On Tue, 2018-06-05 at 19:20 +0200, Cyril Brulebois wrote: > Hi, > > Ben Hutchings <ben@decadent.org.uk> (2018-05-13): > > This sounds similar to the regression seen in stable affecting gnome- > > session. libICE has a function for generating the MIT-MAGIC-COOKIE-1 > > cookie, which now uses libbsd's arc4random_buf() which now uses > > getrandom(). getrandom() is documented to block (by default) when not > > much entropy is available, but until recently it could return too early > > after boot. > > > > I don't know whether the cookie is generated this way when d-i starts > > the X server. > > > > I opened bug #898088 against libbsd because arc4random_buf() is > > documented to return "very quickly". > > Many thanks for the initial pointer and the libbsd bugfix. Unfortunately > that wasn't our issue at all (we don't use libICE and there's no udeb > for it anyway). I've documented my findings in a blog post: > https://debamax.com/blog/2018/05/25/debugging-black-screen-in-debian-installer/ > > > Summary: The issue is rather that we have fontconfig needing UUIDs for > font handling, calling uuid_generate_random() from util-linux, which > uses getrandom(), trying a few times with a little delay between each > attempts when EAGAIN has been returned, before finally falling back to > reading from /dev/{u,}random. [...] plymouth had the same problem, fixed in version 0.9.3-3. You should be able to avoid it by running fc-cache as part of the initramfs build process. Ben. -- Ben Hutchings Larkinson's Law: All laws are basically false.
Attachment:
signature.asc
Description: This is a digitally signed message part