Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 19 May 2019 16:11:41 +0000
From:      bugzilla-noreply@freebsd.org
To:        bugs@FreeBSD.org
Subject:   [Bug 219216] sched_bind() blocks if the entropy pool is starved
Message-ID:  <bug-219216-227-r9NtMNEKVr@https.bugs.freebsd.org/bugzilla/>
In-Reply-To: <bug-219216-227@https.bugs.freebsd.org/bugzilla/>
References:  <bug-219216-227@https.bugs.freebsd.org/bugzilla/>

next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D219216

--- Comment #10 from Conrad Meyer <cem@freebsd.org> ---
rand_harvestq has had CPU reduced significantly in 12.x timeframe =E2=80=94=
 previously
it was attempting to obtain far more entropy and far more often than it
actually needed.  This was in part due to conservative estimates (r338293) =
and
in part due to bugs in random read (r338292, r339492).

r287023, r334450 may also be relevant.

If you have a concurrent reader, rand_harvestq will fight it for the global
Fortuna state mutex on head today (and spin while waiting).  You are somewh=
at
saved by READ_RANDOM_UIO dropping the Fortuna lock every 4kB (PAGE_SIZE).  I
have a revision proposed to reduce this conflict further:

https://people.freebsd.org/~cem/rand_harvestq.png
https://reviews.freebsd.org/D20313

--=20
You are receiving this mail because:
You are the assignee for the bug.=



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-219216-227-r9NtMNEKVr>