Files
coredns/plugin/cache/setup.go
Nitin Nizhawan 87ccb6f90e plugin/cache: add prefer_positive stale policy (#8378)
* plugin/cache: add prefer_positive stale policy

Add an opt-in serve_stale_policy that prefers an eligible success-cache
answer over denial-cache entries while serve_stale is enabled. Preserve the
existing ncache-first behavior when the policy is absent.

Also classify SOA-backed CNAME NODATA responses in the cache so incomplete
answers cannot be selected as positive stale responses.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 25da81ab-92dd-4663-b480-efd6262090c6
Signed-off-by: Nitin Nizhawan <nnizhawan@microsoft.com>

* plugin/cache: retain last-known-good positive answers

Keep an answering success-cache item reachable when a later NOERROR or
referral response overwrites the visible cache key without answering the
question. This lets prefer_positive survive empty responses, referrals, and
additional-only data while leaving policy-off lookup behavior unchanged.

Return the exact accepted verify refresh item instead of re-reading an
ambiguous cache key, avoiding expired TTL wraparound for uncacheable replies.
Add regression coverage for non-answer refreshes, NODATA, SERVFAIL, NOTIMP,
stale-window expiry, and bounded verify reply shaping.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 25da81ab-92dd-4663-b480-efd6262090c6
Signed-off-by: Nitin Nizhawan <nnizhawan@microsoft.com>

* plugin/cache: validate preferred stale answers

Reject truncated, DNSSEC-expired, mismatched-class, unrelated ANY, and ambiguous CNAME refreshes before replacing a stale last-known-good answer. Precompute answer eligibility when cache items are created so prefer_positive hits avoid repeated CNAME walks.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 25da81ab-92dd-4663-b480-efd6262090c6
Signed-off-by: Nitin Nizhawan <nnizhawan@microsoft.com>

---------

Signed-off-by: Nitin Nizhawan <nnizhawan@microsoft.com>
Co-authored-by: Nitin Nizhawan <nnizhawan@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 25da81ab-92dd-4663-b480-efd6262090c6
2026-08-14 01:18:39 -07:00

