[PATCH] pci/controller: Use dev_fwnode()

[PATCH] pci/controller: Use dev_fwnode()

[Jiri Slaby (SUSE)]

irq_domain_create_simple() takes fwnode as the first argument. It can be
extracted from the struct device using dev_fwnode() helper instead of
using of_node with of_fwnode_handle().

So use the dev_fwnode() helper.

Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: linux-pci@vger.kernel.org
---
 drivers/pci/controller/mobiveil/pcie-mobiveil-host.c | 5 ++---
 drivers/pci/controller/pcie-mediatek-gen3.c          | 3 +--
 2 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c b/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
index a600f46ee3c3..cd44ddb698ea 100644
--- a/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
+++ b/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
@@ -464,9 +464,8 @@ static int mobiveil_pcie_init_irq_domain(struct mobiveil_pcie *pcie)
 	struct mobiveil_root_port *rp = &pcie->rp;
 
 	/* setup INTx */
-	rp->intx_domain = irq_domain_create_linear(of_fwnode_handle(dev->of_node), PCI_NUM_INTX,
-						   &intx_domain_ops, pcie);
-
+	rp->intx_domain = irq_domain_create_linear(dev_fwnode(dev), PCI_NUM_INTX, &intx_domain_ops,
+						   pcie);
 	if (!rp->intx_domain) {
 		dev_err(dev, "Failed to get a INTx IRQ domain\n");
 		return -ENOMEM;
diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
index b55f5973414c..5464b4ae5c20 100644
--- a/drivers/pci/controller/pcie-mediatek-gen3.c
+++ b/drivers/pci/controller/pcie-mediatek-gen3.c
@@ -756,8 +756,7 @@ static int mtk_pcie_init_irq_domains(struct mtk_gen3_pcie *pcie)
 	/* Setup MSI */
 	mutex_init(&pcie->lock);
 
-	pcie->msi_bottom_domain = irq_domain_create_linear(of_fwnode_handle(node),
-							   PCIE_MSI_IRQS_NUM,
+	pcie->msi_bottom_domain = irq_domain_create_linear(dev_fwnode(dev), PCIE_MSI_IRQS_NUM,
 							   &mtk_msi_bottom_domain_ops, pcie);
 	if (!pcie->msi_bottom_domain) {
 		dev_err(dev, "failed to create MSI bottom domain\n");
-- 
2.49.0

Re: (subset) [PATCH] pci/controller: Use dev_fwnode()

[Manivannan Sadhasivam]

On Wed, 11 Jun 2025 12:43:44 +0200, Jiri Slaby (SUSE) wrote:
> irq_domain_create_simple() takes fwnode as the first argument. It can be
> extracted from the struct device using dev_fwnode() helper instead of
> using of_node with of_fwnode_handle().
> 
> So use the dev_fwnode() helper.
> 
> 
> [...]

Applied, thanks!

[1/1] pci/controller: Use dev_fwnode()
      commit: 91afe85f4b282712231df9061a4528db51ce3c3b

Best regards,
-- 
Manivannan Sadhasivam <mani@kernel.org>

Re: [PATCH] pci/controller: Use dev_fwnode()

[Bjorn Helgaas]

[+cc Nam]

On Wed, Jun 11, 2025 at 12:43:44PM +0200, Jiri Slaby (SUSE) wrote:
> irq_domain_create_simple() takes fwnode as the first argument. It can be
> extracted from the struct device using dev_fwnode() helper instead of
> using of_node with of_fwnode_handle().
> 
> So use the dev_fwnode() helper.
> 
> Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: linux-pci@vger.kernel.org
> ---
>  drivers/pci/controller/mobiveil/pcie-mobiveil-host.c | 5 ++---
>  drivers/pci/controller/pcie-mediatek-gen3.c          | 3 +--

I think the pcie-mediatek-gen3.c part of this is no longer relevant
after Nam's series [1].  This pcie-mediatek-gen3.c was the only thing
on the pci/controller/mediatek-gen3 branch, so I'm going to drop that
for now.

