[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170619210742.GA554@bhelgaas-glaptop.roam.corp.google.com>
Date: Mon, 19 Jun 2017 16:07:43 -0500
From: Bjorn Helgaas <helgaas@...nel.org>
To: haiyangz@...rosoft.com
Cc: bhelgaas@...gle.com, linux-pci@...r.kernel.org, kys@...rosoft.com,
sthemmin@...rosoft.com, olaf@...fle.de, vkuznets@...hat.com,
driverdev-devel@...uxdriverproject.org,
linux-kernel@...r.kernel.org, Christoph Hellwig <hch@...radead.org>
Subject: Re: [PATCH] pci-hyperv: Use only 16 bit integer for PCI domain
[+cc Christoph]
On Wed, May 24, 2017 at 01:39:15PM -0700, Haiyang Zhang wrote:
> From: Haiyang Zhang <haiyangz@...rosoft.com>
>
> This patch uses the lower 16 bits of the serial number as PCI
> domain, otherwise some drivers may not be able to handle it.
>
> Besides Nvidia drivers, we also found X.org, and DPDK handle
> only 16 bit PCI domain.
If you've sent patches to X.org and DPDK, please includes URLs to
them.
Christoph pointed out the conflict with VMD: vmd_find_free_domain()
allocates domains starting at 0x10000 to avoid the 16-bit domains
returned by ACPI _SEG.
I think we need a solution that works for both Nvidia/Hyper-V and VMD.
As it is, it looks like this will fix one place but break things
elsewhere.
If you believe that this will not break VMD, please explain.
Bjorn
> Signed-off-by: Haiyang Zhang <haiyangz@...rosoft.com>
> ---
> drivers/pci/host/pci-hyperv.c | 4 +++-
> 1 files changed, 3 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/pci/host/pci-hyperv.c b/drivers/pci/host/pci-hyperv.c
> index 8493638..51a815d 100644
> --- a/drivers/pci/host/pci-hyperv.c
> +++ b/drivers/pci/host/pci-hyperv.c
> @@ -1335,9 +1335,11 @@ static void put_pcichild(struct hv_pci_dev *hpdev,
> * can have shorter names than based on the bus instance UUID.
> * Only the first device serial number is used for domain, so the
> * domain number will not change after the first device is added.
> + * The lower 16 bits of the serial number is used, otherwise some
> + * drivers may not be able to handle it.
> */
> if (list_empty(&hbus->children))
> - hbus->sysdata.domain = desc->ser;
> + hbus->sysdata.domain = desc->ser & 0xFFFF;
> list_add_tail(&hpdev->list_entry, &hbus->children);
> spin_unlock_irqrestore(&hbus->device_list_lock, flags);
> return hpdev;
> --
> 1.7.1
>
Powered by blists - more mailing lists