gcc/libgo/go/runtime/os_linux.go

// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package runtime

import (
	"runtime/internal/sys"
	"unsafe"
)

type mOS struct {
	unused byte
}

func futex(addr unsafe.Pointer, op int32, val uint32, ts, addr2 unsafe.Pointer, val3 uint32) int32 {
	return int32(syscall(_SYS_futex, uintptr(addr), uintptr(op), uintptr(val), uintptr(ts), uintptr(addr2), uintptr(val3)))
}

// For sched_getaffinity use the system call rather than the libc call,
// because the system call returns the number of entries set by the kernel.
func sched_getaffinity(pid _pid_t, cpusetsize uintptr, mask *byte) int32 {
	return int32(syscall(_SYS_sched_getaffinity, uintptr(pid), cpusetsize, uintptr(unsafe.Pointer(mask)), 0, 0, 0))
}

// Linux futex.
//
//	futexsleep(uint32 *addr, uint32 val)
//	futexwakeup(uint32 *addr)
//
// Futexsleep atomically checks if *addr == val and if so, sleeps on addr.
// Futexwakeup wakes up threads sleeping on addr.
// Futexsleep is allowed to wake up spuriously.

const (
	_FUTEX_PRIVATE_FLAG = 128
	_FUTEX_WAIT_PRIVATE = 0 | _FUTEX_PRIVATE_FLAG
	_FUTEX_WAKE_PRIVATE = 1 | _FUTEX_PRIVATE_FLAG
)

// Atomically,
//	if(*addr == val) sleep
// Might be woken up spuriously; that's allowed.
// Don't sleep longer than ns; ns < 0 means forever.
//go:nosplit
func futexsleep(addr *uint32, val uint32, ns int64) {
	// Some Linux kernels have a bug where futex of
	// FUTEX_WAIT returns an internal error code
	// as an errno. Libpthread ignores the return value
	// here, and so can we: as it says a few lines up,
	// spurious wakeups are allowed.
	if ns < 0 {
		futex(unsafe.Pointer(addr), _FUTEX_WAIT_PRIVATE, val, nil, nil, 0)
		return
	}

	var ts timespec
	ts.setNsec(ns)
	futex(unsafe.Pointer(addr), _FUTEX_WAIT_PRIVATE, val, unsafe.Pointer(&ts), nil, 0)
}

// If any procs are sleeping on addr, wake up at most cnt.
//go:nosplit
func futexwakeup(addr *uint32, cnt uint32) {
	ret := futex(unsafe.Pointer(addr), _FUTEX_WAKE_PRIVATE, cnt, nil, nil, 0)
	if ret >= 0 {
		return
	}

	// I don't know that futex wakeup can return
	// EAGAIN or EINTR, but if it does, it would be
	// safe to loop and call futex again.
	systemstack(func() {
		print("futexwakeup addr=", addr, " returned ", ret, "\n")
	})

	*(*int32)(unsafe.Pointer(uintptr(0x1006))) = 0x1006
}

func getproccount() int32 {
	// This buffer is huge (8 kB) but we are on the system stack
	// and there should be plenty of space (64 kB).
	// Also this is a leaf, so we're not holding up the memory for long.
	// See golang.org/issue/11823.
	// The suggested behavior here is to keep trying with ever-larger
	// buffers, but we don't have a dynamic memory allocator at the
	// moment, so that's a bit tricky and seems like overkill.
	const maxCPUs = 64 * 1024
	var buf [maxCPUs / 8]byte
	r := sched_getaffinity(0, unsafe.Sizeof(buf), &buf[0])
	if r < 0 {
		return 1
	}
	n := int32(0)
	for _, v := range buf[:r] {
		for v != 0 {
			n += int32(v & 1)
			v >>= 1
		}
	}
	if n == 0 {
		n = 1
	}
	return n
}