356 lines
9.0 KiB
Go

package cache
import (
"errors"
"fmt"
"strconv"
"strings"
"time"
"github.com/coredns/caddy"
"github.com/coredns/coredns/core/dnsserver"
"github.com/coredns/coredns/plugin"
"github.com/coredns/coredns/plugin/pkg/cache"
clog "github.com/coredns/coredns/plugin/pkg/log"
)
var log = clog.NewWithPlugin("cache")
func init() { plugin.Register("cache", setup) }
func setup(c *caddy.Controller) error {
ca, err := cacheParse(c)
if err != nil {
return plugin.Error("cache", err)
}
c.OnStartup(func() error {
ca.viewMetricLabel = dnsserver.GetConfig(c).ViewName
return nil
})
dnsserver.GetConfig(c).AddPlugin(func(next plugin.Handler) plugin.Handler {
ca.Next = next
return ca
})
return nil
}
func cacheParse(c *caddy.Controller) (*Cache, error) {
ca := New()
j := 0
for c.Next() {
if j > 0 {
return nil, plugin.ErrOnce
}
j++
// cache [ttl] [zones..]
args := c.RemainingArgs()
if len(args) > 0 {
// first args may be just a number, then it is the ttl, if not it is a zone
ttl, err := strconv.Atoi(args[0])
if err == nil {
// Reserve 0 (and smaller for future things)
if ttl <= 0 {
return nil, fmt.Errorf("cache TTL can not be zero or negative: %d", ttl)
}
ca.pttl = time.Duration(ttl) * time.Second
ca.nttl = time.Duration(ttl) * time.Second
args = args[1:]
}
}
origins := plugin.OriginsFromArgsOrServerBlock(args, c.ServerBlockKeys)
serveStaleConfigured := false
serveStalePolicyConfigured := false
// Refinements? In an extra block.
for c.NextBlock() {
switch c.Val() {
// first number is cap, second is an new ttl
case Success:
args := c.RemainingArgs()
if len(args) == 0 {
return nil, c.ArgErr()
}
pcap, err := strconv.Atoi(args[0])
if err != nil {
return nil, err
}
ca.pcap = pcap
if len(args) > 1 {
pttl, err := strconv.Atoi(args[1])
if err != nil {
return nil, err
}
// Reserve 0 (and smaller for future things)
if pttl <= 0 {
return nil, fmt.Errorf("cache TTL can not be zero or negative: %d", pttl)
}
ca.pttl = time.Duration(pttl) * time.Second
if len(args) > 2 {
minpttl, err := strconv.Atoi(args[2])
if err != nil {
return nil, err
}
// Reserve < 0
if minpttl < 0 {
return nil, fmt.Errorf("cache min TTL can not be negative: %d", minpttl)
}
ca.minpttl = time.Duration(minpttl) * time.Second
}
}
case Denial:
args := c.RemainingArgs()
if len(args) == 0 {
return nil, c.ArgErr()
}
ncap, err := strconv.Atoi(args[0])
if err != nil {
return nil, err
}
ca.ncap = ncap
if len(args) > 1 {
nttl, err := strconv.Atoi(args[1])
if err != nil {
return nil, err
}
// Reserve 0 (and smaller for future things)
if nttl <= 0 {
return nil, fmt.Errorf("cache TTL can not be zero or negative: %d", nttl)
}
ca.nttl = time.Duration(nttl) * time.Second
if len(args) > 2 {
minnttl, err := strconv.Atoi(args[2])
if err != nil {
return nil, err
}
// Reserve < 0
if minnttl < 0 {
return nil, fmt.Errorf("cache min TTL can not be negative: %d", minnttl)
}
ca.minnttl = time.Duration(minnttl) * time.Second
}
}
case "prefetch":
args := c.RemainingArgs()
if len(args) == 0 || len(args) > 3 {
return nil, c.ArgErr()
}
amount, err := strconv.Atoi(args[0])
if err != nil {
return nil, err
}
if amount < 0 {
return nil, fmt.Errorf("prefetch amount should be positive: %d", amount)
}
ca.prefetch = amount
if len(args) > 1 {
dur, err := time.ParseDuration(args[1])
if err != nil {
return nil, err
}
ca.duration = dur
}
if len(args) > 2 {
pct := args[2]
if x := pct[len(pct)-1]; x != '%' {
return nil, fmt.Errorf("last character of percentage should be `%%`, but is: %q", x)
}
pct = pct[:len(pct)-1]
num, err := strconv.Atoi(pct)
if err != nil {
return nil, err
}
if num < 10 || num > 90 {
return nil, fmt.Errorf("percentage should fall in range [10, 90]: %d", num)
}
ca.percentage = num
}
case "serve_stale":
serveStaleConfigured = true
args := c.RemainingArgs()
if len(args) > 5 {
return nil, c.ArgErr()
}
ca.staleUpTo = 1 * time.Hour
ca.staleTTL = 0
ca.staleRecheck = 0
if len(args) > 0 {
d, err := time.ParseDuration(args[0])
if err != nil {
return nil, err
}
if d < 0 {