The pcie-mobiveil-host.c part is still queued on
pci/controller/mobiveil for v6.17.

[1] https://patch.msgid.link/bfbd2e375269071b69e1aa85e629ee4b7c99518f.1750858083.git.namcao@linutronix.de

>  2 files changed, 3 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c b/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
> index a600f46ee3c3..cd44ddb698ea 100644
> --- a/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
> +++ b/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c
> @@ -464,9 +464,8 @@ static int mobiveil_pcie_init_irq_domain(struct mobiveil_pcie *pcie)
>  	struct mobiveil_root_port *rp = &pcie->rp;
>  
>  	/* setup INTx */
> -	rp->intx_domain = irq_domain_create_linear(of_fwnode_handle(dev->of_node), PCI_NUM_INTX,
> -						   &intx_domain_ops, pcie);
> -
> +	rp->intx_domain = irq_domain_create_linear(dev_fwnode(dev), PCI_NUM_INTX, &intx_domain_ops,
> +						   pcie);
>  	if (!rp->intx_domain) {
>  		dev_err(dev, "Failed to get a INTx IRQ domain\n");
>  		return -ENOMEM;
> diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> index b55f5973414c..5464b4ae5c20 100644
> --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> @@ -756,8 +756,7 @@ static int mtk_pcie_init_irq_domains(struct mtk_gen3_pcie *pcie)
>  	/* Setup MSI */
>  	mutex_init(&pcie->lock);
>  
> -	pcie->msi_bottom_domain = irq_domain_create_linear(of_fwnode_handle(node),
> -							   PCIE_MSI_IRQS_NUM,
> +	pcie->msi_bottom_domain = irq_domain_create_linear(dev_fwnode(dev), PCIE_MSI_IRQS_NUM,
>  							   &mtk_msi_bottom_domain_ops, pcie);
>  	if (!pcie->msi_bottom_domain) {
>  		dev_err(dev, "failed to create MSI bottom domain\n");
> -- 
> 2.49.0
> 

Re: [PATCH] pci/controller: Use dev_fwnode()

[Nam Cao]

On Tue, Jul 15, 2025 at 01:49:17PM -0500, Bjorn Helgaas wrote:
> On Wed, Jun 11, 2025 at 12:43:44PM +0200, Jiri Slaby (SUSE) wrote:
> > irq_domain_create_simple() takes fwnode as the first argument. It can be
> > extracted from the struct device using dev_fwnode() helper instead of
> > using of_node with of_fwnode_handle().
> > 
> > So use the dev_fwnode() helper.
> > 
> > Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> > Cc: Bjorn Helgaas <bhelgaas@google.com>
> > Cc: linux-pci@vger.kernel.org
> > ---
> >  drivers/pci/controller/mobiveil/pcie-mobiveil-host.c | 5 ++---
> >  drivers/pci/controller/pcie-mediatek-gen3.c          | 3 +--
> 
> I think the pcie-mediatek-gen3.c part of this is no longer relevant
> after Nam's series [1].

fwnode is still needed after my patch. As part of
struct irq_domain_info info = { ... }

You could squash this one into my patch. I personally would leave it be.
But fine to me either way.

> This pcie-mediatek-gen3.c was the only thing on the
> pci/controller/mediatek-gen3 branch, so I'm going to drop that for now.
> 
> The pcie-mobiveil-host.c part is still queued on
> pci/controller/mobiveil for v6.17.
> 
> [1] https://patch.msgid.link/bfbd2e375269071b69e1aa85e629ee4b7c99518f.1750858083.git.namcao@linutronix.de

Best regards,
Nam

Re: [PATCH] pci/controller: Use dev_fwnode()

[Bjorn Helgaas]

[+cc Arnd]

