From: garbanzo AT hooked DOT net (Alex) Subject: Re: WARNING: Serious Pentium Bug 7 Nov 1997 22:57:58 -0800 Message-ID: References: <199711080455 DOT UAA21222 AT stars DOT cisco DOT com> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII To: Tim Iverson Cc: gnu-win32 AT cygnus DOT com On Fri, 7 Nov 1997, Tim Iverson wrote: > Just to keep everyone from getting too scared, this really would only matter > for secure sites with multiple users -- that particular sequence is a > nonsense instruction. The only folks using it would be those deliberately > wanting to bring the system down. Well, I'd consider it a pretty big bug, especailly because anybody can run it, if they can run binary files. For instance, if my ISP used a P5 for their mail server, I could use procmail to freeze it, forcing some sort of [expensive] intervention. Or, trojan horses *yikes*. I can see it now. warez4free.exe... If NTFS partitions could be mounted async (or damaged similarly from a power cycle without proper shut down), problems could arise. The only people really exploiting the NetBIOS OOB bug would be people trying to down a system too.. (didn't stop many people) :/ > Here it is decoded: lock cmpxchg8b eax > > BTW, I haven't tested it, so I don't know if this is even true. It should > generate an illegal instruction exception, since the operand needs to be a > memory reference. Yes, it worked for me, all too well (not tested under 95). There was a small rumor circulating on the FreeBSD-Current list that CPUs with >=12 stepping and 0x52c ids would be immune (doesn't seem to be true, for me). gcc compiled the code (that I got my hands on) flawlessly, spitting out only a warning. - alex - 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".