Since 509123f935, we've been leaking sockets
in the filesystem on platforms where abstract sockets aren't supported.
That change relied on Go to cleanup our sockets for us, which Go will happily
do as long as we make sure to close the listener, which we weren't previously
doing unless to signal the plugin to terminate.
This change adds a deferred call to `PluginServer.Close()`, which makes sure we
close the plugin server at the end of the plugin execution, so that we never exit
without cleaning up.
Signed-off-by: Laura Brehm <laurabrehm@hey.com>
Go's `net` package [will unlink][1] for us, as long as we used Listen &
friends to create the Unix socket.
Go will even skip the unlink when the socket appears to be abstract
(starts with a NUL, represented by an @), though we must be cautious to
only create sockets with an abstract address on platforms that actually
support it -- this caused [several][2] [bugs][3] before.
[1]: https://pkg.go.dev/net#UnixListener.SetUnlinkOnClose
[2]: https://github.com/docker/cli/pull/4783
[3]: https://github.com/docker/cli/pull/4863
Signed-off-by: Bjorn Neergaard <bjorn.neergaard@docker.com>
This changes things to rely on a plugin server that manages all
connections made to the server.
An optional handler can be passed into the server when the caller wants
to do extra things with the connection.
It is the caller's responsibility to close the server.
When the server is closed, first all existing connections are closed
(and new connections are prevented).
Now the signal loop only needs to close the server and not deal with
`net.Conn`'s directly (or double-indirects as the case was before this
change).
The socket, when present in the filesystem, is no longer unlinked
eagerly, as reconnections require it to be present for the lifecycle of
the plugin server.
Co-authored-by: Bjorn Neergaard <bjorn.neergaard@docker.com>
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
Signed-off-by: Bjorn Neergaard <bjorn.neergaard@docker.com>
The "github.com/docker/distribution" module moved to the distribution
org ("github.com/docker/distribution/v3"), and the new module deprecated
and removed the uuid package in favor of Google's UUID package.
While we still depend on the old module through packages and as an indirect
dependency, we may want to try avoid using it.
This patch replaces the use for the socket package, and replaces it for a
local utility, taking the same approach as `stringid.GenerateRandomID()`,
which should be random enough for this purpose.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Seems that OpenBSD behaves like darwin and requires to unlink all
socket, after it was used.
Tested on OpenBSD 7.4
Signed-off-by: Kirill A. Korinsky <kirill@korins.ky>
Adds a new plugin to the e2e plugins that simulates an older
plugin binary and a test suite to ensure older plugin binaries
keep behaving the same with newer CLI versions.
Signed-off-by: Laura Brehm <laurabrehm@hey.com>
As macOS does not support the abstract socket namespace, use a temporary
socket in $TMPDIR to connect with the plugin. Ensure this socket is
cleaned up even in the case of crash/ungraceful termination by removing
it after the first connection is accepted.
Co-authored-by: Laura Brehm <laurabrehm@hey.com>
Signed-off-by: Bjorn Neergaard <bjorn.neergaard@docker.com>
Laura Brehm
Bjorn Neergaard
Brian Goff
Sebastiaan van Stijn
Kirill A. Korinsky
Tonis Tiigi