一、基本的数据结构
先看一下基本的内存数据结构:
//runtime/mheap.go
// minPhysPageSize is a lower-bound on the physical page size. The
// true physical page size may be larger than this. In contrast,
// sys.PhysPageSize is an upper-bound on the physical page size.
const minPhysPageSize = 4096
type mheap struct {
lock mutex
free [_MaxMHeapList]mSpanList // free lists of given length
freelarge mSpanList // free lists length >= _MaxMHeapList
busy [_MaxMHeapList]mSpanList // busy lists of large objects of given length
busylarge mSpanList // busy lists of large objects length >= _MaxMHeapList
sweepgen uint32 // sweep generation, see comment in mspan
sweepdone uint32 // all spans are swept
// allspans is a slice of all mspans ever created. Each mspan
// appears exactly once.
//
// The memory for allspans is manually managed and can be
// reallocated and move as the heap grows.
//
// In general, allspans is protected by mheap_.lock, which
// prevents concurrent access as well as freeing the backing
// store. Accesses during STW might not hold the lock, but
// must ensure that allocation cannot happen around the
// access (since that may free the backing store).
allspans []*mspan // all spans out there
// spans is a lookup table to map virtual address page IDs to *mspan.
// For allocated spans, their pages map to the span itself.
// For free spans, only the lowest and highest pages map to the span itself.
// Internal pages map to an arbitrary span.
// For pages that have never been allocated, spans entries are nil.
//
// This is backed by a reserved region of the address space so
// it can grow without moving. The memory up to len(spans) is
// mapped. cap(spans) indicates the total reserved memory.
spans []*mspan
// sweepSpans contains two mspan stacks: one of swept in-use
// spans, and one of unswept in-use spans. These two trade
// roles on each GC cycle. Since the sweepgen increases by 2
// on each cycle, this means the swept spans are in
// sweepSpans[sweepgen/2%2] and the unswept spans are in
// sweepSpans[1-sweepgen/2%2]. Sweeping pops spans from the
// unswept stack and pushes spans that are still in-use on the
// swept stack. Likewise, allocating an in-use span pushes it
// on the swept stack.
sweepSpans [2]gcSweepBuf
_ uint32 // align uint64 fields on 32-bit for atomics
// Proportional sweep
pagesInUse uint64 // pages of spans in stats _MSpanInUse; R/W with mheap.lock
spanBytesAlloc uint64 // bytes of spans allocated this cycle; updated atomically
pagesSwept uint64 // pages swept this cycle; updated atomically
sweepPagesPerByte float64 // proportional sweep ratio; written with lock, read without
// TODO(austin): pagesInUse should be a uintptr, but the 386
// compiler can't 8-byte align fields.
// Malloc stats.
largefree uint64 // bytes freed for large objects (>maxsmallsize)
nlargefree uint64 // number of frees for large objects (>maxsmallsize)
nsmallfree [_NumSizeClasses]uint64 // number of frees for small objects (<=maxsmallsize)
// range of addresses we might see in the heap
bitmap uintptr // Points to one byte past the end of the bitmap
bitmap_mapped uintptr
arena_start uintptr
arena_used uintptr // always mHeap_Map{Bits,Spans} before updating
arena_end uintptr
arena_reserved bool
// central free lists for small size classes.
// the padding makes sure that the MCentrals are
// spaced CacheLineSize bytes apart, so that each MCentral.lock
// gets its own cache line.
central [_NumSizeClasses]struct {
mcentral mcentral
pad [sys.CacheLineSize]byte
}
spanalloc fixalloc // allocator for span*
cachealloc fixalloc // allocator for mcache*
specialfinalizeralloc fixalloc // allocator for specialfinalizer*
specialprofilealloc fixalloc // allocator for specialprofile*
speciallock mutex // lock for special record allocators.
}
type mspan struct {
next *mspan // next span in list, or nil if none
prev *mspan // previous span in list, or nil if none
list *mSpanList // For debugging. TODO: Remove.
startAddr uintptr // address of first byte of span aka s.base()
npages uintptr // number of pages in span
stackfreelist gclinkptr // list of free stacks, avoids overloading freelist
// freeindex is the slot index between 0 and nelems at which to begin scanning
// for the next free object in this span.
// Each allocation scans allocBits starting at freeindex until it encounters a 0
// indicating a free object. freeindex is then adjusted so that subsequent scans begin
// just past the the newly discovered free object.
