device identity #614
device identity #614
Conversation
|
Can you drop commits adding gui, instead of adding+removing gui? the PR is quite big already... |
we might need to ask to attach rest
piotrbartman
force-pushed
the
qdev
branch
2 times, most recently
from
349e2de
to
3d3a72a
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #614 +/- ##
==========================================
+ Coverage 69.33% 69.45% +0.12%
==========================================
Files 58 58
Lines 11946 12275 +329
==========================================
+ Hits 8283 8526 +243
- Misses 3663 3749 +86
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
| def confirm_device_attachment(device, frontends) -> str: | ||
| try: | ||
| # pylint: disable=consider-using-with | ||
| proc = subprocess.Popen( | ||
| ["attach-confirm", device.backend_domain.name, | ||
| device.port_id, device.description, | ||
| *[f.name for f in frontends.keys()]], | ||
| stdout=subprocess.PIPE, stderr=subprocess.PIPE | ||
| ) | ||
| (target_name, _) = proc.communicate() | ||
| return target_name.decode() | ||
| except Exception as exc: | ||
| print("attach-confirm", exc, file=sys.stderr) | ||
| return "" |
There was a problem hiding this comment.
There are a few problems with this function:
- It must be async, since it may block for some time, and it's not acceptable to block the whole qubesd for this time. In fact, attach-confirm probably won't work this way at all if it tries to talk to qubesd, since it's blocked on waiting for attach-confirm...
- subprocess.Popen -> asyncio variant
- The tool name is IMO too generic for a tool in a common /usr/bin/
- The tool belongs to desktop-linux-manager repo, which looks like a layering violation - dom0 code should also work without any of the GUI frontends installed in dom0.
- Extension of the above: this also will need adjustment to the GUI domain threat model: verify the response is one of allowed ones (on the
frontendslist?)
There was a problem hiding this comment.
I see the code of attach-confirm already is async. Maybe simply put its code here, instead of calling external program? It means you will need to keep the params dict format in sync, but changes there needs to be done in compatible way anyway (due to the GUI domain case, where both ends may be updated independently). Plus, you won't need to make external get_system_info() call, as by running inside qubesd you already have all the info here.
implements: QubesOS/qubes-issues/issues/9325