Hi all! This letter maybe doesn't belong to this list.. but it's not so long. I've found on ROSE's egroups page ROSEDIAG and wanted to run it on my AMD. Real mode good, QEMM good, but Win95 and EMM386 crashed (resp. hung). The bug is in RDTSC instruction. On AMD is RDTSC (and maybe RDPMC too, I can't test it) "totaly" privileged. It means it can be executed without emulation only on CPL0 or under real mode EVEN if is TSD 0 (bit in CR4) (resp. PCE 1). In other words: AMD behaves like Intel with TSD permanently set to 1 (resp. PCE to 0). In archive is emulator for Win9x and primitive MSR read/write utils. EliCZ AMD bugs I've found: 1. Setting DR6 bits is not exactly the same as on Intel (doesn't matter) 2. K6 DRx bug (may matter when you're using a debugger) 3. 0F1h opcode doesn't cause exception 1 (doesn't matter, because it's undocummented) 4. RDTSC is privileged .. ------------------------------------------------------------------------ eGroup home: http://www.eGroups.com/list/exelist Free Web-based e-mail groups by eGroups.com
Attachment:
2MSR.zip
Description: Zip compressed data