On Mon, Mar 23, 2015 at 04:03:59PM +1100, Gavin Shan wrote:
> On Mon, Mar 23, 2015 at 02:43:03PM +1100, David Gibson wrote:
> >On Mon, Mar 23, 2015 at 12:56:36PM +1100, Gavin Shan wrote:
> >> On Mon, Mar 23, 2015 at 12:39:45PM +1100, David Gibson wrote:
> >> >On Sat, Mar 21, 2015 at 06:58:45AM +1100, Gavin Shan wrote:
> >> >> The patch adds one more EEH sub-command (VFIO_EEH_PE_INJECT_ERR)
> >> >> to inject the specified EEH error, which is represented by
> >> >> (struct vfio_eeh_pe_err), to the indicated PE for testing purpose.
> >> >>
> >> >> Signed-off-by: Gavin Shan <gwshan@xxxxxxxxxxxxxxxxxx>
> >> >
> >> >Reviewed-by: David Gibson <david@xxxxxxxxxxxxxxxxxxxxx>
> >> >
> >> >> ---
> >> >> Documentation/vfio.txt | 12 ++++++++++++
> >> >> drivers/vfio/vfio_spapr_eeh.c | 10 ++++++++++
> >> >> include/uapi/linux/vfio.h | 36 +++++++++++++++++++++++++++++++++++-
> >> >> 3 files changed, 57 insertions(+), 1 deletion(-)
> >> >>
> >> >> diff --git a/Documentation/vfio.txt b/Documentation/vfio.txt
> >> >> index 96978ec..c6e11a3 100644
> >> >> --- a/Documentation/vfio.txt
> >> >> +++ b/Documentation/vfio.txt
> >> >> @@ -385,6 +385,18 @@ The code flow from the example above should be slightly changed:
> >> >>
> >> >> ....
> >> >>
> >> >> + /* Inject EEH error, which is expected to be caused by 32-bits
> >> >> + * config load.
> >> >> + */
> >> >> + pe_op.op = VFIO_EEH_PE_INJECT_ERR;
> >> >> + pe_op.err.type = VFIO_EEH_ERR_TYPE_32;
> >> >> + pe_op.err.func = VFIO_EEH_ERR_FUNC_LD_CFG_ADDR;
> >> >> + pe_op.err.addr = 0ul;
> >> >> + pe_op.err.mask = 0ul;
> >> >> + ioctl(container, VFIO_EEH_PE_OP, &pe_op);
> >> >> +
> >> >> + ....
> >> >> +
> >> >> /* When 0xFF's returned from reading PCI config space or IO BARs
> >> >> * of the PCI device. Check the PE's state to see if that has been
> >> >> * frozen.
> >> >> diff --git a/drivers/vfio/vfio_spapr_eeh.c b/drivers/vfio/vfio_spapr_eeh.c
> >> >> index 5fa42db..38edeb4 100644
> >> >> --- a/drivers/vfio/vfio_spapr_eeh.c
> >> >> +++ b/drivers/vfio/vfio_spapr_eeh.c
> >> >> @@ -85,6 +85,16 @@ long vfio_spapr_iommu_eeh_ioctl(struct iommu_group *group,
> >> >> case VFIO_EEH_PE_CONFIGURE:
> >> >> 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)
> >> >> + return -EINVAL;
> >> >> + if (copy_from_user(&op, (void __user *)arg, minsz))
> >> >> + return -EFAULT;
> >> >> +
> >> >> + ret = eeh_pe_inject_err(pe, op.err.type, op.err.func,
> >> >> + op.err.addr, op.err.mask);
> >> >> + break;
> >> >> default:
> >> >> ret = -EINVAL;
> >> >> }
> >> >> diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
> >> >> index 82889c3..f68e962 100644
> >> >> --- a/include/uapi/linux/vfio.h
> >> >> +++ b/include/uapi/linux/vfio.h
> >> >> @@ -468,12 +468,23 @@ struct vfio_iommu_spapr_tce_info {
> >> >> * - unfreeze IO/DMA for frozen PE;
> >> >> * - read PE state;
> >> >> * - reset PE;
> >> >> - * - configure PE.
> >> >> + * - configure PE;
> >> >> + * - inject EEH error.
