qemu/hw/isa/i82378.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

157 lines
4.2 KiB
C
Raw Normal View History

/*
* QEMU Intel i82378 emulation (PCI to ISA bridge)
*
* Copyright (c) 2010-2011 Hervé Poussineau
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include "hw/pci/pci_device.h"
#include "hw/irq.h"
#include "hw/intc/i8259.h"
#include "hw/timer/i8254.h"
#include "migration/vmstate.h"
#include "hw/audio/pcspk.h"
#include "qom/object.h"
#define TYPE_I82378 "i82378"
OBJECT_DECLARE_SIMPLE_TYPE(I82378State, I82378)
struct I82378State {
PCIDevice parent_obj;
qemu_irq cpu_intr;
qemu_irq *isa_irqs_in;
};
static const VMStateDescription vmstate_i82378 = {
.name = "pci-i82378",
.version_id = 0,
.minimum_version_id = 0,
.fields = (const VMStateField[]) {
VMSTATE_PCI_DEVICE(parent_obj, I82378State),
VMSTATE_END_OF_LIST()
},
};
Revert "hw/isa/i82378: Remove intermediate IRQ forwarder" To be 'usable', QDev objects (which are QOM objects) must be 1/ initialized (at this point their properties can be modified), then 2/ realized (properties are consumed). Some devices (objects) might depend on other devices. When creating the 'QOM composition tree', parent objects can't be 'realized' until all their children are. We might also have circular dependencies. A common circular dependency occurs with IRQs. Device (A) has an output IRQ wired to device (B), and device (B) has one to device (A). When (A) is realized and connects its IRQ to an unrealized (B), the IRQ handler on (B) is not yet created. QEMU pass IRQ between objects as pointer. When (A) poll (B)'s IRQ, it is NULL. Later (B) is realized and its IRQ pointers are populated, but (A) keeps a reference to a NULL pointer. A common pattern to bypass this circular limitation is to use 'proxy' objects. Proxy (P) is created (and realized) before (A) and (B). Then (A) and (B) can be created in different order, it doesn't matter: (P) pointers are already populated. Commit cef2e7148e ("hw/isa/i82378: Remove intermediate IRQ forwarder") neglected the QOM/QDev circular dependency issue, and removed the 'proxy' between the southbridge, its PCI functions and the interrupt controller, resulting in PCI functions wiring output IRQs to 'NULL', leading to guest failures (IRQ never delivered) [1] [2]. Since we are entering feature freeze, it is safer to revert the offending patch until we figure a way to strengthen our APIs. [1] https://lore.kernel.org/qemu-devel/928a8552-ab62-9e6c-a492-d6453e338b9d@redhat.com/ [2] https://lore.kernel.org/qemu-devel/cover.1677628524.git.balaton@eik.bme.hu/ This reverts commit cef2e7148e32d61338de0220619d308bf42af770. Reported-by: Thomas Huth <thuth@redhat.com> Inspired-by: Bernhard Beschow <shentey@gmail.com> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
2023-03-07 14:10:08 +03:00
static void i82378_request_out0_irq(void *opaque, int irq, int level)
{
I82378State *s = opaque;
qemu_set_irq(s->cpu_intr, level);
}
static void i82378_request_pic_irq(void *opaque, int irq, int level)
{
DeviceState *dev = opaque;
I82378State *s = I82378(dev);
qemu_set_irq(s->isa_irqs_in[irq], level);
}
static void i82378_realize(PCIDevice *pci, Error **errp)
{
DeviceState *dev = DEVICE(pci);
I82378State *s = I82378(dev);
uint8_t *pci_conf;
ISABus *isabus;
ISADevice *pit;
ISADevice *pcspk;
pci_conf = pci->config;
pci_set_word(pci_conf + PCI_COMMAND,
PCI_COMMAND_IO | PCI_COMMAND_MEMORY | PCI_COMMAND_MASTER);
pci_set_word(pci_conf + PCI_STATUS,
PCI_STATUS_DEVSEL_MEDIUM);
pci_config_set_interrupt_pin(pci_conf, 1); /* interrupt pin 0 */
isabus = isa_bus_new(dev, get_system_memory(),