//
// If freeindex == nelem, this span has no free objects.
//
// allocBits is a bitmap of objects in this span.
// If n >= freeindex and allocBits[n/8] & (1<<(n%8)) is 0
// then object n is free;
// otherwise, object n is allocated. Bits starting at nelem are
// undefined and should never be referenced.
//
// Object n starts at address n*elemsize + (start << pageShift).
freeindex uintptr
// TODO: Look up nelems from sizeclass and remove this field if it
// helps performance.
nelems uintptr // number of object in the span.
// Cache of the allocBits at freeindex. allocCache is shifted
// such that the lowest bit corresponds to the bit freeindex.
// allocCache holds the complement of allocBits, thus allowing
// ctz (count trailing zero) to use it directly.
// allocCache may contain bits beyond s.nelems; the caller must ignore
// these.
allocCache uint64
// allocBits and gcmarkBits hold pointers to a span's mark and
// allocation bits. The pointers are 8 byte aligned.
// There are three arenas where this data is held.
// free: Dirty arenas that are no longer accessed
// and can be reused.
// next: Holds information to be used in the next GC cycle.
// current: Information being used during this GC cycle.
// previous: Information being used during the last GC cycle.
// A new GC cycle starts with the call to finishsweep_m.
// finishsweep_m moves the previous arena to the free arena,
// the current arena to the previous arena, and
// the next arena to the current arena.
// The next arena is populated as the spans request
// memory to hold gcmarkBits for the next GC cycle as well
// as allocBits for newly allocated spans.
//
// The pointer arithmetic is done "by hand" instead of using
// arrays to avoid bounds checks along critical performance
// paths.
// The sweep will free the old allocBits and set allocBits to the
// gcmarkBits. The gcmarkBits are replaced with a fresh zeroed
// out memory.
allocBits *uint8
gcmarkBits *uint8
// sweep generation:
// if sweepgen == h->sweepgen - 2, the span needs sweeping
// if sweepgen == h->sweepgen - 1, the span is currently being swept
// if sweepgen == h->sweepgen, the span is swept and ready to use
// h->sweepgen is incremented by 2 after every GC
sweepgen uint32
divMul uint16 // for divide by elemsize - divMagic.mul
baseMask uint16 // if non-0, elemsize is a power of 2, & this will get object allocation base
allocCount uint16 // capacity - number of objects in freelist
sizeclass uint8 // size class
incache bool // being used by an mcache
state mSpanState // mspaninuse etc
needzero uint8 // needs to be zeroed before allocation
divShift uint8 // for divide by elemsize - divMagic.shift
divShift2 uint8 // for divide by elemsize - divMagic.shift2
elemsize uintptr // computed from sizeclass or from npages
unusedsince int64 // first time spotted by gc in mspanfree state
npreleased uintptr // number of pages released to the os
limit uintptr // end of data in span
speciallock mutex // guards specials list
specials *special // linked list of special records sorted by offset.
}
type gcBitsHeader struct {
free uintptr // free is the index into bits of the next free byte.
next uintptr // *gcBits triggers recursive type bug. (issue 14620)
}
//go:notinheap
type gcBits struct {
// gcBitsHeader // side step recursive type bug (issue 14620) by including fields by hand.
free uintptr // free is the index into bits of the next free byte.
next *gcBits
bits [gcBitsChunkBytes - gcBitsHeaderBytes]uint8
}
var gcBitsArenas struct {
lock mutex
free *gcBits
next *gcBits
current *gcBits
previous *gcBits
}
从上面的代码再比对Go的内存管理模型,就可以知道对应的内容。最基本的是Span,然后是Arena,GC是Bitmap。这在mehap的数据结构中都体现出来了。同样,为了保持和物理页的同步,一开始设置了minPhysPageSize = 4K,一般在X86上默认页的大小就是4K,当然在一些其它机器和最新的一些大页管理中,还是有不同的,但这并不妨碍最小的定义。
//runtime/mstats.go
// Statistics.
// If you edit this structure, also edit type MemStats below.
// Their layouts must match exactly.
//
// For detailed descriptions see the documentation for MemStats.
// Fields that differ from MemStats are further documented here.
//
// Many of these fields are updated on the fly, while others are only
// updated when updatememstats is called.
type mstats struct {
// General statistics.
alloc uint64 // bytes allocated and not yet freed
total_alloc uint64 // bytes allocated (even if freed)
sys uint64 // bytes obtained from system (should be sum of xxx_sys below, no locking, approximate)
nlookup uint64 // number of pointer lookups
nmalloc uint64 // number of mallocs
nfree uint64 // number of frees
// Statistics about malloc heap.
