| Message ID | 20211223173015.22251-7-prabhakar.mahadev-lad.rj@bp.renesas.com (mailing list archive) |
|---|---|
| State | New |
| Delegated to: | Kieran Bingham |
| Headers | show |
| Series | media: Use platform_get_irq*() variants to fetch IRQ's | expand |
On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: > > platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static > allocation of IRQ resources in DT core code, this causes an issue > when using hierarchical interrupt domains using "interrupts" property > in the node as this bypasses the hierarchical setup and messes up the > irq chaining. > > In preparation for removal of static setup of IRQ resource from DT core > code use platform_get_irq_optional(). > > Also this patch propagates error code in case devm_request_irq() > fails instead of returing -EINVAL. returning ... > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > + res_irq->start = irq; > + res_irq->end = irq; > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; If you convert DEFINE_RES_NAMED() to return a compound literal, then you may use it here like res_irq = DEFINE_RES_NAMED(...); or even do like this if (dev_of_node(...)) res_irq = DEFINE_RES_IRQ_NAMED(...) else res_irq = DEFINE_RES_IRQ(...); res_irq->flags |= irq_get_trigger_type(irq); I'm not sure why you can't simply use the NAMED variant in both cases (yes, I see that of_node_full_name() will return something, not NULL). ... > + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { > + if (err < 0) > + goto vpif_unregister; Needs a better error checking, i.e. consider 0 as no-IRQ (equivalent to -ENXIO (note, OF code never returns 0 as valid vIRQ). > res_idx++; > } ... > + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { > + if (err < 0) > + goto vpif_unregister; Ditto.
Hi Andy, Thank you for the review. On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: > > > > platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static > > allocation of IRQ resources in DT core code, this causes an issue > > when using hierarchical interrupt domains using "interrupts" property > > in the node as this bypasses the hierarchical setup and messes up the > > irq chaining. > > > > In preparation for removal of static setup of IRQ resource from DT core > > code use platform_get_irq_optional(). > > > > Also this patch propagates error code in case devm_request_irq() > > fails instead of returing -EINVAL. > > returning > > ... > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > + res_irq->start = irq; > > + res_irq->end = irq; > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > you may use it here like > > res_irq = DEFINE_RES_NAMED(...); > > or even do like this > > if (dev_of_node(...)) > res_irq = DEFINE_RES_IRQ_NAMED(...) > else > res_irq = DEFINE_RES_IRQ(...); > res_irq->flags |= irq_get_trigger_type(irq); > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() changing this macos just for this single user tree wide doesn't make sense. Let me know if you think otherwise. > I'm not sure why you can't simply use the NAMED variant in both cases > (yes, I see that of_node_full_name() will return something, not NULL). > > ... > > > + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { > > + if (err < 0) > > + goto vpif_unregister; > > Needs a better error checking, i.e. consider 0 as no-IRQ (equivalent > to -ENXIO (note, OF code never returns 0 as valid vIRQ). > Will fix that. > > res_idx++; > > } > > ... > > > + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { > > + if (err < 0) > > + goto vpif_unregister; > > Ditto. > Will fix that. Cheers, Prabhakar
On Tue, Jan 4, 2022 at 7:23 PM Lad, Prabhakar <prabhakar.csengg@gmail.com> wrote: > On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: ... > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > > + res_irq->start = irq; > > > + res_irq->end = irq; > > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > > you may use it here like > > > > res_irq = DEFINE_RES_NAMED(...); > > > > or even do like this > > > > if (dev_of_node(...)) > > res_irq = DEFINE_RES_IRQ_NAMED(...) > > else > > res_irq = DEFINE_RES_IRQ(...); > > res_irq->flags |= irq_get_trigger_type(irq); > > > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() > changing this macos just for this single user tree wide doesn't make > sense. Let me know if you think otherwise. Converting them to produce compound literal is straightforward and does not require changes in the users. But on the other hand it allows you to use it and convert existing users to use that form directly. You may conduct research on how macros in the property.h were morphing towards that. > > I'm not sure why you can't simply use the NAMED variant in both cases > > (yes, I see that of_node_full_name() will return something, not NULL).
