l3_cache_size seems incorrect for Zen CPUs

[brunocodutra]

The documentation for l3cache_size is not very clear, but I believe the value is reported in multiples of 512KB, not 1KB, as assumed by cache-size.

Specifies the L3 cache size range

(L3Size[31:18] * 512KB) <= L3 cache size < ((L3Size[31:18]+1) * 512KB).

Indeed, l3cache_size it's implemented as get_bits(self.edx, 18, 31) as u16 and should be scaled by 512KB according to the expression above.

On my Ryzen 1700 machine, l3_cache_size returns 32 * 1024, however lscpu reports that the L3 cache is 16MB (see below), which is exactly 512 times the size reported by l3_cache_size and thus consistent with my interpretation that raw-cpuid reports the L3 cache size on AMD machines in multiples of 512KB, not 1KB.

Architecture:            x86_64
  CPU op-mode(s):        32-bit, 64-bit
  Address sizes:         43 bits physical, 48 bits virtual
  Byte Order:            Little Endian
CPU(s):                  16
  On-line CPU(s) list:   0-15
Vendor ID:               AuthenticAMD
  Model name:            AMD Ryzen 7 1700 Eight-Core Processor
    CPU family:          23
    Model:               1
    Thread(s) per core:  2
    Core(s) per socket:  8
    Socket(s):           1
    Stepping:            1
    Frequency boost:     disabled
    CPU(s) scaling MHz:  53%
    CPU max MHz:         3650.0000
    CPU min MHz:         1550.0000
    BogoMIPS:            7302.68
    Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pg
                         e mca cmov pat pse36 clflush mmx fxsr sse sse2 ht s
                         yscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constan
                         t_tsc rep_good nopl nonstop_tsc cpuid extd_apicid a
                         perfmperf rapl pni pclmulqdq monitor ssse3 fma cx16
                          sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdra
                         nd lahf_lm cmp_legacy svm extapic cr8_legacy abm ss
                         e4a misalignsse 3dnowprefetch osvw skinit wdt tce t
                         opoext perfctr_core perfctr_nb bpext perfctr_llc mw
                         aitx cpb hw_pstate ssbd ibpb vmmcall fsgsbase bmi1 
                         avx2 smep bmi2 rdseed adx smap clflushopt sha_ni xs
                         aveopt xsavec xgetbv1 xsaves clzero irperf xsaveerp
                         tr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_
                         clean flushbyasid decodeassists pausefilter pfthres
                         hold avic v_vmsave_vmload vgif overflow_recov succo
                         r smca sev
Virtualization features: 
  Virtualization:        AMD-V
Caches (sum of all):     
  L1d:                   256 KiB (8 instances)
  L1i:                   512 KiB (8 instances)
  L2:                    4 MiB (8 instances)
  L3:                    16 MiB (2 instances)
NUMA:                    
  NUMA node(s):          1
  NUMA node0 CPU(s):     0-15
Vulnerabilities:         
  Itlb multihit:         Not affected
  L1tf:                  Not affected
  Mds:                   Not affected
  Meltdown:              Not affected
  Mmio stale data:       Not affected
  Retbleed:              Mitigation; untrained return thunk; SMT vulnerable
  Spec store bypass:     Mitigation; Speculative Store Bypass disabled via p
                         rctl
  Spectre v1:            Mitigation; usercopy/swapgs barriers and __user poi
                         nter sanitization
  Spectre v2:            Mitigation; Retpolines, IBPB conditional, STIBP dis
                         abled, RSB filling
  Srbds:                 Not affected
  Tsx async abort:       Not affected