// Protected by mheap.lock
//
// In mstats, heap_sys and heap_inuse includes stack memory,
// while in MemStats stack memory is separated out from the
// heap stats.
heap_alloc uint64 // bytes allocated and not yet freed (same as alloc above)
heap_sys uint64 // virtual address space obtained from system
heap_idle uint64 // bytes in idle spans
heap_inuse uint64 // bytes in non-idle spans
heap_released uint64 // bytes released to the os
heap_objects uint64 // total number of allocated objects
// TODO(austin): heap_released is both useless and inaccurate
// in its current form. It's useless because, from the user's
// and OS's perspectives, there's no difference between a page
// that has not yet been faulted in and a page that has been
// released back to the OS. We could fix this by considering
// newly mapped spans to be "released". It's inaccurate
// because when we split a large span for allocation, we
// "unrelease" all pages in the large span and not just the
// ones we split off for use. This is trickier to fix because
// we currently don't know which pages of a span we've
// released. We could fix it by separating "free" and
// "released" spans, but then we have to allocate from runs of
// free and released spans.
// Statistics about allocation of low-level fixed-size structures.
// Protected by FixAlloc locks.
stacks_inuse uint64 // this number is included in heap_inuse above; differs from MemStats.StackInuse
stacks_sys uint64 // only counts newosproc0 stack in mstats; differs from MemStats.StackSys
mspan_inuse uint64 // mspan structures
mspan_sys uint64
mcache_inuse uint64 // mcache structures
mcache_sys uint64
buckhash_sys uint64 // profiling bucket hash table
gc_sys uint64
other_sys uint64
// Statistics about garbage collector.
// Protected by mheap or stopping the world during GC.
next_gc uint64 // goal heap_live for when next GC ends; ^0 if disabled
last_gc uint64 // last gc (in absolute time)
pause_total_ns uint64
pause_ns [256]uint64 // circular buffer of recent gc pause lengths
pause_end [256]uint64 // circular buffer of recent gc end times (nanoseconds since 1970)
numgc uint32
numforcedgc uint32 // number of user-forced GCs
gc_cpu_fraction float64 // fraction of CPU time used by GC
enablegc bool
debuggc bool
// Statistics about allocation size classes.
by_size [_NumSizeClasses]struct {
size uint32
nmalloc uint64
nfree uint64
}
// Statistics below here are not exported to MemStats directly.
tinyallocs uint64 // number of tiny allocations that didn't cause actual allocation; not exported to go directly
// gc_trigger is the heap size that triggers marking.
//
// When heap_live ≥ gc_trigger, the mark phase will start.
// This is also the heap size by which proportional sweeping
// must be complete.
gc_trigger uint64
// heap_live is the number of bytes considered live by the GC.
// That is: retained by the most recent GC plus allocated
// since then. heap_live <= heap_alloc, since heap_alloc
// includes unmarked objects that have not yet been swept (and
// hence goes up as we allocate and down as we sweep) while
// heap_live excludes these objects (and hence only goes up
// between GCs).
//
// This is updated atomically without locking. To reduce
// contention, this is updated only when obtaining a span from
// an mcentral and at this point it counts all of the
// unallocated slots in that span (which will be allocated
// before that mcache obtains another span from that
// mcentral). Hence, it slightly overestimates the "true" live
// heap size. It's better to overestimate than to
// underestimate because 1) this triggers the GC earlier than
// necessary rather than potentially too late and 2) this
// leads to a conservative GC rate rather than a GC rate that
// is potentially too low.
//
// Whenever this is updated, call traceHeapAlloc() and
// gcController.revise().
heap_live uint64
// heap_scan is the number of bytes of "scannable" heap. This
// is the live heap (as counted by heap_live), but omitting
// no-scan objects and no-scan tails of objects.
//
// Whenever this is updated, call gcController.revise().
heap_scan uint64
// heap_marked is the number of bytes marked by the previous
// GC. After mark termination, heap_live == heap_marked, but
// unlike heap_live, heap_marked does not change until the
// next mark termination.
heap_marked uint64
}
var memstats mstats
// A MemStats records statistics about the memory allocator.
type MemStats struct {
// General statistics.
// Alloc is bytes of allocated heap objects.
