linux-kernel - Re: [PATCH] usb: gadget: Increase the limit of USB_GADGET_VBUS

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] [day] [month] [year] [list]

Message-ID: <628ea5a8-8ef0-452f-b2b5-4325f98b3bd3@rowland.harvard.edu>
Date: Thu, 20 Feb 2025 10:15:47 -0500
From: Alan Stern <stern@...land.harvard.edu>
To: Prashanth K <prashanth.k@....qualcomm.com>
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Simona Vetter <simona.vetter@...ll.ch>,
	Takashi Iwai <tiwai@...e.de>, linux-usb@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] usb: gadget: Increase the limit of USB_GADGET_VBUS_DRAW
 to 900mA

On Thu, Feb 20, 2025 at 09:47:58AM +0530, Prashanth K wrote:
> 
> 
> On 22-01-25 12:56 am, Alan Stern wrote:
> > On Tue, Jan 21, 2025 at 10:19:08PM +0530, Prashanth K wrote:
> >>
> >>
> >> On 21-01-25 08:36 pm, Alan Stern wrote:
> >>> On Tue, Jan 21, 2025 at 09:50:08AM +0530, Prashanth K wrote:
> >>>>
> >>>>
> >>>> On 20-01-25 08:17 pm, Alan Stern wrote:
> >>>>> On Mon, Jan 20, 2025 at 04:47:02PM +0530, Prashanth K wrote:
> >>>>>> Currently CONFIG_USB_GADGET_VBUS_DRAW limits the maximum current
> >>>>>> drawn from Vbus to be up to 500mA. However USB gadget operating
> >>>>>> in SuperSpeed or higher can draw up to 900mA. Also, MaxPower in
> >>>>>> ConfigFS takes its default value from this config. Hence increase
> >>>>>> the allowed range of CONFIG_USB_GADGET_VBUS_DRAW to 900mA.
> >>>>>
> >>>>> Is this the sort of thing that really needs to be a Kconfig option?  Why 
> >>>>> not make the decision at runtime, based on the needs of the gadget or 
> >>>>> function drivers and the connection speed?
> >>>>>
> >>>>> Alan Stern
> >>>>>
> >>>>
> >>>> Right, set_config() in composite.c does this in runtime based on the
> >>>> values of MaxPower (from configFS), VBUS_DRAW defconfig and speed.
> >>>> If we don't set MaxPower from configFS, this config helps to set it
> >>>> during compile time. In fact MaxPower in configFS takes its default
> >>>> value from CONFIG_USB_GADGET_VBUS_DRAW . Sent this patch because Kconfig
> >>>> has this limitation where it's only allowing values upto 500mA.
> >>>
> >>> Why does MaxPower need to be set at compile time?  Why not set it at 
> >>> runtime instead?
> >>>
> >>> If MaxPower gets set at runtime then it can take its default value to be 
> >>> 500 mA or 900 mA depending on the connection speed.  There will be no 
> >>> need for CONFIG_USB_GAGDGET_VBUS_DRAW.
> >>>
> >>
> >> Yes, agreed. Can we mark CONFIG_USB_GAGDGET_VBUS_DRAW as legacy and
> >> maybe also avoid configfs/composite from using it?
> > 
> > Indeed, the whole idea is to avoid using CONFIG_USB_GADGET_VBUS_DRAW in 
> > configfs and composite.
> > 
> > If nothing will still be using it, just remove it entirely.  No need to 
> > mark it as legacy.
> > 
> > Alan Stern
> 
> Hi Alan, small doubt, I see that gadget/legacy/webcam.c is a super-speed
> gadget which uses CONFIG_USB_GADGET_VBUS_DRAW. I'm quite not really sure
> if anyone uses it now, but if someone uses it, then wouldn't my patch be
> applicable there?

A quick search shows that the legacy/gmidi.c driver also uses it.

Still, if these are the only drivers that uses 
CONFIG_USB_GADGET_VBUS_DRAW, it means that parameter isn't really very 
useful.

The best solution I can think of is to make vbus_draw a module parameter 
in for the two drivers, with the default set to 500.  That's more 
flexible than relying on a Kconfig parameter anyway.  Of course, this 
means the .MaxPower field in the webcam_config_driver and midi_config 
structures will have to be set dynamically at runtime.

Alan Stern

> Noticed this while preparing a patch to remove dependency of VBUS_DRAW
> config from configfs/composite layer. Its ready, will send that after
> some testing.
> 
> Regards,
> Prashanth K