Hi Andy, On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Tue, Jan 4, 2022 at 7:23 PM Lad, Prabhakar > <prabhakar.csengg@gmail.com> wrote: > > On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > > > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: > > ... > > > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > > > + res_irq->start = irq; > > > > + res_irq->end = irq; > > > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > > > > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > > > you may use it here like > > > > > > res_irq = DEFINE_RES_NAMED(...); > > > > > > or even do like this > > > > > > if (dev_of_node(...)) > > > res_irq = DEFINE_RES_IRQ_NAMED(...) > > > else > > > res_irq = DEFINE_RES_IRQ(...); > > > res_irq->flags |= irq_get_trigger_type(irq); > > > > > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() > > changing this macos just for this single user tree wide doesn't make > > sense. Let me know if you think otherwise. > > Converting them to produce compound literal is straightforward and > does not require changes in the users. But on the other hand it allows > you to use it and convert existing users to use that form directly. > You may conduct research on how macros in the property.h were morphing > towards that. > Thank you for the pointer. I did the below change for this. diff --git a/include/linux/ioport.h b/include/linux/ioport.h index 8359c50f9988..da1208e8f164 100644 --- a/include/linux/ioport.h +++ b/include/linux/ioport.h @@ -153,7 +153,7 @@ enum { /* helpers to define resources */ #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ - { \ + (struct resource) { \ .start = (_start), \ .end = (_start) + (_size) - 1, \ .name = (_name), \ But there are some instances which need to be touched, for example vexpress-sysreg.c [1]. Are you OK with files to be changed? [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mfd/vexpress-sysreg.c?h=v5.16-rc8#n65 Cheers, Prabhakar
On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar <prabhakar.csengg@gmail.com> wrote: > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > On Tue, Jan 4, 2022 at 7:23 PM Lad, Prabhakar > > <prabhakar.csengg@gmail.com> wrote: > > > On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko > > > <andy.shevchenko@gmail.com> wrote: > > > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > > > > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: > > > > ... > > > > > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > > > > + res_irq->start = irq; > > > > > + res_irq->end = irq; > > > > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > > > > > > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > > > > you may use it here like > > > > > > > > res_irq = DEFINE_RES_NAMED(...); > > > > > > > > or even do like this > > > > > > > > if (dev_of_node(...)) > > > > res_irq = DEFINE_RES_IRQ_NAMED(...) > > > > else > > > > res_irq = DEFINE_RES_IRQ(...); > > > > res_irq->flags |= irq_get_trigger_type(irq); > > > > > > > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() > > > changing this macos just for this single user tree wide doesn't make > > > sense. Let me know if you think otherwise. > > > > Converting them to produce compound literal is straightforward and > > does not require changes in the users. But on the other hand it allows > > you to use it and convert existing users to use that form directly. > > You may conduct research on how macros in the property.h were morphing > > towards that. > > > Thank you for the pointer. I did the below change for this. > > diff --git a/include/linux/ioport.h b/include/linux/ioport.h > index 8359c50f9988..da1208e8f164 100644 > --- a/include/linux/ioport.h > +++ b/include/linux/ioport.h > @@ -153,7 +153,7 @@ enum { > > /* helpers to define resources */ > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > - { \ > + (struct resource) { \ Yep, that's it. > .start = (_start), \ > .end = (_start) + (_size) - 1, \ > .name = (_name), \ > > But there are some instances which need to be touched, for example > vexpress-sysreg.c [1]. Are you OK with files to be changed? Nice! That's exactly my point and you can sell it to the community because there are already users of it like this. Yes, I'm fine, but it seems it needs to be done treewide in one patch. Btw, how many of those already in use? > [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mfd/vexpress-sysreg.c?h=v5.16-rc8#n65
