ghw
is a Go library providing hardware inspection and discovery for Linux and
Windows. There currently exists partial support for MacOSX.
-
No root privileges needed for discovery
ghw
goes the extra mile to be useful without root priveleges. We query for host hardware information as directly as possible without relying on shellouts to programs likedmidecode
that require root privileges to execute.Elevated privileges are indeed required to query for some information, but
ghw
will never error out if blocked from reading that information. Instead,ghw
will print a warning message about the information that could not be retrieved. You may disable these warning messages with theGHW_DISABLE_WARNINGS
environment variable. -
Well-documented code and plenty of example code
The code itself should be well-documented with lots of usage examples.
-
Interfaces should be consistent across modules
Each module in the library should be structured in a consistent fashion, and the structs returned by various library functions should have consistent attribute and method names.
ghw
is a tool for gathering information about your hardware's capacity
and capabilities.
It is important to point out that ghw
does NOT report information that is
temporary or variable. It is NOT a system monitor nor is it an appropriate
tool for gathering data points for metrics that change over time. If you are
looking for a system that tracks usage of CPU, memory, network I/O or disk
I/O, there are plenty of great open source tools that do this! Check out the
Prometheus project for a great example.
ghw
has functions that return an Info
object about a particular hardware
domain (e.g. CPU, Memory, Block storage, etc).
Use the following functions in ghw
to inspect information about the host
hardware:
ghw.CPU()
ghw.Memory()
ghw.Block()
(block storage)ghw.Topology()
(processor architecture, NUMA topology and memory cache hierarchy)ghw.Network()
ghw.PCI()
ghw.GPU()
(graphical processing unit)ghw.Accelerator()
(processing accelerators, AI)ghw.Chassis()
ghw.BIOS()
ghw.Baseboard()
ghw.Product()
The ghw.CPU()
function returns a ghw.CPUInfo
struct that contains
information about the CPUs on the host system.
ghw.CPUInfo
contains the following fields:
ghw.CPUInfo.TotalCores
has the total number of physical cores the host system containsghw.CPUInfo.TotalHardwareThreads
has the total number of hardware threads the host system containsghw.CPUInfo.Processors
is an array ofghw.Processor
structs, one for each physical processor package contained in the host
Each ghw.Processor
struct contains a number of fields:
ghw.Processor.ID
is the physical processoruint32
ID according to the systemghw.Processor.TotalCores
is the number of physical cores in the processor packageghw.Processor.TotalHardwareThreads
is the number of hardware threads in the processor packageghw.Processor.Vendor
is a string containing the vendor nameghw.Processor.Model
is a string containing the vendor's model nameghw.Processor.Capabilities
(Linux only) is an array of strings indicating the features the processor has enabledghw.Processor.Cores
(Linux only) is an array ofghw.ProcessorCore
structs that are packed onto this physical processor
A ghw.ProcessorCore
has the following fields:
ghw.ProcessorCore.ID
is theuint32
identifier that the host gave this core. Note that this does not necessarily equate to a zero-based index of the core within a physical package. For example, the core IDs for an Intel Core i7 are 0, 1, 2, 8, 9, and 10ghw.ProcessorCore.TotalHardwareThreads
is the number of hardware threads associated with the coreghw.ProcessorCore.LogicalProcessors
is an array of ints representing the logical processor IDs assigned to any processing unit for the core. These are sometimes called the "thread siblings". Logical processor IDs are the zero-based index of the processor on the host and are not related to the core ID.
package main
import (
"fmt"
"math"
"strings"
"github.com/jaypipes/ghw"
)
func main() {
cpu, err := ghw.CPU()
if err != nil {
fmt.Printf("Error getting CPU info: %v", err)
}
fmt.Printf("%v\n", cpu)
for _, proc := range cpu.Processors {
fmt.Printf(" %v\n", proc)
for _, core := range proc.Cores {
fmt.Printf(" %v\n", core)
}
if len(proc.Capabilities) > 0 {
// pretty-print the (large) block of capability strings into rows
// of 6 capability strings
rows := int(math.Ceil(float64(len(proc.Capabilities)) / float64(6)))
for row := 1; row < rows; row = row + 1 {
rowStart := (row * 6) - 1
rowEnd := int(math.Min(float64(rowStart+6), float64(len(proc.Capabilities))))
rowElems := proc.Capabilities[rowStart:rowEnd]
capStr := strings.Join(rowElems, " ")
if row == 1 {
fmt.Printf(" capabilities: [%s\n", capStr)
} else if rowEnd < len(proc.Capabilities) {
fmt.Printf(" %s\n", capStr)
} else {
fmt.Printf(" %s]\n", capStr)
}
}
}
}
}
Example output from my personal workstation:
cpu (1 physical package, 6 cores, 12 hardware threads)
physical package #0 (6 cores, 12 hardware threads)
processor core #0 (2 threads), logical processors [0 6]
processor core #1 (2 threads), logical processors [1 7]
processor core #2 (2 threads), logical processors [2 8]
processor core #3 (2 threads), logical processors [3 9]
processor core #4 (2 threads), logical processors [4 10]
processor core #5 (2 threads), logical processors [5 11]
capabilities: [msr pae mce cx8 apic sep
mtrr pge mca cmov pat pse36
clflush dts acpi mmx fxsr sse
sse2 ss ht tm pbe syscall
nx pdpe1gb rdtscp lm constant_tsc arch_perfmon
pebs bts rep_good nopl xtopology nonstop_tsc
cpuid aperfmperf pni pclmulqdq dtes64 monitor
ds_cpl vmx est tm2 ssse3 cx16
xtpr pdcm pcid sse4_1 sse4_2 popcnt
aes lahf_lm pti retpoline tpr_shadow vnmi
flexpriority ept vpid dtherm ida arat]
The ghw.Memory()
function returns a ghw.MemoryInfo
struct that contains
information about the RAM on the host system.