On Wed, Jul 16, 2025 at 09:59:42AM +0200, Nam Cao wrote:
> On Tue, Jul 15, 2025 at 01:49:17PM -0500, Bjorn Helgaas wrote:
> > On Wed, Jun 11, 2025 at 12:43:44PM +0200, Jiri Slaby (SUSE) wrote:
> > > irq_domain_create_simple() takes fwnode as the first argument. It can be
> > > extracted from the struct device using dev_fwnode() helper instead of
> > > using of_node with of_fwnode_handle().
> > > 
> > > So use the dev_fwnode() helper.
> > > 
> > > Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> > > Cc: Bjorn Helgaas <bhelgaas@google.com>
> > > Cc: linux-pci@vger.kernel.org
> > > ---
> > >  drivers/pci/controller/mobiveil/pcie-mobiveil-host.c | 5 ++---
> > >  drivers/pci/controller/pcie-mediatek-gen3.c          | 3 +--
> > 
> > I think the pcie-mediatek-gen3.c part of this is no longer relevant
> > after Nam's series [1].
> 
> fwnode is still needed after my patch. As part of
> struct irq_domain_info info = { ... }
> 
> You could squash this one into my patch. I personally would leave it be.
> But fine to me either way.

Oh, I think I see what happened:

  - Jiri replaced of_fwnode_handle() with dev_fwnode() in the
    irq_domain_create_linear() call [1]

  - On top of that, Nam replaced irq_domain_create_linear() with
    msi_create_parent_irq_domain(), and moved the dev_fwnode() to the
    struct irq_domain_info [2]

  - I rebuilt pci/next with Nam's series merged *before* Jiri's,
    resulting in a conflict (of_fwnode_handle() was still used in the
    irq_domain_create_linear() call) which I resolved by using
    dev_fwnode() when building struct irq_domain_info [3]

I think the result [4] is OK, but it's not ideal because a
dev_fwnode() conversion got inserted into Nam's patch without
explanation.

So I think I'll put Jiri's patches (along with Arnd's similar altera
patch [5]) on a branch and merge them before Nam's.

Jiri, question for you: even after all this, there are still several
uses in drivers/pci/ of of_fwnode_handle() to extract the
fwnode_handle for a struct device * [6,7,8,9,10,11,12].

Should these also be changed?

Bjorn

[1] https://lore.kernel.org/r/20250611104348.192092-16-jirislaby@kernel.org
[2] https://patch.msgid.link/bfbd2e375269071b69e1aa85e629ee4b7c99518f.1750858083.git.namcao@linutronix.de
[3] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/commit/?id=dd6fad415071
[4] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-mediatek-gen3.c?id=beedc9eb3114#n750
[5] https://lore.kernel.org/r/20250611104348.192092-2-jirislaby@kernel.org

[6] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/dwc/pcie-designware-host.c?id=beedc9eb3114#n217
[7] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c?id=beedc9eb3114#n442
[8] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-altera-msi.c?id=beedc9eb3114#n169
[9] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-mediatek.c?id=beedc9eb3114#n490
[10] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-xilinx-dma-pl.c?id=beedc9eb3114#n468
[11] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-xilinx-nwl.c?id=beedc9eb3114#n501
[12] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/plda/pcie-plda-host.c?id=beedc9eb3114#n156

Re: [PATCH] pci/controller: Use dev_fwnode()

[Bjorn Helgaas]

Jiri, question for you below about more possible drivers/pci/
conversions to use dev_fwnode() for struct device * cases.

Would like to get this in for v6.17 if these should be changed.

