Currently 64-bit PCI BARs are unconditionally mapped by BIOS right over 4G + RamSizeOver4G location, which doesn't allow to reserve extra space before 64-bit PCI window. For memory hotplug an extra RAM space might be reserved after present 64-bit RAM end and BIOS should map 64-bit PCI BARs after it.
Introduce "etc/pcimem64-start" romfile to provide BIOS a hint where it should start mapping of 64-bit PCI BARs. If romfile is missing BIOS reverts to legacy behavior and starts mapping right after high memory.
Signed-off-by: Igor Mammedov imammedo@redhat.com --- src/fw/pciinit.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)
diff --git a/src/fw/pciinit.c b/src/fw/pciinit.c index b29db99..f056633 100644 --- a/src/fw/pciinit.c +++ b/src/fw/pciinit.c @@ -18,6 +18,8 @@ #include "paravirt.h" // RamSize #include "string.h" // memset #include "util.h" // pci_setup +#include "byteorder.h" // le64_to_cpu +#include "romfile.h" // romfile_loadint
#define PCI_DEVICE_MEM_MIN 0x1000 #define PCI_BRIDGE_IO_MIN 0x1000 @@ -764,6 +766,8 @@ static void pci_bios_map_devices(struct pci_bus *busses) { if (pci_bios_init_root_regions(busses)) { struct pci_region r64_mem, r64_pref; + u64 base64 = le64_to_cpu(romfile_loadint("etc/pcimem64-start", + 0x100000000ULL + RamSizeOver4G)); r64_mem.list.first = NULL; r64_pref.list.first = NULL; pci_region_migrate_64bit_entries(&busses[0].r[PCI_REGION_TYPE_MEM], @@ -779,7 +783,7 @@ static void pci_bios_map_devices(struct pci_bus *busses) u64 align_mem = pci_region_align(&r64_mem); u64 align_pref = pci_region_align(&r64_pref);
- r64_mem.base = ALIGN(0x100000000LL + RamSizeOver4G, align_mem); + r64_mem.base = ALIGN(base64, align_mem); r64_pref.base = ALIGN(r64_mem.base + sum_mem, align_pref); pcimem64_start = r64_mem.base; pcimem64_end = r64_pref.base + sum_pref;
On Fr, 2013-10-11 at 12:19 +0200, Igor Mammedov wrote:
Currently 64-bit PCI BARs are unconditionally mapped by BIOS right over 4G + RamSizeOver4G location, which doesn't allow to reserve extra space before 64-bit PCI window. For memory hotplug an extra RAM space might be reserved after present 64-bit RAM end and BIOS should map 64-bit PCI BARs after it.
Introduce "etc/pcimem64-start" romfile to provide BIOS a hint where it should start mapping of 64-bit PCI BARs. If romfile is missing BIOS reverts to legacy behavior and starts mapping right after high memory.
Looks good overall.
struct pci_region r64_mem, r64_pref;
u64 base64 = le64_to_cpu(romfile_loadint("etc/pcimem64-start",0x100000000ULL + RamSizeOver4G));
Should we sanity-check base64 maybe? Make sure it isn't smaller than 0x100000000ULL + RamSizeOver4G?
cheers, Gerd
On Fri, 11 Oct 2013 14:35:50 +0200 Gerd Hoffmann kraxel@redhat.com wrote:
On Fr, 2013-10-11 at 12:19 +0200, Igor Mammedov wrote:
Currently 64-bit PCI BARs are unconditionally mapped by BIOS right over 4G + RamSizeOver4G location, which doesn't allow to reserve extra space before 64-bit PCI window. For memory hotplug an extra RAM space might be reserved after present 64-bit RAM end and BIOS should map 64-bit PCI BARs after it.
Introduce "etc/pcimem64-start" romfile to provide BIOS a hint where it should start mapping of 64-bit PCI BARs. If romfile is missing BIOS reverts to legacy behavior and starts mapping right after high memory.
Looks good overall.
struct pci_region r64_mem, r64_pref;
u64 base64 = le64_to_cpu(romfile_loadint("etc/pcimem64-start",0x100000000ULL + RamSizeOver4G));Should we sanity-check base64 maybe? Make sure it isn't smaller than 0x100000000ULL + RamSizeOver4G?
If it would be less than 0x100000000ULL + RamSizeOver4G then it would mean that QEMU for some reason decided to overshadow present RAM with PCI hole.
It would add a couple of extra bytes to BIOS and point where it should fail or at least print a warning that nobody will read at runtime. I think it's better to assert on QEMU side and abort even before guest started if it's considered as error.
cheers, Gerd
-
Gerd Hoffmann -
Igor Mammedov