const (
	_AT_NULL   = 0  // End of vector
	_AT_PAGESZ = 6  // System physical page size
	_AT_HWCAP  = 16 // hardware capability bit vector
	_AT_RANDOM = 25 // introduced in 2.6.29
	_AT_HWCAP2 = 26 // hardware capability bit vector 2
)

var procAuxv = []byte("/proc/self/auxv\x00")

var addrspace_vec [1]byte

//extern mincore
func mincore(addr unsafe.Pointer, n uintptr, dst *byte) int32

func sysargs(argc int32, argv **byte) {
	n := argc + 1

	// skip over argv, envp to get to auxv
	for argv_index(argv, n) != nil {
		n++
	}

	// skip NULL separator
	n++

	// now argv+n is auxv
	auxv := (*[1 << 28]uintptr)(add(unsafe.Pointer(argv), uintptr(n)*sys.PtrSize))
	if sysauxv(auxv[:]) != 0 {
		return
	}
	// In some situations we don't get a loader-provided
	// auxv, such as when loaded as a library on Android.
	// Fall back to /proc/self/auxv.
	fd := open(&procAuxv[0], 0 /* O_RDONLY */, 0)
	if fd < 0 {
		// On Android, /proc/self/auxv might be unreadable (issue 9229), so we fallback to
		// try using mincore to detect the physical page size.
		// mincore should return EINVAL when address is not a multiple of system page size.
		const size = 256 << 10 // size of memory region to allocate
		p, err := mmap(nil, size, _PROT_READ|_PROT_WRITE, _MAP_ANON|_MAP_PRIVATE, -1, 0)
		if err != 0 {
			return
		}
		var n uintptr
		for n = 4 << 10; n < size; n <<= 1 {
			err := mincore(unsafe.Pointer(uintptr(p)+n), 1, &addrspace_vec[0])
			if err == 0 {
				physPageSize = n
				break
			}
		}
		if physPageSize == 0 {
			physPageSize = size
		}
		munmap(p, size)
		return
	}
	var buf [128]uintptr
	n = read(fd, noescape(unsafe.Pointer(&buf[0])), int32(unsafe.Sizeof(buf)))
	closefd(fd)
	if n < 0 {
		return
	}
	// Make sure buf is terminated, even if we didn't read
	// the whole file.
	buf[len(buf)-2] = _AT_NULL
	sysauxv(buf[:])
}

func sysauxv(auxv []uintptr) int {
	var i int
	for ; auxv[i] != _AT_NULL; i += 2 {
		tag, val := auxv[i], auxv[i+1]
		switch tag {
		case _AT_RANDOM:
			// The kernel provides a pointer to 16-bytes
			// worth of random data.
			startupRandomData = (*[16]byte)(unsafe.Pointer(val))[:]

			setRandomNumber(uint32(startupRandomData[4]) | uint32(startupRandomData[5])<<8 |
				uint32(startupRandomData[6])<<16 | uint32(startupRandomData[7])<<24)

		case _AT_PAGESZ:
			physPageSize = val
		}

		archauxv(tag, val)

		// Commented out for gccgo for now.
		// vdsoauxv(tag, val)
	}
	return i / 2
}

var sysTHPSizePath = []byte("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size\x00")

func getHugePageSize() uintptr {
	var numbuf [20]byte
	fd := open(&sysTHPSizePath[0], 0 /* O_RDONLY */, 0)
	if fd < 0 {
		return 0
	}
	n := read(fd, noescape(unsafe.Pointer(&numbuf[0])), int32(len(numbuf)))
	closefd(fd)
	if n <= 0 {
		return 0
	}
	l := n - 1 // remove trailing newline
	v, ok := atoi(slicebytetostringtmp(numbuf[:l]))
	if !ok || v < 0 {
		v = 0
	}
	if v&(v-1) != 0 {
		// v is not a power of 2
		return 0
	}
	return uintptr(v)
}

func osinit() {
	ncpu = getproccount()
	physHugePageSize = getHugePageSize()
}