hyperstart: fix several mem leak

[keloyang]

Signed-off-by: y00316549 yangshukui@huawei.com

[gnawux]

looks the portmapping case failed, right?

[Jimmy-Xu]

Can one of the admins verify this patch?

[gnawux]

ok to test, hykins

[Jimmy-Xu]

Can one of the admins verify this patch?

[sohu0106]

memory leak call trace:
hyper_ctlmsg_handle--->hyper_new_container--->hyper_setup_container--->hyper_setup_container_rootfs--->container_setup_modules
and
hyper_ctlmsg_handle--->hyper_new_container--->hyper_setup_container--->container_prepare_rootfs_dev--->hyper_rescan_scsi
each new container leak memory in pod, vm's memory will exhaust
ping @laijs

[bergwolf]

retest this please, @hykins

[bergwolf]

LGTM, thanks!

[attritionorg]

Can someone (@keloyang, @bergwolf) confirm this is exploitable to conduct a DoS? I ask because CVE-2018-10205 was assigned to this issue, but there is no indication it is a vulnerability. Thanks!

[bergwolf]

@attritionorg No, I don't think one can use the issue to create a DoS. For one thing, it only affects one VM. For another thing, it only happens once per VM to set up the kernel modules, and once per new container to scan the scsi bus.

[attritionorg]

@bergwolf thanks for clarifying!