turbostat: Incorrect physical unit for PL4

turbostat: Incorrect physical unit for PL4

[zhangpengjie (A)]

Hi Len and linux-pm community, I was reviewing the PL4 output in turbostat
and noticed a confusing behavior regarding its physical unit.

According to Intel's hardware specifications, there is a contradiction in
the documentation itself:The overview describes it as: "Package-level
maximum power limit (in Watts). It is a proactive, instantaneous limit."
However, the actual bitfield definition states: "The PL4 value in bits
[12:0] is specified in strict 0.125 A increments."

Currently, in tools/power/x86/turbostat/turbostat.c, the code parses it 
as follows:

fprintf(outf, "cpu%d: PKG Limit #4: %f Watts (%slocked)\n", cpu, ((msr 
 >> 0) & 0x1FFF) * rapl_power_units, (msr >> 31) & 1 ? "" : "UN");

There is a significant issue here:
Because the manual broadly labels PL4 in "Watts", turbostat incorrectly
multiplies the raw Ampere value by rapl_power_units and displays the result
as "Watts". As a result, the tool outputs a PL4 value that is often 
absurdly
lower than the PL1 and PL2 values. Since PL4 is supposed to be the absolute
maximum instantaneous limit, displaying it as a value smaller than PL1/PL2
causes significant confusion for users monitoring system power.

Could you please take a look at this and consider clarifying the output
to reflect the actual Amps?

Best regards,

Pengjie Zhang <zhangpengjie2@huawei.com>