On some machines (s390x would never be your guess, would it?) there is a permanent
lack of random data entropy which is quite a sever issue, but nobody seems to be
alerted by the fact. So (as expected) it turns out that we cannot just wait
forever for enough random data entropy to be gathered by kernel. Thus these two
patches implement a timeout that forces the installation to continue even if
there's not enough entropy for LUKS format creation. A warning is logged in such
case.
The new look of the EntropyDialog can be seen in the following screenshot:
http://vpodzime.fedorapeople.org/entropy_wait_dialog_timeout.png
Is the time information more or less unrelated to the entropy % too confusing?
Ideas? Suggestions? All are welcome, just please note that we should avoid
adding options/switches tweaking the behaviour because we don't want to commit
to supporting such options/switches in the future and it's hard to predict how
the situation around requiring/enforcing enough entropy will evolve. Obviously
the ideal solution is to patch kernel to gather enough entropy even on s390x,
make virt-rng the default etc. so we should stand on "the right side" and help
pushing such solutions further.
Vratislav Podzimek (1):
Add timeout to callbacks waiting for enough entropy (#1073679)
pyanaconda/constants.py | 3 +++
pyanaconda/ui/gui/spokes/lib/entropy_dialog.glade | 25 +++++++++++++++---
pyanaconda/ui/gui/spokes/lib/entropy_dialog.py | 25 +++++++++++++++---
pyanaconda/ui/lib/entropy.py | 32 +++++++++++++++++------
4 files changed, 71 insertions(+), 14 deletions(-)
--
1.9.3