # User Olaf Hering <olaf@aepfle.de>
> # Date 1310741871 -7200
> # Node ID e0ff4eea0432e0af3210e090a47414a0126e9904
> # Parent  d0dcdddf5285eba0605a95dfda79b794803fa733
> xentrace: replace ASSERT with printk in __trace_var
> 
> If trace_var gets called with large extra_data, do not crash the hypervisor.
> Instead print a warning and truncate the buffer.
> 
> Signed-off-by: Olaf Hering <olaf@aepfle.de>
> 
> diff -r d0dcdddf5285 -r e0ff4eea0432 xen/common/trace.c
> --- a/xen/common/trace.c
> +++ b/xen/common/trace.c
> @@ -683,7 +683,10 @@ void __trace_var(u32 event, bool_t cycle
>      if ( (extra % sizeof(u32)) != 0 )
>          extra_word++;
>      
> -    ASSERT(extra_word <= TRACE_EXTRA_MAX);
> +    if ( unlikely(extra_word > TRACE_EXTRA_MAX) )
> +        printk(XENLOG_WARNING "xentrace: event %x extra_data %u too large.\n",
> +                event, extra);
> +
>      extra_word = min_t(int, extra_word, TRACE_EXTRA_MAX);
>  
>      /* Round size up to nearest word */