linux-kernel - Re: [PATCH v7 19/19] staging: r8188eu: remove shared buffer for usb requests

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:   Fri, 17 Sep 2021 17:18:13 +0200
From:   Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To:     Pavel Skripkin <paskripkin@...il.com>
Cc:     Larry Finger <Larry.Finger@...inger.net>,
        Phillip Potter <phil@...lpotter.co.uk>,
        linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org,
        David Laight <david.Laight@...lab.com>,
        Dan Carpenter <dan.carpenter@...cle.com>,
        "Fabio M. De Francesco" <fmdefrancesco@...il.com>
Subject: Re: [PATCH v7 19/19] staging: r8188eu: remove shared buffer for usb
 requests

On Fri, Sep 17, 2021 at 06:03:52PM +0300, Pavel Skripkin wrote:
> On 9/17/21 17:55, Greg Kroah-Hartman wrote:
> > On Fri, Sep 17, 2021 at 09:18:37AM +0200, Fabio M. De Francesco wrote:
> > > From: Pavel Skripkin <paskripkin@...il.com>
> > > 
> > > This driver used shared buffer for usb requests. It led to using
> > > mutexes, i.e no usb requests can be done in parallel.
> > > 
> > > USB requests can be fired in parallel since USB Core allows it. In
> > > order to allow them, remove usb_vendor_req_buf from dvobj_priv (since
> > > USB I/O is the only user of it) and remove also usb_vendor_req_mutex
> > > (since there is nothing to protect).
> > 
> > Ah, you are removing this buffer, nice!
> > 
> > But, just because the USB core allows multiple messages to be sent to a
> > device at the same time, does NOT mean that the device itself can handle
> > that sort of a thing.
> > 
> > Keeping that lock might be a good idea, until you can prove otherwise.
> > You never know, maybe there's never any contention at all for it because
> > these accesses are all done in a serial fashion and the lock
> > grab/release is instant.  But if that is not the case, you might really
> > get a device confused here by throwing multiple control messages at it
> > in ways that it is not set up to handle at all.
> > 
> > So please do not drop the lock.
> > 
> > More comments below.
> > 
> 
> We have tested this change. I've tested it in qemu with TP-Link TL-WN722N v2
> / v3 [Realtek RTL8188EUS], and Fabio has tested it on his host for like a
> whole evening.
> 
> I agree, that our testing does not cover all possible cases and I can't say
> it was "good stress testing", so, I think, we need some comments from
> maintainers.

Ok, then make it a single patch that does nothing but remove the lock so
that we can revert it later when problems show up :)

thanks,

greg k-h