[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170504214303.GB9648@bhelgaas-glaptop.roam.corp.google.com>
Date: Thu, 4 May 2017 16:43:03 -0500
From: Bjorn Helgaas <helgaas@...nel.org>
To: Yinghai Lu <yinghai@...nel.org>
Cc: Bjorn Helgaas <bhelgaas@...gle.com>,
David Miller <davem@...emloft.net>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>,
Wei Yang <weiyang@...ux.vnet.ibm.com>,
Khalid Aziz <khalid.aziz@...cle.com>,
linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 10/13] PCI: Only treat non-pref mmio64 as pref if all
bridges have MEM_64
On Thu, Apr 20, 2017 at 10:04:57PM -0700, Yinghai Lu wrote:
> If any bridge up to root only have 32bit pref mmio, We don't need to
> treat device non-pref mmio64 as as pref mmio64.
I don't understand the reasoning here.
Apparently it is only safe to put a non-prefetchable PCIe BAR in a
prefetchable window if all upstream bridges have a 64-bit prefetchable
window? Why is that? Why is it important that the prefetchable
window be 64-bit? I don't see this mentioned in the implementation
note.
Does this mean "PCI: Check pref compatible bit for mem64 resource of
PCIe device" can make unsafe assignments until we apply this patch?
If so, I don't want that sort of bisection hole.
What bad thing happens without this patch?
> We need to move pci_bridge_check_ranges calling early.
> For parent bridges pref mmio BAR may not allocated by BIOS, res flags
> is still 0, we need to have it correct set before we check them for
> child device resources.
>
> -v2: check all bus resources instead of just res[15].
>
> Signed-off-by: Yinghai Lu <yinghai@...nel.org>
> Tested-by: Khalid Aziz <khalid.aziz@...cle.com>
> ---
> drivers/pci/setup-bus.c | 31 +++++++++++++++++++++++++++++--
> 1 file changed, 29 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c
> index 3de66e6..b3fd314 100644
> --- a/drivers/pci/setup-bus.c
> +++ b/drivers/pci/setup-bus.c
> @@ -735,6 +735,29 @@ int pci_claim_bridge_resource(struct pci_dev *bridge, int i)
> return -EINVAL;
> }
>
> +static bool pci_up_path_over_pref_mem64(struct pci_bus *bus)
> +{
> + if (pci_is_root_bus(bus))
> + return true;
> +
> + if (bus->self) {
> + int i;
> + bool found = false;
> + struct resource *res;
> +
> + pci_bus_for_each_resource(bus, res, i)
> + if (res->flags & IORESOURCE_MEM_64) {
> + found = true;
> + break;
> + }
> +
> + if (!found)
> + return false;
> + }
> +
> + return pci_up_path_over_pref_mem64(bus->parent);
> +}
> +
> int pci_resource_pref_compatible(const struct pci_dev *dev,
> struct resource *res)
> {
> @@ -743,7 +766,8 @@ int pci_resource_pref_compatible(const struct pci_dev *dev,
>
> if ((res->flags & IORESOURCE_MEM) &&
> (res->flags & IORESOURCE_MEM_64) &&
> - dev->on_all_pcie_path)
> + dev->on_all_pcie_path &&
> + pci_up_path_over_pref_mem64(dev->bus))
> return res->flags | IORESOURCE_PREFETCH;
>
> return res->flags;
> @@ -1236,6 +1260,10 @@ void __pci_bus_size_bridges(struct pci_bus *bus, struct list_head *realloc_head)
> struct resource *b_res;
> int ret;
>
> + if (!pci_is_root_bus(bus) &&
> + (bus->self->class >> 8) == PCI_CLASS_BRIDGE_PCI)
> + pci_bridge_check_ranges(bus);
> +
> list_for_each_entry(dev, &bus->devices, bus_list) {
> struct pci_bus *b = dev->subordinate;
> if (!b)
> @@ -1263,7 +1291,6 @@ void __pci_bus_size_bridges(struct pci_bus *bus, struct list_head *realloc_head)
> break;
>
> case PCI_CLASS_BRIDGE_PCI:
> - pci_bridge_check_ranges(bus);
> if (bus->self->is_hotplug_bridge) {
> additional_io_size = pci_hotplug_io_size;
> additional_mem_size = pci_hotplug_mem_size;
> --
> 2.9.3
>
Powered by blists - more mailing lists