| 1 | package hostbox |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "github.com/carlosframework/platform/internal/park" |
| 7 | "github.com/carlosframework/platform/internal/registry" |
| 8 | "sync" |
| 9 | "time" |
| 10 | |
| 11 | "github.com/carlosframework/platform/internal/hostclaim" |
| 12 | "github.com/carlosframework/platform/internal/hostguard" |
| 13 | "github.com/carlosframework/platform/internal/hostwrite" |
| 14 | ) |
| 15 | |
| 16 | var ErrChanged = errors.New("hostbox: selected authority changed") |
| 17 | var ErrUnsupported = errors.New("hostbox: unresolved lifecycle contract") |
| 18 | |
| 19 | type Selection struct { |
| 20 | HomeRegion, FleetHomeRegion string |
| 21 | Host, Action, Box, Region string |
| 22 | Captured hostguard.Captured |
| 23 | Kind, Op, Phase, Generation string |
| 24 | Session uint64 |
| 25 | // installation marks the work Reconciler.installWithWork selected for |
| 26 | // one installation (Installation). Only this package sets it. |
| 27 | installation bool |
| 28 | } |
| 29 | |
| 30 | // Installation reports whether the selection is the work an installation |
| 31 | // selected for itself (Reconciler.installWithWork, the path to the remote |
| 32 | // route registration and the local row write), not ordinary work on an |
| 33 | // installed host. A policy may admit what only an installation can need, |
| 34 | // such as a registry row the installation has not written yet (harness G10, |
| 35 | // 2026-10-21); no caller outside this package can set it. |
| 36 | func (s Selection) Installation() bool { return s.installation } |
| 37 | |
| 38 | // installationWork marks w, a selection SelectAt made, as installation |
| 39 | // work. Any other work is returned unmarked, so it is refused whatever an |
| 40 | // installation-only admission would allow. |
| 41 | func installationWork(w hostwrite.Work) hostwrite.Work { |
| 42 | selected, ok := w.(selectedWork) |
| 43 | if !ok { |
| 44 | return w |
| 45 | } |
| 46 | selected.selection.installation = true |
| 47 | return selected |
| 48 | } |
| 49 | |
| 50 | type WorkSource struct { |
| 51 | Timer func(context.Context, hostwrite.Work, registry.Timer, string, bool) (*registry.Timer, error) |
| 52 | FleetHomeRegion string |
| 53 | HomeRegion string |
| 54 | Prepare func(context.Context, hostclaim.Observed) error |
| 55 | Guard hostguard.Guard |
| 56 | Box, Region string |
| 57 | Now func() time.Time |
| 58 | // Policy checks the action's role/phase and the selected config/release. |
| 59 | Policy func(context.Context, *hostclaim.Claim, Selection, hostwrite.Identity) error |
| 60 | // Local checks quarantine, execution deadlines and failed cleanup under flock. |
| 61 | Local func(context.Context, *hostclaim.Claim, Selection) error |
| 62 | // Continue reports whether a selection last checked at one op phase may |
| 63 | // keep working at another (hostclaim.SessionObserves). Nil keeps main's rule: |
| 64 | // any phase change ends the selection. Only the moves C1 integration sets |
| 65 | // it, and only with CARLOS_HOST_CLAIM and CARLOS_MOVES_PHASE1 both on. |
| 66 | Continue func(kind, from, to string) bool |
| 67 | } |
| 68 | |
| 69 | type selectedWork struct { |
| 70 | source WorkSource |
| 71 | selection Selection |
| 72 | // checked is the op phase of the last claim a With check accepted. The |
| 73 | // selection's own Phase stays the selected one for Policy and Local. |
| 74 | checked *checkedPhase |
| 75 | } |
| 76 | |
| 77 | type checkedPhase struct { |
| 78 | mu sync.Mutex |
| 79 | phase string |
| 80 | } |
| 81 | |
| 82 | func (s WorkSource) Select(ctx context.Context, host, action string) (hostwrite.Work, error) { |
| 83 | if ctx == nil || s.Box == "" || s.Region == "" || action == "" || s.Guard.Claims == nil || s.Policy == nil || s.Local == nil || s.Now == nil { |
| 84 | return nil, ErrUnsupported |
| 85 | } |
| 86 | captured, err := s.Guard.Capture(host) |
| 87 | if err != nil { |
| 88 | return nil, err |
| 89 | } |
| 90 | obs, err := s.Guard.Claims.Read(ctx, host) |
| 91 | if err != nil { |
| 92 | return nil, err |
| 93 | } |