Bjorn

On Wed, Jul 16, 2025 at 09:49:38AM -0500, Bjorn Helgaas wrote:
> On Wed, Jul 16, 2025 at 09:59:42AM +0200, Nam Cao wrote:
> > On Tue, Jul 15, 2025 at 01:49:17PM -0500, Bjorn Helgaas wrote:
> > > On Wed, Jun 11, 2025 at 12:43:44PM +0200, Jiri Slaby (SUSE) wrote:
> > > > irq_domain_create_simple() takes fwnode as the first argument. It can be
> > > > extracted from the struct device using dev_fwnode() helper instead of
> > > > using of_node with of_fwnode_handle().
> > > > 
> > > > So use the dev_fwnode() helper.
> > > > 
> > > > Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> > > > Cc: Bjorn Helgaas <bhelgaas@google.com>
> > > > Cc: linux-pci@vger.kernel.org
> > > > ---
> > > >  drivers/pci/controller/mobiveil/pcie-mobiveil-host.c | 5 ++---
> > > >  drivers/pci/controller/pcie-mediatek-gen3.c          | 3 +--
> > > 
> > > I think the pcie-mediatek-gen3.c part of this is no longer relevant
> > > after Nam's series [1].
> > 
> > fwnode is still needed after my patch. As part of
> > struct irq_domain_info info = { ... }
> > 
> > You could squash this one into my patch. I personally would leave it be.
> > But fine to me either way.
> 
> Oh, I think I see what happened:
> 
>   - Jiri replaced of_fwnode_handle() with dev_fwnode() in the
>     irq_domain_create_linear() call [1]
> 
>   - On top of that, Nam replaced irq_domain_create_linear() with
>     msi_create_parent_irq_domain(), and moved the dev_fwnode() to the
>     struct irq_domain_info [2]
> 
>   - I rebuilt pci/next with Nam's series merged *before* Jiri's,
>     resulting in a conflict (of_fwnode_handle() was still used in the
>     irq_domain_create_linear() call) which I resolved by using
>     dev_fwnode() when building struct irq_domain_info [3]
> 
> I think the result [4] is OK, but it's not ideal because a
> dev_fwnode() conversion got inserted into Nam's patch without
> explanation.
> 
> So I think I'll put Jiri's patches (along with Arnd's similar altera
> patch [5]) on a branch and merge them before Nam's.
> 
> Jiri, question for you: even after all this, there are still several
> uses in drivers/pci/ of of_fwnode_handle() to extract the
> fwnode_handle for a struct device * [6,7,8,9,10,11,12].
> 
> Should these also be changed?
> 
> Bjorn
> 
> [1] https://lore.kernel.org/r/20250611104348.192092-16-jirislaby@kernel.org
> [2] https://patch.msgid.link/bfbd2e375269071b69e1aa85e629ee4b7c99518f.1750858083.git.namcao@linutronix.de
> [3] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/commit/?id=dd6fad415071
> [4] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-mediatek-gen3.c?id=beedc9eb3114#n750
> [5] https://lore.kernel.org/r/20250611104348.192092-2-jirislaby@kernel.org
> 
> [6] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/dwc/pcie-designware-host.c?id=beedc9eb3114#n217
> [7] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/mobiveil/pcie-mobiveil-host.c?id=beedc9eb3114#n442
> [8] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-altera-msi.c?id=beedc9eb3114#n169
> [9] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-mediatek.c?id=beedc9eb3114#n490
> [10] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-xilinx-dma-pl.c?id=beedc9eb3114#n468
> [11] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/pcie-xilinx-nwl.c?id=beedc9eb3114#n501
> [12] https://git.kernel.org/cgit/linux/kernel/git/pci/pci.git/tree/drivers/pci/controller/plda/pcie-plda-host.c?id=beedc9eb3114#n156

Re: [PATCH] pci/controller: Use dev_fwnode()

[Jiri Slaby]

On 21. 07. 25, 19:08, Bjorn Helgaas wrote:
> Jiri, question for you below about more possible drivers/pci/
> conversions to use dev_fwnode() for struct device * cases.

Sorry, I am a way too occupied :/.

> Would like to get this in for v6.17 if these should be changed.

