stream: fix panic caused by failing to get a transport for a retry attempt

[canguler]

To not set new attempt when newAttemptLocked fails to get a transport.

related to #2952 #2954

[gyuho]

@canguler Sorry, it still panics

--- FAIL: TestCtlV3GetRevokedCRL (5.64s)
    ctl_v3_kv_test.go:68: expected reset connection on put, got read /dev/ptmx: input/output error (expected "OK", got ["

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x18 pc=0x9e944e]

goroutine 1 [running]:
go.etcd.io/etcd/vendor/google.golang.org/grpc.(*clientStream).finish(0xc0000b37a0, 0xe657e0, 0xc000178000)
        /go/src/go.etcd.io/etcd/vendor/google.golang.org/grpc/stream.go:824 +0x12e
go.etcd.io/etcd/vendor/google.golang.org/grpc.newClientStream(0xe7c5c0, 0xc0002e2870, 0x14e0520, 0xc000169c00, 0xd4ade3, 0x14, 0xc000260a00, 0x3, 0x4, 0x0, ...)
        /go/src/go.etcd.io/etcd/vendor/google.golang.org/grpc/stream.go:285 +0xbc8
go.etcd.io/etcd/vendor/google.golang.org/grpc.invoke(0xe7c580, 0xc000248c60, 0xd4ade3, 0x14, 0xcf0040, 0xc0002609c0, 0xcdaaa0, 0xc0002f0320, 0xc000169c00, 0xc000260a00, ...)
        /go/src/go.etcd.io/etcd/vendor/google.golang.org/grpc/call.go:66 +0x9c
go.etcd.io/etcd/clientv3.(*Client).unaryClientInterceptor.func1(0xe7c580, 0xc000248c60, 0xd4ade3, 0x14, 0xcf0040, 0xc0002609c0, 0xcdaaa0, 0xc0002f0320, 0xc000169c00, 0xd7f0c8, ...)
        /go/src/go.etcd.io/etcd/clientv3/retry_interceptor.go:57 +0x458
go.etcd.io/etcd/vendor/google.golang.org/grpc.(*ClientConn).Invoke(0xc000169c00, 0xe7c580, 0xc000248c60, 0xd4ade3, 0x14, 0xcf0040, 0xc0002609c0, 0xcdaaa0, 0xc0002f0320, 0x14e09e0, ...)
        /go/src/go.etcd.io/etcd/vendor/google.golang.org/grpc/call.go:35 +0x10a
go.etcd.io/etcd/vendor/google.golang.org/grpc.Invoke(...)
        /go/src/go.etcd.io/etcd/vendor/google.golang.org/grpc/call.go:60
go.etcd.io/etcd/etcdserver/etcdserverpb.(*kVClient).Put(0xc00000e620, 0xe7c580, 0xc000248c60, 0xc0002609c0, 0x14e09e0, 0x3, 0x3, 0xcdd1a0, 0x1, 0xc0002609c0)
        /go/src/go.etcd.io/etcd/etcdserver/etcdserverpb/rpc.pb.go:3477 +0xd3
go.etcd.io/etcd/clientv3.(*retryKVClient).Put(0xc0002f02b0, 0xe7c580, 0xc000248c60, 0xc0002609c0, 0x14e09e0, 0x3, 0x3, 0x8, 0xc0002ee208, 0xc00023b8c8)
        /go/src/go.etcd.io/etcd/clientv3/retry.go:109 +0x7c
go.etcd.io/etcd/clientv3.(*kv).Do(0xc0002e2660, 0xe7c580, 0xc000248c60, 0x2, 0xc0002ee200, 0x1, 0x8, 0x0, 0x0, 0x0, ...)
        /go/src/go.etcd.io/etcd/clientv3/kv.go:156 +0x3b6
go.etcd.io/etcd/clientv3.(*kv).Put(0xc0002e2660, 0xe7c580, 0xc000248c60, 0x7ffe31026e54, 0x1, 0x7ffe31026e56, 0x1, 0x14fdce0, 0x0, 0x0, ...)
        /go/src/go.etcd.io/etcd/clientv3/kv.go:114 +0x116
go.etcd.io/etcd/etcdctl/ctlv3/command.putCommandFunc(0xc00024e280, 0xc00024b030, 0x2, 0x7)
        /go/src/go.etcd.io/etcd/etcdctl/ctlv3/command/put_command.go:71 +0x178
go.etcd.io/etcd/vendor/github.com/spf13/cobra.(*Command).execute(0xc00024e280, 0xc000256dc0, 0x7, 0xa, 0xc00024e280, 0xc000256dc0)
        /go/src/go.etcd.io/etcd/vendor/github.com/spf13/cobra/command.go:766 +0x2ae
go.etcd.io/etcd/vendor/github.com/spf13/cobra.(*Command).ExecuteC(0x14d5200, 0xb, 0xc000054d88, 0xb79701)
        /go/src/go.etcd.io/etcd/vendor/github.com/spf13/cobra/command.go:852 +0x2ec
go.etcd.io/etcd/vendor/github.com/spf13/cobra.(*Command).Execute(...)
        /go/src/go.etcd.io/etcd/vendor/github.com/spf13/cobra/command.go:800
go.etcd.io/etcd/etcdctl/ctlv3.Start()
        /go/src/go.etcd.io/etcd/etcdctl/ctlv3/ctl_nocov.go:25 +0x81
main.main()
        /go/src/go.etcd.io/etcd/etcdctl/main.go:35 +0x77

[canguler]

Hi @gyuho,

I think your commit is not exactly the same. In finish method, I moved an existing code block whereas you have added a new block. Can you please fix that and retry?

[gyuho]

@canguler Ah, you are right. Let me try again.

[canguler]

@gyuho,

Any update?

[gyuho]

@canguler I ran it against this branch, and it seems to fix the panic issue. If anything else comes up, I will open a new issue. Thanks!

[menghanl]

@gyuho This should also fix #2954, can you give it a try? Thanks!

[gyuho]

@menghanl Yes. If anything else comes up, will create a new issue. Thanks for quick fix!

[hexfusion]

@menghanl any chance we can get this as part of 1.23?

[menghanl]

@hexfusion Yes, this will go into 1.23.

Also, the change that caused this is not in any release yet, so the past releases should be good.