linux-kernel - [PATCH] regmap: do not call regmap_debugfs_init() from regmap_attach

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]

Date:   Mon, 26 Jul 2021 09:36:27 +0200
From:   Matthias Schiffer <matthias.schiffer@...tq-group.com>
To:     Mark Brown <broonie@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     David Lechner <david@...hnology.com>, linux-kernel@...r.kernel.org,
        Matthias Schiffer <matthias.schiffer@...tq-group.com>
Subject: [PATCH] regmap: do not call regmap_debugfs_init() from regmap_attach_dev()

regmap_debugfs_init() should never be called twice for the same regmap,
as it initializes various fields of the regmap struct, including list
heads and mutices. A visible symptom are messages like:

    debugfs: Directory 'dummy-iomuxc-gpr@...4000' with parent 'regmap'
    already present!

This happened whenever regmap_attach_dev() was called for an existing
regmap. Remove the call from regmap_attach_dev() and change
__regmap_init() so that regmap_debugfs_init() is called exactly once.

Signed-off-by: Matthias Schiffer <matthias.schiffer@...tq-group.com>
Fixes: 9b947a13e7f6 ("regmap: use debugfs even when no device")
---
 drivers/base/regmap/regmap.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

As mentioned in my previous mail [1], I believe that the duplicate call to
regmap_debugfs_init() is not the only issue with regmap_attach_dev().

Every single user of regmap_attach_dev() outside of __regmap_init() is
using it to attach a device to a syscon regmap obtained using one of the
syscon_*() functions. AIUI, syscon regmaps do not belong to a single
device, so calling regmap_attach_dev() seems wrong to me. My (possibly
lacking) understanding of the semantics aside, the fact that
regmap_attach_dev() is setting fields on the shared regmap without any
kind of locking is at least suspicious.

Maybe regmap_attach_dev() shouldn't be exported from the regmap code at
all, removing all calls to the functions from drivers?

[1] https://lkml.org/lkml/2021/7/19/500

diff --git a/drivers/base/regmap/regmap.c b/drivers/base/regmap/regmap.c
index fe3e38dd5324..27625a1330ac 100644
--- a/drivers/base/regmap/regmap.c
+++ b/drivers/base/regmap/regmap.c
@@ -630,8 +630,6 @@ int regmap_attach_dev(struct device *dev, struct regmap *map,
 	if (ret)
 		return ret;

-	regmap_debugfs_init(map);
-
 	/* Add a devres resource for dev_get_regmap() */
 	m = devres_alloc(dev_get_regmap_release, sizeof(*m), GFP_KERNEL);
 	if (!m) {
@@ -1192,10 +1190,10 @@ struct regmap *__regmap_init(struct device *dev,
 		ret = regmap_attach_dev(dev, map, config);
 		if (ret != 0)
 			goto err_regcache;
-	} else {
-		regmap_debugfs_init(map);
 	}

+	regmap_debugfs_init(map);
+
 	return map;

 err_regcache:
-- 
2.17.1