> >> >> */
> >> >> +struct vfio_eeh_pe_err {
> >> >> + __u32 type;
> >> >> + __u32 func;
> >> >> + __u64 addr;
> >> >> + __u64 mask;
> >> >> +};
> >> >> +
> >> >> struct vfio_eeh_pe_op {
> >> >> __u32 argsz;
> >> >> __u32 flags;
> >> >> __u32 op;
> >> >> + union {
> >> >> + struct vfio_eeh_pe_err err;
> >> >> + };
> >> >> };
> >> >>
> >> >> #define VFIO_EEH_PE_DISABLE 0 /* Disable EEH functionality */
> >> >> @@ -490,6 +501,29 @@ struct vfio_eeh_pe_op {
> >> >> #define VFIO_EEH_PE_RESET_HOT 6 /* Assert hot reset */
> >> >> #define VFIO_EEH_PE_RESET_FUNDAMENTAL 7 /* Assert fundamental reset */
> >> >> #define VFIO_EEH_PE_CONFIGURE 8 /* PE configuration */
> >> >> +#define VFIO_EEH_PE_INJECT_ERR 9 /* Inject EEH error */
> >> >> +#define VFIO_EEH_ERR_TYPE_32 0 /* 32-bits EEH error type */
> >> >> +#define VFIO_EEH_ERR_TYPE_64 1 /* 64-bits EEH error type */
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_MEM_ADDR 0 /* Memory load */
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_MEM_DATA 1
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_IO_ADDR 2 /* IO load */
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_IO_DATA 3
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_CFG_ADDR 4 /* Config load */
> >> >> +#define VFIO_EEH_ERR_FUNC_LD_CFG_DATA 5
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_MEM_ADDR 6 /* Memory store */
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_MEM_DATA 7
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_IO_ADDR 8 /* IO store */
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_IO_DATA 9
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_CFG_ADDR 10 /* Config store */
> >> >> +#define VFIO_EEH_ERR_FUNC_ST_CFG_DATA 11
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_RD_ADDR 12 /* DMA read */
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_RD_DATA 13
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_RD_MASTER 14
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_RD_TARGET 15
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_WR_ADDR 16 /* DMA write */
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_WR_DATA 17
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_WR_MASTER 18
> >> >> +#define VFIO_EEH_ERR_FUNC_DMA_WR_TARGET 19
> >> >
> >> >These all seem to duplicate (under slightly different names) the
> >> >defines you just added in arch/powerpc/include/asm/eeh.h. It would
> >> >make more sense to only have the uapi copy, surely.
> >> >
> >>
> >> David, Thanks for your time. The defines in eeh.h is used not only
> >> for mapping those constants defined in vfio.h, but also for validating
> >> arguments from debugfs entry "err_injct". So it'd better to keep
> >> them separate: one set is used by userspace, another set is used
> >> by host kernel.
> >
> >I don't understand why this is an argument for duplicating them. They
> >have the same meanings and values, why can't you use the uapi versions
> >everywhere?
> >
>
> We can drop the constants defined in eeh.h and just use those defines
> in vfio.h. But I prefer separate constants for some ovbious reasons
> except the one I mentioned above :)
>
> - We already had separate sets of constants for EEH and VFIO. For example,
> VFIO_EEH_PE_STATE_* in include/uapi/vfio.h always have counter part in
> arch/powerpc/asm/include/eeh.h
Oh.. I see your point.
Yes, the EEH internal code shouldn't need to use the VFIO defined
constants.
But it's still silly to duplicate, I think the correct think would be
to expose the EEH constants in uapi, and use those for VFIO as well.
> - The UAPI interface is expected to be stable, or back-compatible when
> introducing changes to it. The defines (at least their names) in eeh.h
> doesn't have to be stable.
But is there any reason not to make them stable. The VFIO case shows
that we need to lock down an interface somewhere. I don't see that
there's an advantage to having a not locked down interface in addition
to a locked down one.
> - I would like keep the code between EEH/VFIO independent to each other
> enough.
>
> Thanks,
> Gavin
>
> >> By the way, Alex Williamson, could you please ACK this if you don't
> >> find obvious errors in the patch. I really want this to be merged
> >> to 4.1.
> >>
> >> Thanks,
> >> Gavin
> >>
> >>
> >>
> >
>
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
Attachment:
pgpiaNlm6uuvL.pgp
Description: PGP signature
