linux-kernel - [PATCH] drm/i915: Set PROBE_PREFER

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>] [day] [month] [year] [list]

Date:   Fri, 28 Oct 2022 14:53:22 -0700
From:   Brian Norris <briannorris@...omium.org>
To:     Jani Nikula <jani.nikula@...ux.intel.com>,
        Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>,
        Rodrigo Vivi <rodrigo.vivi@...el.com>,
        Tvrtko Ursulin <tvrtko.ursulin@...ux.intel.com>
Cc:     Feng Tang <feng.tang@...el.com>,
        Jani Nikula <jani.nikula@...el.com>,
        Chris Wilson <chris@...is-wilson.co.uk>,
        intel-gfx@...ts.freedesktop.org, dri-devel@...ts.freedesktop.org,
        Ville Syrjälä 
        <ville.syrjala@...ux.intel.com>, linux-kernel@...r.kernel.org,
        Daniel Vetter <daniel.vetter@...ll.ch>,
        Brian Norris <briannorris@...omium.org>,
        Daniel Vetter <daniel@...ll.ch>,
        David Airlie <airlied@...il.com>
Subject: [PATCH] drm/i915: Set PROBE_PREFER_ASYNCHRONOUS

This driver often takes a good 100ms to start, but in some particularly
bad cases takes more than 1 second.

In surveying risk for this driver, I poked around for cross-device
shared state, which can be a source of race conditions. GVT support
(intel_gvt_devices) seems potentially suspect, but it has an appropriate
mutex, and doesn't seem to care about ordering -- if devices are present
when the kvmgt module loads, they'll get picked up; and if they probe
later than kvmgt, they'll attach then.

Additionally, I see past discussions about this patch [1], which
concluded that there were problems at the time due to the way
hdac_i915.c called request_module("i915") and expected it to complete
probing [2]. Work has since been merged [3] to fix that problem.

This driver was pinpointed as part of a survey of drivers that take more
than 100ms in their initcall (i.e., are built in, and probing
synchronously) on a lab of ChromeOS systems.

[1] [RFC] i915: make the probe asynchronous
    https://lore.kernel.org/all/20180604053219.2040-1-feng.tang@intel.com/

[2] https://lore.kernel.org/all/s5hin4d1e3f.wl-tiwai@suse.de/

[3] Commit f9b54e1961c7 ("ALSA: hda/i915: Allow delayed i915 audio
    component binding")

Signed-off-by: Brian Norris <briannorris@...omium.org>
---

 drivers/gpu/drm/i915/i915_pci.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c
index 38460a0bd7cb..1cb1f87aea86 100644
--- a/drivers/gpu/drm/i915/i915_pci.c
+++ b/drivers/gpu/drm/i915/i915_pci.c
@@ -1371,7 +1371,10 @@ static struct pci_driver i915_pci_driver = {
 	.probe = i915_pci_probe,
 	.remove = i915_pci_remove,
 	.shutdown = i915_pci_shutdown,
-	.driver.pm = &i915_pm_ops,
+	.driver = {
+		.pm = &i915_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
 };

 int i915_pci_register_driver(void)
-- 
2.38.1.273.g43a17bfeac-goog