Files
got/download/engine.go
Hojun-Cho e3505855c1 httpdl: fix progress-reporter race on name/out; harden mirror & resume validation
Findings from a Rob-Pike-lens review (bugs/races/network), each verified
against the code before fixing:

- httpdl: name/out are now atomic.Pointer[string] -- the engine publishes a
  download to the reporter before Run resolves the name, so Stat raced the
  write (bt already did this)
- httpdl: a malformed --proxy fails loudly instead of silently bypassing it
- httpdl: a 206 must carry a matching Content-Range; a 200 in segmented mode is
  fatal so it fails over instead of burning the retry budget
- httpdl: single-stream mirror failover validates the range before appending;
  ErrTooSlow only when the error is a real ctx cancellation
- httpdl: idle guard tracks progress by timestamp (no Reset/Stop race, no
  sticky fired flag)
- httpdl/control: reject a resume file whose segments don't tile [0,total)
- bt: clamp the listen-port range; verify on-disk data before choosing pieces
  under --check-integrity
- cli: reject size overflow; show --seed-time=MIN; clamp --select-file range
- progress/engine/main: clamp ETA against int64 overflow; show queued
  downloads as waiting; join the reporter on exit instead of a 20ms sleep
2026-06-20 23:28:26 +09:00

112 lines
3.1 KiB
Go

package download
import (
"context"
"sync"
)
// Result pairs a finished download with its outcome and a final stat snapshot.
type Result struct {
Name string
Index int // position in the slice passed to Run, so callers can restore input order
Err error
Stat Stat
}
// Engine runs a set of downloads, at most maxConcurrent at a time, and exposes
// a live snapshot of the running ones for the progress renderer. A buffered
// channel of slots enforces the concurrency limit, and one goroutine per
// download replaces an explicit scheduler tick.
type Engine struct {
maxConcurrent int
mu sync.Mutex
active []Download
}
// NewEngine returns an engine that runs at most maxConcurrent downloads at once.
func NewEngine(maxConcurrent int) *Engine {
if maxConcurrent < 1 {
maxConcurrent = 1
}
return &Engine{maxConcurrent: maxConcurrent}
}
// Run starts every download, honouring the concurrency limit, and returns once
// all of them have finished or ctx is cancelled. There is one Result per
// download; results arrive in completion order, but each Result carries its
// input Index so callers can sort back to the original slice order.
func (e *Engine) Run(ctx context.Context, downloads []Download) []Result {
slots := make(chan struct{}, e.maxConcurrent)
results := make(chan Result, len(downloads))
var wg sync.WaitGroup
for i, d := range downloads {
wg.Add(1)
go func(i int, d Download) {
defer wg.Done()
// Track immediately so a download still queued for a concurrency slot
// is visible to the progress reporter as Waiting, rather than hidden
// until it starts running.
e.track(d)
defer e.untrack(d)
// Acquire a concurrency slot, or bail if we're shutting down
// before this download ever started.
select {
case slots <- struct{}{}:
case <-ctx.Done():
results <- Result{Name: d.Name(), Index: i, Err: ctx.Err(), Stat: d.Stat()}
return
}
defer func() { <-slots }()
err := d.Run(ctx)
results <- Result{Name: d.Name(), Index: i, Err: err, Stat: d.Stat()}
}(i, d)
}
wg.Wait()
close(results)
out := make([]Result, 0, len(downloads))
for r := range results {
out = append(out, r)
}
return out
}
// Snapshot returns a Stat for every tracked download — those running plus those
// still queued for a concurrency slot (reported as Waiting).
func (e *Engine) Snapshot() []Stat {
e.mu.Lock()
defer e.mu.Unlock()
out := make([]Stat, 0, len(e.active))
for _, d := range e.active {
out = append(out, d.Stat())
}
return out
}
func (e *Engine) track(d Download) {
e.mu.Lock()
e.active = append(e.active, d)
e.mu.Unlock()
}
// untrack removes d from the active set. Every Download implementation is a
// pointer type, so the == compares identities — never the fields of a value,
// which could panic on a non-comparable one. (Keying on Stat().ID is not an
// option: a torrent's ID changes from source to infohash once metadata loads.)
func (e *Engine) untrack(d Download) {
e.mu.Lock()
defer e.mu.Unlock()
for i, x := range e.active {
if x == d {
e.active = append(e.active[:i], e.active[i+1:]...)
return
}
}
}