1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
|
// Copyright 2018 The gVisor Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//go:build amd64
// +build amd64
package systrap
import (
"fmt"
"strings"
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/seccomp"
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/platform/systrap/sysmsg"
)
const (
// initRegsRipAdjustment is the size of the syscall instruction.
initRegsRipAdjustment = 2
)
// resetSysemuRegs sets up emulation registers.
//
// This should be called prior to calling sysemu.
func (s *subprocess) resetSysemuRegs(regs *arch.Registers) {
regs.Cs = s.sysmsgInitRegs.Cs
regs.Ss = s.sysmsgInitRegs.Ss
regs.Ds = s.sysmsgInitRegs.Ds
regs.Es = s.sysmsgInitRegs.Es
regs.Fs = s.sysmsgInitRegs.Fs
regs.Gs = s.sysmsgInitRegs.Gs
}
// createSyscallRegs sets up syscall registers.
//
// This should be called to generate registers for a system call.
func createSyscallRegs(initRegs *arch.Registers, sysno uintptr, args ...arch.SyscallArgument) arch.Registers {
// Copy initial registers.
regs := *initRegs
// Set our syscall number.
regs.Rax = uint64(sysno)
if len(args) >= 1 {
regs.Rdi = args[0].Uint64()
}
if len(args) >= 2 {
regs.Rsi = args[1].Uint64()
}
if len(args) >= 3 {
regs.Rdx = args[2].Uint64()
}
if len(args) >= 4 {
regs.R10 = args[3].Uint64()
}
if len(args) >= 5 {
regs.R8 = args[4].Uint64()
}
if len(args) >= 6 {
regs.R9 = args[5].Uint64()
}
return regs
}
// updateSyscallRegs updates registers after finishing sysemu.
func updateSyscallRegs(regs *arch.Registers) {
// Ptrace puts -ENOSYS in rax on syscall-enter-stop.
regs.Rax = regs.Orig_rax
}
// syscallReturnValue extracts a sensible return from registers.
func syscallReturnValue(regs *arch.Registers) (uintptr, error) {
rval := int64(regs.Rax)
if rval < 0 {
return 0, unix.Errno(-rval)
}
return uintptr(rval), nil
}
func dumpRegs(regs *arch.Registers) string {
var m strings.Builder
fmt.Fprintf(&m, "Registers:\n")
fmt.Fprintf(&m, "\tR15\t = %016x\n", regs.R15)
fmt.Fprintf(&m, "\tR14\t = %016x\n", regs.R14)
fmt.Fprintf(&m, "\tR13\t = %016x\n", regs.R13)
fmt.Fprintf(&m, "\tR12\t = %016x\n", regs.R12)
fmt.Fprintf(&m, "\tRbp\t = %016x\n", regs.Rbp)
fmt.Fprintf(&m, "\tRbx\t = %016x\n", regs.Rbx)
fmt.Fprintf(&m, "\tR11\t = %016x\n", regs.R11)
fmt.Fprintf(&m, "\tR10\t = %016x\n", regs.R10)
fmt.Fprintf(&m, "\tR9\t = %016x\n", regs.R9)
fmt.Fprintf(&m, "\tR8\t = %016x\n", regs.R8)
fmt.Fprintf(&m, "\tRax\t = %016x\n", regs.Rax)
fmt.Fprintf(&m, "\tRcx\t = %016x\n", regs.Rcx)
fmt.Fprintf(&m, "\tRdx\t = %016x\n", regs.Rdx)
fmt.Fprintf(&m, "\tRsi\t = %016x\n", regs.Rsi)
fmt.Fprintf(&m, "\tRdi\t = %016x\n", regs.Rdi)
fmt.Fprintf(&m, "\tOrig_rax = %016x\n", regs.Orig_rax)
fmt.Fprintf(&m, "\tRip\t = %016x\n", regs.Rip)
fmt.Fprintf(&m, "\tCs\t = %016x\n", regs.Cs)
fmt.Fprintf(&m, "\tEflags\t = %016x\n", regs.Eflags)
fmt.Fprintf(&m, "\tRsp\t = %016x\n", regs.Rsp)
fmt.Fprintf(&m, "\tSs\t = %016x\n", regs.Ss)
fmt.Fprintf(&m, "\tFs_base\t = %016x\n", regs.Fs_base)
fmt.Fprintf(&m, "\tGs_base\t = %016x\n", regs.Gs_base)
fmt.Fprintf(&m, "\tDs\t = %016x\n", regs.Ds)
fmt.Fprintf(&m, "\tEs\t = %016x\n", regs.Es)
fmt.Fprintf(&m, "\tFs\t = %016x\n", regs.Fs)
fmt.Fprintf(&m, "\tGs\t = %016x\n", regs.Gs)
return m.String()
}
// adjustInitregsRip adjust the current register RIP value to
// be just before the system call instruction execution
func (t *thread) adjustInitRegsRip() {
t.initRegs.Rip -= initRegsRipAdjustment
}
// Pass the expected PPID to the child via R15 when creating stub process.
