linux-kernel - Re: [PATCH 2/2] driver: core: add probe control driver

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: <20240918145516.GA1560032-robh@kernel.org>
Date: Wed, 18 Sep 2024 09:55:16 -0500
From: Rob Herring <robh@...nel.org>
To: Nayeemahmed Badebade <nayeemahmed.badebade@...y.com>
Cc: krzk+dt@...nel.org, conor+dt@...nel.org, gregkh@...uxfoundation.org,
	rafael@...nel.org, yoshihiro.toyama@...y.com,
	linux-kernel@...r.kernel.org, devicetree@...r.kernel.org
Subject: Re: [PATCH 2/2] driver: core: add probe control driver

On Tue, Sep 17, 2024 at 02:25:50PM +0530, Nayeemahmed Badebade wrote:
> Hi Rob,
> 
> Thank you for taking the time to check our patch and provide
> valuable feedback. We appreciate your comments/suggestions.
> 
> Please find our reply to your comments.
> 
> On Thu, Sep 12, 2024 at 03:46:34PM -0500, Rob Herring wrote:
> > On Wed, Sep 11, 2024 at 07:53:19PM +0530, Nayeemahmed Badebade wrote:
> > > Probe control driver framework allows deferring the probes of a group of
> > > devices to an arbitrary time, giving the user control to trigger the probes
> > > after boot. This is useful for deferring probes from builtin drivers that
> > > are not required during boot and probe when user wants after boot.
> > 
> > This seems like the wrong way around to me. Why not define what you want 
> > to probe first or some priority order? I could see use for kernel to 
> > probe whatever is the console device first. Or the rootfs device... You 
> > don't need anything added to DT for those.
> > 
> > Of course, there's the issue that Linux probes are triggered bottom-up 
> > rather than top-down.
> > 
> 
> Our intention is to only postpone some driver probes not required during
> boot, similar to https://elinux.org/Deferred_Initcalls. But instead of
> delaying initcall execution(which requires initmem to be kept and not
> freed during boot) we are trying to delay driver probes as this is much
> simpler.
> 
> > 
> > > This is achieved by adding a dummy device aka probe control device node
> > > as provider to a group of devices(consumer nodes) in platform's device
> > > tree. Consumers are the devices we want to probe after boot.
> > 
> > There's the obvious question of then why not make those devices modules 
> > instead of built-in?
> > 
> 
> Yes we can use modules for this, but there are drivers that cannot be
> built as modules and this framework is specifically for such scenario.
> Example: drivers/pci/controller/dwc/pci-imx6.c

Then fix the driver to work as a module. Or to use async probe which is 
not the default and is opt-in per driver.

> 
> > > 
> > > To establish control over consumer device probes, each consumer device node
> > > need to refer the probe control provider node by the phandle.
> > > 'probe-control-supply' property is used for this.
> > > 
> > > Example:
> > >     // The node below defines a probe control device/provider node
> > >     prb_ctrl_dev_0: prb_ctrl_dev_0 {
> > >         compatible = "linux,probe-control";
> > >     };
> > > 
> > >     // The node below is the consumer device node that refers to provider
> > >     // node by its phandle and a result will not be probed until provider
> > >     // node is probed.
> > >     pcie@...c000 {
> > >         reg = <0x01ffc000 0x04000>, <0x01f00000 0x80000>;
> > >         #address-cells = <3>;
> > >         #size-cells = <2>;
> > >         device_type = "pci";
> > >         ranges = <0x81000000 0 0          0x01f80000 0 0x00010000>,
> > >                  <0x82000000 0 0x01000000 0x01000000 0 0x00f00000>;
> > > 
> > >         probe-control-supply = <&prb_ctrl_dev_0>;
> > >     };
> > 
> > Sorry, but this isn't going to happen in DT.
> > 
> 
> You mean we cannot add custom properties like this to an existing
> device node in DT?

Sure, you can add properties. It happens all the time. This is too tied 
to some OS implementation/behavior and therefore is not appropriate for 
DT.

Rob