[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v5 1/2] pcie: Add a simple PCIe ACS (Access Cont
From: |
Alex Williamson |
Subject: |
Re: [Qemu-devel] [PATCH v5 1/2] pcie: Add a simple PCIe ACS (Access Control Services) helper function |
Date: |
Sat, 16 Feb 2019 11:17:41 -0700 |
On Sat, 16 Feb 2019 17:51:11 +0100
Knut Omang <address@hidden> wrote:
> Implementing an ACS capability on downstream ports and multifunction
> endpoints indicates isolation and IOMMU visibility to a finer
> granularity. This creates smaller IOMMU groups in the guest and thus
> more flexibility in assigning endpoints to guest userspace or an L2
> guest.
>
> Signed-off-by: Knut Omang <address@hidden>
> ---
> hw/pci/pcie.c | 39 +++++++++++++++++++++++++++++++++++++++-
> include/hw/pci/pcie.h | 6 ++++++-
> include/hw/pci/pcie_regs.h | 4 ++++-
> 3 files changed, 49 insertions(+)
>
> diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> index 230478f..6afc37a 100644
> --- a/hw/pci/pcie.c
> +++ b/hw/pci/pcie.c
> @@ -906,3 +906,42 @@ void pcie_ats_init(PCIDevice *dev, uint16_t offset)
>
> pci_set_word(dev->wmask + dev->exp.ats_cap + PCI_ATS_CTRL, 0x800f);
> }
> +
> +/* ACS (Access Control Services) */
> +void pcie_acs_init(PCIDevice *dev, uint16_t offset)
> +{
> + bool is_downstream = pci_is_express_downstream_port(dev);
> + uint16_t cap_bits = 0;
> +
> + /* For endpoints, only multifunction devs may have an ACS capability: */
> + assert(is_downstream ||
> + (dev->cap_present & QEMU_PCI_CAP_MULTIFUNCTION) ||
> + PCI_FUNC(dev->devfn));
> +
> + pcie_add_capability(dev, PCI_EXT_CAP_ID_ACS, PCI_ACS_VER, offset,
> + PCI_ACS_SIZEOF);
> + dev->exp.acs_cap = offset;
> +
> + if (is_downstream) {
> + /* Downstream ports must implement SV, TB, RR, CR, and UF (with
> + * caveats on the latter three that we ignore for simplicity).
> + * Endpoints may also implement a subset of ACS capabilities,
> + * but these are optional if the endpoint does not support
> + * peer-to-peer between functions and thus omitted here.
> + * Downstream switch ports must also implement DT, while this
> + * is optional for root ports, so we set that as well:
> + */
Again...
https://git.qemu.org/?p=qemu.git;a=blob;f=CODING_STYLE;hb=HEAD#l127
Personally I'd add DT in the original list rather than mention is
separately, the caveats are pretty similar.
> + cap_bits = PCI_ACS_SV | PCI_ACS_TB | PCI_ACS_RR |
> + PCI_ACS_CR | PCI_ACS_UF | PCI_ACS_DT;
> + }
> +
> + pci_set_word(dev->config + offset + PCI_ACS_CAP, cap_bits);
> + pci_set_word(dev->wmask + offset + PCI_ACS_CTRL, cap_bits);
> +}
> +
> +void pcie_acs_reset(PCIDevice *dev)
> +{
> + if (dev->exp.acs_cap) {
> + pci_set_word(dev->config + dev->exp.acs_cap + PCI_ACS_CTRL, 0);
> + }
> +}
> diff --git a/include/hw/pci/pcie.h b/include/hw/pci/pcie.h
> index 5b82a0d..e30334d 100644
> --- a/include/hw/pci/pcie.h
> +++ b/include/hw/pci/pcie.h
> @@ -79,6 +79,9 @@ struct PCIExpressDevice {
>
> /* Offset of ATS capability in config space */
> uint16_t ats_cap;
> +
> + /* ACS */
> + uint16_t acs_cap;
> };
>
> #define COMPAT_PROP_PCP "power_controller_present"
> @@ -128,6 +131,9 @@ void pcie_add_capability(PCIDevice *dev,
> uint16_t offset, uint16_t size);
> void pcie_sync_bridge_lnk(PCIDevice *dev);
>
> +void pcie_acs_init(PCIDevice *dev, uint16_t offset);
> +void pcie_acs_reset(PCIDevice *dev);
> +
> void pcie_ari_init(PCIDevice *dev, uint16_t offset, uint16_t nextfn);
> void pcie_dev_ser_num_init(PCIDevice *dev, uint16_t offset, uint64_t
> ser_num);
> void pcie_ats_init(PCIDevice *dev, uint16_t offset);
> diff --git a/include/hw/pci/pcie_regs.h b/include/hw/pci/pcie_regs.h
> index ad4e780..1db86b0 100644
> --- a/include/hw/pci/pcie_regs.h
> +++ b/include/hw/pci/pcie_regs.h
> @@ -175,4 +175,8 @@ typedef enum PCIExpLinkWidth {
> PCI_ERR_COR_INTERNAL | \
> PCI_ERR_COR_HL_OVERFLOW)
>
> +/* ACS */
> +#define PCI_ACS_VER 0x1
> +#define PCI_ACS_SIZEOF 8
> +
> #endif /* QEMU_PCIE_REGS_H */