From: weiqigao AT a DOT crl DOT com (Weiqi Gao) Subject: Re: WARNING: Serious Pentium Bug 8 Nov 1997 18:38:25 -0800 Message-ID: <34651EC8.7C63B98A.cygnus.gnu-win32@a.crl.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit To: gnu-win32 AT cygnus DOT com Corey Minyard wrote: > > Anonymous writes: > > > > > There is a SERIOUS bug in all pentium CPUs. The following > > code will crash any machine running on a pentium CPU, MMX or no > > MMX, any speed, regardless of OS (crash as in instant seize, hard > > reboot the only cure): > > > > char x [5] = { 0xf0, 0x0f, 0xc7, 0xc8 }; > > > > main () > > { > > void (*f)() = x; > > f(); > > } > > > > This require no special permissions to run, it works fine with > > average-joe-userspace permissions. I have verified this, it works. > > Demand a new CPU from Intel. > > > > I tested it on an Intel Pentium 166 MMX and on a Cyrix 686L 166. It > did lock up the Pentium but the 686 correctly reported an illegal > instruction. I tried it under Linux and NT, So it is not just a > Microsoft OS problem. This does look pretty serious. > > Demand a new CPU? I'm not sure what Intel will do about this one. > > Who is this and how long did you look for this? Perhaps Cyrix > or AMD :-? I think that finding this would be a pretty momentous > thing, I'm not sure I would go anonymous. A few observations: 1. After receiving the Anonymous post from GNU-Win32 list, I tested the code on my machine and it did lock up my machine (Pentium 200MHz) under Windows NT 4.0 SP3, Windows 95, Linux 2.1.12, and Windows 98 Beta. GNU-Win32 Beta 18, DJGPP 2.01 and MSVC 5.0 all produced the offending code. However on my other machine (Pentium 83MHz overdrive for 483 motherboards, Windows 98 Beta and DJGPP 2.01) a SIGILL was trapped. 2. This topic dominated the comp.sys.intel news group, which is an appropriate place for the discussion. 3. Intel has posted in that group saying essentially -We noticed the discussion here, and is investigating.- 4. Several people (including Robert Colins, who wrote the Pentium II math bug article for Dr. Dobb's Journal) hase came forward and stated that the bug was known for sometime now. The reason they did not publicize it on the internet is their concern for the security of Pentium based multi-user servers. 5. Several test programs (including a four byte .COM program) was posted which works under Windows. 6. The bug was not present on AMD and Cyrix chips. -- Weiqi Gao weiqigao AT a DOT crl DOT com - For help on using this list (especially unsubscribing), send a message to "gnu-win32-request AT cygnus DOT com" with one line of text: "help".