linux-kernel - [PATCH RESEND] i2c: core: Fix possible memleak in i2c_register

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: <20230928012709.1247208-1-lizetao1@huawei.com>
Date:   Thu, 28 Sep 2023 09:27:09 +0800
From:   Li Zetao <lizetao1@...wei.com>
To:     <wsa@...nel.org>, <khali@...ux-fr.org>, <gregkh@...e.de>
CC:     <lizeta1@...wei.com>, <linux-i2c@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, Li Zetao <lizetao1@...wei.com>
Subject: [PATCH RESEND] i2c: core: Fix possible memleak in i2c_register_adapter()

There is a memory leak reported by kmemleak:

unreferenced object 0xffff88818be6dcb8 (size 8):
  comm "modprobe", pid 1022129, jiffies 4363911608 (age 43.838s)
    hex dump (first 8 bytes):
      69 32 63 2d 30 00 6b a5                          i2c-0.k.
    backtrace:
      [<ffffffff812f588f>] __kmem_cache_alloc_node+0x2cf/0x4b0
      [<ffffffff81283c43>] __kmalloc_node_track_caller+0x53/0x140
      [<ffffffff81604b0a>] kvasprintf+0x6a/0xd0
      [<ffffffff81604be7>] kvasprintf_const+0x77/0xa0
      [<ffffffff81a00173>] kobject_set_name_vargs+0x23/0x90
      [<ffffffff817bf743>] dev_set_name+0x53/0x70
      [<ffffffffa023ebb2>] i2c_register_adapter+0x112/0x6c0 [i2c_core]
      [<ffffffffa023f268>] i2c_add_adapter+0x78/0xc0 [i2c_core]
      [<ffffffffa0300182>] piix4_add_adapter+0x132/0x210 [i2c_piix4]
      [<ffffffffa0300535>] piix4_probe+0x2d5/0x5f4 [i2c_piix4]
      ...

The root cause was traced to an error handing path in
i2c_register_adapter() when device_register() fails. After
calling dev_set_name() which called by i2c_register_adapter(),
the put_device() should be used to give up the device reference
in error handling path in order to free "kobj->name" alloced
in dev_set_name().

Fix it by calling put_device() when device_register() fails in
i2c_register_adapter(). In addition, when i2c_adapter device
is released, i2c_adapter_dev_release() is called, in which
"dev_released" must be initialized, otherwise a kernel bug
will be triggered.

Fixes: b119c6c952a0 ("i2c: __must_check fixes (core drivers)")
Signed-off-by: Li Zetao <lizetao1@...wei.com>
---
 drivers/i2c/i2c-core-base.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
index 60746652fd52..cc4a20465456 100644
--- a/drivers/i2c/i2c-core-base.c
+++ b/drivers/i2c/i2c-core-base.c
@@ -1514,6 +1514,9 @@ static int i2c_register_adapter(struct i2c_adapter *adap)
 	res = device_register(&adap->dev);
 	if (res) {
 		pr_err("adapter '%s': can't register device (%d)\n", adap->name, res);
+		init_completion(&adap->dev_released);
+		put_device(&adap->dev);
+		wait_for_completion(&adap->dev_released);
 		goto out_list;
 	}
 
-- 
2.34.1