isa: Clean up error handling around isa_bus_new() We can have at most one ISA bus. If you try to create another one, isa_bus_new() complains to stderr and returns null. isa_bus_new() is called in two contexts, machine's init() and device's realize() methods. Since complaining to stderr is not proper in the latter context, convert isa_bus_new() to Error. Machine's init(): * mips_jazz_init(), called from the init() methods of machines "magnum" and "pica" * mips_r4k_init(), the init() method of machine "mips" * pc_init1() called from the init() methods of non-q35 PC machines * typhoon_init(), called from clipper_init(), the init() method of machine "clipper" These callers always create the first ISA bus, hence isa_bus_new() can't fail. Simply pass &error_abort. Device's realize(): * i82378_realize(), of PCI device "i82378" * ich9_lpc_realize(), of PCI device "ICH9-LPC" * pci_ebus_realize(), of PCI device "ebus" * piix3_realize(), of PCI device "pci-piix3", abstract parent of "PIIX3" and "PIIX3-xen" * piix4_realize(), of PCI device "PIIX4" * vt82c686b_realize(), of PCI device "VT82C686B" Propagate the error. Note that these devices are typically created only by machine init() methods with qdev_init_nofail() or similar. If we screwed up and created an ISA bus before that call, we now give up right away. Before, we'd hobble on, and typically die in isa_bus_irqs(). Similar if someone finds a way to hot-plug one of these critters. Cc: Richard Henderson <rth@twiddle.net> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: "Hervé Poussineau" <hpoussin@reactos.org> Cc: Aurelien Jarno <aurelien@aurel32.net> Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> Signed-off-by: Markus Armbruster <armbru@pond.sub.org> Reviewed-by: Marcel Apfelbaum <marcel@redhat.com> Reviewed-by: Hervé Poussineau <hpoussin@reactos.org> Reviewed-by: Michael S. Tsirkin <mst@redhat.com> Message-Id: <1450370121-5768-11-git-send-email-armbru@redhat.com>
2015-12-17 19:35:18 +03:00
pci_address_space_io(pci), errp);
if (!isabus) {
return;
}
/* This device has:
2 82C59 (irq)
1 82C54 (pit)
2 82C37 (dma)
NMI
Utility Bus Support Registers
All devices accept byte access only, except timer
*/
/* 2 82C59 (irq) */
Revert "hw/isa/i82378: Remove intermediate IRQ forwarder" To be 'usable', QDev objects (which are QOM objects) must be 1/ initialized (at this point their properties can be modified), then 2/ realized (properties are consumed). Some devices (objects) might depend on other devices. When creating the 'QOM composition tree', parent objects can't be 'realized' until all their children are. We might also have circular dependencies. A common circular dependency occurs with IRQs. Device (A) has an output IRQ wired to device (B), and device (B) has one to device (A). When (A) is realized and connects its IRQ to an unrealized (B), the IRQ handler on (B) is not yet created. QEMU pass IRQ between objects as pointer. When (A) poll (B)'s IRQ, it is NULL. Later (B) is realized and its IRQ pointers are populated, but (A) keeps a reference to a NULL pointer. A common pattern to bypass this circular limitation is to use 'proxy' objects. Proxy (P) is created (and realized) before (A) and (B). Then (A) and (B) can be created in different order, it doesn't matter: (P) pointers are already populated. Commit cef2e7148e ("hw/isa/i82378: Remove intermediate IRQ forwarder") neglected the QOM/QDev circular dependency issue, and removed the 'proxy' between the southbridge, its PCI functions and the interrupt controller, resulting in PCI functions wiring output