| Message ID | 1420563093-26026-1-git-send-email-balbi@ti.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Jan 06, 2015 at 10:51:33AM -0600, Felipe Balbi wrote: > commit 55601c9f2467 (arm: omap: intc: switch over > to linear irq domain) introduced a regression with > SDMA legacy driver because that driver strictly depends > on INTC's IRQs starting at NR_IRQs. Aparently > irq_domain_add_linear() won't guarantee that, since we see > a 7 IRQs difference when booting with and without the > commit cited above. > > Until arch/arm/plat-omap/dma.c is properly fixed, we > must maintain OMAP2/3 using irq_domain_add_legacy(). > > A FIXME note was added so people know to delete that > code once that legacy DMA driver is fixed up. > > Fixes: 55601c9f2467 (arm: omap: intc: switch over to linear irq domain) > Signed-off-by: Felipe Balbi <balbi@ti.com> I tested this on N950 with 3.19-rc3, and /proc/interrupts looks sane and also the "In-band Error" is gone. Tested-by: Aaro Koskinen <aaro.koskinen@iki.fi> BTW, I guess people still using 3.18.x will get the wrong IRQ for DMA, so maybe you should consider adding also Cc: stable... A. > --- > drivers/irqchip/irq-omap-intc.c | 26 +++++++++++++++++++++----- > 1 file changed, 21 insertions(+), 5 deletions(-) > > diff --git a/drivers/irqchip/irq-omap-intc.c b/drivers/irqchip/irq-omap-intc.c > index 3c970259c0eb..6ef88f56cf8d 100644 > --- a/drivers/irqchip/irq-omap-intc.c > +++ b/drivers/irqchip/irq-omap-intc.c > @@ -263,7 +263,7 @@ static int __init omap_init_irq_of(struct device_node *node) > return ret; > } > > -static int __init omap_init_irq_legacy(u32 base) > +static int __init omap_init_irq_legacy(u32 base, struct device_node *node) > { > int j, irq_base; > > @@ -277,7 +277,7 @@ static int __init omap_init_irq_legacy(u32 base) > irq_base = 0; > } > > - domain = irq_domain_add_legacy(NULL, omap_nr_irqs, irq_base, 0, > + domain = irq_domain_add_legacy(node, omap_nr_irqs, irq_base, 0, > &irq_domain_simple_ops, NULL); > > omap_irq_soft_reset(); > @@ -301,10 +301,26 @@ static int __init omap_init_irq(u32 base, struct device_node *node) > { > int ret; > > - if (node) > + /* > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > + * depends is still not ready for linear IRQ domains; because of that > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > + * linear IRQ Domain until that driver is finally fixed. > + */ > + if (of_device_is_compatible(node, "ti,omap2-intc") || > + of_device_is_compatible(node, "ti,omap3-intc")) { > + struct resource res; > + > + if (of_address_to_resource(node, 0, &res)) > + return -ENOMEM; > + > + base = res.start; > + ret = omap_init_irq_legacy(base, node); > + } else if (node) { > ret = omap_init_irq_of(node); > - else > - ret = omap_init_irq_legacy(base); > + } else { > + ret = omap_init_irq_legacy(base, NULL); > + } > > if (ret == 0) > omap_irq_enable_protection(); > -- > 2.2.0 > -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
* Aaro Koskinen <aaro.koskinen@iki.fi> [150106 09:51]: > On Tue, Jan 06, 2015 at 10:51:33AM -0600, Felipe Balbi wrote: > > commit 55601c9f2467 (arm: omap: intc: switch over > > to linear irq domain) introduced a regression with > > SDMA legacy driver because that driver strictly depends > > on INTC's IRQs starting at NR_IRQs. Aparently > > irq_domain_add_linear() won't guarantee that, since we see > > a 7 IRQs difference when booting with and without the > > commit cited above. > > > > Until arch/arm/plat-omap/dma.c is properly fixed, we > > must maintain OMAP2/3 using irq_domain_add_legacy(). > > > > A FIXME note was added so people know to delete that > > code once that legacy DMA driver is fixed up. > > > > Fixes: 55601c9f2467 (arm: omap: intc: switch over to linear irq domain) > > Signed-off-by: Felipe Balbi <balbi@ti.com> > > I tested this on N950 with 3.19-rc3, and /proc/interrupts looks sane > and also the "In-band Error" is gone. > > Tested-by: Aaro