On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > <prabhakar.csengg@gmail.com> wrote: > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > On Tue, Jan 4, 2022 at 7:23 PM Lad, Prabhakar > > > <prabhakar.csengg@gmail.com> wrote: > > > > On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko > > > > <andy.shevchenko@gmail.com> wrote: > > > > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > > > > > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: ... > > > > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > > > > > + res_irq->start = irq; > > > > > > + res_irq->end = irq; > > > > > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > > > > > > > > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > > > > > you may use it here like > > > > > > > > > > res_irq = DEFINE_RES_NAMED(...); > > > > > > > > > > or even do like this > > > > > > > > > > if (dev_of_node(...)) > > > > > res_irq = DEFINE_RES_IRQ_NAMED(...) > > > > > else > > > > > res_irq = DEFINE_RES_IRQ(...); > > > > > res_irq->flags |= irq_get_trigger_type(irq); > > > > > > > > > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() > > > > changing this macos just for this single user tree wide doesn't make > > > > sense. Let me know if you think otherwise. > > > > > > Converting them to produce compound literal is straightforward and > > > does not require changes in the users. But on the other hand it allows > > > you to use it and convert existing users to use that form directly. > > > You may conduct research on how macros in the property.h were morphing > > > towards that. > > > > > Thank you for the pointer. I did the below change for this. > > > > diff --git a/include/linux/ioport.h b/include/linux/ioport.h > > index 8359c50f9988..da1208e8f164 100644 > > --- a/include/linux/ioport.h > > +++ b/include/linux/ioport.h > > @@ -153,7 +153,7 @@ enum { > > > > /* helpers to define resources */ > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > - { \ > > + (struct resource) { \ > > Yep, that's it. > > > .start = (_start), \ > > .end = (_start) + (_size) - 1, \ > > .name = (_name), \ > > > > But there are some instances which need to be touched, for example > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > Nice! That's exactly my point and you can sell it to the community > because there are already users of it like this. > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > Btw, how many of those already in use? Actually you don't need to change that. It's an array of resources and everything should be kept as is there. > > [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mfd/vexpress-sysreg.c?h=v5.16-rc8#n65
Hi Andy, On Thu, Jan 6, 2022 at 2:15 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > > <prabhakar.csengg@gmail.com> wrote: > > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > > > <andy.shevchenko@gmail.com> wrote: > > > > On Tue, Jan 4, 2022 at 7:23 PM Lad, Prabhakar > > > > <prabhakar.csengg@gmail.com> wrote: > > > > > On Sat, Dec 25, 2021 at 5:32 PM Andy Shevchenko > > > > > <andy.shevchenko@gmail.com> wrote: > > > > > > On Sat, Dec 25, 2021 at 3:04 AM Lad Prabhakar > > > > > > <prabhakar.mahadev-lad.rj@bp.renesas.com> wrote: > > ... > > > > > > > > + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); > > > > > > > + res_irq->start = irq; > > > > > > > + res_irq->end = irq; > > > > > > > + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; > > > > > > > > > > > > If you convert DEFINE_RES_NAMED() to return a compound literal, then > > > > > > you may use it here like > > > > > > > > > > > > res_irq = DEFINE_RES_NAMED(...); > > > > > > > > > > > > or even do like this > > > > > > > > > > > > if (dev_of_node(...)) > > > > > > res_irq = DEFINE_RES_IRQ_NAMED(...) > > > > > > else > > > > > > res_irq = DEFINE_RES_IRQ(...); > > > > > > res_irq->flags |= irq_get_trigger_type(irq); > > > > > > > > > > > There are quite a few users of DEFINE_RES_IRQ_NAMED()/DEFINE_RES_IRQ() > > > > > changing this macos just for this single user tree wide doesn't make > > > > > sense. Let me know if you think otherwise. > > > > > > > > Converting them to produce compound literal is straightforward and > > > > does not require changes in the users. But on the other hand it allows > > > > you to use it and convert existing users to use that form directly. > > > > You may conduct research on how macros in the property.h were morphing > > > > towards that. > > > > > > > Thank you for the pointer. I did the below change for this. > > > > > > diff --git a/include/linux/ioport.h b/include/linux/ioport.h > > > index 8359c50f9988..da1208e8f164 100644 > > > --- a/include/linux/ioport.h > > > +++ b/include/linux/ioport.h > > > @@ -153,7 +153,7 @@ enum { > > > > > > /* helpers to define resources */ > > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > > - { \ > > > + (struct resource) { \ > > > > Yep, that's it. > > > > > .start = (_start), \ > > > .end = (_start) + (_size) - 1, \ > > > .name = (_name), \ > > > > > > But there are some instances which need to be touched, for example > > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > > > Nice! That's exactly my point and you can sell it to the community > > because there are already users of it like this. > > > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > > Btw, how many of those already in use? > > Actually you don't need to change that. It's an array of resources and > everything should be kept as is there. > I do get below build failures, with the above literal change for vexpress-sysreg.c. drivers/mfd/vexpress-sysreg.c: At top level: drivers/mfd/vexpress-sysreg.c:64:37: error: initialiser element is not constant 64 | .resources = (struct resource []) { | ^ drivers/mfd/vexpress-sysreg.c:64:37: note: (near initialisation for ‘vexpress_sysreg_cells[0]’) drivers/mfd/vexpress-sysreg.c:73:37: error: initialiser element is not constant 73 | .resources = (struct resource []) { | ^ drivers/mfd/vexpress-sysreg.c:73:37: note: (near initialisation for ‘vexpress_sysreg_cells[1]’) drivers/mfd/vexpress-sysreg.c:82:37: error: initialiser element is not constant 82 | .resources = (struct resource []) { | ^ drivers/mfd/vexpress-sysreg.c:82:37: note: (near initialisation for ‘vexpress_sysreg_cells[2]’) drivers/mfd/vexpress-sysreg.c:90:37: error: initialiser element is not constant 90 | .resources = (struct resource []) { | ^ drivers/mfd/vexpress-sysreg.c:90:37: note: (near initialisation for ‘vexpress_sysreg_cells[3]’) drivers/mfd/vexpress-sysreg.c:93:2: warning: missing initialiser for field ‘ignore_resource_conflicts’ of ‘struct mfd_cell’ [-Wmissing-field-initializers] 93 | } Cheers, Prabhakar
On Thu, Jan 6, 2022 at 5:27 PM Lad, Prabhakar <prabhakar.csengg@gmail.com> wrote: > On Thu, Jan 6, 2022 at 2:15 PM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > > > <prabhakar.csengg@gmail.com> wrote: > > > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko ... > > > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > > > - { \ > > > > + (struct resource) { \ > > > > > > Yep, that's it. > > > > > > > .start = (_start), \ > > > > .end = (_start) + (_size) - 1, \ > > > > .name = (_name), \ > > > > > > > > But there are some instances which need to be touched, for example > > > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > > > > > Nice! That's exactly my point and you can sell it to the community > > > because there are already users of it like this. > > > > > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > > > Btw, how many of those already in use? > > > > Actually you don't need to change that. It's an array of resources and > > everything should be kept as is there. > > > I do get below build failures, with the above literal change for > vexpress-sysreg.c. > > drivers/mfd/vexpress-sysreg.c: At top level: > drivers/mfd/vexpress-sysreg.c:64:37: error: initialiser element is not constant > 64 | .resources = (struct resource []) { > | ^ > drivers/mfd/vexpress-sysreg.c:64:37: note: (near initialisation for > ‘vexpress_sysreg_cells[0]’) > drivers/mfd/vexpress-sysreg.c:73:37: error: initialiser element is not constant > 73 | .resources = (struct resource []) { > | ^ > drivers/mfd/vexpress-sysreg.c:73:37: note: (near initialisation for > ‘vexpress_sysreg_cells[1]’) > drivers/mfd/vexpress-sysreg.c:82:37: error: initialiser element is not constant > 82 | .resources = (struct resource []) { > | ^ > drivers/mfd/vexpress-sysreg.c:82:37: note: (near initialisation for > ‘vexpress_sysreg_cells[2]’) > drivers/mfd/vexpress-sysreg.c:90:37: error: initialiser element is not constant > 90 | .resources = (struct resource []) { > | ^ > drivers/mfd/vexpress-sysreg.c:90:37: note: (near initialisation for > ‘vexpress_sysreg_cells[3]’) > drivers/mfd/vexpress-sysreg.c:93:2: warning: missing initialiser for > field ‘ignore_resource_conflicts’ of ‘struct mfd_cell’ > [-Wmissing-field-initializers] > 93 | } Hmm... Interesting, so I suppose the fix is to drop (struct resource []) parts from the driver?
