@@ -724,7 +724,8 @@ static long tce_iommu_ioctl(void *iommu_data,
ret = 1;
break;
default:
- ret = vfio_spapr_iommu_eeh_ioctl(NULL, cmd, arg);
+ ret = vfio_spapr_iommu_eeh_ioctl(NULL, cmd,
+ arg, NULL, 0);
break;
}
@@ -953,17 +954,55 @@ static long tce_iommu_ioctl(void *iommu_data,
return 0;
case VFIO_EEH_PE_OP: {
- struct tce_iommu_group *tcegrp;
- int eeh_enabled;
+ struct tce_iommu_group *tmp, *tcegrp;
+ struct vfio_eeh_pe_op op;
+ int enabled, flag;
+
+ enabled = vfio_spapr_pci_eeh_enabled();
+ if (enabled == VFIO_EEH_DISABLED)
+ return -ENOTTY;
- eeh_enabled = vfio_spapr_pci_eeh_enabled();
- if (eeh_enabled == VFIO_EEH_DISABLED)
+ /* Get the specified version */
+ minsz = offsetofend(struct vfio_eeh_pe_op, flags);
+ if (copy_from_user(&op, (void __user *)arg, minsz))
+ return -EFAULT;
+ flag = (op.flags & VFIO_EEH_ENABLED_MASK);
+ if (flag > enabled)
return -ENOTTY;
+ else if (flag == VFIO_EEH_DISABLED)
+ flag = VFIO_EEH_ENABLED_V1;
+
+ if (flag == VFIO_EEH_ENABLED_V1)
+ minsz = offsetofend(struct vfio_eeh_pe_op, op);
+ else if (flag == VFIO_EEH_ENABLED_V2)
+ minsz = offsetofend(struct vfio_eeh_pe_op, groupid);
+ if (copy_from_user(&op, (void __user *)arg, minsz))
+ return -EFAULT;
+
+ if (op.argsz < minsz)
+ return -EINVAL;
+
+ if (flag == VFIO_EEH_ENABLED_V2) {
+ tcegrp = NULL;
+ list_for_each_entry(tmp, &container->group_list, next) {
+ if (tmp->grp &&
+ iommu_group_id(tmp->grp) == op.groupid) {
+ tcegrp = tmp;
+ break;
+ }
+ }
+
+ if (!tcegrp)
+ return -ENODEV;
+
+ return vfio_spapr_iommu_eeh_ioctl(tcegrp->grp,
+ cmd, arg, &op, flag);
+ }
ret = 0;
list_for_each_entry(tcegrp, &container->group_list, next) {
ret = vfio_spapr_iommu_eeh_ioctl(tcegrp->grp,
- cmd, arg);
+ cmd, arg, &op, flag);
if (ret)
return ret;
}
@@ -20,7 +20,7 @@
int vfio_spapr_pci_eeh_enabled(void)
{
- return VFIO_EEH_ENABLED_V1;
+ return VFIO_EEH_ENABLED_V2;
}
EXPORT_SYMBOL_GPL(vfio_spapr_pci_eeh_enabled);
@@ -38,11 +38,12 @@ void vfio_spapr_pci_eeh_release(struct pci_dev *pdev)
EXPORT_SYMBOL_GPL(vfio_spapr_pci_eeh_release);
long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
- unsigned int cmd, unsigned long arg)
+ unsigned int cmd, unsigned long arg,
+ void *parm, int flag)
{
struct eeh_pe *pe;
- struct vfio_eeh_pe_op op;
- unsigned long minsz;
+ struct vfio_eeh_pe_op *op;
+ unsigned long src, dst, len;
long ret = -EINVAL;
switch (cmd) {
@@ -54,17 +55,12 @@ long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
ret = 0;
break;
case VFIO_EEH_PE_OP:
+ op = (struct vfio_eeh_pe_op *)parm;
pe = eeh_iommu_group_to_pe(group);
if (!pe)
return -ENODEV;
- minsz = offsetofend(struct vfio_eeh_pe_op, op);
- if (copy_from_user(&op, (void __user *)arg, minsz))
- return -EFAULT;
- if (op.argsz < minsz || op.flags)
- return -EINVAL;
-
- switch (op.op) {
+ switch (op->op) {
case VFIO_EEH_PE_DISABLE:
ret = eeh_pe_set_option(pe, EEH_OPT_DISABLE);
break;
@@ -93,14 +89,25 @@ long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
ret = eeh_pe_configure(pe);
break;
case VFIO_EEH_PE_INJECT_ERR:
- minsz = offsetofend(struct vfio_eeh_pe_op, err.mask);
- if (op.argsz < minsz)
+ if (flag == VFIO_EEH_ENABLED_V1)
+ src = offsetofend(struct vfio_eeh_pe_op, op);
+ else if (flag == VFIO_EEH_ENABLED_V2)
+ src = offsetofend(struct vfio_eeh_pe_op,
+ groupid);
+ else
+ return -ENOTTY;
+
+ len = sizeof(struct vfio_eeh_pe_err);
+ if (op->argsz < src + len)
return -EINVAL;
- if (copy_from_user(&op, (void __user *)arg, minsz))
+
+ dst = offsetofend(struct vfio_eeh_pe_op, groupid);
+ if (copy_from_user(parm + dst,
+ (void __user *)(arg + src), len))
return -EFAULT;
- ret = eeh_pe_inject_err(pe, op.err.type, op.err.func,
- op.err.addr, op.err.mask);
+ ret = eeh_pe_inject_err(pe, op->err.type, op->err.func,
+ op->err.addr, op->err.mask);
break;
default:
ret = -EINVAL;
@@ -95,8 +95,8 @@ extern int vfio_spapr_pci_eeh_enabled(void);
extern void vfio_spapr_pci_eeh_open(struct pci_dev *pdev);
extern void vfio_spapr_pci_eeh_release(struct pci_dev *pdev);
extern long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
- unsigned int cmd,
- unsigned long arg);
+ unsigned int cmd, unsigned long arg,
+ void *param, int flag);
#else
static inline int vfio_spapr_pci_eeh_enabled(void)
{
@@ -113,7 +113,8 @@ static inline void vfio_spapr_pci_eeh_release(struct pci_dev *pdev)
static inline long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
unsigned int cmd,
- unsigned long arg)
+ unsigned long arg,
+ void *param, int flag)
{
return -ENOTTY;
}
@@ -497,6 +497,7 @@ struct vfio_iommu_spapr_tce_info {
*/
#define VFIO_EEH_DISABLED 0
#define VFIO_EEH_ENABLED_V1 1
+#define VFIO_EEH_ENABLED_V2 2
struct vfio_eeh_pe_err {
__u32 type;
@@ -508,7 +509,9 @@ struct vfio_eeh_pe_err {
struct vfio_eeh_pe_op {
__u32 argsz;
__u32 flags;
+#define VFIO_EEH_ENABLED_MASK 0xFF
__u32 op;
+ __u32 groupid;
union {
struct vfio_eeh_pe_err err;
};
Currently, EEH module works based on the assumption that every container has only one attached IOMMU group. It's not true any more. So the userland has to specify the IOMMU group (PE) to which the requested EEH operation is applied. This exposes "v2" interface for the userland to specify IOMMU group (PE) ID when requesting EEH operation. Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com> --- drivers/vfio/vfio_iommu_spapr_tce.c | 51 ++++++++++++++++++++++++++++++++----- drivers/vfio/vfio_spapr_eeh.c | 39 ++++++++++++++++------------ include/linux/vfio.h | 7 ++--- include/uapi/linux/vfio.h | 3 +++ 4 files changed, 75 insertions(+), 25 deletions(-)