ghw.MemoryInfo
contains the following fields:
ghw.MemoryInfo.TotalPhysicalBytes
contains the amount of physical memory on the hostghw.MemoryInfo.TotalUsableBytes
contains the amount of memory the system can actually use. Usable memory accounts for things like the kernel's resident memory size and some reserved system bits. Please note this value is NOT the amount of memory currently in use by processes in the system. See [the discussion][#physical-versus-usage-memory] about the difference.ghw.MemoryInfo.SupportedPageSizes
is an array of integers representing the size, in bytes, of memory pages the system supportsghw.MemoryInfo.Modules
is an array of pointers toghw.MemoryModule
structs, one for each physical DIMM. Currently, this information is only included on Windows, with Linux support planned.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
memory, err := ghw.Memory()
if err != nil {
fmt.Printf("Error getting memory info: %v", err)
}
fmt.Println(memory.String())
}
Example output from my personal workstation:
memory (24GB physical, 24GB usable)
There has been some confusion regarding the difference between the total physical bytes versus total usable bytes of memory.
Some of this confusion has been due to a misunderstanding of the term "usable".
As mentioned above, ghw
does inspection of the
system's capacity.
A host computer has two capacities when it comes to RAM. The first capacity is
the amount of RAM that is contained in all memory banks (DIMMs) that are
attached to the motherboard. ghw.MemoryInfo.TotalPhysicalBytes
refers to this
first capacity.
There is a (usually small) amount of RAM that is consumed by the bootloader
before the operating system is started (booted). Once the bootloader has booted
the operating system, the amount of RAM that may be used by the operating
system and its applications is fixed. ghw.MemoryInfo.TotalUsableBytes
refers
to this second capacity.
You can determine the amount of RAM that the bootloader used (that is not made
available to the operating system) by subtracting
ghw.MemoryInfo.TotalUsableBytes
from ghw.MemoryInfo.TotalPhysicalBytes
:
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
memory, err := ghw.Memory()
if err != nil {
fmt.Printf("Error getting memory info: %v", err)
}
phys := memory.TotalPhysicalBytes
usable := memory.TotalUsableBytes
fmt.Printf("The bootloader consumes %d bytes of RAM\n", phys - usable)
}
Example output from my personal workstation booted into a Windows10 operating system with a Linux GRUB bootloader:
The bootloader consumes 3832720 bytes of RAM
The ghw.Block()
function returns a ghw.BlockInfo
struct that contains
information about the block storage on the host system.
ghw.BlockInfo
contains the following fields:
ghw.BlockInfo.TotalSizeBytes
contains the amount of physical block storage on the host.ghw.BlockInfo.Disks
is an array of pointers toghw.Disk
structs, one for each disk found by the system
Each ghw.Disk
struct contains the following fields:
ghw.Disk.Name
contains a string with the short name of the disk, e.g. "sda"ghw.Disk.SizeBytes
contains the amount of storage the disk providesghw.Disk.PhysicalBlockSizeBytes
contains the size of the physical blocks used on the disk, in bytes. This is typically the minimum amount of data that will be written in a single write operation for the disk.ghw.Disk.IsRemovable
contains a boolean indicating if the disk drive is removableghw.Disk.DriveType
is the type of drive. It is of typeghw.DriveType
which has aghw.DriveType.String()
method that can be called to return a string representation of the bus. This string will beHDD
,FDD
,ODD
, orSSD
, which correspond to a hard disk drive (rotational), floppy drive, optical (CD/DVD) drive and solid-state drive.ghw.Disk.StorageController
is the type of storage controller. It is of typeghw.StorageController
which has aghw.StorageController.String()
method that can be called to return a string representation of the bus. This string will beSCSI
,IDE
,virtio
,MMC
, orNVMe
ghw.Disk.BusPath
(Linux, Darwin only) is the filepath to the bus used by the disk.ghw.Disk.NUMANodeID
(Linux only) is the numeric index of the NUMA node this disk is local to, or -1 if the host system is not a NUMA system or is not Linux.ghw.Disk.Vendor
contains a string with the name of the hardware vendor for the diskghw.Disk.Model
contains a string with the vendor-assigned disk model nameghw.Disk.SerialNumber
contains a string with the disk's serial numberghw.Disk.WWN
contains a string with the disk's World Wide Nameghw.Disk.Partitions
contains an array of pointers toghw.Partition
structs, one for each partition on the disk
Each ghw.Partition
struct contains these fields:
ghw.Partition.Name
contains a string with the short name of the partition, e.g.sda1
ghw.Partition.Label
contains the label for the partition itself. On Linux systems, this is derived from theID_PART_ENTRY_NAME
udev entry for the partition.ghw.Partition.FilesystemLabel
contains the label for the filesystem housed on the partition. On Linux systems, this is derived from theID_FS_NAME
udev entry for the partition.ghw.Partition.SizeBytes
contains the amount of storage the partition providesghw.Partition.MountPoint
contains a string with the partition's mount point, or""
if no mount point was discoveredghw.Partition.Type
contains a string indicated the filesystem type for the partition, or""
if the system could not determine the typeghw.Partition.IsReadOnly
is a bool indicating the partition is read-onlyghw.Partition.Disk
is a pointer to theghw.Disk
object associated with the partition.ghw.Partition.UUID
is a string containing the partition UUID on Linux, the partition UUID on MacOS and nothing on Windows. On Linux systems, this is derived from theID_PART_ENTRY_UUID
udev entry for the partition.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
block, err := ghw.Block()
if err != nil {
fmt.Printf("Error getting block storage info: %v", err)
}
fmt.Printf("%v\n", block)
for _, disk := range block.Disks {
fmt.Printf(" %v\n", disk)
for _, part := range disk.Partitions {
fmt.Printf(" %v\n", part)
}
}
}
Example output from my personal workstation:
block storage (1 disk, 2TB physical storage)
sda HDD (2TB) SCSI [@pci-0000:04:00.0-scsi-0:1:0:0 (node #0)] vendor=LSI model=Logical_Volume serial=600508e000000000f8253aac9a1abd0c WWN=0x600508e000000000f8253aac9a1abd0c
/dev/sda1 (100MB)
/dev/sda2 (187GB)
/dev/sda3 (449MB)
/dev/sda4 (1KB)
/dev/sda5 (15GB)
/dev/sda6 (2TB) [ext4] mounted@/
NOTE:
ghw
looks in the udev runtime database for some information. If you are usingghw
in a container, remember to bind mount/dev/disk
and/run
into your container, otherwiseghw
won't be able to query the udev DB or sysfs paths for information.