Hi Andy, On Thu, Jan 6, 2022 at 4:02 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 5:27 PM Lad, Prabhakar > <prabhakar.csengg@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 2:15 PM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko > > > <andy.shevchenko@gmail.com> wrote: > > > > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > > > > <prabhakar.csengg@gmail.com> wrote: > > > > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > > ... > > > > > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > > > > - { \ > > > > > + (struct resource) { \ > > > > > > > > Yep, that's it. > > > > > > > > > .start = (_start), \ > > > > > .end = (_start) + (_size) - 1, \ > > > > > .name = (_name), \ > > > > > > > > > > But there are some instances which need to be touched, for example > > > > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > > > > > > > Nice! That's exactly my point and you can sell it to the community > > > > because there are already users of it like this. > > > > > > > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > > > > Btw, how many of those already in use? > > > > > > Actually you don't need to change that. It's an array of resources and > > > everything should be kept as is there. > > > > > I do get below build failures, with the above literal change for > > vexpress-sysreg.c. > > > > drivers/mfd/vexpress-sysreg.c: At top level: > > drivers/mfd/vexpress-sysreg.c:64:37: error: initialiser element is not constant > > 64 | .resources = (struct resource []) { > > | ^ > > drivers/mfd/vexpress-sysreg.c:64:37: note: (near initialisation for > > ‘vexpress_sysreg_cells[0]’) > > drivers/mfd/vexpress-sysreg.c:73:37: error: initialiser element is not constant > > 73 | .resources = (struct resource []) { > > | ^ > > drivers/mfd/vexpress-sysreg.c:73:37: note: (near initialisation for > > ‘vexpress_sysreg_cells[1]’) > > drivers/mfd/vexpress-sysreg.c:82:37: error: initialiser element is not constant > > 82 | .resources = (struct resource []) { > > | ^ > > drivers/mfd/vexpress-sysreg.c:82:37: note: (near initialisation for > > ‘vexpress_sysreg_cells[2]’) > > drivers/mfd/vexpress-sysreg.c:90:37: error: initialiser element is not constant > > 90 | .resources = (struct resource []) { > > | ^ > > drivers/mfd/vexpress-sysreg.c:90:37: note: (near initialisation for > > ‘vexpress_sysreg_cells[3]’) > > drivers/mfd/vexpress-sysreg.c:93:2: warning: missing initialiser for > > field ‘ignore_resource_conflicts’ of ‘struct mfd_cell’ > > [-Wmissing-field-initializers] > > 93 | } > > Hmm... Interesting, so I suppose the fix is to drop (struct resource > []) parts from the driver? > A bit more than that like something below: diff --git a/drivers/mfd/vexpress-sysreg.c b/drivers/mfd/vexpress-sysreg.c index aaf24af287dd..eab82619ec31 100644 --- a/drivers/mfd/vexpress-sysreg.c +++ b/drivers/mfd/vexpress-sysreg.c @@ -61,35 +61,27 @@ static struct mfd_cell vexpress_sysreg_cells[] = { .name = "basic-mmio-gpio", .of_compatible = "arm,vexpress-sysreg,sys_led", .num_resources = 1, - .resources = (struct resource []) { - DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), - }, + .resources = &DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), .platform_data = &vexpress_sysreg_sys_led_pdata, .pdata_size = sizeof(vexpress_sysreg_sys_led_pdata), }, { .name = "basic-mmio-gpio", .of_compatible = "arm,vexpress-sysreg,sys_mci", .num_resources = 1, - .resources = (struct resource []) { - DEFINE_RES_MEM_NAMED(SYS_MCI, 0x4, "dat"), - }, + .resources = &DEFINE_RES_MEM_NAMED(SYS_MCI, 0x4, "dat"), .platform_data = &vexpress_sysreg_sys_mci_pdata, .pdata_size = sizeof(vexpress_sysreg_sys_mci_pdata), }, { .name = "basic-mmio-gpio", .of_compatible = "arm,vexpress-sysreg,sys_flash", .num_resources = 1, - .resources = (struct resource []) { - DEFINE_RES_MEM_NAMED(SYS_FLASH, 0x4, "dat"), - }, + .resources = &DEFINE_RES_MEM_NAMED(SYS_FLASH, 0x4, "dat"), .platform_data = &vexpress_sysreg_sys_flash_pdata, .pdata_size = sizeof(vexpress_sysreg_sys_flash_pdata), }, { .name = "vexpress-syscfg", .num_resources = 1, - .resources = (struct resource []) { - DEFINE_RES_MEM(SYS_MISC, 0x4c), - }, + .resources = &DEFINE_RES_MEM(SYS_MISC, 0x4c), } }; Cheers, Prabhakar
On Thu, Jan 6, 2022 at 6:11 PM Lad, Prabhakar <prabhakar.csengg@gmail.com> wrote: > On Thu, Jan 6, 2022 at 4:02 PM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 5:27 PM Lad, Prabhakar > > <prabhakar.csengg@gmail.com> wrote: > > > On Thu, Jan 6, 2022 at 2:15 PM Andy Shevchenko > > > <andy.shevchenko@gmail.com> wrote: > > > > On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko > > > > <andy.shevchenko@gmail.com> wrote: > > > > > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > > > > > <prabhakar.csengg@gmail.com> wrote: > > > > > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > > > > ... > > > > > > > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > > > > > - { \ > > > > > > + (struct resource) { \ > > > > > > > > > > Yep, that's it. > > > > > > > > > > > .start = (_start), \ > > > > > > .end = (_start) + (_size) - 1, \ > > > > > > .name = (_name), \ > > > > > > > > > > > > But there are some instances which need to be touched, for example > > > > > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > > > > > > > > > Nice! That's exactly my point and you can sell it to the community > > > > > because there are already users of it like this. > > > > > > > > > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > > > > > Btw, how many of those already in use? > > > > > > > > Actually you don't need to change that. It's an array of resources and > > > > everything should be kept as is there. > > > > > > > I do get below build failures, with the above literal change for > > > vexpress-sysreg.c. > > > > > > drivers/mfd/vexpress-sysreg.c: At top level: > > > drivers/mfd/vexpress-sysreg.c:64:37: error: initialiser element is not constant > > > 64 | .resources = (struct resource []) { > > > | ^ > > > drivers/mfd/vexpress-sysreg.c:64:37: note: (near initialisation for > > > ‘vexpress_sysreg_cells[0]’) > > Hmm... Interesting, so I suppose the fix is to drop (struct resource > > []) parts from the driver? > > > A bit more than that like something below: > - .resources = (struct resource []) { > - DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), > - }, > + .resources = &DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), This is not an equivalent change. The warning is about const qualifier. Can it rather be const struct resource [] ?
