Windows: Support setting permissions on Unix domain sockets

[sunjayBhatia]

In the Envoy API for Pipe addresses, we explicitly support setting the file mode on Unix domain socket files, see:

envoy/api/envoy/config/core/v3/address.proto

Line 30 in ece1b6a

uint32 mode = 2 [(validate.rules).uint32 = {lte: 511}];

This file mode pattern does not translate well to Windows, to get similar granular permission control we would need to leverage Windows ACLs.

This issue is for discussion around this topic and to consider different options, some that come to mind immediately:

• As with abstract Unix domain sockets, simply do not support setting a file mode
• Attempt to translate the file mode to something that is passable on Windows
  • See parts of Windows: Fix address_imp_test #11334 as an attempt
  • See https://devblogs.microsoft.com/commandline/windowswsl-interop-with-af_unix/ for an interesting case study between Unix domain sockets on WSL + Windows
• Revamp the mode field of the API to transition to a more portable representation

[sunjayBhatia]

cc @wrowe @davinci26 @nigriMSFT

[sunjayBhatia]

The test that demonstrates this most succinctly is //test/common/network:address_impl_test -> PipeInstanceTest.BasicPermission

We have an underlying issue that we are not able to even stat the socket file, once that is solved, the file mode itself is the bigger question

[sunjayBhatia]

We are considering disabling that individual test for now on Windows before this issue is resolved

[sunjayBhatia]

related general issue: #9571

[davinci26]

Is setting granular file permissions are requirement for a high priority scenario or is it a good to have? Is there a way to see what the customer demand for this is?

What I am thinking is that we could have this API in Windows with the current limitation. Basically only setting the socket read-only or write and write. And then offer an additional module just for win32 ACL. This is similar to the approach Python is taking.

Is there a way we could add win32 ACL support as plugin?

[sunjayBhatia]

further discussion: https://envoyproxy.slack.com/archives/CNAK09BSB/p1590785118071500

We are seeing differences in behavior about stat() on a unix domain socket

[sunjayBhatia]

as per @wrowe catching this: golang/go#33357 (comment)

[stale]

This issue has been automatically marked as stale because it has not had activity in the last 30 days. It will be closed in the next 7 days unless it is tagged "help wanted" or other activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically closed because it has not had activity in the last 37 days. If this issue is still valid, please ping a maintainer and ask them to label it as "help wanted". Thank you for your contributions.

[wrowe]

I'd suggest this issue remains unresolved and relevant