On Jun 29,  7:29pm, gallegos@csxxi.net.mx (Heron Gallegos) wrote:
-- Subject: Re: VIA C7 CPU and crypto capabilities SOLVED partially

| Thanks, Christos for the patch...
| 
| I changed this line
| #defne VIA_ACE (CPUID_VIA_HAS_PADLOCK|CPUID_VIA_DO_PADLOCK)
| #define VIA_ACE (CPUID_VIA_HAS_ACE|CPUID_VIA_DO_ACE)
| because I supose this is what you want to do, and
| CPUID_VIA_HAS_PADLOCK and CPUID_VIA_DO_PADLOCK not exists.

Yup this is right.
| 
| The line aprint_verbose("%s: features3 %s\n", cpuname, buf);
| does not works. I supose at this point msgbuf is not initialized.
| 
| I untared a clean tree from syssrc.tgz
| (I am using /pub/NetBSD-daily/HEAD/200706200002Z/i386)
| I did a "make clean" and "make cleandir" in compile/POLARIS.
| I did a "cd /sys/arch; patch < /tmp/patch"
| I built a kernel with the following changes:
| 7,8c7,8
| < NetBSD 4.99.20 (GENERIC) #0: Wed Jun 20 12:12:20 PDT 2007
| <       builds@wb29:/home/builds/ab/HEAD/i386/200706200002Z-obj/home/builds/ab/HEAD/src/sys/arch/i386/compile/GENERIC
| ---
| > NetBSD 4.99.20 (POLARIS) #27: Fri Jun 29 14:44:26 CDT 2007
| >       gallegos@polaris:/usr/src/sys/arch/i386/compile/POLARIS
| 17c17
| < cpu0: VIA C3 (686-class), 799.97 MHz, id 0x6a9
| ---
| > cpu0: VIA C7 Esther (686-class), 799.98 MHz, id 0x6a9
| 21a22
| > cpu0: padlock features 3fcc<RNG,AES,AES/CTR,SHA1/SHA256,RSA>
| 27,31c28,30
| < cpu0: using thermal monitor 2
| < cpu0: Enhanced SpeedStep (1004 mV) 533 MHz
| < cpu0: unknown Enhanced SpeedStep CPU.
| < cpu0: using only highest, current and lowest power states.
| < cpu0: Enhanced SpeedStep frequencies available (MHz): 667 533 533
| ---
| > cpu0: enabling thermal monitor 1 ... failed!
| > cpu0: failed to enable thermal monitoring!
| > cpu0: Enhanced SpeedStep disabled by BIOS
| 80c79
| < acpitz0: critical 80.0C passive 45.0C
| ---
| > acpitz0: critical 80.0C passive 47.0C
| 168a168
| > pckbdprobe: reset error 5
| 171a172
| > pmsprobe: reset error 5
| 173a175
| > PadLock: registered support for AES_CBC

That's good.

| The mouse and kbd was unplugged. (working with LAN in both dmesg)
| 
| I have seen, every time I test a new kernel, the part of TM2 and
| SpeedStep fails when are executed rdmsr()/wrmsr() in the function
| via_cpu_probe() of identcpu.c
| 
| I will do some tests with and without the "options VIA_PADLOCK"

Thanks a lot. Let me know how it goes and I will commit the changes.

christos