[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH] vnc: tight: Fix crash after 2GB of output

From: Michael Tokarev
Subject: Re: [Qemu-devel] [PATCH] vnc: tight: Fix crash after 2GB of output
Date: Fri, 04 Mar 2011 10:34:49 +0300
User-agent: Mozilla/5.0 (X11; U; Linux i686 (x86_64); en-US; rv: Gecko/20101227 Icedove/3.0.11

04.03.2011 03:57, Roland Dreier wrote:
> From: Roland Dreier <address@hidden>
> If one leaves a VNC session with tight compression running for long
> enough, Qemu crashes.  This is because of the computation
>     bytes = zstream->total_out - previous_out;
> in tight_compress_data, where zstream->total_out is a uLong but
> previous_out is an int.  As soon as zstream->total_out gets past
> INT_MAX (ie 2GB), previous_out becomes negative and therefore the
> result of the subtraction, bytes, becomes a huge positive number that
> causes havoc for obvious reasons when passed as a length to
> vnc_write().

Excellent work, Roland!  No doubt I wasn't able to hit this bug
locally - I tried many things, but I never waited long enough
to hit this 2Gb border ;)

> The fix for this is simple: keep previous_out as a uLong too, which
> avoids any problems with sign conversion or truncation.

This looks wrong to me.  On 32bit x86 uLong is 32bits.  Yes
it's unsigned there, but it's still 32bits.  And sure thing
it _will_ overflow again, not at 2Gb but now at 4Gb, and the
next total_out will start from zero again, so that now bytes
will be a large integer again because "next" (new total_out)
will be less than "current" (previous_out).

If the whole total_out is actually needed, that counter should
be kept separate and added to when necessary.  But this
total_out should be reset to avoid letting it large values
in the first place.

total_out isn't used by zlib internally, so if the resulting
"total" counter is not needed in qemu, we can just zero-out
the total_out in this function before calling zlib, and
use the resulting value directly as "bytes", without
saving its previous value in previous_out.  Something like
the attached patch does.

By the way, the same thing is used in ui/vnc-enc-zlib.c too,
and it looks like that place has the same issue as well.

Cc'ng to Corentin Chary since he made lots of work in that
area recently.


> Signed-off-by: Roland Dreier <address@hidden>
> ---
>  ui/vnc-enc-tight.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> diff --git a/ui/vnc-enc-tight.c b/ui/vnc-enc-tight.c
> index af45edd..59ec0e3 100644
> --- a/ui/vnc-enc-tight.c
> +++ b/ui/vnc-enc-tight.c
> @@ -829,7 +829,7 @@ static int tight_compress_data(VncState *vs, int 
> stream_id, size_t bytes,
>                                 int level, int strategy)
>  {
>      z_streamp zstream = &vs->tight.stream[stream_id];
> -    int previous_out;
> +    uLong previous_out;
>      if (bytes < VNC_TIGHT_MIN_TO_COMPRESS) {
>          vnc_write(vs, vs->tight.tight.buffer, vs->tight.tight.offset);

Attachment: vnc-tight-2g-overflow.diff
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]