| Message ID | 1244481435-17224-1-git-send-email-weidong.han@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Monday 08 June 2009, Weidong Han wrote: > When hot remove an assigned device, segmentation fault was triggered > by qemu_free(&pci_dev->qdev) in pci_unregister_device(). > pci_register_device() doesn't initialize or set pci_dev->qdev. For an > assigned device, qdev variable isn't touched at all. So segmentation > fault happens when to free a non-initialized qdev. Better would be to just disable hot remove for devices still using the legacy pci_register_device API. Paul -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Paul Brook wrote: > On Monday 08 June 2009, Weidong Han wrote: >> When hot remove an assigned device, segmentation fault was triggered >> by qemu_free(&pci_dev->qdev) in pci_unregister_device(). >> pci_register_device() doesn't initialize or set pci_dev->qdev. For an >> assigned device, qdev variable isn't touched at all. So segmentation >> fault happens when to free a non-initialized qdev. > > Better would be to just disable hot remove for devices still using > the legacy pci_register_device API. > PCI passthrough uses pci_register_device to register assigned device to qemu. Is there newer API to do so? Regards, Weidong -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tuesday 09 June 2009, Han, Weidong wrote: > Paul Brook wrote: > > On Monday 08 June 2009, Weidong Han wrote: > >> When hot remove an assigned device, segmentation fault was triggered > >> by qemu_free(&pci_dev->qdev) in pci_unregister_device(). > >> pci_register_device() doesn't initialize or set pci_dev->qdev. For an > >> assigned device, qdev variable isn't touched at all. So segmentation > >> fault happens when to free a non-initialized qdev. > > > > Better would be to just disable hot remove for devices still using > > the legacy pci_register_device API. > > PCI passthrough uses pci_register_device to register assigned device to > qemu. Is there newer API to do so? Yes. See e.g. LSI scsi emulation. Paul -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 06/09/09 16:51, Paul Brook wrote: > On Tuesday 09 June 2009, Han, Weidong wrote: >> Paul Brook wrote: >>> On Monday 08 June 2009, Weidong Han wrote: >>>> When hot remove an assigned device, segmentation fault was triggered >>>> by qemu_free(&pci_dev->qdev) in pci_unregister_device(). >>>> pci_register_device() doesn't initialize or set pci_dev->qdev. For an >>>> assigned device, qdev variable isn't touched at all. So segmentation >>>> fault happens when to free a non-initialized qdev. >>> Better would be to just disable hot remove for devices still using >>> the legacy pci_register_device API. >> PCI passthrough uses pci_register_device to register assigned device to >> qemu. Is there newer API to do so? > > Yes. See e.g. LSI scsi emulation. Well. Except that you can't (yet) register pci config read/write callbacks using the qdev-based API. cheers, Gerd -- To unsubscribe from this list: send the line "unsubscribe kvm" 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/hw/pci.c b/hw/pci.c index 25581a4..77d63d8 100644 --- a/hw/pci.c +++ b/hw/pci.c @@ -377,7 +377,7 @@ int pci_unregister_device(PCIDevice *pci_dev) qemu_free_irqs(pci_dev->irq); pci_irq_index--; pci_dev->bus->devices[pci_dev->devfn] = NULL; - qdev_free(&pci_dev->qdev); + qemu_free(pci_dev); return 0; }
When hot remove an assigned device, segmentation fault was triggered by qemu_free(&pci_dev->qdev) in pci_unregister_device(). pci_register_device() doesn't initialize or set pci_dev->qdev. For an assigned device, qdev variable isn't touched at all. So segmentation fault happens when to free a non-initialized qdev. Paul, you introduced the code to free qdev in pci_unregiser_device. Did you miss something? Following patch changes the code back to free pci_dev, and fixes the hot remove issue. Signed-off-by: Weidong Han <weidong.han@intel.com> --- hw/pci.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)