On Fri, 13 May 2011 14:48:01 +0200, Luca Tettamanti wrote: > On Thu, May 12, 2011 at 05:49:44PM -0600, Harry G McGavran Jr wrote: > > I went back through the logs to make sure there was nothing > > about the latest i8k in them. Actually I found only ONE > > line in syslog: > > > > May 11 11:07:12 XX kernel: [173400.479332] [<ffffffffa00613f2>] > > i8k_hwmon_show_fan+0x32/0x40 [i8k] > > Ok, this is consistent with the hung program reported by Jeff, there's > an error at kernel level while the sensor is being read. > > AMD64 Architecture Manual says that: > > (pushf) In 64-bit mode, this instruction defaults to a 64-bit operand > size and there is no prefix available to encode a 32-bit operand size. D'oh. I suspected something like this, but given that pushfq exists, I assumed pushf had to be different. I guess I shouldn't have assumed... > > so my patch may be corrupting the stack. Lets try like this: > > diff --git a/drivers/char/i8k.c b/drivers/char/i8k.c > index d72433f..ee01716 100644 > --- a/drivers/char/i8k.c > +++ b/drivers/char/i8k.c > @@ -139,8 +139,8 @@ static int i8k_smm(struct smm_regs *regs) > "movl %%edi,20(%%rax)\n\t" > "popq %%rdx\n\t" > "movl %%edx,0(%%rax)\n\t" > - "lahf\n\t" > - "shrl $8,%%eax\n\t" > + "pushfq\n\t" > + "popq %%rax\n\t" > "andl $1,%%eax\n" > :"=a"(rc) > : "a"(regs) > > > If it still doesn't work we should involve someone who actually has a > clue about x86 asm ;) Hopefully it will work OK this time. I've updated the driver at: http://khali.linux-fr.org/devel/misc/i8k/ Harry, Jeff, care to try again? Thanks. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
