On Mon, Jul 04, 2011 at 12:59:09PM +0200, Gerd Hoffmann wrote:
This patch adds a second device scan to the pci initialization, which counts the memory bars of the various sizes and types. Then it calculates the sizes and the packing of the prefetchable and non-prefetchable pci memory windows and prints the results.
The patch doesn't actually map the devices to make debugging easier.
Looks like the patch has a couple of minor issues (see below). But, otherwise I'm okay with it. I'll commit it if there are no further comments.
+static struct pci_bus {
- struct {
/* pci region stats */u32 count[32 - PCI_MEM_INDEX_SHIFT];u32 sum, max;/* seconday bus region sizes */u32 size;/* pci region assignments */u32 bases[32 - PCI_MEM_INDEX_SHIFT];u32 base;- } r[PCI_REGION_TYPE_COUNT];
+} *busses;
Having bus->r[x] and dev->r[x] is a bit confusing. Maybe rename dev->r to dev->bars?
[...] @@ -440,6 +508,223 @@ pci_bios_init_bus(void) { u8 pci_bus = 0; pci_bios_init_bus_rec(0 /* host bus */, &pci_bus); + busses_count = pci_bus + 1; + busses = malloc_tmphigh(busses_count * sizeof(struct pci_bus)); +}
You should be able to move this to after pci_probe() and do:
busses = malloc_tmp(sizeof(*buses) * (MaxPCIBus + 1));
Note, if you use malloc_tmphigh, then "qemu -m 1" wont boot.
[...]
+static void pci_bios_check_device(struct pci_bus *bus, struct pci_device *dev) +{
- u16 bdf = dev->bdf;
- u32 limit;
- int i,type;
- if (dev->class == PCI_CLASS_BRIDGE_PCI) {
if (dev->secondary_bus >= busses_count) {/* should never trigger */dprintf(1, "PCI: bus count too small (%d), skipping bus #%d\n",busses_count, dev->secondary_bus);return;}struct pci_bus *s = busses + dev->secondary_bus;pci_bios_check_device_in_bus(dev->secondary_bus);for (type = 0; type < PCI_REGION_TYPE_COUNT; type++) {s->r[type].size = pci_size_roundup(s->r[type].sum);
Out of curiosity - is this pci_size_roundup() needed? You wont get the best memory usage this way.
[...]
+static void pci_bios_map_device_in_bus(int bus) +{
- struct pci_device *pci;
- foreachpci(pci) {
if (pci->rootbus != bus)continue;
rootbus isn't correct here.
[...]
- /* matches acpi-dsdt.dsl */
- u32 start = 0xe0000000;
- u32 end = BUILD_IOAPIC_ADDR;
This should use BUILD_PCIMEM_START instead of 0xe0000000. If there is a change from 0xf0000000 to 0xe0000000 it should be in a separate patch.
-Kevin