//
// This is the same as HeapAlloc (see below).
Alloc uint64
// TotalAlloc is cumulative bytes allocated for heap objects.
//
// TotalAlloc increases as heap objects are allocated, but
// unlike Alloc and HeapAlloc, it does not decrease when
// objects are freed.
TotalAlloc uint64
// Sys is the total bytes of memory obtained from the OS.
//
// Sys is the sum of the XSys fields below. Sys measures the
// virtual address space reserved by the Go runtime for the
// heap, stacks, and other internal data structures. It's
// likely that not all of the virtual address space is backed
// by physical memory at any given moment, though in general
// it all was at some point.
Sys uint64
// Lookups is the number of pointer lookups performed by the
// runtime.
//
// This is primarily useful for debugging runtime internals.
Lookups uint64
// Mallocs is the cumulative count of heap objects allocated.
// The number of live objects is Mallocs - Frees.
Mallocs uint64
// Frees is the cumulative count of heap objects freed.
Frees uint64
// Heap memory statistics.
//
// Interpreting the heap statistics requires some knowledge of
// how Go organizes memory. Go divides the virtual address
// space of the heap into "spans", which are contiguous
// regions of memory 8K or larger. A span may be in one of
// three states:
//
// An "idle" span contains no objects or other data. The
// physical memory backing an idle span can be released back
// to the OS (but the virtual address space never is), or it
// can be converted into an "in use" or "stack" span.
//
// An "in use" span contains at least one heap object and may
// have free space available to allocate more heap objects.
//
// A "stack" span is used for goroutine stacks. Stack spans
// are not considered part of the heap. A span can change
// between heap and stack memory; it is never used for both
// simultaneously.
// HeapAlloc is bytes of allocated heap objects.
//
// "Allocated" heap objects include all reachable objects, as
// well as unreachable objects that the garbage collector has
// not yet freed. Specifically, HeapAlloc increases as heap
// objects are allocated and decreases as the heap is swept
// and unreachable objects are freed. Sweeping occurs
// incrementally between GC cycles, so these two processes
// occur simultaneously, and as a result HeapAlloc tends to
// change smoothly (in contrast with the sawtooth that is
// typical of stop-the-world garbage collectors).
HeapAlloc uint64
// HeapSys is bytes of heap memory obtained from the OS.
//
// HeapSys measures the amount of virtual address space
// reserved for the heap. This includes virtual address space
// that has been reserved but not yet used, which consumes no
// physical memory, but tends to be small, as well as virtual
// address space for which the physical memory has been
// returned to the OS after it became unused (see HeapReleased
// for a measure of the latter).
//
// HeapSys estimates the largest size the heap has had.
HeapSys uint64
// HeapIdle is bytes in idle (unused) spans.
//
// Idle spans have no objects in them. These spans could be
// (and may already have been) returned to the OS, or they can
// be reused for heap allocations, or they can be reused as
// stack memory.
//
// HeapIdle minus HeapReleased estimates the amount of memory
// that could be returned to the OS, but is being retained by
// the runtime so it can grow the heap without requesting more
// memory from the OS. If this difference is significantly
// larger than the heap size, it indicates there was a recent
// transient spike in live heap size.
HeapIdle uint64
// HeapInuse is bytes in in-use spans.
//
// In-use spans have at least one object in them. These spans
// can only be used for other objects of roughly the same
// size.
//
// HeapInuse minus HeapAlloc esimates the amount of memory
// that has been dedicated to particular size classes, but is
// not currently being used. This is an upper bound on
// fragmentation, but in general this memory can be reused
// efficiently.
HeapInuse uint64
// HeapReleased is bytes of physical memory returned to the OS.
//
// This counts heap memory from idle spans that was returned
// to the OS and has not yet been reacquired for the heap.
HeapReleased uint64
// HeapObjects is the number of allocated heap objects.
//
// Like HeapAlloc, this increases as objects are allocated and
// decreases as the heap is swept and unreachable objects are
// freed.
HeapObjects uint64
// Stack memory statistics.
//
// Stacks are not considered part of the heap, but the runtime
// can reuse a span of heap memory for stack memory, and
// vice-versa.
// StackInuse is bytes in stack spans.
//
// In-use stack spans have at least one stack in them. These
// spans can only be used for other stacks of the same size.
//
// There is no StackIdle because unused stack spans are
// returned to the heap (and hence counted toward HeapIdle).
