Re: [PATCH v4 2/7] hw/pci/pci.c: Don't leak PCIBus::irq_count[] in pci_bus_irqs()

[Bernhard Beschow]

Am 3. April 2023 07:41:19 UTC schrieb Bernhard Beschow <shentey@gmail.com>:
>When calling pci_bus_irqs() multiple times on the same object without calling
>pci_bus_irqs_cleanup() in between PCIBus::irq_count[] is currently leaked.
>Let's fix this because Xen will do just that in a few commits, and because
>calling pci_bus_irqs_cleanup() in between seems fragile and cumbersome.
>
>Note that pci_bus_irqs_cleanup() now has to NULL irq_count such that
>pci_bus_irqs() doesn't do a double free.
>
>Signed-off-by: Bernhard Beschow <shentey@gmail.com>

Ping PCI maintainers

>---
> hw/pci/pci.c | 2 ++
> 1 file changed, 2 insertions(+)
>
>diff --git a/hw/pci/pci.c b/hw/pci/pci.c
>index def5000e7b..be1c5d16ec 100644
>--- a/hw/pci/pci.c
>+++ b/hw/pci/pci.c
>@@ -558,6 +558,7 @@ void pci_bus_irqs(PCIBus *bus, pci_set_irq_fn set_irq,
>     bus->set_irq = set_irq;
>     bus->irq_opaque = irq_opaque;
>     bus->nirq = nirq;
>+    g_free(bus->irq_count);
>     bus->irq_count = g_malloc0(nirq * sizeof(bus->irq_count[0]));
> }
> 
>@@ -573,6 +574,7 @@ void pci_bus_irqs_cleanup(PCIBus *bus)
>     bus->irq_opaque = NULL;
>     bus->nirq = 0;
>     g_free(bus->irq_count);
>+    bus->irq_count = NULL;
> }
> 
> PCIBus *pci_register_root_bus(DeviceState *parent, const char *name,