Handle EOF to prevent file descriptor leak

[georgethebeatle]

Currently the EOF that the server gets when a client connection is
closed is being ignored, which means that the goroutine that is
processing the client connection will never exit. If it never exits
it will never close the underlying unix socket and this would lead
to a file descriptor leak.

Signed-off-by: Georgi Sabev georgethebeatle@gmail.com

Reprodction steps:

# create a long running task
ctr image pull docker.io/library/alpine:latest
ctr run --rm docker.io/library/alpine:latest test sleep 9999

# count number of open sockets
ls -la /proc/$(pidof containerd-shim)/fd | grep socket | wc -l
2

# restart containerd to make it reconnect to the shim 
kill $(pidof containerd)
containerd & 

# observe that socket number increased. It increases by 1 with every containerd restart
ls -la /proc/$(pidof containerd-shim)/fd | grep socket | wc -l
3

[crosbymichael]

LGTM

[estesp]

LGTM

[stevvooe]

The err == io.ErrUnexpectedEO won't work unless you ignore it above.

[georgethebeatle]

It would not have worked if there was a return statement in the if above, but there isn't. It would still be better to not log the ErrUnecpectedEOF though so we have created a new pr: #38. Pls, take a look.