From patchwork Fri Apr 19 14:57:23 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Peter Maydell X-Patchwork-Id: 238023 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from lists.gnu.org (lists.gnu.org [208.118.235.17]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPS id 595C32C01B5 for ; Sat, 20 Apr 2013 01:01:57 +1000 (EST) Received: from localhost ([::1]:58941 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UTCp1-00063G-K2 for incoming@patchwork.ozlabs.org; Fri, 19 Apr 2013 11:01:55 -0400 Received: from eggs.gnu.org ([208.118.235.92]:46250) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UTCl2-0007fj-EE for qemu-devel@nongnu.org; Fri, 19 Apr 2013 10:58:00 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UTCkr-0007Vr-Ps for qemu-devel@nongnu.org; Fri, 19 Apr 2013 10:57:48 -0400 Received: from 1.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.d.1.0.0.b.8.0.1.0.0.2.ip6.arpa ([2001:8b0:1d0::1]:56618 helo=mnementh.archaic.org.uk) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UTCkr-0007TS-Fk for qemu-devel@nongnu.org; Fri, 19 Apr 2013 10:57:37 -0400 Received: from pm215 by mnementh.archaic.org.uk with local (Exim 4.72) (envelope-from ) id 1UTCkh-0003Pf-No; Fri, 19 Apr 2013 15:57:27 +0100 From: Peter Maydell To: Anthony Liguori Date: Fri, 19 Apr 2013 15:57:23 +0100 Message-Id: <1366383447-13082-8-git-send-email-peter.maydell@linaro.org> X-Mailer: git-send-email 1.7.2.5 In-Reply-To: <1366383447-13082-1-git-send-email-peter.maydell@linaro.org> References: <1366383447-13082-1-git-send-email-peter.maydell@linaro.org> X-detected-operating-system: by eggs.gnu.org: Error: Malformed IPv6 address (bad octet value). X-Received-From: 2001:8b0:1d0::1 Cc: qemu-devel@nongnu.org, Paul Brook Subject: [Qemu-devel] [PATCH 07/11] versatile_pci: Implement the correct PCI IRQ mapping X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Implement the correct IRQ mapping for the Versatile PCI controller; it differs between realview and versatile boards, but the previous QEMU implementation was correct only for the first PCI card on a versatile board, since we weren't swizzling IRQs based on the slot number. Since this change would otherwise break any uses of PCI on Linux kernels which have an equivalent bug (since they have effectively only been tested against QEMU, not real hardware), we implement a mechanism for automatically detecting those broken kernels and switching back to the old mapping. This works by looking at the values the kernel writes to the PCI_INTERRUPT_LINE register in the config space, which is effectively the interrupt number the kernel expects the device to be using. Signed-off-by: Peter Maydell --- hw/pci-host/versatile.c | 105 ++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 99 insertions(+), 6 deletions(-) diff --git a/hw/pci-host/versatile.c b/hw/pci-host/versatile.c index 8f8612c..5d543a9 100644 --- a/hw/pci-host/versatile.c +++ b/hw/pci-host/versatile.c @@ -13,6 +13,28 @@ #include "hw/pci/pci_host.h" #include "exec/address-spaces.h" +/* Old and buggy versions of QEMU used the wrong mapping from + * PCI IRQs to system interrupt lines. Unfortunately the Linux + * kernel also had the corresponding bug in setting up interrupts + * (so older kernels work on QEMU and not on real hardware). + * We automatically detect these broken kernels and flip back + * to the broken irq mapping by spotting guest writes to the + * PCI_INTERRUPT_LINE register to see where the guest thinks + * interrupts are going to be routed. So we start in state + * ASSUME_OK on reset, and transition to either BROKEN or + * FORCE_OK at the first write to an INTERRUPT_LINE register for + * a slot where broken and correct interrupt mapping would differ. + * Once in either BROKEN or FORCE_OK we never transition again; + * this allows a newer kernel to use the INTERRUPT_LINE + * registers arbitrarily once it has indicated that it isn't + * broken in its init code somewhere. + */ +enum { + PCI_VPB_IRQMAP_ASSUME_OK, + PCI_VPB_IRQMAP_BROKEN, + PCI_VPB_IRQMAP_FORCE_OK, +}; + typedef struct { PCIHostState parent_obj; @@ -26,6 +48,9 @@ typedef struct { /* Constant for life of device: */ int realview; + + /* Variable state: */ + uint8_t irq_mapping; } PCIVPBState; #define TYPE_VERSATILE_PCI "versatile_pci" @@ -44,14 +69,27 @@ static inline uint32_t vpb_pci_config_addr(hwaddr addr) static void pci_vpb_config_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) { - pci_data_write(opaque, vpb_pci_config_addr(addr), val, size); + PCIVPBState *s = opaque; + if (!s->realview && (addr & 0xff) == PCI_INTERRUPT_LINE + && s->irq_mapping == PCI_VPB_IRQMAP_ASSUME_OK) { + uint8_t devfn = addr >> 8; + if ((PCI_SLOT(devfn) % PCI_NUM_PINS) != 2) { + if (val == 27) { + s->irq_mapping = PCI_VPB_IRQMAP_BROKEN; + } else { + s->irq_mapping = PCI_VPB_IRQMAP_FORCE_OK; + } + } + } + pci_data_write(&s->pci_bus, vpb_pci_config_addr(addr), val, size); } static uint64_t pci_vpb_config_read(void *opaque, hwaddr addr, unsigned size) { + PCIVPBState *s = opaque; uint32_t val; - val = pci_data_read(opaque, vpb_pci_config_addr(addr), size); + val = pci_data_read(&s->pci_bus, vpb_pci_config_addr(addr), size); return val; } @@ -63,7 +101,47 @@ static const MemoryRegionOps pci_vpb_config_ops = { static int pci_vpb_map_irq(PCIDevice *d, int irq_num) { - return irq_num; + PCIVPBState *s = container_of(d->bus, PCIVPBState, pci_bus); + + if (s->irq_mapping == PCI_VPB_IRQMAP_BROKEN) { + /* Legacy broken IRQ mapping for compatibility with old and + * buggy Linux guests + */ + return irq_num; + } + + /* Slot to IRQ mapping for RealView Platform Baseboard 926 backplane + * name slot IntA IntB IntC IntD + * A 31 IRQ28 IRQ29 IRQ30 IRQ27 + * B 30 IRQ27 IRQ28 IRQ29 IRQ30 + * C 29 IRQ30 IRQ27 IRQ28 IRQ29 + * Slot C is for the host bridge; A and B the peripherals. + * Our output irqs 0..3 correspond to the baseboard's 27..30. + * + * This mapping function takes account of an oddity in the PB926 + * board wiring, where the FPGA's P_nINTA input is connected to + * the INTB connection on the board PCI edge connector, P_nINTB + * is connected to INTC, and so on, so everything is one number + * further round from where you might expect. + */ + return pci_swizzle_map_irq_fn(d, irq_num + 2); +} + +static int pci_vpb_rv_map_irq(PCIDevice *d, int irq_num) +{ + /* Slot to IRQ mapping for RealView EB and PB1176 backplane + * name slot IntA IntB IntC IntD + * A 31 IRQ50 IRQ51 IRQ48 IRQ49 + * B 30 IRQ49 IRQ50 IRQ51 IRQ48 + * C 29 IRQ48 IRQ49 IRQ50 IRQ51 + * Slot C is for the host bridge; A and B the peripherals. + * Our output irqs 0..3 correspond to the baseboard's 48..51. + * + * The PB1176 and EB boards don't have the PB926 wiring oddity + * described above; P_nINTA connects to INTA, P_nINTB to INTB + * and so on, which is why this mapping function is different. + */ + return pci_swizzle_map_irq_fn(d, irq_num + 3); } static void pci_vpb_set_irq(void *opaque, int irq_num, int level) @@ -73,6 +151,13 @@ static void pci_vpb_set_irq(void *opaque, int irq_num, int level) qemu_set_irq(pic[irq_num], level); } +static void pci_vpb_reset(DeviceState *d) +{ + PCIVPBState *s = PCI_VPB(d); + + s->irq_mapping = PCI_VPB_IRQMAP_ASSUME_OK; +} + static void pci_vpb_init(Object *obj) { PCIHostState *h = PCI_HOST_BRIDGE(obj); @@ -95,13 +180,20 @@ static void pci_vpb_realize(DeviceState *dev, Error **errp) { PCIVPBState *s = PCI_VPB(dev); SysBusDevice *sbd = SYS_BUS_DEVICE(dev); + pci_map_irq_fn mapfn; int i; for (i = 0; i < 4; i++) { sysbus_init_irq(sbd, &s->irq[i]); } - pci_bus_irqs(&s->pci_bus, pci_vpb_set_irq, pci_vpb_map_irq, s->irq, 4); + if (s->realview) { + mapfn = pci_vpb_rv_map_irq; + } else { + mapfn = pci_vpb_map_irq; + } + + pci_bus_irqs(&s->pci_bus, pci_vpb_set_irq, mapfn, s->irq, 4); /* ??? Register memory space. */ @@ -110,10 +202,10 @@ static void pci_vpb_realize(DeviceState *dev, Error **errp) * 1 : PCI config window * 2 : PCI IO window */ - memory_region_init_io(&s->mem_config, &pci_vpb_config_ops, &s->pci_bus, + memory_region_init_io(&s->mem_config, &pci_vpb_config_ops, s, "pci-vpb-selfconfig", 0x1000000); sysbus_init_mmio(sbd, &s->mem_config); - memory_region_init_io(&s->mem_config2, &pci_vpb_config_ops, &s->pci_bus, + memory_region_init_io(&s->mem_config2, &pci_vpb_config_ops, s, "pci-vpb-config", 0x1000000); sysbus_init_mmio(sbd, &s->mem_config2); @@ -159,6 +251,7 @@ static void pci_vpb_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); dc->realize = pci_vpb_realize; + dc->reset = pci_vpb_reset; } static const TypeInfo pci_vpb_info = {