StackInuse uint64
// StackSys is bytes of stack memory obtained from the OS.
//
// StackSys is StackInuse, plus any memory obtained directly
// from the OS for OS thread stacks (which should be minimal).
StackSys uint64
// Off-heap memory statistics.
//
// The following statistics measure runtime-internal
// structures that are not allocated from heap memory (usually
// because they are part of implementing the heap). Unlike
// heap or stack memory, any memory allocated to these
// structures is dedicated to these structures.
//
// These are primarily useful for debugging runtime memory
// overheads.
// MSpanInuse is bytes of allocated mspan structures.
MSpanInuse uint64
// MSpanSys is bytes of memory obtained from the OS for mspan
// structures.
MSpanSys uint64
// MCacheInuse is bytes of allocated mcache structures.
MCacheInuse uint64
// MCacheSys is bytes of memory obtained from the OS for
// mcache structures.
MCacheSys uint64
// BuckHashSys is bytes of memory in profiling bucket hash tables.
BuckHashSys uint64
// GCSys is bytes of memory in garbage collection metadata.
GCSys uint64
// OtherSys is bytes of memory in miscellaneous off-heap
// runtime allocations.
OtherSys uint64
// Garbage collector statistics.
// NextGC is the target heap size of the next GC cycle.
//
// The garbage collector's goal is to keep HeapAlloc ≤ NextGC.
// At the end of each GC cycle, the target for the next cycle
// is computed based on the amount of reachable data and the
// value of GOGC.
NextGC uint64
// LastGC is the time the last garbage collection finished, as
// nanoseconds since 1970 (the UNIX epoch).
LastGC uint64
// PauseTotalNs is the cumulative nanoseconds in GC
// stop-the-world pauses since the program started.
//
// During a stop-the-world pause, all goroutines are paused
// and only the garbage collector can run.
PauseTotalNs uint64
// PauseNs is a circular buffer of recent GC stop-the-world
// pause times in nanoseconds.
//
// The most recent pause is at PauseNs[(NumGC+255)%256]. In
// general, PauseNs[N%256] records the time paused in the most
// recent N%256th GC cycle. There may be multiple pauses per
// GC cycle; this is the sum of all pauses during a cycle.
PauseNs [256]uint64
// PauseEnd is a circular buffer of recent GC pause end times,
// as nanoseconds since 1970 (the UNIX epoch).
//
// This buffer is filled the same way as PauseNs. There may be
// multiple pauses per GC cycle; this records the end of the
// last pause in a cycle.
PauseEnd [256]uint64
// NumGC is the number of completed GC cycles.
NumGC uint32
// NumForcedGC is the number of GC cycles that were forced by
// the application calling the GC function.
NumForcedGC uint32
// GCCPUFraction is the fraction of this program's available
// CPU time used by the GC since the program started.
//
// GCCPUFraction is expressed as a number between 0 and 1,
// where 0 means GC has consumed none of this program's CPU. A
// program's available CPU time is defined as the integral of
// GOMAXPROCS since the program started. That is, if
// GOMAXPROCS is 2 and a program has been running for 10
// seconds, its "available CPU" is 20 seconds. GCCPUFraction
// does not include CPU time used for write barrier activity.
//
// This is the same as the fraction of CPU reported by
// GODEBUG=gctrace=1.
GCCPUFraction float64
// EnableGC indicates that GC is enabled. It is always true,
// even if GOGC=off.
EnableGC bool
// DebugGC is currently unused.
DebugGC bool
// BySize reports per-size class allocation statistics.
//
// BySize[N] gives statistics for allocations of size S where
// BySize[N-1].Size < S ≤ BySize[N].Size.
//
// This does not report allocations larger than BySize[60].Size.
BySize [61]struct {
// Size is the maximum byte size of an object in this
// size class.
Size uint32
// Mallocs is the cumulative count of heap objects
// allocated in this size class. The cumulative bytes
// of allocation is Size*Mallocs. The number of live
// objects in this size class is Mallocs - Frees.
Mallocs uint64
// Frees is the cumulative count of heap objects freed
// in this size class.
Frees uint64
}
}
这两个数据结构体在上面的说明里提到了,一损俱损,一荣俱荣,不能单独搞一个修改。他们的主要目的在于提供当前内存管理的数据状态。各种统计信息数据都在这两个数据结构体里有体现。注释已经很清楚,不再说明。
//runtime/mcache.go
//go:notinheap
type mcache struct {
// The following members are accessed on every malloc,
// so they are grouped here for better caching.
next_sample int32 // trigger heap sample after allocating this many bytes
local_scan uintptr // bytes of scannable heap allocated
// Allocator cache for tiny objects w/o pointers.
