[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20180205181758.GA22738@usblab-sd-06.qualcomm.com>
Date: Mon, 5 Feb 2018 10:17:58 -0800
From: Jack Pham <jackp@...eaurora.org>
To: William Wu <william.wu@...k-chips.com>
Cc: gregkh@...uxfoundation.org, felipe.balbi@...ux.intel.com,
linux-kernel@...r.kernel.org, linux-usb@...r.kernel.org,
linux-rockchip@...ts.infradead.org, frank.wang@...k-chips.com,
huangtao@...k-chips.com, daniel.meng@...k-chips.com,
fml@...k-chips.com
Subject: Re: [PATCH] usb: gadget: f_fs: get the correct address of comp_desc
Hi William,
On Mon, Feb 05, 2018 at 07:33:38PM +0800, William Wu wrote:
> Refer to the USB 3.0 spec '9.6.7 SuperSpeed Endpoint Companion',
> the companion descriptor follows the standard endpoint descriptor.
> This descriptor is only defined for SuperSpeed endpoints. The
> f_fs driver gets the address of the companion descriptor via
> 'ds + USB_DT_ENDPOINT_SIZE', and actually, the ds variable is
> a pointer to the struct usb_endpoint_descriptor, so the offset
> of the companion descriptor which we get is USB_DT_ENDPOINT_SIZE *
> sizeof(struct usb_endpoint_descriptor), the wrong offset is 63
> bytes. This cause out-of-bound with the following error log if
> CONFIG_KASAN and CONFIG_SLUB_DEBUG is enabled on Rockchip RK3399
> Evaluation Board.
>
> android_work: sent uevent USB_STATE=CONNECTED
> configfs-gadget gadget: super-speed config #1: b
> ==================================================================
> BUG: KASAN: slab-out-of-bounds in ffs_func_set_alt+0x230/0x398
> Read of size 1 at addr ffffffc0ce2d0b10 by task irq/224-dwc3/364
> Memory state around the buggy address:
> ffffffc0ce2d0a00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> ffffffc0ce2d0a80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> >ffffffc0ce2d0b00: 00 04 fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> ^
> ffffffc0ce2d0b80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> ffffffc0ce2d0c00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> ==================================================================
> Disabling lock debugging due to kernel taint
> android_work: sent uevent USB_STATE=CONFIGURED
>
> This patch adds struct usb_endpoint_descriptor * -> u8 * type conversion
> for ds variable, then we can get the correct address of comp_desc
> with offset USB_DT_ENDPOINT_SIZE bytes.
>
> Signed-off-by: William Wu <william.wu@...k-chips.com>
> ---
> drivers/usb/gadget/function/f_fs.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
> index 6756472..f13ead0 100644
> --- a/drivers/usb/gadget/function/f_fs.c
> +++ b/drivers/usb/gadget/function/f_fs.c
> @@ -1882,8 +1882,8 @@ static int ffs_func_eps_enable(struct ffs_function *func)
> ep->ep->desc = ds;
>
> if (needs_comp_desc) {
> - comp_desc = (struct usb_ss_ep_comp_descriptor *)(ds +
> - USB_DT_ENDPOINT_SIZE);
> + comp_desc = (struct usb_ss_ep_comp_descriptor *)
> + ((u8 *)ds + USB_DT_ENDPOINT_SIZE);
> ep->ep->maxburst = comp_desc->bMaxBurst + 1;
> ep->ep->comp_desc = comp_desc;
> }
Please see my alternative fix for this. I proposed changing this
function to use config_ep_by_speed() instead.
https://www.spinics.net/lists/linux-usb/msg165149.html
Jack
--
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
Powered by blists - more mailing lists