[Mesa-dev] [Bug 105274] Buffer overflow in gallium/auxiliary/hud/hud_cpufreq.c

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Tue Feb 27 23:00:39 UTC 2018


--- Comment #3 from Gert Wollny <gw.fossdev at gmail.com> ---
The sscanf will accept the number and not check the rest of the string, yes,
but later in the code you have

  char basename[256];
  snprintf(basename, sizeof(basename), 
           "/sys/devices/system/cpu/%s", dp->d_name);

so basename is limited to 256 chars including the terminator 0, and then 

  snprintf(fn, sizeof(fn), "%s/cpufreq/scaling_cur_freq", basename);

with sizeof(fn)== 128, which means fn is not longer then 128 byte, just like it
is defined in cpufreq_info::sysfs_filename, hence no buffer overflow there.

      if (stat(fn, &stat_buf) < 0)

must pass, which means whatever string was created before, it must still point
to a valid file name that at this point is


and this limits the size of ?? == dp->d_name to less than 80 characters (This,
by the way, also means one has to create this special file, it is not
sufficient to add an arbitrary string to  /sys/devices/system/cpu/cpuN). So
yes, one can get a buffer overflow in cpufreq_info::name but it can not
overwrite anything important, because given the memory layout of cpufreq_info
it may temporarily overwrite cpufreq_info::cpu_index and a part of
cpufreq_info::sysfs_filename, but these are set later in add_object anyway, and
because of cpufreq_info::sysfs_filename being a null terminated string,
cpufreq_info::name would also always be null-termianted, limiting the impact of
its later use.

IMHO the better fix is snprintf, because it creates a null terminated string
and doesn't do any post-hoc error handling. Unless someone beats me to it I'll
send a patch tomorrow.

You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180227/740cb180/attachment.html>

More information about the mesa-dev mailing list