Messages in this thread | | | Date | Mon, 20 Aug 2012 23:43:54 +0800 | From | Jiang Liu <> | Subject | Re: [PATCH v1] PCI,IA64: free associated resources when removing host bridges |
| |
On 08/16/2012 04:06 AM, Bjorn Helgaas wrote: >> --- >> >> This patch applies to >> git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git pci/next-3.5 > > x86 does the exact same thing in arch/x86/pci/acpi.c (some of the data > structures are slightly different and ia64 does have the extra > memory-mapped I/O port space regions, but they are essentially > similar). Can you rework this a bit so it looks more like the x86 > code, e.g., use the same function names and code when possible? > > Someday the x86 and ia64 code should be converged, and if the code > looks more similar, that's more likely to happen. Hi Bjorn, My original plan was to follow x86 implementation, but found there's a race condition with current "pci_set_host_bridge_release()" design. So the IA64 version is implemented in the way to fix the design flaw related to pci_set_host_bridge_release(). I will send out another patches to fix the design flaw and to enhance x86 version to follow IA64.
>> +static void shutdown_pci_controller(struct pci_host_bridge *bridge) >> +{ >> + unsigned int i; >> + struct resource *resource; >> + struct iospace_resource *iospace; >> + struct pci_controller *controller = bridge->release_data; >> + >> + if (!controller) >> + return; > > I'd remove this test because controller will be NULL only if there's a > programming error, and having the test here would cover up such > errors. Over defensive, will remove it.
Regards! Gerry
| |