SHA256
1
0
forked from pool/docker
docker/bsc1037436-0001-client-check-tty-before-creating-exec-job.patch
Miquel Sabate Sola e7663235ed Accepting request 494283 from home:cyphar:containers
- Add a partial fix for boo#1038493.
- Fixed bsc#1037436 where execids were being leaked due to bad error handling.
  This is a backport of https://github.com/docker/cli/pull/52.
   + bsc1037436-0001-client-check-tty-before-creating-exec-job.patch

OBS-URL: https://build.opensuse.org/request/show/494283
OBS-URL: https://build.opensuse.org/package/show/Virtualization:containers/docker?expand=0&rev=184
2017-05-10 14:04:27 +00:00

70 lines
2.5 KiB
Diff

From c117441b1a74affb013a42ee8225d69ecfaf4d72 Mon Sep 17 00:00:00 2001
From: Aleksa Sarai <asarai@suse.de>
Date: Tue, 9 May 2017 23:31:46 +1000
Subject: [PATCH] client: check tty before creating exec job
This is necessary in order to avoid execId leaks in the case where a
`docker exec -it` is run without a terminal available for the client.
You can reproduce this issue by running the following command many
times.
% nohup docker exec -it some_container true
The container `some_container` will have execIDs that will never
normally be cleaned up (because the client died before they were
started).
In addition, this patch adds a docker-inspect step to ensure that we
give "container does not exist" errors consistently.
[SUSE: Fixes bsc#1037436.]
Signed-off-by: Valentin Rothberg <vrothberg@suse.com>
Signed-off-by: Aleksa Sarai <asarai@suse.de>
---
cli/command/container/exec.go | 21 +++++++++++++++------
1 file changed, 15 insertions(+), 6 deletions(-)
diff --git a/cli/command/container/exec.go b/cli/command/container/exec.go
index 676708c77b91..d85113259242 100644
--- a/cli/command/container/exec.go
+++ b/cli/command/container/exec.go
@@ -79,6 +79,19 @@ func runExec(dockerCli *command.DockerCli, opts *execOptions, container string,
ctx := context.Background()
client := dockerCli.Client()
+ // We need to check the tty _before_ we do the ContainerExecCreate, because
+ // otherwise if we error out we will leak execIDs on the server (and
+ // there's no easy way to clean those up). But also in order to make "not
+ // exist" errors take precedence we do a dummy inspect first.
+ if _, err := client.ContainerInspect(ctx, container); err != nil {
+ return err
+ }
+ if !execConfig.Detach {
+ if err := dockerCli.In().CheckTty(execConfig.AttachStdin, execConfig.Tty); err != nil {
+ return err
+ }
+ }
+
response, err := client.ContainerExecCreate(ctx, container, *execConfig)
if err != nil {
return err
@@ -90,12 +103,8 @@ func runExec(dockerCli *command.DockerCli, opts *execOptions, container string,
return nil
}
- //Temp struct for execStart so that we don't need to transfer all the execConfig
- if !execConfig.Detach {
- if err := dockerCli.In().CheckTty(execConfig.AttachStdin, execConfig.Tty); err != nil {
- return err
- }
- } else {
+ // Temp struct for execStart so that we don't need to transfer all the execConfig.
+ if execConfig.Detach {
execStartCheck := types.ExecStartCheck{
Detach: execConfig.Detach,
Tty: execConfig.Tty,
--
2.12.2