Koskinen <aaro.koskinen@iki.fi> > > BTW, I guess people still using 3.18.x will get the wrong IRQ for DMA, > so maybe you should consider adding also Cc: stable... Yeah this one should be cc stable. Fixes the error for me too: Tested-by: Tony Lindgren <tony@atomide.com> > > --- > > drivers/irqchip/irq-omap-intc.c | 26 +++++++++++++++++++++----- > > 1 file changed, 21 insertions(+), 5 deletions(-) > > > > diff --git a/drivers/irqchip/irq-omap-intc.c b/drivers/irqchip/irq-omap-intc.c > > index 3c970259c0eb..6ef88f56cf8d 100644 > > --- a/drivers/irqchip/irq-omap-intc.c > > +++ b/drivers/irqchip/irq-omap-intc.c > > @@ -263,7 +263,7 @@ static int __init omap_init_irq_of(struct device_node *node) > > return ret; > > } > > > > -static int __init omap_init_irq_legacy(u32 base) > > +static int __init omap_init_irq_legacy(u32 base, struct device_node *node) > > { > > int j, irq_base; > > > > @@ -277,7 +277,7 @@ static int __init omap_init_irq_legacy(u32 base) > > irq_base = 0; > > } > > > > - domain = irq_domain_add_legacy(NULL, omap_nr_irqs, irq_base, 0, > > + domain = irq_domain_add_legacy(node, omap_nr_irqs, irq_base, 0, > > &irq_domain_simple_ops, NULL); > > > > omap_irq_soft_reset(); > > @@ -301,10 +301,26 @@ static int __init omap_init_irq(u32 base, struct device_node *node) > > { > > int ret; > > > > - if (node) > > + /* > > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > > + * depends is still not ready for linear IRQ domains; because of that > > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > > + * linear IRQ Domain until that driver is finally fixed. > > + */ > > + if (of_device_is_compatible(node, "ti,omap2-intc") || > > + of_device_is_compatible(node, "ti,omap3-intc")) { > > + struct resource res; > > + > > + if (of_address_to_resource(node, 0, &res)) > > + return -ENOMEM; > > + > > + base = res.start; > > + ret = omap_init_irq_legacy(base, node); > > + } else if (node) { > > ret = omap_init_irq_of(node); > > - else > > - ret = omap_init_irq_legacy(base); > > + } else { > > + ret = omap_init_irq_legacy(base, NULL); > > + } > > > > if (ret == 0) > > omap_irq_enable_protection(); > > -- > > 2.2.0 > > -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Jan 06, 2015 at 10:51:33AM -0600, Felipe Balbi wrote: > + /* > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > + * depends is still not ready for linear IRQ domains; because of that > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > + * linear IRQ Domain until that driver is finally fixed. "finally fixed" or finally killed off like it really needs to be, once all users of it are killed. We've been trying to do this for, what, three years now... I finally pushed a WARN_ON() into that code to make it obvious to anyone who uses omap_request_dma() that they really need to update their code. Here's the list of references to that symbol which *still* need to be fixed so that we can kill the legacy DMA driver: drivers/media/platform/omap/omap_vout_vrfb.c: ret = omap_request_dma(vout->vrfb_dma_tx.dev_id, "VRFB DMA TX", drivers/media/platform/omap3isp/isphist.c: ret = omap_request_dma(OMAP24XX_DMA_NO_DEVICE, "DMA_ISP_HIST", drivers/media/platform/soc_camera/omap1_camera.c: err = omap_request_dma(OMAP_DMA_CAMERA_IF_RX, DRIVER_NAME, drivers/mtd/onenand/omap2.c: r = omap_request_dma(0, pdev->dev.driver->name, drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, drivers/usb/musb/tusb6010_omap.c: ret = omap_request_dma(chdat->sync_dev, dev_name, drivers/usb/musb/tusb6010_omap.c: ret = omap_request_dma(tusb_dma->sync_dev, "TUSB shared",
