github/acrn-hypervisor
1
0
Fork 0
mirror of https://github.com/projectacrn/acrn-hypervisor.git synced 2025-06-20 12:42:54 +00:00
Code Issues Projects Releases Wiki Activity

dm: check if PTM root is enabled

Browse Source 
Since hv is going to ensure that PTM is always enabled on root port if it is
PTM root capable, add a sanity check whether PTM is enabled on root port.  If
not, don't allow guest to change the setting of physical PTM (Guest OS
can still enable PTM in virtual root port however PTM does not function properly
since it is not enabled on physical root port).
Also revisit a few comments to add assumption that acrn only support a
simple PTM hierarch emulation i.e., EP (ptm requestor) is directly connected
to root port (ptm root), or ptm requestor is rcie).

V3:
 - When users tries to enable PTM while PTM root is not enabled on
physical root port, allow user to launch VM and enable PTM in virtual
root port.  This is not going to enable PTM on physical root port.

Tracked-On: #5915
Signed-off-by: Rong Liu <rong.l.liu@intel.com>
This commit is contained in:
Rong Liu 2021-05-24 00:50:44 +00:00 committed by wenlingz
parent d51ac9d368
commit 01b5aa8cb7
1 changed files with 37 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
devicemodel/hw/pci/ptm.c
View File

@ -42,25 +42,23 @@ static int ptm_root_port_secondary_bus;
/* get ptm capability register value */
static int
get_ptm_cap(struct pci_device *pdev, int *offset)
get_ptm_reg_value(struct pci_device *pdev, int reg)
{
int pos;
uint32_t cap;
uint32_t reg_val;
pos = pci_find_ext_cap(pdev, PCIZ_PTM);
if (!pos) {
*offset = 0;
return 0;
}
*offset = pos;
pci_device_cfg_read_u32(pdev, &cap, pos + PCIR_PTM_CAP);
pci_device_cfg_read_u32(pdev, &reg_val, pos + reg);
pr_notice("<PTM>-%s: device [%x:%x.%x]: ptm-cap=0x%x, offset=0x%x.\n",
__func__, pdev->bus, pdev->dev, pdev->func, cap, *offset);
pr_notice("<PTM>-%s: device [%x:%x.%x]: ptm pos=0x%x, ptm reg val=0x%x.\n",
__func__, pdev->bus, pdev->dev, pdev->func, pos, reg_val);
return cap;
return reg_val;
}
/* add virtual root port to hv */
@ -106,7 +104,7 @@ add_vroot_port(struct vmctx *ctx, struct passthru_dev *ptdev, struct pci_device
int ptm_probe(struct vmctx *ctx, struct passthru_dev *ptdev, int *vrp_sec_bus)
{
int error = 0;
int pos, pcie_type, cap, rp_ptm_offset, device_ptm_offset;
int pos, pcie_type, cap, rp_ptm_offset;
struct pci_device *phys_dev = ptdev->phys_dev;
struct pci_device *rp;
struct pci_device_info rp_info = {};
@ -133,13 +131,17 @@ int ptm_probe(struct vmctx *ctx, struct passthru_dev *ptdev, int *vrp_sec_bus)
pcie_type = pci_get_pcie_type(phys_dev);
/* Assume that PTM requestor can be enabled on pci ep or rcie.
* If PTM is enabled on an ep, PTM hierarchy requires it has an upstream
* PTM root. Based on available HW platforms, currently PTM root
* capability is implemented in pci root port.
/* The following sanity checks are based on these assumptions:
* 1. PTM requestor can be enabled on pci ep or rcie.
* 2. HW implements this simple PTM hierarchy: PTM requestor (EP) is
* directly connected to PTM root (root port), or requestor itself is RCIE.
* 3. There is no intermediate nodes (such as switch) in between the PTM
* root and PTM requestor.
* 4. HW only implements one PCI domain in the system and only one PTM
* domain implemented in this PCI domain
*/
if (pcie_type == PCIEM_TYPE_ENDPOINT) {
cap = get_ptm_cap(phys_dev, &device_ptm_offset);
cap = get_ptm_reg_value(phys_dev, PCIR_PTM_CAP);
if (!(cap & PCIM_PTM_CAP_REQ)) {
pr_err("%s Error: %x:%x.%x must be PTM requestor.\n", __func__,
phys_dev->bus, phys_dev->dev, phys_dev->func);
@ -153,18 +155,19 @@ int ptm_probe(struct vmctx *ctx, struct passthru_dev *ptdev, int *vrp_sec_bus)
return -ENODEV;
}
cap = get_ptm_cap(rp, &rp_ptm_offset);
/* check whether root port is PTM root-capable */
cap = get_ptm_reg_value(rp, PCIR_PTM_CAP);
if (!(cap & PCIM_PTM_CAP_ROOT)) {
pr_err("%s Error: root port %x:%x.%x of %x:%x.%x is not PTM root.\n",
__func__, rp->bus, rp->dev,
rp->func, phys_dev->bus, phys_dev->dev, phys_dev->func);
pr_err("%s Error: root port %x:%x.%x of %x:%x.%x is not PTM root capable.\n",
__func__, rp->bus, rp->dev, rp->func,
phys_dev->bus, phys_dev->dev, phys_dev->func);
return -EINVAL;
}
/* check whether more than one devices are connected to the root port.
* if more than one devices are connected to the root port, we flag
* this as error and won't enable PTM. We do this just because we
* don't have this hw configuration and won't be able totest this case.
* don't have this hw configuration and won't be able to test this case.
*/
error = pci_device_get_bridge_buses(rp, &(rp_info.primary_bus),
&(rp_info.secondary_bus), &(rp_info.subordinate_bus));
@ -175,11 +178,23 @@ int ptm_probe(struct vmctx *ctx, struct passthru_dev *ptdev, int *vrp_sec_bus)
return -EINVAL;
}
// add virtual root port
*vrp_sec_bus = add_vroot_port(ctx, ptdev, rp, rp_ptm_offset);
/* hv is responsible to ensure that PTM is enabled on hw root port if
* root port is PTM root-capable. If PTM root is not enabled already in physical
* root port before guest launch, guest OS can only enable it in root port's virtual
* config space and PTM may not function as desired.
*/
cap = get_ptm_reg_value(rp, PCIR_PTM_CTRL);
if (!(cap & PCIM_PTM_CTRL_ENABLE) || !(cap & PCIM_PTM_CTRL_ROOT_SELECT)) {
pr_err("%s Warning: guest is not allowed to enable PTM on root port %x:%x.%x.\n",
__func__, rp->bus, rp->dev, rp->func);
}
rp_ptm_offset = pci_find_ext_cap(rp, PCIZ_PTM);
/* add virtual root port */
*vrp_sec_bus = add_vroot_port(ctx, ptdev, rp, rp_ptm_offset);
} else if (pcie_type == PCIEM_TYPE_ROOT_INT_EP) {
// No need to emulate root port if ptm requestor is RCIE
/* Do NOT emulate root port if ptm requestor is RCIE */
pr_notice("%s: ptm requestor is root complex integrated device.\n",
__func__);
} else {