// See "Tiny allocator" comment in malloc.go.
// tiny points to the beginning of the current tiny block, or
// nil if there is no current tiny block.
//
// tiny is a heap pointer. Since mcache is in non-GC'd memory,
// we handle it by clearing it in releaseAll during mark
// termination.
tiny uintptr
tinyoffset uintptr
local_tinyallocs uintptr // number of tiny allocs not counted in other stats
// The rest is not accessed on every malloc.
alloc [_NumSizeClasses]*mspan // spans to allocate from
stackcache [_NumStackOrders]stackfreelist
// Local allocator stats, flushed during GC.
local_nlookup uintptr // number of pointer lookups
local_largefree uintptr // bytes freed for large objects (>maxsmallsize)
local_nlargefree uintptr // number of frees for large objects (>maxsmallsize)
local_nsmallfree [_NumSizeClasses]uintptr // number of frees for small objects (<=maxsmallsize)
}
//runtime/mcentral.go
// The MCentral doesn't actually contain the list of free objects; the MSpan does.
// Each MCentral is two lists of MSpans: those with free objects (c->nonempty)
// and those that are completely allocated (c->empty).
// Central list of free objects of a given size.
//
//go:notinheap
type mcentral struct {
lock mutex
sizeclass int32
nonempty mSpanList // list of spans with a free object, ie a nonempty free list
empty mSpanList // list of spans with no free objects (or cached in an mcache)
}
//runtime/mfixalloc.go
// FixAlloc is a simple free-list allocator for fixed size objects.
// Malloc uses a FixAlloc wrapped around sysAlloc to manages its
// MCache and MSpan objects.
//
// Memory returned by fixalloc.alloc is zeroed by default, but the
// caller may take responsibility for zeroing allocations by setting
// the zero flag to false. This is only safe if the memory never
// contains heap pointers.
//
// The caller is responsible for locking around FixAlloc calls.
// Callers can keep state in the object but the first word is
// smashed by freeing and reallocating.
//
// Consider marking fixalloc'd types go:notinheap.
type fixalloc struct {
size uintptr
first func(arg, p unsafe.Pointer) // called first time p is returned
arg unsafe.Pointer
list *mlink
chunk unsafe.Pointer
nchunk uint32
inuse uintptr // in-use bytes now
stat *uint64
zero bool // zero allocations
}
而管理Span的数据结构为mcentral,它通过mcache这个数据结构来为线程申请内存时的缓存,这样就需要再操作锁的过程。而上面的mheap来管理所有的堆,这样一大一小,就把内存的管理搞定了。
二、流程代码分析
看完后基本的数据结构代码,就可以看一下内存管理的的流程了,从init开始:
//mgc.go
func gcinit() {
if unsafe.Sizeof(workbuf{}) != _WorkbufSize {
throw("size of Workbuf is suboptimal")
}
_ = setGCPercent(readgogc())
memstats.gc_trigger = heapminimum
// Compute the goal heap size based on the trigger:
// trigger = marked * (1 + triggerRatio)
// marked = trigger / (1 + triggerRatio)
// goal = marked * (1 + GOGC/100)
// = trigger / (1 + triggerRatio) * (1 + GOGC/100)
//当下一次 GC 结束后,堆内存的目标
memstats.next_gc = uint64(float64(memstats.gc_trigger) / (1 + gcController.triggerRatio) * (1 + float64(gcpercent)/100))
if gcpercent < 0 {
memstats.next_gc = ^uint64(0)
}
work.startSema = 1
work.markDoneSema = 1
}
//go:linkname setGCPercent runtime/debug.setGCPercent
func setGCPercent(in int32) (out int32) {
lock(&mheap_.lock)
out = gcpercent
if in < 0 {
in = -1
}
gcpercent = in
heapminimum = defaultHeapMinimum * uint64(gcpercent) / 100
if gcController.triggerRatio > float64(gcpercent)/100 {
gcController.triggerRatio = float64(gcpercent) / 100
}
// This is either in gcinit or followed by a STW GC, both of
// which will reset other stats like memstats.gc_trigger and
// memstats.next_gc to appropriate values.