Hi, On Tue, Jan 06, 2015 at 06:05:32PM +0000, Russell King - ARM Linux wrote: > On Tue, Jan 06, 2015 at 10:51:33AM -0600, Felipe Balbi wrote: > > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > > + * depends is still not ready for linear IRQ domains; because of that > > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > > + * linear IRQ Domain until that driver is finally fixed. > > "finally fixed" or finally killed off like it really needs to be, once > all users of it are killed. > > We've been trying to do this for, what, three years now... I finally > pushed a WARN_ON() into that code to make it obvious to anyone who > uses omap_request_dma() that they really need to update their code. > Here's the list of references to that symbol which *still* need to be > fixed so that we can kill the legacy DMA driver: > > drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, > drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, I only learned about this after the WARN_ON() appeared in 3.17 (just couple months ago), and it's on my TODO list... A. -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
* Russell King - ARM Linux <linux@arm.linux.org.uk> [150106 10:08]: > On Tue, Jan 06, 2015 at 10:51:33AM -0600, Felipe Balbi wrote: > > + /* > > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > > + * depends is still not ready for linear IRQ domains; because of that > > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > > + * linear IRQ Domain until that driver is finally fixed. > > "finally fixed" or finally killed off like it really needs to be, once > all users of it are killed. > > We've been trying to do this for, what, three years now... I finally > pushed a WARN_ON() into that code to make it obvious to anyone who > uses omap_request_dma() that they really need to update their code. > > Here's the list of references to that symbol which *still* need to be > fixed so that we can kill the legacy DMA driver: > > drivers/media/platform/omap/omap_vout_vrfb.c: ret = omap_request_dma(vout->vrfb_dma_tx.dev_id, "VRFB DMA TX", > drivers/media/platform/omap3isp/isphist.c: ret = omap_request_dma(OMAP24XX_DMA_NO_DEVICE, "DMA_ISP_HIST", > drivers/media/platform/soc_camera/omap1_camera.c: err = omap_request_dma(OMAP_DMA_CAMERA_IF_RX, DRIVER_NAME, > drivers/mtd/onenand/omap2.c: r = omap_request_dma(0, pdev->dev.driver->name, AFAIK we should just remove DMA support from the drivers above. Nobody seems to be interested in doing anything about them. > drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, > drivers/usb/gadget/udc/omap_udc.c: status = omap_request_dma(dma_channel, OK so Aaro picked this one. > drivers/usb/musb/tusb6010_omap.c: ret = omap_request_dma(chdat->sync_dev, dev_name, > drivers/usb/musb/tusb6010_omap.c: ret = omap_request_dma(tusb_dma->sync_dev, "TUSB shared", I'll update this one. FYI, I already have some work-in-progress MUSB DMA patches that allow building in all the MUSB DMA glue layers. I just need to finish that series for v3.20: https://git.kernel.org/cgit/linux/kernel/git/tmlind/linux-omap.git/log/?h=musb-dma-2014-11-25-v2 So converting tusb6010 over to the dmaengine API would be the next logical step after that series. Probably not going to happen before v3.21 though.. Regards, Tony -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Am 06.01.2015 um 17:51 schrieb Felipe Balbi: > commit 55601c9f2467 (arm: omap: intc: switch over > to linear irq domain) introduced a regression with > SDMA legacy driver because that driver strictly depends > on INTC's IRQs starting at NR_IRQs. Aparently > irq_domain_add_linear() won't guarantee that, since we see > a 7 IRQs difference when booting with and without the > commit cited above. > > Until arch/arm/plat-omap/dma.c is properly fixed, we > must maintain OMAP2/3 using irq_domain_add_legacy(). > > A FIXME note was added so people know to delete that > code once that legacy DMA driver is fixed up. > > Fixes: 55601c9f2467 (arm: omap: intc: switch over to linear irq domain) > Signed-off-by: Felipe Balbi <balbi@ti.com> > --- > drivers/irqchip/irq-omap-intc.c | 26 +++++++++++++++++++++----- > 1 file changed, 21 insertions(+), 5 deletions(-) > > diff --git a/drivers/irqchip/irq-omap-intc.c b/drivers/irqchip/irq-omap-intc.c > index 3c970259c0eb..6ef88f56cf8d 