linux-kernel - Re: [PATCH 2/3] HID: steam: add serial number information.

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]

Message-ID: <20180216090243.GB18755@casa>
Date:   Fri, 16 Feb 2018 10:02:43 +0100
From:   Rodrigo Rivas Costa <rodrigorivascosta@...il.com>
To:     Benjamin Tissoires <benjamin.tissoires@...hat.com>
Cc:     Jiri Kosina <jikos@...nel.org>,
        lkml <linux-kernel@...r.kernel.org>, linux-input@...r.kernel.org
Subject: Re: [PATCH 2/3] HID: steam: add serial number information.

On Fri, Feb 16, 2018 at 09:44:34AM +0100, Benjamin Tissoires wrote:
> > I have an issue with this one. The problem is that using
> > hid_report_len() on the feature report returns 64. But I must call
> > hid_hw_raw_request() with 65 or it will fail with EOVERFLOW.
> >
> > Currently I'm allocating a buffer of 65 bytes and all is well.
> > If I change to hid_alloc_report_buf(), the current implementation
> > allocates (64+7), so I'm still safe. But I'm worried that the extra
> > bytes are not guaranteed and a future implementation could return
> > exactly 64 bytes, leaving me 1 byte short.
> >
> > About why an array of 65 is required for a report of size 64, I think it
> > is related to hid_report->id == 0 (so hid_report_enum->numbered == 0).
> 
> That's the other way around actually. If you are just using the output
> of hid_report_len(), it will take into account the extra byte for the
> report ID.
> *But*, given the way implement() is working (see the comment in the
> implementation of hid_alloc_report()), you need to have up to 7 extra
> bytes to not have the EOVERFLOW.
> 
> So if we ever change the implement() function (which is *really*
> unlikely), we will have to make sure hid_alloc_report() still works,
> so you are on the safe side if you use hid_alloc_report().

Ok, I'll do that. The weird thing, however, is that:

	hid_hw_raw_request(steam->hid_dev, 0x00,
		buf, hid_report_len(r), /* 64 */
		HID_FEATURE_REPORT, HID_REQ_GET_REPORT);

fails with EOVERFLOW. I have to use:

	hid_hw_raw_request(steam->hid_dev, 0x00,
		buf, 65
		HID_FEATURE_REPORT, HID_REQ_GET_REPORT);

which just feels wrong to me.

And looking around drivers/hid/*.c I see that most calls to
hid_hw_raw_request(..., HID_REQ_GET_REPORT) use a buffer allocated with
{devm_,}kzalloc() and a constant length, never using
hid_alloc_report_buf() or hid_report_len().

Maybe there is a bug in hid_hw_raw_request() and it should add 1 to the
given buffer len? But then, custom buffer allocations will overflow by
one!

Rodrigo.