Message ID | 20170430192955.9098-1-james.r.harris@intel.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Sun, Apr 30, 2017 at 12:29:55PM -0700, Jim Harris wrote: > Some userspace drivers and frameworks only poll and do not > require interrupts to be available and enabled on the > PCI device. So remove the requirement that an IRQ is > assigned. If an IRQ is not assigned and a userspace > driver tries to read()/write(), the generic uio > framework will just return -EIO. > > This allows binding uio_pci_generic to devices which > cannot get an IRQ assigned, such as an NVMe controller > behind Intel Volume Management Device (VMD), since VMD > does not support INTx interrupts. > > Signed-off-by: Jim Harris <james.r.harris@intel.com> Without interrupts, why do you want uio? All it does is forward interrupts to userspace. > --- > drivers/uio/uio_pci_generic.c | 20 +++++++++----------- > 1 file changed, 9 insertions(+), 11 deletions(-) > > diff --git a/drivers/uio/uio_pci_generic.c b/drivers/uio/uio_pci_generic.c > index d0b508b68f3c..81c59b4f8552 100644 > --- a/drivers/uio/uio_pci_generic.c > +++ b/drivers/uio/uio_pci_generic.c > @@ -66,14 +66,7 @@ static int probe(struct pci_dev *pdev, > return err; > } > > - if (!pdev->irq) { > - dev_warn(&pdev->dev, "No IRQ assigned to device: " > - "no support for interrupts?\n"); > - pci_disable_device(pdev); > - return -ENODEV; > - } > - > - if (!pci_intx_mask_supported(pdev)) { > + if (pci->irq && !pci_intx_mask_supported(pdev)) { > err = -ENODEV; > goto err_verify; > } > @@ -86,10 +79,15 @@ static int probe(struct pci_dev *pdev, > > gdev->info.name = "uio_pci_generic"; > gdev->info.version = DRIVER_VERSION; > - gdev->info.irq = pdev->irq; > - gdev->info.irq_flags = IRQF_SHARED; > - gdev->info.handler = irqhandler; > gdev->pdev = pdev; > + if (pdev->irq) { > + gdev->info.irq = pdev->irq; > + gdev->info.irq_flags = IRQF_SHARED; > + gdev->info.handler = irqhandler; > + } else { > + dev_warn(&pdev->dev, "No IRQ assigned to device: " > + "no support for interrupts?\n"); > + } > > err = uio_register_device(&pdev->dev, &gdev->info); > if (err) > -- > 2.12.2
> On Apr 30, 2017, at 6:17 PM, Michael S. Tsirkin <mst@redhat.com> wrote: > > On Sun, Apr 30, 2017 at 12:29:55PM -0700, Jim Harris wrote: >> Some userspace drivers and frameworks only poll and do not >> require interrupts to be available and enabled on the >> PCI device. So remove the requirement that an IRQ is >> assigned. If an IRQ is not assigned and a userspace >> driver tries to read()/write(), the generic uio >> framework will just return -EIO. >> >> This allows binding uio_pci_generic to devices which >> cannot get an IRQ assigned, such as an NVMe controller >> behind Intel Volume Management Device (VMD), since VMD >> does not support INTx interrupts. >> >> Signed-off-by: Jim Harris <james.r.harris@intel.com> > > Without interrupts, why do you want uio? All it does is > forward interrupts to userspace. Primarily so that a driver is bound to the device while using it from userspace. It can also be helpful to have the uioX sysfs symlinks to the underlying PCI config/resources. Some userspace drivers/frameworks such as DPDK and SPDK primarily run in polled-only mode - not using interrupts at all. But DPDK does provide a facility to drop into interrupt mode which uio facilitates. So looking ahead to NVMe devices behind a VMD endpoint, we want to be able to at least bind uio to these NVMe devices - DPDK/SPDK just will not be able to drop into interrupt mode on these devices. This is different obviously for vfio where we can allocate MSI/MSIX vectors. But for dev systems with IOMMU disabled, or passing a VMD endpoint to a guest VM, uio support is still desired. > >> --- >> drivers/uio/uio_pci_generic.c | 20 +++++++++----------- >> 1 file changed, 9 insertions(+), 11 deletions(-) >> >> diff --git a/drivers/uio/uio_pci_generic.c b/drivers/uio/uio_pci_generic.c >> index d0b508b68f3c..81c59b4f8552 100644 >> --- a/drivers/uio/uio_pci_generic.c >> +++ b/drivers/uio/uio_pci_generic.c >> @@ -66,14 +66,7 @@ static int probe(struct pci_dev *pdev, >> return err; >> } >> >> - if (!pdev->irq) { >> - dev_warn(&pdev->dev, "No IRQ assigned to device: " >> - "no support for interrupts?