linux-kernel - [PATCH v2 0/8] Asynchronous device/driver probing support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1427757610-27882-1-git-send-email-dmitry.torokhov@gmail.com>
Date:	Mon, 30 Mar 2015 16:20:02 -0700
From:	Dmitry Torokhov <dmitry.torokhov@...il.com>
To:	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	"Luis R . Rodriguez" <mcgrof@...e.com>, Tejun Heo <tj@...nel.org>
Cc:	linux-kernel@...r.kernel.org,
	Arjan van de Ven <arjan@...ux.intel.com>,
	Rusty Russell <rusty@...tcorp.com.au>,
	Olof Johansson <olof@...om.net>,
	Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
Subject: [PATCH v2 0/8] Asynchronous device/driver probing support

This series is a combination of changes proposed by Luis a couple months
ago and implementation used by Chrome OS. The issue we are trying to solve
here is "slow" devices and drivers spending "too much time" in their probe()
methods and it affects:

- overall kernel boot process when drivers are compiled into the kernel
  and slow devices stall entire boot progress;
- systemd desire to time out module loading process.

Unlike Luis' proposal we do make use of asycn_schedule() infrastructure
instead of using a dedicated workqueue, so all  existing synchronization
points in kernel that wait for device registration still work the same.
Also, the asynchronous probing is done not only during driver registration
(i.e. when devices are probed asynchronously only if they are registered
before the driver), but also during device registration and deferred probe
handling. This way slow devices do not stall kernel boot even when drivers
are compiled into the kernel.

The last patch is for adventurous people to try and force
fully-asynchronous boot. It works for me with limited success - I can boot
Rockhip-based box to userspace as long as I force serial to be sychronously
probed and ignore the fact that most devices are using "dummy" regulators
as regulator subsystem really expects regulators to be registered in
orderly fashion on OF-based systems.

Changes from v1:

- Changed verbage in change logs and code to emphasise that
  PROBE_PREFER_ASYNCHRONOUS is a temporary measure and the end goal is
  to enable asynchronous probing by default, as requested by Tejun.

Thanks,
Dmitry

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/