IRQs to 'NULL', leading to guest failures (IRQ never delivered) [1] [2]. Since we are entering feature freeze, it is safer to revert the offending patch until we figure a way to strengthen our APIs. [1] https://lore.kernel.org/qemu-devel/928a8552-ab62-9e6c-a492-d6453e338b9d@redhat.com/ [2] https://lore.kernel.org/qemu-devel/cover.1677628524.git.balaton@eik.bme.hu/ This reverts commit cef2e7148e32d61338de0220619d308bf42af770. Reported-by: Thomas Huth <thuth@redhat.com> Inspired-by: Bernhard Beschow <shentey@gmail.com> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
2023-03-07 14:10:08 +03:00
s->isa_irqs_in = i8259_init(isabus,
qemu_allocate_irq(i82378_request_out0_irq,
s, 0));
isa_bus_register_input_irqs(isabus, s->isa_irqs_in);
/* 1 82C54 (pit) */
pit = i8254_pit_init(isabus, 0x40, 0, NULL);
/* speaker */
pcspk = isa_new(TYPE_PC_SPEAKER);
object_property_set_link(OBJECT(pcspk), "pit", OBJECT(pit), &error_fatal);
if (!isa_realize_and_unref(pcspk, isabus, errp)) {
return;
}
/* 2 82C37 (dma) */
isa_create_simple(isabus, "i82374");
}
static void i82378_init(Object *obj)
{
DeviceState *dev = DEVICE(obj);
I82378State *s = I82378(obj);
qdev_init_gpio_out(dev, &s->cpu_intr, 1);
qdev_init_gpio_in(dev, i82378_request_pic_irq, 16);
}
static void i82378_class_init(ObjectClass *klass, void *data)
{
PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
DeviceClass *dc = DEVICE_CLASS(klass);
k->realize = i82378_realize;
k->vendor_id = PCI_VENDOR_ID_INTEL;
k->device_id = PCI_DEVICE_ID_INTEL_82378;
k->revision = 0x03;
k->class_id = PCI_CLASS_BRIDGE_ISA;
dc->vmsd = &vmstate_i82378;
set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
}
static const TypeInfo i82378_type_info = {
.name = TYPE_I82378,
.parent = TYPE_PCI_DEVICE,
.instance_size = sizeof(I82378State),
.instance_init = i82378_init,
.class_init = i82378_class_init,
pci: Add INTERFACE_CONVENTIONAL_PCI_DEVICE to Conventional PCI devices Add INTERFACE_CONVENTIONAL_PCI_DEVICE to all direct subtypes of TYPE_PCI_DEVICE, except: 1) The ones that already have INTERFACE_PCIE_DEVICE set: * base-xhci * e1000e * nvme * pvscsi * vfio-pci * virtio-pci * vmxnet3 2) base-pci-bridge Not all PCI bridges are Conventional PCI devices, so INTERFACE_CONVENTIONAL_PCI_DEVICE is added only to the subtypes that are actually Conventional PCI: * dec-21154-p2p-bridge * i82801b11-bridge * pbm-bridge * pci-bridge The direct subtypes of base-pci-bridge not touched by this patch are: * xilinx-pcie-root: Already marked as PCIe-only. * pcie-pci-bridge: Already marked as PCIe-only. * pcie-port: all non-abstract subtypes of pcie-port are already marked as PCIe-only devices. 3) megasas-base Not all megasas devices are Conventional PCI devices, so the interface names are added to the subclasses registered by megasas_register_types(), according to information in the megasas_devices[] array. "megasas-gen2" already implements INTERFACE_PCIE_DEVICE, so add INTERFACE_CONVENTIONAL_PCI_DEVICE only to "megasas". Acked-by: Alberto Garcia <berto@igalia.com> Acked-by: John Snow <jsnow@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Signed-off-by: Eduardo Habkost <ehabkost@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Acked-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Marcel Apfelbaum <marcel@redhat.com> Reviewed-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
2017-09-27 22:56:34 +03:00
.interfaces = (InterfaceInfo[]) {
{ INTERFACE_CONVENTIONAL_PCI_DEVICE },
{ },
},
};
static void i82378_register_types(void)
{
type_register_static(&i82378_type_info);
}
type_init(i82378_register_types)