func initChildProcessPPID(initregs *arch.Registers, ppid int32) {
// Rbx has to be set to 1 when creating stub process.
initregs.Rbx = _NEW_STUB
}
// patchSignalInfo patches the signal info to account for hitting the seccomp
// filters from vsyscall emulation, specified below. We allow for SIGSYS as a
// synchronous trap, but patch the structure to appear like a SIGSEGV with the
// Rip as the faulting address.
//
// Note that this should only be called after verifying that the signalInfo has
// been generated by the kernel.
// Returns true if the signal info was patched, false otherwise.
func maybePatchSignalInfo(regs *arch.Registers, signalInfo *linux.SignalInfo) bool {
if signalInfo.Addr() < linux.VSyscallStartAddr ||
signalInfo.Addr() >= linux.VSyscallEndAddr {
return false
}
// The syscall event was triggered from vsyscall emulation.
signalInfo.Signo = int32(linux.SIGSEGV)
// Unwind the kernel emulation, if any has occurred. A SIGSYS is delivered
// with the si_call_addr field pointing to the current RIP. This field
// aligns with the si_addr field for a SIGSEGV, so we don't need to touch
// anything there. We do need to unwind emulation however, so we set the
// instruction pointer to the faulting value, and "unpop" the stack.
regs.Rip = signalInfo.Addr()
regs.Rsp -= 8
return true
}
// enableCpuidFault enables cpuid-faulting.
//
// This may fail on older kernels or hardware, so we just disregard the result.
// Host CPUID will be enabled.
//
// This is safe to call in an afterFork context.
//
//go:nosplit
//go:norace
func enableCpuidFault() {
unix.RawSyscall6(unix.SYS_ARCH_PRCTL, linux.ARCH_SET_CPUID, 0, 0, 0, 0, 0)
}
// appendArchSeccompRules append architecture specific seccomp rules when creating BPF program.
// Ref attachedThread() for more detail.
func appendArchSeccompRules(rules []seccomp.RuleSet) []seccomp.RuleSet {
return append(rules, []seccomp.RuleSet{
// Rules for trapping vsyscall access.
{
Rules: seccomp.MakeSyscallRules(map[uintptr]seccomp.SyscallRule{
unix.SYS_GETTIMEOFDAY: seccomp.MatchAll{},
unix.SYS_TIME: seccomp.MatchAll{},
unix.SYS_GETCPU: seccomp.MatchAll{}, // SYS_GETCPU was not defined in package syscall on amd64.
}),
Action: linux.SECCOMP_RET_TRAP,
Vsyscall: true,
},
{
Rules: seccomp.MakeSyscallRules(map[uintptr]seccomp.SyscallRule{
unix.SYS_ARCH_PRCTL: seccomp.Or{
seccomp.PerArg{seccomp.EqualTo(linux.ARCH_SET_CPUID), seccomp.EqualTo(0)},
seccomp.PerArg{seccomp.EqualTo(linux.ARCH_SET_FS)},
seccomp.PerArg{seccomp.EqualTo(linux.ARCH_GET_FS)},
},
}),
Action: linux.SECCOMP_RET_ALLOW,
},
}...)
}
func restoreArchSpecificState(ctx *sysmsg.ThreadContext, ac *arch.Context64) {
}
func setArchSpecificRegs(sysThread *sysmsgThread, regs *arch.Registers) {
// Set the start function and initial stack.
regs.PtraceRegs.Rip = uint64(stubSysmsgStart + uintptr(sysmsg.Sighandler_blob_offset____export_start))
regs.PtraceRegs.Rsp = uint64(sysmsg.StackAddrToSyshandlerStack(sysThread.sysmsgPerThreadMemAddr()))
// Set gs_base; this is the only time we set it and we don't expect it to ever
// change for any thread.
regs.Gs_base = sysThread.msg.Self
}
func retrieveArchSpecificState(ctx *sysmsg.ThreadContext, ac *arch.Context64) {
}
func archSpecificSysmsgThreadInit(sysThread *sysmsgThread) {
}
|