On Mon, Dec 5, 2011 at 2:08 AM, Michael S. Tsirkin <m...@redhat.com> wrote: > On Sun, Dec 04, 2011 at 04:47:17PM -0700, Cam Macdonell wrote: >> On Sun, Dec 4, 2011 at 3:20 AM, Michael S. Tsirkin <m...@redhat.com> wrote: >> > On Fri, Dec 02, 2011 at 04:34:21PM -0700, Cam Macdonell wrote: >> >> Based on a git bisect, this patch breaks msi-x interrupt delivery in >> >> the ivshmem device. >> > >> > I think the following should fix it. Compiled-only - >> > could you pls check? If yes let's apply to the stable branch. >> >> Thanks for the patch Michael. >> >> It addresses the need for msix_write_config() to be called, but the >> addition of the msix_reset() is causing a reset of the vectors after >> they've been initialized in pci_ivshmem_init(). So, interrupts still >> aren't delivered with this patch applied as it is. >> >> In particular, a reset occurs after pci_ivshmem_init runs, so the >> msix_entry_used array is reset to 0s, which causes the interrupt >> delivery to fail. >> >> If I comment out the msix_reset(), then interrupts are delivered. >> Would the reset be caused by a bug in the guest driver? or do I need >> to reconfigure the msix after reset? I'm unclear as to the proper >> behaviour after a reset. >> >> Thanks, >> Cam >> > > I didn't anticipate this, virtio only configures vectors when msi-x is > enabled. > For 1.0 it's safest to do the same and just re-enable after reset. > > So we'll end up with something like the following - does it help?
Yes, this fixes interrupt delivery. > > Side note: exit(1) is not the best way to handle > errors in the init function. I think you should add error_report, then > goto err on failures to init, cleanup what you have set up > meanwhile and return an error code. Thanks, I'll fix that. > > diff --git a/hw/ivshmem.c b/hw/ivshmem.c > index 242fbea..c58f4d3 100644 > --- a/hw/ivshmem.c > +++ b/hw/ivshmem.c > @@ -500,11 +500,29 @@ static void ivshmem_read(void *opaque, const uint8_t * > buf, int flags) > return; > } > > +/* Select the MSI-X vectors used by device. > + * ivshmem maps events to vectors statically, so > + * we just enable all vectors on init and after reset. */ > +static void ivshmem_use_msix(IVShmemState * s) > +{ > + int i; > + > + if (!msix_present(&s->dev)) { > + return; > + } > + > + for (i = 0; i < s->vectors; i++) { > + msix_vector_use(&s->dev, i); > + } > +} > + > static void ivshmem_reset(DeviceState *d) > { > IVShmemState *s = DO_UPCAST(IVShmemState, dev.qdev, d); > > s->intrstatus = 0; > + msix_reset(&s->dev); > + ivshmem_use_msix(s); > return; > } > > @@ -535,12 +553,8 @@ static uint64_t ivshmem_get_size(IVShmemState * s) { > return value; > } > > -static void ivshmem_setup_msi(IVShmemState * s) { > - > - int i; > - > - /* allocate the MSI-X vectors */ > - > +static void ivshmem_setup_msi(IVShmemState * s) > +{ > memory_region_init(&s->msix_bar, "ivshmem-msix", 4096); > if (!msix_init(&s->dev, s->vectors, &s->msix_bar, 1, 0)) { > pci_register_bar(&s->dev, 1, PCI_BASE_ADDRESS_SPACE_MEMORY, > @@ -551,13 +565,10 @@ static void ivshmem_setup_msi(IVShmemState * s) { > exit(1); > } > > - /* 'activate' the vectors */ > - for (i = 0; i < s->vectors; i++) { > - msix_vector_use(&s->dev, i); > - } > - > /* allocate Qemu char devices for receiving interrupts */ > s->eventfd_table = g_malloc0(s->vectors * sizeof(EventfdEntry)); > + > + ivshmem_use_msix(s); > } > > static void ivshmem_save(QEMUFile* f, void *opaque) > @@ -610,6 +621,13 @@ static int ivshmem_load(QEMUFile* f, void *opaque, int > version_id) > return 0; > } > > +static void ivshmem_write_config(PCIDevice *pci_dev, uint32_t address, > + uint32_t val, int len) > +{ > + pci_default_write_config(pci_dev, address, val, len); > + msix_write_config(pci_dev, address, val, len); > +} > + > static int pci_ivshmem_init(PCIDevice *dev) > { > IVShmemState *s = DO_UPCAST(IVShmemState, dev, dev); > @@ -734,6 +752,8 @@ static int pci_ivshmem_init(PCIDevice *dev) > > } > > + s->dev.config_write = ivshmem_write_config; > + > return 0; > } > >