NOTE: Topology support is currently Linux-only. Windows support is planned.
The ghw.Topology()
function returns a ghw.TopologyInfo
struct that contains
information about the host computer's architecture (NUMA vs. SMP), the host's
NUMA node layout and processor-specific memory caches.
The ghw.TopologyInfo
struct contains two fields:
ghw.TopologyInfo.Architecture
contains an enum with the valueghw.NUMA
orghw.SMP
depending on what the topology of the system isghw.TopologyInfo.Nodes
is an array of pointers toghw.TopologyNode
structs, one for each topology node (typically physical processor package) found by the system
Each ghw.TopologyNode
struct contains the following fields:
ghw.TopologyNode.ID
is the system'suint32
identifier for the nodeghw.TopologyNode.Memory
is aghw.MemoryArea
struct describing the memory attached to this node.ghw.TopologyNode.Cores
is an array of pointers toghw.ProcessorCore
structs that are contained in this nodeghw.TopologyNode.Caches
is an array of pointers toghw.MemoryCache
structs that represent the low-level caches associated with processors and cores on the systemghw.TopologyNode.Distance
is an array of distances between NUMA nodes as reported by the system.
ghw.MemoryArea
describes a collection of physical RAM on the host.
In the simplest and most common case, all system memory fits in a single memory area. In more complex host systems, like NUMA systems, many memory areas may be present in the host system (e.g. one for each NUMA cell).
The ghw.MemoryArea
struct contains the following fields:
ghw.MemoryArea.TotalPhysicalBytes
contains the amount of physical memory associated with this memory area.ghw.MemoryArea.TotalUsableBytes
contains the amount of memory of this memory area the system can actually use. Usable memory accounts for things like the kernel's resident memory size and some reserved system bits. Please note this value is NOT the amount of memory currently in use by processes in the system. See [the discussion][#physical-versus-usage-memory] about the difference.
See above in the CPU section for information about the
ghw.ProcessorCore
struct and how to use and query it.
Each ghw.MemoryCache
struct contains the following fields:
ghw.MemoryCache.Type
is an enum that contains one ofghw.DATA
,ghw.INSTRUCTION
orghw.UNIFIED
depending on whether the cache stores CPU instructions, program data, or bothghw.MemoryCache.Level
is a positive integer indicating how close the cache is to the processor. The lower the number, the closer the cache is to the processor and the faster the processor can access its contentsghw.MemoryCache.SizeBytes
is an integer containing the number of bytes the cache can containghw.MemoryCache.LogicalProcessors
is an array of integers representing the logical processors that use the cache
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
topology, err := ghw.Topology()
if err != nil {
fmt.Printf("Error getting topology info: %v", err)
}
fmt.Printf("%v\n", topology)
for _, node := range topology.Nodes {
fmt.Printf(" %v\n", node)
for _, cache := range node.Caches {
fmt.Printf(" %v\n", cache)
}
}
}
Example output from my personal workstation:
topology SMP (1 nodes)
node #0 (6 cores)
L1i cache (32 KB) shared with logical processors: 3,9
L1i cache (32 KB) shared with logical processors: 2,8
L1i cache (32 KB) shared with logical processors: 11,5
L1i cache (32 KB) shared with logical processors: 10,4
L1i cache (32 KB) shared with logical processors: 0,6
L1i cache (32 KB) shared with logical processors: 1,7
L1d cache (32 KB) shared with logical processors: 11,5
L1d cache (32 KB) shared with logical processors: 10,4
L1d cache (32 KB) shared with logical processors: 3,9
L1d cache (32 KB) shared with logical processors: 1,7
L1d cache (32 KB) shared with logical processors: 0,6
L1d cache (32 KB) shared with logical processors: 2,8
L2 cache (256 KB) shared with logical processors: 2,8
L2 cache (256 KB) shared with logical processors: 3,9
L2 cache (256 KB) shared with logical processors: 0,6
L2 cache (256 KB) shared with logical processors: 10,4
L2 cache (256 KB) shared with logical processors: 1,7
L2 cache (256 KB) shared with logical processors: 11,5
L3 cache (12288 KB) shared with logical processors: 0,1,10,11,2,3,4,5,6,7,8,9
The ghw.Network()
function returns a ghw.NetworkInfo
struct that contains
information about the host computer's networking hardware.
