Message-ID: <130064e95124f32a40618620450016bec0a96ffd.camel@mailbox.org>
Date: Thu, 11 Dec 2025 15:44:57 +0100
From: Philipp Stanner <phasta@...lbox.org>
To: Yao Zi <me@...ao.cc>, phasta@...nel.org, Bjorn Helgaas
 <helgaas@...nel.org>,  "Russell King (Oracle)" <linux@...linux.org.uk>,
 Thomas Gleixner <tglx@...utronix.de>
Cc: Bjorn Helgaas <bhelgaas@...gle.com>, Andrew Lunn
 <andrew+netdev@...n.ch>,  "David S. Miller" <davem@...emloft.net>, Eric
 Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo
 Abeni <pabeni@...hat.com>, Frank <Frank.Sae@...or-comm.com>, Heiner
 Kallweit <hkallweit1@...il.com>, Vladimir Oltean <vladimir.oltean@....com>,
 Choong Yong Liang <yong.liang.choong@...ux.intel.com>, Chen-Yu Tsai
 <wens@...e.org>, Jisheng Zhang <jszhang@...nel.org>, Furong Xu
 <0x1207@...il.com>,  linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
 Mingcong Bai <jeffbai@...c.io>, Kexy Biscuit <kexybiscuit@...c.io>, Runhua
 He <hua@...c.io>,  Xi Ruoyao <xry111@...111.site>
Subject: Re: [PATCH net-next v3 2/3] net: stmmac: Add glue driver for
 Motorcomm YT6801 ethernet controller

On Thu, 2025-12-11 at 14:23 +0000, Yao Zi wrote:
> On Mon, Dec 08, 2025 at 10:54:36AM +0100, Philipp Stanner wrote:
> > On Fri, 2025-12-05 at 16:16 -0600, Bjorn Helgaas wrote:
> > > [+to Philipp, Thomas for MSI devres question]
> > > 
> > > On Fri, Dec 05, 2025 at 09:34:54AM +0000, Russell King (Oracle) wrote:
> > > > On Fri, Dec 05, 2025 at 05:31:34AM +0000, Yao Zi wrote:
> > > > > On Mon, Nov 24, 2025 at 07:06:12PM +0000, Russell King (Oracle) wrote:
> > > > > > On Mon, Nov 24, 2025 at 04:32:10PM +0000, Yao Zi wrote:
> 
> ...
> 
> > > > This looks very non-intuitive, and the documentation for
> > > > pci_alloc_irq_vectors() doesn't help:
> > > > 
> > > >  * Upon a successful allocation, the caller should use pci_irq_vector()
> > > >  * to get the Linux IRQ number to be passed to request_threaded_irq().
> > > >  * The driver must call pci_free_irq_vectors() on cleanup.
> > > >    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > > > 
> > > > because if what you say is correct (and it looks like it is) then this
> > > > line is blatently incorrect.
> > 
> > True, this line is false. It should probably state "If you didn't
> > enable your PCI device with pcim_enable_device(), you must call
> > pci_free_irq_vectors() on cleanup."
> > 
> > If it's not a bug, one could keep the docu that way or at least phrase
> > it in a way so that no additional users start relying on that hybrid
> > mechanism.
> 
> Thanks for the clarification, would you mind me sending a patch to fix
> the description, and also mention the automatic clean-up behavior
> shouldn't be relied anymore in new code?

If I would mind if *you* send such a patch? I for sure wouldn't mind,
that's the entire idea of open source development ^^

I can of course review it if you +Cc me.

> 
> ...
> 
> > The good news is that it's the last remainder of PCI hybrid devres and
> > getting rid of it would allow for removal of some additional code, too
> > (e.g., is_enabled bit and pcim_pin_device()).
> > 
> > The bad news is that it's not super trivial to remove. I looked into it
> > about two times and decided I can't invest that time currently. You
> > need to go over all drivers again to see who uses pcim_enable_device(),
> > then add free_irq_vecs() for them all and so on…
> 
> Do you think adding an implementation of pcim_alloc_irq_vectors(), that
> always call pci_free_irq_vectors() regardless whether the PCI device is
> managed, will help the conversion?
> 
> This will make it more trival to rewrite drivers depending on the
> automatic clean-up behavior: since calling pci_free_irq_vectors()
> several times is okay, we could simply change pci_alloc_irq_vectors() to
> pcim_alloc_irq_vectors(), without considering where to call
> pci_free_irq_vectors().
> 
> Introducing pcim_alloc_irq_vectors() will also help newly-introduced
> drivers to reduce duplicated code to handle resource clean-up.

That's in fact how I have cleaned up the hybrid nature that was present
until this year in pci_request_region() et al.:

https://lore.kernel.org/linux-pci/20250519112959.25487-2-phasta@kernel.org/

It's one way to do it. First port everyone who relies on managed
behavior to a pcim_ function, and once they're all ported, remove the
hybrid nature from the pci_ function.

So that works, yes. The real question is just whether one wants a pcim_
function for the irq vectors. My personal impression is that this looks
like a useful feature; but my expertise with MSI is a bit limited.
There's also this strange kernel-wide msi_enabled global bool.. 

I guess the best way to find out is to try implementing it.

In case of doubt, the boring unmanaged pci_ version is the safe choice.
One contributor around here has once called the managed versions "the
crazy devres voodoo" :p

(BTW, just to be sure, pcim_ functions must not be interconnected with
pcim_enable() in the future anymore, nor shall they use global state.
All PCI devres functionality should purely work on the device file
through pure devres. The pcim_enable() interconnection cam only from
the hybrid feature.)


P.

> 
> > If you give me a pointer I can provide a TODO entry. In any case, feel
> > free to set me as a reviewer!
> 
> > Regards
> > Philipp
> 
> Regards,
> Yao Zi

Powered by blists - more mailing lists