netdev - Re: [PATCH net] genetlink: take netlink table lock when (un)registering

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

Date:   Sat, 27 Jun 2020 17:26:37 -0700 (PDT)
From:   David Miller <davem@...emloft.net>
To:     stranche@...eaurora.org
Cc:     netdev@...r.kernel.org, pshelar@....org, subashab@...eaurora.org
Subject: Re: [PATCH net] genetlink: take netlink table lock when
 (un)registering

From: Sean Tranchetti <stranche@...eaurora.org>
Date: Fri, 26 Jun 2020 18:31:03 -0600

> @@ -328,6 +325,10 @@ int genl_register_family(struct genl_family *family)
>  	if (err)
>  		return err;
>  
> +	/* Acquire netlink table lock before any GENL-specific locks to ensure
> +	 * sync with any netlink operations making calls into the GENL code.
> +	 */
> +	netlink_table_grab();
>  	genl_lock_all();

This locking sequence is illegal, and if you tested this change with the
proper lock debugging options enabled you wouldn't have been able to
even boot a machine without it OOPS'ing.

This code was essentially not tested as far as I am concerned.

netlink_table_grab() takes an atomic lock (write_lock_irq), so it
creates an atomic section.  But then we immediately call
genl_lock_all() which takes multiple sleepable locks (a semaphore and
a mutex).

You'll have to find another way to fix this bug and I would like to ask
that you do so in a way that keeps all of these code paths sleepable
and does not do any GFP_ATOMIC conversions.

Thank you.