The ghw.NetworkInfo
struct contains one field:
ghw.NetworkInfo.NICs
is an array of pointers toghw.NIC
structs, one for each network interface controller found for the systen
Each ghw.NIC
struct contains the following fields:
ghw.NIC.Name
is the system's identifier for the NICghw.NIC.MACAddress
is the Media Access Control (MAC) address for the NIC, if anyghw.NIC.IsVirtual
is a boolean indicating if the NIC is a virtualized deviceghw.NIC.Capabilities
(Linux only) is an array of pointers toghw.NICCapability
structs that can describe the things the NIC supports. These capabilities match the returned values from theethtool -k <DEVICE>
call on Linux as well as the AutoNegotiation and PauseFrameUse capabilities fromethtool
.ghw.NIC.PCIAddress
(Linux only) is the PCI device address of the device backing the NIC. this is not-nil only if the backing device is indeed a PCI device; more backing devices (e.g. USB) will be added in future versions.ghw.NIC.Speed
(Linux only) is a string showing the current link speed. On Linux, this field will be present even ifethtool
is not available.ghw.NIC.Duplex
(Linux only) is a string showing the current link duplex. On Linux, this field will be present even ifethtool
is not available.ghw.NIC.SupportedLinkModes
(Linux only) is a string slice containing a list of supported link modes, e.g. "10baseT/Half", "1000baseT/Full".ghw.NIC.SupportedPorts
(Linux only) is a string slice containing the list of supported port types, e.g. "MII", "TP", "FIBRE", "Twisted Pair".ghw.NIC.SupportedFECModes
(Linux only) is a string slice containing a list of supported Forward Error Correction (FEC) Modes.ghw.NIC.AdvertisedLinkModes
(Linux only) is a string slice containing the link modes being advertised during auto negotiation.ghw.NIC.AdvertisedFECModes
(Linux only) is a string slice containing the Forward Error Correction (FEC) modes advertised during auto negotiation.
The ghw.NICCapability
struct contains the following fields:
ghw.NICCapability.Name
is the string name of the capability (e.g. "tcp-segmentation-offload")ghw.NICCapability.IsEnabled
is a boolean indicating whether the capability is currently enabled/active on the NICghw.NICCapability.CanEnable
is a boolean indicating whether the capability may be enabled
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
net, err := ghw.Network()
if err != nil {
fmt.Printf("Error getting network info: %v", err)
}
fmt.Printf("%v\n", net)
for _, nic := range net.NICs {
fmt.Printf(" %v\n", nic)
enabledCaps := make([]int, 0)
for x, cap := range nic.Capabilities {
if cap.IsEnabled {
enabledCaps = append(enabledCaps, x)
}
}
if len(enabledCaps) > 0 {
fmt.Printf(" enabled capabilities:\n")
for _, x := range enabledCaps {
fmt.Printf(" - %s\n", nic.Capabilities[x].Name)
}
}
}
}
Example output from my personal laptop:
net (3 NICs)
docker0
enabled capabilities:
- tx-checksumming
- tx-checksum-ip-generic
- scatter-gather
- tx-scatter-gather
- tx-scatter-gather-fraglist
- tcp-segmentation-offload
- tx-tcp-segmentation
- tx-tcp-ecn-segmentation
- tx-tcp-mangleid-segmentation
- tx-tcp6-segmentation
- udp-fragmentation-offload
- generic-segmentation-offload
- generic-receive-offload
- tx-vlan-offload
- highdma
- tx-lockless
- netns-local
- tx-gso-robust
- tx-fcoe-segmentation
- tx-gre-segmentation
- tx-gre-csum-segmentation
- tx-ipxip4-segmentation
- tx-ipxip6-segmentation
- tx-udp_tnl-segmentation
- tx-udp_tnl-csum-segmentation
- tx-gso-partial
- tx-sctp-segmentation
- tx-esp-segmentation
- tx-vlan-stag-hw-insert
enp58s0f1
enabled capabilities:
- rx-checksumming
- generic-receive-offload
- rx-vlan-offload
- tx-vlan-offload
- highdma
- auto-negotiation
wlp59s0
enabled capabilities:
- scatter-gather
- tx-scatter-gather
- generic-segmentation-offload
- generic-receive-offload
- highdma
- netns-local
ghw
contains a PCI database inspection and querying facility that allows
developers to not only gather information about devices on a local PCI bus but
also query for information about hardware device classes, vendor and product
information.
NOTE: Parsing of the PCI-IDS file database is provided by the separate github.com/jaypipes/pcidb library. You can read that library's README for more information about the various structs that are exposed on the
ghw.PCIInfo
struct.
The ghw.PCI()
function returns a ghw.PCIInfo
struct that contains
information about the host computer's PCI devices.
The ghw.PCIInfo
struct contains one field:
ghw.PCIInfo.Devices
is a slice of pointers toghw.PCIDevice
structs that describe the PCI devices on the host system
NOTE: PCI products are often referred to by their "device ID". We use the term "product ID" in
ghw
because it more accurately reflects what the identifier is for: a specific product line produced by the vendor.
The ghw.PCIDevice
struct has the following fields:
ghw.PCIDevice.Vendor
is a pointer to apcidb.Vendor
struct that describes the device's primary vendor. This will always be non-nil.ghw.PCIDevice.Product
is a pointer to apcidb.Product
struct that describes the device's primary product. This will always be non-nil.ghw.PCIDevice.Subsystem
is a pointer to apcidb.Product
struct that describes the device's secondary/sub-product. This will always be non-nil.ghw.PCIDevice.Class
is a pointer to apcidb.Class
struct that describes the device's class. This will always be non-nil.ghw.PCIDevice.Subclass
is a pointer to apcidb.Subclass
struct that describes the device's subclass. This will always be non-nil.ghw.PCIDevice.ProgrammingInterface
is a pointer to apcidb.ProgrammingInterface
struct that describes the device subclass' programming interface. This will always be non-nil.ghw.PCIDevice.Driver
is a string representing the device driver the system is using to handle this device. Can be empty string if this information is not available. If the information is not available, this does not mean the device is not functioning, but rather thatghw
was not able to retrieve driver information.
The ghw.PCIAddress
(which is an alias for the ghw.pci.address.Address
struct) contains the PCI address fields. It has a ghw.PCIAddress.String()
method that returns the canonical Domain:Bus:Device.Function ([D]BDF)
representation of this Address.
The ghw.PCIAddress
struct has the following fields:
ghw.PCIAddress.Domain
is a string representing the PCI domain component of the address.ghw.PCIAddress.Bus
is a string representing the PCI bus component of the address.ghw.PCIAddress.Device
is a string representing the PCI device component of the address.ghw.PCIAddress.Function
is a string representing the PCI function component of the address.