Hi Andy, On Thu, Jan 6, 2022 at 4:28 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 6:11 PM Lad, Prabhakar > <prabhakar.csengg@gmail.com> wrote: > > On Thu, Jan 6, 2022 at 4:02 PM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > On Thu, Jan 6, 2022 at 5:27 PM Lad, Prabhakar > > > <prabhakar.csengg@gmail.com> wrote: > > > > On Thu, Jan 6, 2022 at 2:15 PM Andy Shevchenko > > > > <andy.shevchenko@gmail.com> wrote: > > > > > On Thu, Jan 6, 2022 at 3:43 PM Andy Shevchenko > > > > > <andy.shevchenko@gmail.com> wrote: > > > > > > On Wed, Jan 5, 2022 at 7:41 PM Lad, Prabhakar > > > > > > <prabhakar.csengg@gmail.com> wrote: > > > > > > > On Wed, Jan 5, 2022 at 9:43 AM Andy Shevchenko > > > > > > ... > > > > > > > > > > #define DEFINE_RES_NAMED(_start, _size, _name, _flags) \ > > > > > > > - { \ > > > > > > > + (struct resource) { \ > > > > > > > > > > > > Yep, that's it. > > > > > > > > > > > > > .start = (_start), \ > > > > > > > .end = (_start) + (_size) - 1, \ > > > > > > > .name = (_name), \ > > > > > > > > > > > > > > But there are some instances which need to be touched, for example > > > > > > > vexpress-sysreg.c [1]. Are you OK with files to be changed? > > > > > > > > > > > > Nice! That's exactly my point and you can sell it to the community > > > > > > because there are already users of it like this. > > > > > > > > > > > > Yes, I'm fine, but it seems it needs to be done treewide in one patch. > > > > > > Btw, how many of those already in use? > > > > > > > > > > Actually you don't need to change that. It's an array of resources and > > > > > everything should be kept as is there. > > > > > > > > > I do get below build failures, with the above literal change for > > > > vexpress-sysreg.c. > > > > > > > > drivers/mfd/vexpress-sysreg.c: At top level: > > > > drivers/mfd/vexpress-sysreg.c:64:37: error: initialiser element is not constant > > > > 64 | .resources = (struct resource []) { > > > > | ^ > > > > drivers/mfd/vexpress-sysreg.c:64:37: note: (near initialisation for > > > > ‘vexpress_sysreg_cells[0]’) > > > > Hmm... Interesting, so I suppose the fix is to drop (struct resource > > > []) parts from the driver? > > > > > A bit more than that like something below: > > > - .resources = (struct resource []) { > > - DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), > > - }, > > + .resources = &DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), > > This is not an equivalent change. > The warning is about const qualifier. Can it rather be const struct > resource [] ? > No, since it's just a single resource, just the below should be OK. .resources = &DEFINE_RES_MEM_NAMED(SYS_LED, 0x4, "dat"), [1] https://elixir.bootlin.com/linux/v5.16-rc8/source/include/linux/mfd/core.h#L108 On the other note I could use the below without changing the macro: if (dev_of_node(...)) res_irq = (struct resource) DEFINE_RES_IRQ_NAMED(...) else res_irq = (struct resource) DEFINE_RES_IRQ(...); res_irq->flags |= irq_get_trigger_type(irq); Cheers, Prabhakar
diff --git a/drivers/media/platform/davinci/vpif.c b/drivers/media/platform/davinci/vpif.c index 5a89d885d0e3..c3c78e1afdda 100644 --- a/drivers/media/platform/davinci/vpif.c +++ b/drivers/media/platform/davinci/vpif.c @@ -20,8 +20,10 @@ #include <linux/err.h> #include <linux/init.h> #include <linux/io.h> +#include <linux/irq.h> #include <linux/kernel.h> #include <linux/module.h> +#include <linux/of.h> #include <linux/platform_device.h> #include <linux/pm_runtime.h> #include <linux/spinlock.h> @@ -428,6 +430,7 @@ static int vpif_probe(struct platform_device *pdev) static struct resource *res_irq; struct platform_device *pdev_capture, *pdev_display; struct device_node *endpoint = NULL; + int irq; vpif_base = devm_platform_ioremap_resource(pdev, 0); if (IS_ERR(vpif_base)) @@ -453,12 +456,20 @@ static int vpif_probe(struct platform_device *pdev) * For DT platforms, manually create platform_devices for * capture/display drivers. */ - res_irq = platform_get_resource(pdev, IORESOURCE_IRQ, 0); + irq = platform_get_irq(pdev, 0); + if (irq < 0) { + pm_runtime_put(&pdev->dev); + return irq; + } + res_irq = devm_kzalloc(&pdev->dev, sizeof(*res_irq), GFP_KERNEL); if (!res_irq) { - dev_warn(&pdev->dev, "Missing IRQ resource.