unlock(&mheap_.lock)
return out
}
func readgogc() int32 {
p := gogetenv("GOGC")
if p == "off" {
return -1
}
if n, ok := atoi32(p); ok {
return n
}
return 100
}
GC的初始化中,首先对缓冲区的大小进行控制,然后对GC的频率进行控制,也即对nextgc和触发进行比例分配(以100为基点“defaultHeapMinimum is the value of heapminimum for GOGC==100”)。在setGCPercent这个函数中可以看到,heapminimum 的计算方式,而这个计算结果就是后面的内存状态数据结构中的触发数据。
在这个版本中,触发、标记和目标被注释。一定要注意到GC设置中的锁保护,这是至关重要的问题,在前面数据结构中,英文注释中也提到了。内存操作,首要就是安全。初始化基本就是对一些参数状态的配置,然后就没有然后了。
初始化完成后,就可以启动GC了,看一下这个函数:
func gcStart(mode gcMode, forceTrigger bool) {
// Since this is called from malloc and malloc is called in
// the guts of a number of libraries that might be holding
// locks, don't attempt to start GC in non-preemptible or
// potentially unstable situations.
mp := acquirem()
if gp := getg(); gp == mp.g0 || mp.locks > 1 || mp.preemptoff != "" {
releasem(mp)
return
}
releasem(mp)
mp = nil
// Pick up the remaining unswept/not being swept spans concurrently
//
// This shouldn't happen if we're being invoked in background
// mode since proportional sweep should have just finished
// sweeping everything, but rounding errors, etc, may leave a
// few spans unswept. In forced mode, this is necessary since
// GC can be forced at any point in the sweeping cycle.
//
// We check the transition condition continuously here in case
// this G gets delayed in to the next GC cycle.
for (mode != gcBackgroundMode || gcShouldStart(forceTrigger)) && gosweepone() != ^uintptr(0) {
sweep.nbgsweep++
}
// Perform GC initialization and the sweep termination
// transition.
//
// If this is a forced GC, don't acquire the transition lock
// or re-check the transition condition because we
// specifically *don't* want to share the transition with
// another thread.
useStartSema := mode == gcBackgroundMode
if useStartSema {
semacquire(&work.startSema, 0)
// Re-check transition condition under transition lock.
if !gcShouldStart(forceTrigger) {
semrelease(&work.startSema)
return
}
}
// For stats, check if this GC was forced by the user.
forced := mode != gcBackgroundMode
// In gcstoptheworld debug mode, upgrade the mode accordingly.
// We do this after re-checking the transition condition so
// that multiple goroutines that detect the heap trigger don't
// start multiple STW GCs.
if mode == gcBackgroundMode {
if debug.gcstoptheworld == 1 {
mode = gcForceMode
} else if debug.gcstoptheworld == 2 {
mode = gcForceBlockMode
}
}
// Ok, we're doing it! Stop everybody else
semacquire(&worldsema, 0)
if trace.enabled {
traceGCStart()
}
if mode == gcBackgroundMode {
gcBgMarkStartWorkers()
}
gcResetMarkState()
now := nanotime()
work.stwprocs, work.maxprocs = gcprocs(), gomaxprocs
work.tSweepTerm = now
work.heap0 = memstats.heap_live
work.pauseNS = 0
work.mode = mode
work.pauseStart = now
systemstack(stopTheWorldWithSema)
// Finish sweep before we start concurrent scan.
systemstack(func() {
finishsweep_m()
})
// clearpools before we start the GC. If we wait they memory will not be
// reclaimed until the next GC cycle.
clearpools()
if mode == gcBackgroundMode { // Do as much work concurrently as possible
gcController.startCycle()
work.heapGoal = memstats.next_gc
// Enter concurrent mark phase and enable
// write barriers.
//
// Because the world is stopped, all Ps will
// observe that write barriers are enabled by
// the time we start the world and begin
// scanning.
//
// It's necessary to enable write barriers
// during the scan phase for several reasons:
//
// They must be enabled for writes to higher
// stack frames before we scan stacks and
// install stack barriers because this is how
// we track writes to inactive stack frames.
// (Alternatively, we could not install stack
// barriers over frame boundaries with
// up-pointers).