NOTE: Older versions (pre-
v0.9.0
) erroneously referred to theDevice
field as theSlot
field. As noted by @pearsonk in #220, this was a misnomer.
The following code snippet shows how to list the PCI devices on the host system and output a simple list of PCI address and vendor/product information:
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
pci, err := ghw.PCI()
if err != nil {
fmt.Printf("Error getting PCI info: %v", err)
}
fmt.Printf("host PCI devices:\n")
fmt.Println("====================================================")
for _, device := range pci.Devices {
vendor := device.Vendor
vendorName := vendor.Name
if len(vendor.Name) > 20 {
vendorName = string([]byte(vendorName)[0:17]) + "..."
}
product := device.Product
productName := product.Name
if len(product.Name) > 40 {
productName = string([]byte(productName)[0:37]) + "..."
}
fmt.Printf("%-12s\t%-20s\t%-40s\n", device.Address, vendorName, productName)
}
}
on my local workstation the output of the above looks like the following:
host PCI devices:
====================================================
0000:00:00.0 Intel Corporation 5520/5500/X58 I/O Hub to ESI Port
0000:00:01.0 Intel Corporation 5520/5500/X58 I/O Hub PCI Express Roo...
0000:00:02.0 Intel Corporation 5520/5500/X58 I/O Hub PCI Express Roo...
0000:00:03.0 Intel Corporation 5520/5500/X58 I/O Hub PCI Express Roo...
0000:00:07.0 Intel Corporation 5520/5500/X58 I/O Hub PCI Express Roo...
0000:00:10.0 Intel Corporation 7500/5520/5500/X58 Physical and Link ...
0000:00:10.1 Intel Corporation 7500/5520/5500/X58 Routing and Protoc...
0000:00:14.0 Intel Corporation 7500/5520/5500/X58 I/O Hub System Man...
0000:00:14.1 Intel Corporation 7500/5520/5500/X58 I/O Hub GPIO and S...
0000:00:14.2 Intel Corporation 7500/5520/5500/X58 I/O Hub Control St...
0000:00:14.3 Intel Corporation 7500/5520/5500/X58 I/O Hub Throttle R...
0000:00:19.0 Intel Corporation 82567LF-2 Gigabit Network Connection
0000:00:1a.0 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1a.1 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1a.2 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1a.7 Intel Corporation 82801JI (ICH10 Family) USB2 EHCI Cont...
0000:00:1b.0 Intel Corporation 82801JI (ICH10 Family) HD Audio Contr...
0000:00:1c.0 Intel Corporation 82801JI (ICH10 Family) PCI Express Ro...
0000:00:1c.1 Intel Corporation 82801JI (ICH10 Family) PCI Express Po...
0000:00:1c.4 Intel Corporation 82801JI (ICH10 Family) PCI Express Ro...
0000:00:1d.0 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1d.1 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1d.2 Intel Corporation 82801JI (ICH10 Family) USB UHCI Contr...
0000:00:1d.7 Intel Corporation 82801JI (ICH10 Family) USB2 EHCI Cont...
0000:00:1e.0 Intel Corporation 82801 PCI Bridge
0000:00:1f.0 Intel Corporation 82801JIR (ICH10R) LPC Interface Contr...
0000:00:1f.2 Intel Corporation 82801JI (ICH10 Family) SATA AHCI Cont...
0000:00:1f.3 Intel Corporation 82801JI (ICH10 Family) SMBus Controller
0000:01:00.0 NEC Corporation uPD720200 USB 3.0 Host Controller
0000:02:00.0 Marvell Technolog... 88SE9123 PCIe SATA 6.0 Gb/s controller
0000:02:00.1 Marvell Technolog... 88SE912x IDE Controller
0000:03:00.0 NVIDIA Corporation GP107 [GeForce GTX 1050 Ti]
0000:03:00.1 NVIDIA Corporation UNKNOWN
0000:04:00.0 LSI Logic / Symbi... SAS2004 PCI-Express Fusion-MPT SAS-2 ...
0000:06:00.0 Qualcomm Atheros AR5418 Wireless Network Adapter [AR50...
0000:08:03.0 LSI Corporation FW322/323 [TrueFire] 1394a Controller
0000:3f:00.0 Intel Corporation UNKNOWN
0000:3f:00.1 Intel Corporation Xeon 5600 Series QuickPath Architectu...
0000:3f:02.0 Intel Corporation Xeon 5600 Series QPI Link 0
0000:3f:02.1 Intel Corporation Xeon 5600 Series QPI Physical 0
0000:3f:02.2 Intel Corporation Xeon 5600 Series Mirror Port Link 0
0000:3f:02.3 Intel Corporation Xeon 5600 Series Mirror Port Link 1
0000:3f:03.0 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:03.1 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:03.4 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:04.0 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:04.1 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:04.2 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:04.3 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:05.0 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:05.1 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:05.2 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:05.3 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:06.0 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:06.1 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:06.2 Intel Corporation Xeon 5600 Series Integrated Memory Co...
0000:3f:06.3 Intel Corporation Xeon 5600 Series Integrated Memory Co...