\n"); pm_runtime_put(&pdev->dev); - return -EINVAL; + return -ENOMEM; } + res_irq->flags = IORESOURCE_IRQ | irq_get_trigger_type(irq); + res_irq->start = irq; + res_irq->end = irq; + res_irq->name = dev_of_node(&pdev->dev) ? of_node_full_name(pdev->dev.of_node) : NULL; pdev_capture = devm_kzalloc(&pdev->dev, sizeof(*pdev_capture), GFP_KERNEL); diff --git a/drivers/media/platform/davinci/vpif_capture.c b/drivers/media/platform/davinci/vpif_capture.c index 8fe55374c5a3..c1a67a221743 100644 --- a/drivers/media/platform/davinci/vpif_capture.c +++ b/drivers/media/platform/davinci/vpif_capture.c @@ -1607,7 +1607,6 @@ static __init int vpif_probe(struct platform_device *pdev) { struct vpif_subdev_info *subdevdata; struct i2c_adapter *i2c_adap; - struct resource *res; int subdev_count; int res_idx = 0; int i, err; @@ -1632,15 +1631,14 @@ static __init int vpif_probe(struct platform_device *pdev) goto vpif_free; } - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, res_idx))) { - err = devm_request_irq(&pdev->dev, res->start, vpif_channel_isr, - IRQF_SHARED, VPIF_DRIVER_NAME, - (void *)(&vpif_obj.dev[res_idx]-> - channel_id)); - if (err) { - err = -EINVAL; + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { + if (err < 0) + goto vpif_unregister; + err = devm_request_irq(&pdev->dev, err, vpif_channel_isr, + IRQF_SHARED, VPIF_DRIVER_NAME, + (void *)(&vpif_obj.dev[res_idx]->channel_id)); + if (err) goto vpif_unregister; - } res_idx++; } diff --git a/drivers/media/platform/davinci/vpif_display.c b/drivers/media/platform/davinci/vpif_display.c index 59f6b782e104..9c552ea3787e 100644 --- a/drivers/media/platform/davinci/vpif_display.c +++ b/drivers/media/platform/davinci/vpif_display.c @@ -1221,7 +1221,6 @@ static __init int vpif_probe(struct platform_device *pdev) { struct vpif_subdev_info *subdevdata; struct i2c_adapter *i2c_adap; - struct resource *res; int subdev_count; int res_idx = 0; int i, err; @@ -1245,13 +1244,13 @@ static __init int vpif_probe(struct platform_device *pdev) goto vpif_free; } - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, res_idx))) { - err = devm_request_irq(&pdev->dev, res->start, vpif_channel_isr, - IRQF_SHARED, VPIF_DRIVER_NAME, - (void *)(&vpif_obj.dev[res_idx]-> - channel_id)); + while ((err = platform_get_irq_optional(pdev, res_idx)) != -ENXIO) { + if (err < 0) + goto vpif_unregister; + err = devm_request_irq(&pdev->dev, err, vpif_channel_isr, + IRQF_SHARED, VPIF_DRIVER_NAME, + (void *)(&vpif_obj.dev[res_idx]->channel_id)); if (err) { - err = -EINVAL; vpif_err("VPIF IRQ request failed\n"); goto vpif_unregister; }
platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static allocation of IRQ resources in DT core code, this causes an issue when using hierarchical interrupt domains using "interrupts" property in the node as this bypasses the hierarchical setup and messes up the irq chaining. In preparation for removal of static setup of IRQ resource from DT core code use platform_get_irq_optional(). Also this patch propagates error code in case devm_request_irq() fails instead of returing -EINVAL. Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> --- drivers/media/platform/davinci/vpif.c | 17 ++++++++++++++--- drivers/media/platform/davinci/vpif_capture.c | 16 +++++++--------- drivers/media/platform/davinci/vpif_display.c | 13 ++++++------- 3 files changed, 27 insertions(+), 19 deletions(-)