Fix cache segfaults due to unaligned atomics on 32b architectures

[nickbp]

Environment where the issue was encountered:

• Raspberry Pi 4: Linux pi-03 4.19.66-v7l+ #1253 SMP Thu Aug 15 12:02:08 BST 2019 armv7l GNU/Linux
• Go version 1.13.1

Per the atomic library docs:

On ARM, x86-32, and 32-bit MIPS, it is the caller's responsibility to arrange for 64-bit alignment of 64-bit words accessed atomically. The first word in a variable or in an allocated struct, array, or slice can be relied upon to be 64-bit aligned.

---

I was able to reproduce the segfault with the following test code (main.go):

package main

import (
	"fmt"
	"time"
	"istio.io/pkg/cache"
)

func set(c cache.Cache, k string, v string) {
	fmt.Printf("Setting %v => %v\n", k, v)
	c.Set(k, v)
}

func get(c cache.Cache, k string) {
	v, ok := c.Get(k)
    if ok {
        fmt.Printf("%v=%v\n", k, v)
    } else {
        fmt.Printf("%v not found\n", k)
    }
}

func test(c cache.Cache) {
	set(c, "one", "foo")

	get(c, "one")
	get(c, "two")

	// Trigger size-based eviction
	set(c, "two", "bar")

	get(c, "one")
	get(c, "two")

	// Trigger TTL-based eviction
	fmt.Printf("Waiting for expiry...\n")
	time.Sleep(1500 * time.Millisecond)

	get(c, "one")
	get(c, "two")
}

func main() {
	fmt.Printf("*** LRU ***\n")
	test(cache.NewLRU(1 * time.Second, 1 * time.Second, 1))

	fmt.Printf("\n*** TTL ***\n")
	test(cache.NewTTL(1 * time.Second, 1 * time.Second))

	fmt.Printf("\n*** PASS ***\n")
}

The above can be built as a static ARM7 binary via the following go build command:

CGO_ENABLED=0 GOARCH=arm GOOS=linux go build -ldflags="-s"

---

Error 1: atomic increment of unaligned within lock in lruCache.go, fix was to just remove the atomic increment since it's within a lock and none of the other stats do this:

# ./main 
*** LRU ***
Setting one => foo
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x11da0]

goroutine 1 [running]:
runtime/internal/atomic.goXadd64(0x188402c, 0x1, 0x0, 0x186a4a8, 0x1866100)
	/usr/lib/go/src/runtime/internal/atomic/atomic_arm.go:103 +0x1c
istio.io/pkg/cache.(*lruCache).SetWithExpiration(0x1884000, 0xc1280, 0x1866128, 0xc1280, 0x1866130, 0x3b9aca00, 0x0)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/cache/lruCache.go:247 +0x1bc
istio.io/pkg/cache.(*lruCache).Set(0x1884000, 0xc1280, 0x1866128, 0xc1280, 0x1866130)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/cache/lruCache.go:224 +0x4c
main.set(0xa6d22008, 0x1866110, 0xd6946, 0x3, 0xd6931, 0x3)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:11 +0x108
main.test(0xa6d22008, 0x1866110)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:24 +0x40
main.main()
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:45 +0xa0

---

Error 2: unaligned baseTimeNanos in ttlCache.go, fix was to move the value to the start of the type to ensure alignment as recommended by atomic docs:

# ./main 
*** LRU ***
Setting one => foo
one=foo
two not found
Setting two => bar
one not found
two=bar
Waiting for expiry...
one not found
two not found

*** TTL ***
Setting one => foo
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x11f68]

goroutine 1 [running]:
runtime/internal/atomic.goLoad64(0x89e04c, 0x16c618, 0xd1850)
	/usr/lib/go/src/runtime/internal/atomic/atomic_arm.go:131 +0x1c
istio.io/pkg/cache.(*ttlCache).SetWithExpiration(0x89e000, 0xc1280, 0x88e040, 0xc1280, 0x88e048, 0x3b9aca00, 0x0)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/cache/ttlCache.go:173 +0x24
istio.io/pkg/cache.(*ttlCache).Set(0x89e000, 0xc1280, 0x88e040, 0xc1280, 0x88e048)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/cache/ttlCache.go:167 +0x4c
main.set(0xa6c37008, 0x88e028, 0xd6946, 0x3, 0xd6931, 0x3)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:11 +0x108
main.test(0xa6c37008, 0x88e028)
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:24 +0x40
main.main()
	/home/nick/code/pkg/.gopath/src/istio.io/pkg/main/main.go:48 +0x12c

---

After both fixes:

# ./main 
*** LRU ***
Setting one => foo
one=foo
two not found
Setting two => bar
one not found
two=bar
Waiting for expiry...
one not found
two not found

*** TTL ***
Setting one => foo
one=foo
two not found
Setting two => bar
one=foo
two=bar
Waiting for expiry...
one not found
two not found

*** PASS ***

[googlebot]

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here with @googlebot I signed it! and we'll verify it.

---

What to do if you already signed the CLA

Individual signers

• It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

• Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
• The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
• The email used to register you as an authorized contributor must also be attached to your GitHub account.

ℹ️ Googlers: Go here for more info.

[istio-testing]

Hi @nickbp. Thanks for your PR.

I'm waiting for a istio member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[nickbp]

@googlebot I signed it!

[googlebot]

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

[nickbp]

This was the only place that was using an atomic increment against c.stats. It wasn't clear why the atomic increment was necessary.

[geeknoid]

Hmmm, seems like a leftover, probably from cloning the ttlCache.go code initially.

[nickbp]

Forgot to point out earlier that this is also within c.Lock()

[geeknoid]

Thanks for the improvements.

[geeknoid]

/ok-to-test

[nickbp]

Anything else needed here? This has been idle for a while.