In addition to the above information, the ghw.PCIInfo
struct has the
following method:
ghw.PCIInfo.GetDevice(address string)
The following code snippet shows how to call the ghw.PCIInfo.GetDevice()
method and use its returned ghw.PCIDevice
struct pointer:
package main
import (
"fmt"
"os"
"github.com/jaypipes/ghw"
)
func main() {
pci, err := ghw.PCI()
if err != nil {
fmt.Printf("Error getting PCI info: %v", err)
}
addr := "0000:00:00.0"
if len(os.Args) == 2 {
addr = os.Args[1]
}
fmt.Printf("PCI device information for %s\n", addr)
fmt.Println("====================================================")
deviceInfo := pci.GetDevice(addr)
if deviceInfo == nil {
fmt.Printf("could not retrieve PCI device information for %s\n", addr)
return
}
vendor := deviceInfo.Vendor
fmt.Printf("Vendor: %s [%s]\n", vendor.Name, vendor.ID)
product := deviceInfo.Product
fmt.Printf("Product: %s [%s]\n", product.Name, product.ID)
subsystem := deviceInfo.Subsystem
subvendor := pci.Vendors[subsystem.VendorID]
subvendorName := "UNKNOWN"
if subvendor != nil {
subvendorName = subvendor.Name
}
fmt.Printf("Subsystem: %s [%s] (Subvendor: %s)\n", subsystem.Name, subsystem.ID, subvendorName)
class := deviceInfo.Class
fmt.Printf("Class: %s [%s]\n", class.Name, class.ID)
subclass := deviceInfo.Subclass
fmt.Printf("Subclass: %s [%s]\n", subclass.Name, subclass.ID)
progIface := deviceInfo.ProgrammingInterface
fmt.Printf("Programming Interface: %s [%s]\n", progIface.Name, progIface.ID)
}
Here's a sample output from my local workstation:
PCI device information for 0000:03:00.0
====================================================
Vendor: NVIDIA Corporation [10de]
Product: GP107 [GeForce GTX 1050 Ti] [1c82]
Subsystem: UNKNOWN [8613] (Subvendor: ASUSTeK Computer Inc.)
Class: Display controller [03]
Subclass: VGA compatible controller [00]
Programming Interface: VGA controller [00]
The ghw.GPU()
function returns a ghw.GPUInfo
struct that contains
information about the host computer's graphics hardware.
The ghw.GPUInfo
struct contains one field:
ghw.GPUInfo.GraphicCards
is an array of pointers toghw.GraphicsCard
structs, one for each graphics card found for the system
Each ghw.GraphicsCard
struct contains the following fields:
ghw.GraphicsCard.Index
is the system's numeric zero-based index for the card on the busghw.GraphicsCard.Address
is the PCI address for the graphics cardghw.GraphicsCard.DeviceInfo
is a pointer to aghw.PCIDevice
struct describing the graphics card. This may benil
if no PCI device information could be determined for the card.ghw.GraphicsCard.Node
is an pointer to aghw.TopologyNode
struct that the GPU/graphics card is affined to. On non-NUMA systems, this will always benil
.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
gpu, err := ghw.GPU()
if err != nil {
fmt.Printf("Error getting GPU info: %v", err)
}
fmt.Printf("%v\n", gpu)
for _, card := range gpu.GraphicsCards {
fmt.Printf(" %v\n", card)
}
}
Example output from my personal workstation:
gpu (1 graphics card)
card #0 @0000:03:00.0 -> class: 'Display controller' vendor: 'NVIDIA Corporation' product: 'GP107 [GeForce GTX 1050 Ti]'
NOTE: You can read more about the fields of the ghw.PCIDevice
struct if you'd like to dig deeper into PCI subsystem and programming interface
information
NOTE: You can read more about the fields of the
ghw.TopologyNode
struct if you'd like to dig deeper into the NUMA/topology
subsystem
The ghw.Accelerator()
function returns a ghw.AcceleratorInfo
struct that contains
information about the host computer's processing accelerator hardware. In this category
we can find used hardware for AI. The hardware detected in this category will be
processing accelerators (PCI class 1200
), 3D controllers (0302
) and Display
controllers (0380
).
The ghw.AcceleratorInfo
struct contains one field:
ghw.AcceleratorInfo.Devices
is an array of pointers toghw.AcceleratorDevice
structs, one for each processing accelerator card found for the system.
Each ghw.AcceleratorDevice
struct contains the following fields:
ghw.AcceleratorDevice.Address
is the PCI address for the processing accelerator card.ghw.AcceleratorDevice.PCIDevice
is a pointer to aghw.PCIDevice
struct. describing the processing accelerator card. This may benil
if no PCI device information could be determined for the card.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
accel, err := ghw.Accelerator()
if err != nil {
fmt.Printf("Error getting processing accelerator info: %v", err)
}
fmt.Printf("%v\n", accel)
for _, card := range accel.Devices {
fmt.Printf(" %v\n", device)
}
}
Example output from a testing machine:
processing accelerators (1 device)
device @0000:00:04.0 -> driver: 'fake_pci_driver' class: 'Processing accelerators' vendor: 'Red Hat, Inc.' product: 'QEMU PCI Test Device'
NOTE: You can read more about the fields of the ghw.PCIDevice
struct if you'd like to dig deeper into PCI subsystem and programming interface
information
The ghw.Chassis()
function returns a ghw.ChassisInfo
struct that contains
information about the host computer's hardware chassis.
The ghw.ChassisInfo
struct contains multiple fields:
ghw.ChassisInfo.AssetTag
is a string with the chassis asset tagghw.ChassisInfo.SerialNumber
is a string with the chassis serial numberghw.ChassisInfo.Type
is a string with the chassis type codeghw.ChassisInfo.TypeDescription
is a string with a description of the chassis typeghw.ChassisInfo.Vendor
is a string with the chassis vendorghw.ChassisInfo.Version
is a string with the chassis version
NOTE: These fields are often missing for non-server hardware. Don't be surprised to see empty string or "None" values.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
chassis, err := ghw.Chassis()
if err != nil {
fmt.Printf("Error getting chassis info: %v", err)
}
fmt.Printf("%v\n", chassis)
}
Example output from my personal workstation:
chassis type=Desktop vendor=System76 version=thelio-r1
NOTE: Some of the values such as serial numbers are shown as unknown because the Linux kernel by default disallows access to those fields if you're not running as root. They will be populated if it runs as root or otherwise you may see warnings like the following:
WARNING: Unable to read chassis_serial: open /sys/class/dmi/id/chassis_serial: permission denied
You can ignore them or use the Disabling warning messages feature to quiet things down.