//
// They must be enabled before assists are
// enabled because they must be enabled before
// any non-leaf heap objects are marked. Since
// allocations are blocked until assists can
// happen, we want enable assists as early as
// possible.
setGCPhase(_GCmark)
gcBgMarkPrepare() // Must happen before assist enable.
gcMarkRootPrepare()
// Mark all active tinyalloc blocks. Since we're
// allocating from these, they need to be black like
// other allocations. The alternative is to blacken
// the tiny block on every allocation from it, which
// would slow down the tiny allocator.
gcMarkTinyAllocs()
// At this point all Ps have enabled the write
// barrier, thus maintaining the no white to
// black invariant. Enable mutator assists to
// put back-pressure on fast allocating
// mutators.
atomic.Store(&gcBlackenEnabled, 1)
// Assists and workers can start the moment we start
// the world.
gcController.markStartTime = now
// Concurrent mark.
systemstack(startTheWorldWithSema)
now = nanotime()
work.pauseNS += now - work.pauseStart
work.tMark = now
} else {
t := nanotime()
work.tMark, work.tMarkTerm = t, t
work.heapGoal = work.heap0
if forced {
memstats.numforcedgc++
}
// Perform mark termination. This will restart the world.
gcMarkTermination()
}
if useStartSema {
semrelease(&work.startSema)
}
}
而哪里开始调用这个函数呢:
//mgc.go
// GC runs a garbage collection and blocks the caller until the
// garbage collection is complete. It may also block the entire
// program.
func GC() {
gcStart(gcForceBlockMode, false)
}
看看注释,“GC运行垃圾收集并阻塞调用程序,直到垃圾收集已完成。它还可能阻塞整个程序。”这是整个世界停止的节奏啊。
还有:
// Allocate an object of size bytes.
// Small objects are allocated from the per-P cache's free lists.
// Large objects (> 32 kB) are allocated straight from the heap.
func mallocgc(size uintptr, typ *_type, needzero bool) unsafe.Pointer {
......
if shouldhelpgc && gcShouldStart(false) {
gcStart(gcBackgroundMode, false)
}
return x
}
//mheap.go
//go:linkname runtime_debug_freeOSMemory runtime/debug.freeOSMemory
func runtime_debug_freeOSMemory() {
gcStart(gcForceBlockMode, false)
systemstack(func() { mheap_.scavenge(-1, ^uint64(0), 0) })
}
//proc.go
func forcegchelper() {
forcegc.g = getg()
for {
lock(&forcegc.lock)
if forcegc.idle != 0 {
throw("forcegc: phase error")
}
atomic.Store(&forcegc.idle, 1)
goparkunlock(&forcegc.lock, "force gc (idle)", traceEvGoBlock, 1)
// this goroutine is explicitly resumed by sysmon
if debug.gctrace > 0 {
println("GC forced")
}
gcStart(gcBackgroundMode, true)
}
}
这几个调用的地方恰好印证了GO中何时启动垃圾回收,一种是在手动调用GC函数时,上面看到了吧;另外一个是在申请内存时,根据堆大小来调用;还是有一个是在强制GC时调用,也就是最后一个。
Go 程序启动后,会在后台run一个线程,定时运行runtime.sysmon 函数,它主要用来检查死锁、运行计时器、调度抢占、以及 GC 等状态。其通过test 函数来判断是否应该进行 GC。由于 GC 可能需要执行时间比较长,所以启动一个强制触发垃圾收集的 Goroutine 执行 forcegchelper 函数。但 forcegchelper 函数一般会被 goparkunlock 函数挂起,直到 sysmon 触发GC 校验通过,才会将该被挂起的 Goroutine 放转身到全局调度队列中等待被调度执行 GC。
在gcStart这个函数中,基本上把所有的GC过程都包含了。包括一些控制方式,这里要是继续分析下去,就是一个超长篇的代码分析了。挪到下面,把控制策略和具体的GC过程逐一分析。
三、总结
理论从哪里来?太祖说过:不是从天上掉下来的。理论是从实践活动中来。但自从人类有了书籍可以传承知识后,后人在学习的过程中,往往忽视了这些理论是从哪里来的。这在文学上往往有意无意被忽略。但是在计算机这种强理论和实践结合的领域上,则不得不重视,结果就是很多人感到不可理解,很高深的样子。所以在前面学习了很多GC的理论,在后面就要把理论和实践相结合起来。每个实践和理论不一定百分百相吻合,一定要特殊之处,但整体的间架结构一定不会有不同。通过理论来指导实践,再通过实践反过来验证并不断抽象新的理论。这才是一个否定之否定的过程,一个追求真理的过程。
世上万物,莫不如是!