On 31/05/2017 11:50, Peter Xu wrote:
> On Wed, May 31, 2017 at 05:42:34PM +0800, Peter Xu wrote:
>> A whitelist for it does not really makes sense. Let's remove it and
>> then we can introduce a blacklist when really needed, with msi_broken.
>> That's patch 1.
> Ok this paragraph does not make sense if not mentioning what's "it"...
>
> Please just read the commit messages of patch 1. It should be much
> better.
I think fixing the leak in case we have to reintroduce msi_(non)broken should
be as simple as
diff --git a/hw/misc/edu.c b/hw/misc/edu.c
index 401039c100..01acacf142 100644
--- a/hw/misc/edu.c
+++ b/hw/misc/edu.c
@@ -343,6 +343,12 @@ static void pci_edu_realize(PCIDevice *pdev, Error **errp)
EduState *edu = DO_UPCAST(EduState, pdev, pdev);
uint8_t *pci_conf = pdev->config;
+ pci_config_set_interrupt_pin(pci_conf, 1);
+
+ if (msi_init(pdev, 0, 1, true, false, errp)) {
+ return;
+ }
+
timer_init_ms(&edu->dma_timer, QEMU_CLOCK_VIRTUAL, edu_dma_timer, edu);
qemu_mutex_init(&edu->thr_mutex);
@@ -350,12 +356,6 @@ static void pci_edu_realize(PCIDevice *pdev, Error **errp)
qemu_thread_create(&edu->thread, "edu", edu_fact_thread,
edu, QEMU_THREAD_JOINABLE);
- pci_config_set_interrupt_pin(pci_conf, 1);
-
- if (msi_init(pdev, 0, 1, true, false, errp)) {
- return;
- }
-
memory_region_init_io(&edu->mmio, OBJECT(edu), &edu_mmio_ops, edu,
"edu-mmio", 1 << 20);
pci_register_bar(pdev, 0, PCI_BASE_ADDRESS_SPACE_MEMORY, &edu->mmio);
Then the two patches can be even squashed together.
Paolo
