Subject: Re: [PATCH] usb: gadget: f_fs: Fix kernel panic for SuperSpeed
To: Jim Lin <jilin@nvidia.com>, balbi@kernel.org
References: <1461321780-3226-1-git-send-email-jilin@nvidia.com>
Cc: linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org
From: Lars-Peter Clausen <lars@metafoo.de>
Message-ID: <571A0951.2030802@metafoo.de>
Date: Fri, 22 Apr 2016 13:21:53 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Icedove/38.7.0
MIME-Version: 1.0
In-Reply-To: <1461321780-3226-1-git-send-email-jilin@nvidia.com>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 944
Lines: 20

On 04/22/2016 12:43 PM, Jim Lin wrote:
> Android N adds os_desc_compat in v2_descriptor by init_functionfs()
> (system/core/adb/usb_linux_client.cpp) to support automatic install
> of MTP driver on Windows for USB device mode.
> 
> Current __ffs_data_do_os_desc() of f_fs.c will check reserved1 field
> and return -EINVAL.
> This results in a second adb_write of usb_linux_client.cpp
> (system/core/adb/) which doesn't have ss_descriptors filled.
> Then later kernel_panic (composite.c) occurs when ss_descriptors
> as a pointer with NULL is being accessed.
> 
> Fix is to ignore the checking on reserved1 field so that first
> adb_write goes successfully with v2_descriptor which has
> ss_descriptors filled.

That sounds like the wrong approach. The kernel should not crash if
ss_descriptors is not filled. I think the right fix is to make sure that the
NULL pointer deref can never happen regardless of which input is supplied by
userspace.