| 94 | if obs.Absent || obs.Host != host || obs.Claim == nil || obs.Claim.Holder == nil || obs.Claim.Host != host { |
| 95 | return nil, ErrChanged |
| 96 | } |
| 97 | if s.Prepare != nil { |
| 98 | if err := s.Prepare(ctx, obs); err != nil { |
| 99 | return nil, err |
| 100 | } |
| 101 | } |
| 102 | c := obs.Claim |
| 103 | if c.Holder.Incarnation != captured.Incarnation || c.Holder.Stamp != captured.Stamp || captured.Stamp == 0 { |
| 104 | return nil, ErrChanged |
| 105 | } |
| 106 | w := selectedWork{source: s, selection: Selection{HomeRegion: s.HomeRegion, FleetHomeRegion: s.FleetHomeRegion, Host: host, Action: action, Box: s.Box, Region: s.Region, Captured: captured, Kind: c.Op.Kind, Op: c.Op.ID, Phase: c.Op.Phase, Generation: c.Holder.Generation, Session: c.Holder.Session}, checked: &checkedPhase{phase: c.Op.Phase}} |
| 107 | return w, nil |
| 108 | } |
| 109 | |
| 110 | func (w selectedWork) With(ctx context.Context, identity hostwrite.Identity, fn func() error) error { |
| 111 | s, selected := w.source, w.selection |
| 112 | return s.Guard.With(ctx, selected.Host, selected.Captured, func(c *hostclaim.Claim) error { |
| 113 | if c.Op.Kind != selected.Kind || c.Op.ID != selected.Op || !s.continues(selected.Kind, w.checked.get(), c.Op.Phase) || c.Holder.Stamp != selected.Captured.Stamp || c.Holder.Session != selected.Session || c.Holder.Generation != selected.Generation { |
| 114 | return ErrChanged |
| 115 | } |
| 116 | switch selected.Action { |
| 117 | case "install", "restore", "spawn", "seed", "writer", "probe": |
| 118 | if until := park.QuietUntil(c.QuietAfter, s.HomeRegion, s.Region); !until.IsZero() && !s.Now().After(until) { |
| 119 | return ErrChanged |
| 120 | } |
| 121 | } |
| 122 | if err := s.Local(ctx, c, selected); err != nil { |
| 123 | return err |
| 124 | } |
| 125 | if err := s.Policy(ctx, c, selected, identity); err != nil { |
| 126 | return err |
| 127 | } |
| 128 | return w.checked.advance(func(from string) bool { return s.continues(selected.Kind, from, c.Op.Phase) }, c.Op.Phase) |
| 129 | }, fn) |
| 130 | } |
| 131 | |
| 132 | func (p *checkedPhase) get() string { |
| 133 | p.mu.Lock() |
| 134 | defer p.mu.Unlock() |
| 135 | return p.phase |
| 136 | } |
| 137 | |
| 138 | // advance records phase as checked if it still continues from the last |
| 139 | // checked phase. A concurrent check that already moved on refuses this one. |
| 140 | func (p *checkedPhase) advance(continues func(from string) bool, phase string) error { |
| 141 | p.mu.Lock() |
| 142 | defer p.mu.Unlock() |
| 143 | if !continues(p.phase) { |
| 144 | return ErrChanged |
| 145 | } |
| 146 | p.phase = phase |
| 147 | return nil |
| 148 | } |
| 149 | |
| 150 | // continues binds the selection to op id, kind, session, stamp and generation. |
| 151 | // Its phase may only take a forward step the configured table allows from the |
| 152 | // last phase a check accepted, so a skipped or backward step ends it. |
| 153 | func (s WorkSource) continues(kind, from, to string) bool { |
| 154 | if s.Continue == nil { |
| 155 | return from == to |
| 156 | } |
| 157 | return s.Continue(kind, from, to) |
| 158 | } |
| 159 | |
| 160 | // Serving permits ordinary host writes only. Operations need their own policy. |
| 161 | func Serving(c *hostclaim.Claim, selected Selection) error { |
| 162 | if c.Holder == nil || c.Op.Kind != hostclaim.KindServing || c.Op.ID != "" || c.Op.Phase != "" || c.Op.Exec != selected.Box { |
| 163 | return ErrChanged |
| 164 | } |
| 165 | for _, box := range c.Lost { |
| 166 | if box == selected.Box { |
| 167 | return ErrChanged |
| 168 | } |
| 169 | } |
| 170 | return nil |
| 171 | } |
| 172 | |
| 173 | // SelectAt captures an already chosen installation without relabelling it. |
| 174 | func (s WorkSource) SelectAt(ctx context.Context, host, action string, captured hostguard.Captured) (hostwrite.Work, error) { |
| 175 | work, err := s.Select(ctx, host, action) |
| 176 | if err != nil { |
| 177 | return nil, err |
| 178 | } |
| 179 | selected := work.(selectedWork) |
| 180 | if selected.selection.Captured != captured { |
| 181 | return nil, ErrChanged |
| 182 | } |
| 183 | return work, nil |
| 184 | } |
| 185 | |
amadan