\n"); >> - pci_disable_device(pdev); >> - return -ENODEV; >> - } >> - >> - if (!pci_intx_mask_supported(pdev)) { >> + if (pci->irq && !pci_intx_mask_supported(pdev)) { >> err = -ENODEV; >> goto err_verify; >> } >> @@ -86,10 +79,15 @@ static int probe(struct pci_dev *pdev, >> >> gdev->info.name = "uio_pci_generic"; >> gdev->info.version = DRIVER_VERSION; >> - gdev->info.irq = pdev->irq; >> - gdev->info.irq_flags = IRQF_SHARED; >> - gdev->info.handler = irqhandler; >> gdev->pdev = pdev; >> + if (pdev->irq) { >> + gdev->info.irq = pdev->irq; >> + gdev->info.irq_flags = IRQF_SHARED; >> + gdev->info.handler = irqhandler; >> + } else { >> + dev_warn(&pdev->dev, "No IRQ assigned to device: " >> + "no support for interrupts?\n"); >> + } >> >> err = uio_register_device(&pdev->dev, &gdev->info); >> if (err) >> -- >> 2.12.2
Hi Jim, [auto build test ERROR on char-misc/char-misc-testing] [also build test ERROR on v4.11 next-20170501] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system] url: https://github.com/0day-ci/linux/commits/Jim-Harris/uio-uio_pci_generic-don-t-fail-probe-if-pdev-irq-NULL/20170501-040733 config: x86_64-rhel (attached as .config) compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901 reproduce: # save the attached .config to linux build tree make ARCH=x86_64 All errors (new ones prefixed by >>): drivers/uio/uio_pci_generic.c: In function 'probe': >> drivers/uio/uio_pci_generic.c:69:6: error: 'pci' undeclared (first use in this function) if (pci->irq && !pci_intx_mask_supported(pdev)) { ^~~ drivers/uio/uio_pci_generic.c:69:6: note: each undeclared identifier is reported only once for each function it appears in vim +/pci +69 drivers/uio/uio_pci_generic.c 63 if (err) { 64 dev_err(&pdev->dev, "%s: pci_enable_device failed: %d\n", 65 __func__, err); 66 return err; 67 } 68 > 69 if (pci->irq && !pci_intx_mask_supported(pdev)) { 70 err = -ENODEV; 71 goto err_verify; 72 } --- 0-DAY kernel test infrastructure Open Source Technology Center https://lists.01.org/pipermail/kbuild-all Intel Corporation
diff --git a/drivers/uio/uio_pci_generic.c b/drivers/uio/uio_pci_generic.c index d0b508b68f3c..81c59b4f8552 100644 --- a/drivers/uio/uio_pci_generic.c +++ b/drivers/uio/uio_pci_generic.c @@ -66,14 +66,7 @@ static int probe(struct pci_dev *pdev, return err; } - if (!pdev->irq) { - dev_warn(&pdev->dev, "No IRQ assigned to device: " - "no support for interrupts?\n"); - pci_disable_device(pdev); - return -ENODEV; - } - - if (!pci_intx_mask_supported(pdev)) { + if (pci->irq && !pci_intx_mask_supported(pdev)) { err = -ENODEV; goto err_verify; } @@ -86,10 +79,15 @@ static int probe(struct pci_dev *pdev, gdev->info.name = "uio_pci_generic"; gdev->info.version = DRIVER_VERSION; - gdev->info.irq = pdev->irq; - gdev->info.irq_flags = IRQF_SHARED; - gdev->info.handler = irqhandler; gdev->pdev = pdev; + if (pdev->irq) { + gdev->info.irq = pdev->irq; + gdev->info.irq_flags = IRQF_SHARED; + gdev->info.handler = irqhandler; + } else { + dev_warn(&pdev->dev, "No IRQ assigned to device: " + "no support for interrupts?\n"); + } err = uio_register_device(&pdev->dev, &gdev->info); if (err)
Some userspace drivers and frameworks only poll and do not require interrupts to be available and enabled on the PCI device. So remove the requirement that an IRQ is assigned. If an IRQ is not assigned and a userspace driver tries to read()/write(), the generic uio framework will just return -EIO. This allows binding uio_pci_generic to devices which cannot get an IRQ assigned, such as an NVMe controller behind Intel Volume Management Device (VMD), since VMD does not support INTx interrupts. Signed-off-by: Jim Harris <james.r.harris@intel.com> --- drivers/uio/uio_pci_generic.c | 20 +++++++++----------- 1 file changed, 9 insertions(+), 11 deletions(-)