glibc tst-getrandom test needs more entropy causing test failures
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Auto Package Testing |
Fix Committed
|
Undecided
|
Unassigned | ||
launchpad-buildd |
Triaged
|
Critical
|
Unassigned | ||
glibc (Ubuntu) |
Fix Released
|
Undecided
|
Unassigned | ||
Bionic |
Fix Released
|
Undecided
|
Unassigned | ||
Focal |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
[Impact]
* Builds and autpkgtests frequently fail due to failing tst-getrandom probably due to insufficient source of entropy on Launchpad. This causes extra work by requiring retriggering failing builds and autopkgtests.
[Test Case]
* Observe tst-getrandom tests being marked XFAIL in build logs.
[Regression Potential]
* The fix does not change run-time behaviour of glibc, just ignores a test. This lets potential breakages in getrandom()'s implementation or in kernel's random source being undetected, but the test results can still be observed in the logs if needed and the random implementation is also most likely exercised in other tests, too.
[Other Info]
The issue can be reproduced locally in a built glibc package tree without haveged or any other good entropy source installed:
rbalint@
Timed out: killed the child process
Termination time: 2020-08-
Last write to standard output: 2020-08-
Timed out: killed the child process
Termination time: 2020-08-
Last write to standard output: 2020-08-
Timed out: killed the child process
Termination time: 2020-08-
Last write to standard output: 2020-08-
Timed out: killed the child process
Termination time: 2020-08-
Last write to standard output: 2020-08-
Timed out: killed the child process
...
With haveged installed the tests pass quickly and without error:
rbalint@
rbalint@
An other way of fixing the issue would be installing better entropy sources in the build and autopkgtest infrastructure.
CVE References
description: | updated |
summary: |
- glibc tests needs more entropy causing tst-getrandom failures + glibc tst-getrandom test needs more entropy causing test failures |
Changed in launchpad-buildd: | |
importance: | Undecided → Critical |
status: | New → Triaged |
I know qemu has some entropy forwarding stuff and I wonder if that could be used in the cloud so instances have enough entropy to fix this
e.g.
I think openstack has support for that, and this maybe should be toggled on? Might be worth checking with IS and security.