It's not necessary, but a good to have cleanup (opposed to the posted 
fixes, which were required). I will switch those eventually, but I don't 
promise 6.17. (If someone does not beat me to it.)

thanks,
-- 
js
suse labs

Re: [PATCH] pci/controller: Use dev_fwnode()

[Bjorn Helgaas]

On Tue, Jul 22, 2025 at 08:24:26AM +0200, Jiri Slaby wrote:
> On 21. 07. 25, 19:08, Bjorn Helgaas wrote:
> > Jiri, question for you below about more possible drivers/pci/
> > conversions to use dev_fwnode() for struct device * cases.
> 
> Sorry, I am a way too occupied :/.
> 
> > Would like to get this in for v6.17 if these should be changed.
> 
> It's not necessary, but a good to have cleanup (opposed to the posted fixes,
> which were required). I will switch those eventually, but I don't promise
> 6.17. (If someone does not beat me to it.)

It's not clear from the commit log:

  irq_domain_create_simple() takes fwnode as the first argument. It can be
  extracted from the struct device using dev_fwnode() helper instead of
  using of_node with of_fwnode_handle().

  So use the dev_fwnode() helper.

why the posted fixes are required (other than Arnd's change to
altera_pcie_init_irq_domain(), which fixes an unused variable warning
when CONFIG_OF is not enabled).

Since it sounds like no changes are required for the other ones I
mentioned, I'm going to leave them alone for now:

  dw_pcie_allocate_domains()
  mobiveil_allocate_msi_domains()
  altera_allocate_domains()
  mtk_pcie_allocate_msi_domains()
  xilinx_pl_dma_pcie_init_msi_irq_domain()
  nwl_pcie_init_msi_irq_domain()
  plda_allocate_msi_domains()

Bjorn

Re: [PATCH] pci/controller: Use dev_fwnode()

[Jiri Slaby]

On 23. 07. 25, 1:20, Bjorn Helgaas wrote:
> On Tue, Jul 22, 2025 at 08:24:26AM +0200, Jiri Slaby wrote:
>> On 21. 07. 25, 19:08, Bjorn Helgaas wrote:
>>> Jiri, question for you below about more possible drivers/pci/
>>> conversions to use dev_fwnode() for struct device * cases.
>>
>> Sorry, I am a way too occupied :/.
>>
>>> Would like to get this in for v6.17 if these should be changed.
>>
>> It's not necessary, but a good to have cleanup (opposed to the posted fixes,
>> which were required). I will switch those eventually, but I don't promise
>> 6.17. (If someone does not beat me to it.)
> 
> It's not clear from the commit log:
> 
>    irq_domain_create_simple() takes fwnode as the first argument. It can be
>    extracted from the struct device using dev_fwnode() helper instead of
>    using of_node with of_fwnode_handle().
> 
>    So use the dev_fwnode() helper.
> 
> why the posted fixes are required (other than Arnd's change to
> altera_pcie_init_irq_domain(), which fixes an unused variable warning
> when CONFIG_OF is not enabled).

Sorry, my bad. These are a cleanup suggested in this series:
https://lore.kernel.org/all/4bc0e1ca-a523-424a-8759-59e353317fba@kernel.org/

I.e. series switching from irq_domain_add_*() (take of_node) to 
irq_domain_create_*() (take fwnode).

These days, fwnode is preferred and if there were no more users of 
of_node in changed functions, the series above even produced warnings 
(Arnd's and others' fixes).

> Since it sounds like no changes are required for the other ones I
> mentioned, I'm going to leave them alone for now:
> 
>    dw_pcie_allocate_domains()
>    mobiveil_allocate_msi_domains()
>    altera_allocate_domains()
>    mtk_pcie_allocate_msi_domains()
>    xilinx_pl_dma_pcie_init_msi_irq_domain()
>    nwl_pcie_init_msi_irq_domain()
>    plda_allocate_msi_domains()

Given fwnode is always used them, it's not necessary to use 
dev_fwnode(). But it'd be a nice cleanup. Provided the list, I started 
the cleanup now :).

thanks,
-- 
js
suse labs