The ghw.BIOS()
function returns a ghw.BIOSInfo
struct that contains
information about the host computer's basis input/output system (BIOS).
The ghw.BIOSInfo
struct contains multiple fields:
ghw.BIOSInfo.Vendor
is a string with the BIOS vendorghw.BIOSInfo.Version
is a string with the BIOS versionghw.BIOSInfo.Date
is a string with the date the BIOS was flashed/created
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
bios, err := ghw.BIOS()
if err != nil {
fmt.Printf("Error getting BIOS info: %v", err)
}
fmt.Printf("%v\n", bios)
}
Example output from my personal workstation:
bios vendor=System76 version=F2 Z5 date=11/14/2018
The ghw.Baseboard()
function returns a ghw.BaseboardInfo
struct that
contains information about the host computer's hardware baseboard.
The ghw.BaseboardInfo
struct contains multiple fields:
ghw.BaseboardInfo.AssetTag
is a string with the baseboard asset tagghw.BaseboardInfo.SerialNumber
is a string with the baseboard serial numberghw.BaseboardInfo.Vendor
is a string with the baseboard vendorghw.BaseboardInfo.Product
is a string with the baseboard name on Linux and Product on Windowsghw.BaseboardInfo.Version
is a string with the baseboard version
NOTE: These fields are often missing for non-server hardware. Don't be surprised to see empty string or "None" values.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
baseboard, err := ghw.Baseboard()
if err != nil {
fmt.Printf("Error getting baseboard info: %v", err)
}
fmt.Printf("%v\n", baseboard)
}
Example output from my personal workstation:
baseboard vendor=System76 version=thelio-r1
NOTE: Some of the values such as serial numbers are shown as unknown because the Linux kernel by default disallows access to those fields if you're not running as root. They will be populated if it runs as root or otherwise you may see warnings like the following:
WARNING: Unable to read board_serial: open /sys/class/dmi/id/board_serial: permission denied
You can ignore them or use the Disabling warning messages feature to quiet things down.
The ghw.Product()
function returns a ghw.ProductInfo
struct that
contains information about the host computer's hardware product line.
The ghw.ProductInfo
struct contains multiple fields:
ghw.ProductInfo.Family
is a string describing the product familyghw.ProductInfo.Name
is a string with the product nameghw.ProductInfo.SerialNumber
is a string with the product serial numberghw.ProductInfo.UUID
is a string with the product UUIDghw.ProductInfo.SKU
is a string with the product stock unit identifier (SKU)ghw.ProductInfo.Vendor
is a string with the product vendorghw.ProductInfo.Version
is a string with the product version
NOTE: These fields are often missing for non-server hardware. Don't be surprised to see empty string, "Default string" or "None" values.
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
product, err := ghw.Product()
if err != nil {
fmt.Printf("Error getting product info: %v", err)
}
fmt.Printf("%v\n", product)
}
Example output from my personal workstation:
product family=Default string name=Thelio vendor=System76 sku=Default string version=thelio-r1
NOTE: Some of the values such as serial numbers are shown as unknown because the Linux kernel by default disallows access to those fields if you're not running as root. They will be populated if it runs as root or otherwise you may see warnings like the following:
WARNING: Unable to read product_serial: open /sys/class/dmi/id/product_serial: permission denied
You can ignore them or use the Disabling warning messages feature to quiet things down.
When ghw
isn't able to retrieve some information, it may print certain
warning messages to stderr
. To disable these warnings, simply set the
GHW_DISABLE_WARNINGS
environs variable:
$ ghwc memory
WARNING:
Could not determine total physical bytes of memory. This may
be due to the host being a virtual machine or container with no
/var/log/syslog file, or the current user may not have necessary
privileges to read the syslog. We are falling back to setting the
total physical amount of memory to the total usable amount of memory
memory (24GB physical, 24GB usable)
$ GHW_DISABLE_WARNINGS=1 ghwc memory
memory (24GB physical, 24GB usable)
You can disable warning programmatically using the WithDisableWarnings
option:
import (
"github.com/jaypipes/ghw"
)
mem, err := ghw.Memory(ghw.WithDisableWarnings())
WithDisableWarnings
is a alias for the WithNullAlerter
option, which in turn
leverages the more general Alerter
feature of ghw.
You may supply a Alerter
to ghw to redirect all the warnings there, like
logger objects (see for example golang's stdlib log.Logger
).
Alerter
is in fact the minimal logging interface ghw needs. To learn more, please check the
option.Alerterinterface and the
ghw.WithAlerter()`
function.
When ghw
looks for information about the host system, it considers /
as its
root mountpoint. So, for example, when looking up CPU information on a Linux
system, ghw.CPU()
will use the path /proc/cpuinfo
.
If you are calling ghw
from a system that has an alternate root mountpoint,
you can either set the GHW_CHROOT
environment variable to that alternate
path, or call one of the functions like ghw.CPU()
or ghw.Memory()
with the
ghw.WithChroot()
modifier.
For example, if you are executing from within an application container that has
bind-mounted the root host filesystem to the mount point /host
, you would set
GHW_CHROOT
to /host
so that ghw
can find /proc/cpuinfo
at
/host/proc/cpuinfo
.