return nil, errors.New("invalid negative duration for serve_stale")
}
ca.staleUpTo = d
}
ca.verifyStale = false
ca.verifyStaleTimeout = 0
if len(args) > 1 {
mode := strings.ToLower(args[1])
if mode != "immediate" && mode != "verify" {
return nil, fmt.Errorf("invalid value for serve_stale refresh mode: %s", mode)
}
ca.verifyStale = mode == "verify"
}
if len(args) > 2 {
if ca.verifyStale {
t, err := time.ParseDuration(args[2])
if err != nil {
return nil, fmt.Errorf("invalid serve_stale verify timeout: %w", err)
}
if t < 0 {
return nil, errors.New("invalid negative timeout for serve_stale verify")
}
ca.verifyStaleTimeout = t
if len(args) > 3 {
ca.staleTTL, err = parseServeStaleTTL(args[3])
if err != nil {
return nil, err
}
}
if len(args) > 4 {
ca.staleRecheck, err = parseServeStaleRecheck(args[4])
if err != nil {
return nil, err
}
}
} else {
if len(args) > 4 {
return nil, c.ArgErr()
}
var err error
ca.staleTTL, err = parseServeStaleTTL(args[2])
if err != nil {
return nil, err
}
if len(args) > 3 {
ca.staleRecheck, err = parseServeStaleRecheck(args[3])
if err != nil {
return nil, err
}
}
}
}
case "serve_stale_policy":
if serveStalePolicyConfigured {
return nil, errors.New("serve_stale_policy can only be specified once")
}
serveStalePolicyConfigured = true
args := c.RemainingArgs()
if len(args) != 1 {
return nil, c.ArgErr()
}
switch strings.ToLower(args[0]) {
case "prefer_positive":
ca.preferPositive = true
default:
return nil, fmt.Errorf("invalid serve_stale_policy: %s", args[0])
}
case "servfail":
args := c.RemainingArgs()
if len(args) != 1 {
return nil, c.ArgErr()
}
d, err := time.ParseDuration(args[0])
if err != nil {
return nil, err
}
if d < 0 {
return nil, errors.New("invalid negative ttl for servfail")
}
if d > 5*time.Minute {
// RFC 2308 prohibits caching SERVFAIL longer than 5 minutes
return nil, errors.New("caching SERVFAIL responses over 5 minutes is not permitted")
}
ca.failttl = d
case "disable":
// disable [success|denial] [zones]...
args := c.RemainingArgs()
if len(args) < 1 {
return nil, c.ArgErr()
}
var zones []string
if len(args) > 1 {
for _, z := range args[1:] { // args[1:] define the list of zones to disable
nz := plugin.Name(z).Normalize()
if nz == "" {
return nil, fmt.Errorf("invalid disabled zone: %s", z)
}
zones = append(zones, nz)
}
} else {
// if no zones specified, default to root
zones = []string{"."}
}
switch args[0] { // args[0] defines which cache to disable
case Denial:
ca.nexcept = zones
case Success:
ca.pexcept = zones
default:
return nil, fmt.Errorf("cache type for disable must be %q or %q", Success, Denial)
}
case "keepttl":
args := c.RemainingArgs()
if len(args) != 0 {
return nil, c.ArgErr()
}
ca.keepttl = true
default:
return nil, c.ArgErr()
}
}
if serveStalePolicyConfigured && !serveStaleConfigured {
return nil, errors.New("serve_stale_policy requires serve_stale")
}
ca.Zones = origins
ca.zonesMetricLabel = strings.Join(origins, ",")
ca.pcache = cache.New[*item](ca.pcap)
ca.ncache = cache.New[*item](ca.ncap)
}
return ca, nil
}
func parseServeStaleTTL(value string) (time.Duration, error) {
ttl, err := time.ParseDuration(value)
if err != nil {
return 0, fmt.Errorf("invalid serve_stale response TTL: %w", err)
}
if ttl < 0 {
return 0, errors.New("invalid negative response TTL for serve_stale")
}
if ttl%time.Second != 0 {
return 0, errors.New("serve_stale response TTL must be a whole number of seconds")
}
if ttl/time.Second > time.Duration(^uint32(0)) {
return 0, errors.New("serve_stale response TTL exceeds the DNS TTL range")
}
return ttl, nil
}
func parseServeStaleRecheck(value string) (time.Duration, error) {
recheck, err := time.ParseDuration(value)
if err != nil {
return 0, fmt.Errorf("invalid serve_stale failure recheck: %w", err)
}
if recheck < 0 {
return 0, errors.New("invalid negative failure recheck for serve_stale")
}
if recheck > 5*time.Minute {
return 0, errors.New("serve_stale failure recheck cannot exceed 5 minutes")
}
return recheck, nil
}