100644 > --- a/drivers/irqchip/irq-omap-intc.c > +++ b/drivers/irqchip/irq-omap-intc.c > @@ -263,7 +263,7 @@ static int __init omap_init_irq_of(struct device_node *node) > return ret; > } > > -static int __init omap_init_irq_legacy(u32 base) > +static int __init omap_init_irq_legacy(u32 base, struct device_node *node) > { > int j, irq_base; > > @@ -277,7 +277,7 @@ static int __init omap_init_irq_legacy(u32 base) > irq_base = 0; > } > > - domain = irq_domain_add_legacy(NULL, omap_nr_irqs, irq_base, 0, > + domain = irq_domain_add_legacy(node, omap_nr_irqs, irq_base, 0, > &irq_domain_simple_ops, NULL); > > omap_irq_soft_reset(); > @@ -301,10 +301,26 @@ static int __init omap_init_irq(u32 base, struct device_node *node) > { > int ret; > > - if (node) > + /* > + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c > + * depends is still not ready for linear IRQ domains; because of that > + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using > + * linear IRQ Domain until that driver is finally fixed. > + */ > + if (of_device_is_compatible(node, "ti,omap2-intc") || > + of_device_is_compatible(node, "ti,omap3-intc")) { > + struct resource res; > + > + if (of_address_to_resource(node, 0, &res)) > + return -ENOMEM; > + > + base = res.start; > + ret = omap_init_irq_legacy(base, node); > + } else if (node) { > ret = omap_init_irq_of(node); > - else > - ret = omap_init_irq_legacy(base); > + } else { > + ret = omap_init_irq_legacy(base, NULL); > + } > > if (ret == 0) > omap_irq_enable_protection(); > Thanks, works on DM3730. Peter -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/irqchip/irq-omap-intc.c b/drivers/irqchip/irq-omap-intc.c index 3c970259c0eb..6ef88f56cf8d 100644 --- a/drivers/irqchip/irq-omap-intc.c +++ b/drivers/irqchip/irq-omap-intc.c @@ -263,7 +263,7 @@ static int __init omap_init_irq_of(struct device_node *node) return ret; } -static int __init omap_init_irq_legacy(u32 base) +static int __init omap_init_irq_legacy(u32 base, struct device_node *node) { int j, irq_base; @@ -277,7 +277,7 @@ static int __init omap_init_irq_legacy(u32 base) irq_base = 0; } - domain = irq_domain_add_legacy(NULL, omap_nr_irqs, irq_base, 0, + domain = irq_domain_add_legacy(node, omap_nr_irqs, irq_base, 0, &irq_domain_simple_ops, NULL); omap_irq_soft_reset(); @@ -301,10 +301,26 @@ static int __init omap_init_irq(u32 base, struct device_node *node) { int ret; - if (node) + /* + * FIXME legacy OMAP DMA driver sitting under arch/arm/plat-omap/dma.c + * depends is still not ready for linear IRQ domains; because of that + * we need to temporarily "blacklist" OMAP2 and OMAP3 devices from using + * linear IRQ Domain until that driver is finally fixed. + */ + if (of_device_is_compatible(node, "ti,omap2-intc") || + of_device_is_compatible(node, "ti,omap3-intc")) { + struct resource res; + + if (of_address_to_resource(node, 0, &res)) + return -ENOMEM; + + base = res.start; + ret = omap_init_irq_legacy(base, node); + } else if (node) { ret = omap_init_irq_of(node); - else - ret = omap_init_irq_legacy(base); + } else { + ret = omap_init_irq_legacy(base, NULL); + } if (ret == 0) omap_irq_enable_protection();
commit 55601c9f2467 (arm: omap: intc: switch over to linear irq domain) introduced a regression with SDMA legacy driver because that driver strictly depends on INTC's IRQs starting at NR_IRQs. Aparently irq_domain_add_linear() won't guarantee that, since we see a 7 IRQs difference when booting with and without the commit cited above. Until arch/arm/plat-omap/dma.c is properly fixed, we must maintain OMAP2/3 using irq_domain_add_legacy(). A FIXME note was added so people know to delete that code once that legacy DMA driver is fixed up. Fixes: 55601c9f2467 (arm: omap: intc: switch over to linear irq domain) Signed-off-by: Felipe Balbi <balbi@ti.com> --- drivers/irqchip/irq-omap-intc.c | 26 +++++++++++++++++++++----- 1 file changed, 21 insertions(+), 5 deletions(-)