Alternately, you can use the ghw.WithChroot()
function like so:
cpu, err := ghw.CPU(ghw.WithChroot("/host"))
All of the ghw
XXXInfo
structs -- e.g. ghw.CPUInfo
-- have two methods
for producing a serialized JSON or YAML string representation of the contained
information:
JSONString()
returns a string containing the information serialized into JSON. It accepts a single boolean parameter indicating whether to use indentation when outputting the stringYAMLString()
returns a string containing the information serialized into YAML
package main
import (
"fmt"
"github.com/jaypipes/ghw"
)
func main() {
mem, err := ghw.Memory()
if err != nil {
fmt.Printf("Error getting memory info: %v", err)
}
fmt.Printf("%s", mem.YAMLString())
}
the above example code prints the following out on my local workstation:
memory:
supported_page_sizes:
- 1073741824
- 2097152
total_physical_bytes: 25263415296
total_usable_bytes: 25263415296
When running inside containers, it can be cumbersome to only override the root
mountpoint. Inside containers, when granting access to the host file systems,
it is common to bind-mount them to a non-standard location, like /sys
on
/host-sys
or /proc
to /host-proc
. It is rarer to mount them to a common
subtree (e.g. /sys
to /host/sys
and /proc
to /host/proc
...)
To better cover this use case, ghw.WithPathOverrides()
can be used to supply
a mapping of directories to mountpoints, like this example shows:
cpu, err := ghw.CPU(ghw.WithPathOverrides(ghw.PathOverrides{
"/proc": "/host-proc",
"/sys": "/host-sys",
}))
NOTE: This feature works in addition and is composable with the
ghw.WithChroot()
function and GHW_CHROOT
environment variable.
The ghw-snapshot
tool can create a snapshot of a host's hardware information.
Please read SNAPSHOT.md
to learn about creating snapshots with
the ghw-snapshot
tool.
You can make ghw
read hardware information from a snapshot created with
ghw-snapshot
using environment variables or programmatically.
Use the GHW_SNAPSHOT_PATH
environment variable to specify the filepath to a
snapshot that ghw
will read to determine hardware information. All the needed
chroot changes will be automatically performed. By default, the snapshot is
unpacked into a temporary directory managed by ghw
. This temporary directory
is automatically deleted when ghw
is finished reading the snapshot.
Three other environment variables are relevant if and only if GHW_SNAPSHOT_PATH
is not empty:
GHW_SNAPSHOT_ROOT
let users specify the directory on which the snapshot should be unpacked. This moves the ownership of that directory fromghw
to users. For this reason,ghw
will not automatically clean up the content unpacked intoGHW_SNAPSHOT_ROOT
.GHW_SNAPSHOT_EXCLUSIVE
tellsghw
that the directory is meant only to contain the given snapshot, thusghw
will not attempt to unpack it unless the directory is empty. You can use bothGHW_SNAPSHOT_ROOT
andGHW_SNAPSHOT_EXCLUSIVE
to make sureghw
unpacks the snapshot only once regardless of how manyghw
packages (e.g. cpu, memory) access it. Set the value of this environment variable to any non-empty string.GHW_SNAPSHOT_PRESERVE
tellsghw
not to clean up the unpacked snapshot. Set the value of this environment variable to any non-empty string.
cpu, err := ghw.CPU(ghw.WithSnapshot(ghw.SnapshotOptions{
Path: "/path/to/linux-amd64-d4771ed3300339bc75f856be09fc6430.tar.gz",
}))
myRoot := "/my/safe/directory"
cpu, err := ghw.CPU(ghw.WithSnapshot(ghw.SnapshotOptions{
Path: "/path/to/linux-amd64-d4771ed3300339bc75f856be09fc6430.tar.gz",
Root: &myRoot,
}))
myOtherRoot := "/my/other/safe/directory"
cpu, err := ghw.CPU(ghw.WithSnapshot(ghw.SnapshotOptions{
Path: "/path/to/linux-amd64-d4771ed3300339bc75f856be09fc6430.tar.gz",
Root: &myOtherRoot,
Exclusive: true,
}))
You can create ghw
snapshots using the ghw-snapshot
tool or
programmatically using the pkg/snapshot
package.
Below is an example of creating a ghw
snapshot using the pkg/snapshot
package.
import (
"fmt"
"os"
"github.com/jaypipes/ghw/pkg/snapshot"
)
// ...
scratchDir, err := os.MkdirTemp("", "ghw-snapshot-*")
if err != nil {
fmt.Printf("Error creating clone directory: %v", err)
}
defer os.RemoveAll(scratchDir)
// this step clones all the files and directories ghw cares about
if err := snapshot.CloneTreeInto(scratchDir); err != nil {
fmt.Printf("error cloning into %q: %v", scratchDir, err)
}
// optionally, you may add extra content into your snapshot.
// ghw will ignore the extra content.
// Glob patterns like `filepath.Glob` are supported.
fileSpecs := []string{
"/proc/cmdline",
}
// options allows the client code to optionally deference symlinks, or copy
// them into the cloned tree as symlinks
var opts *snapshot.CopyFileOptions
if err := snapshot.CopyFilesInto(fileSpecs, scratchDir, opts); err != nil {
fmt.Printf("error cloning extra files into %q: %v", scratchDir, err)
}
// automates the creation of the gzipped tarball out of the given tree.
if err := snapshot.PackFrom("my-snapshot.tgz", scratchDir); err != nil {
fmt.Printf("error packing %q into %q: %v", scratchDir, *output, err)
}
By default ghw
may call external programs, for example ethtool
, to learn
about hardware capabilities. In some rare circumstances it may be useful to
opt out from this behaviour and rely only on the data provided by
pseudo-filesystems, like sysfs.
The most common use case is when we want to read a snapshot from ghw
. In
these cases the information provided by tools will be inconsistent with the
data from the snapshot - since they will be run on a different host than the
host the snapshot was created for.
To prevent ghw
from calling external tools, set the GHW_DISABLE_TOOLS
environment variable to any value, or, programmatically, use the
ghw.WithDisableTools()
function. The default behaviour of ghw is to call
external tools when available.
WARNING: on all platforms, disabling external tools make ghw return less data. Unless noted otherwise, there is no fallback flow if external tools are disabled. On MacOSX/Darwin, disabling external tools disables block support entirely
Contributions to ghw
are welcomed! Fork the repo on GitHub
and submit a pull request with your proposed changes. Or, feel free to log an
issue for a feature request or bug report.