On Fri, Feb 20, 2015 at 09:50:45AM +1100, Gavin Shan wrote:
> On Wed, Feb 18, 2015 at 07:28:29PM -0600, Michael Roth wrote:
> >Quoting Gavin Shan (2015-02-15 23:32:09)
> >> On Mon, Feb 16, 2015 at 12:52:48PM +1100, David Gibson wrote:
> >> >On Mon, Feb 16, 2015 at 10:16:01AM +1100, Gavin Shan wrote:
> >> >> The emulation for EEH RTAS requests from guest isn't covered
> >> >> by QEMU yet and the patch implements them.
> >> >>
> >> >> The patch defines constants used by EEH RTAS calls and adds
> >> >> callbacks sPAPRPHBClass::{eeh_set_option, eeh_get_state, eeh_reset,
> >> >> eeh_configure}, which are going to be used as follows:
> >> >>
> >> >> * RTAS calls are received in spapr_pci.c, sanity check is done
> >> >> there.
> >> >> * RTAS handlers handle what they can. If there is something it
> >> >> cannot handle and the corresponding sPAPRPHBClass callback is
> >> >> defined, it is called.
> >> >> * Those callbacks are only implemented for VFIO now. They do ioctl()
> >> >> to the IOMMU container fd to complete the calls. Error codes from
> >> >> that ioctl() are transferred back to the guest.
> >
> >
> >
> >> >>
> >> >> [aik: defined RTAS tokens for EEH RTAS calls]
> >> >> Signed-off-by: Gavin Shan <gws...@linux.vnet.ibm.com>
> >> >> ---
> >> >> hw/ppc/spapr_pci.c | 281
> >> >> ++++++++++++++++++++++++++++++++++++++++++++
> >> >> include/hw/pci-host/spapr.h | 4 +
> >> >> include/hw/ppc/spapr.h | 43 ++++++-
> >> >> 3 files changed, 326 insertions(+), 2 deletions(-)
> >> >>
> >> >> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> >> >> index cebdeb3..29b071d 100644
> >> >> --- a/hw/ppc/spapr_pci.c
> >> >> +++ b/hw/ppc/spapr_pci.c
> >> >> @@ -406,6 +406,268 @@ static void
> >> >> rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu,
> >> >> rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */
> >> >> }
> >> >>
> >> >> +static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
> >> >> + sPAPREnvironment *spapr,
> >> >> + uint32_t token, uint32_t nargs,
> >> >> + target_ulong args, uint32_t nret,
> >> >> + target_ulong rets)
> >> >> +{
> >> >> + sPAPRPHBState *sphb;
> >> >> + sPAPRPHBClass *spc;
> >> >> + uint32_t addr, option;
> >> >> + uint64_t buid;
> >> >> + int ret;
> >> >> +
> >> >> + if ((nargs != 4) || (nret != 1)) {
> >> >> + goto param_error_exit;
> >> >> + }
> >> >> +
> >> >> + buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
> >> >> + addr = rtas_ld(args, 0);
> >> >> + option = rtas_ld(args, 3);
> >> >> +
> >> >> + sphb = find_phb(spapr, buid);
> >> >> + if (!sphb) {
> >> >> + goto param_error_exit;
> >> >> + }
> >> >> +
> >> >> + spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
> >> >> + if (!spc->eeh_set_option) {
> >> >> + goto param_error_exit;
> >> >> + }
> >> >> +
> >> >> + /*
> >> >> + * The EEH functionality is enabled on basis of PCI device,
> >> >> + * instead of PE. We need check the validity of the PCI
> >> >> + * device address.
> >> >> + */
> >> >> + if (option == RTAS_EEH_ENABLE &&
> >> >> + !find_dev(spapr, buid, addr)) {
> >> >> + goto param_error_exit;
> >> >> + }
> >> >
> >> >You're still breaking your layering by doing checks dependent on the
> >> >specific option both here and in the callback.
> >> >
> >> >What I meant by my comments on the previous version was that this
> >> >find_dev() test should also move into the eeh_set_option callback.
> >> >Obviously that means adding addr into the parameters - but surely if
> >> >the addr has any meaning whatsoever, it must be at least potentially
> >> >needed by the callback anyway.
> >> >
> >>
> >> Ok. Either simply dropping the check here, or moving find_dev() to
> >> sPAPRPHBClass::eeh_set_option() as you suggested. However, there're more
> >> things needed for sPAPRPHBClass::eeh_set_option() to do the check as
> >> follows.
> >> David, could you help to confirm which way you prefer?
> >>
> >> - Rename find_dev() to spapr_find_pci_dev() and make it public. It will be
> >> called in spapr_pci_vfio.c
> >> - Add one field sPAPRPHBState::spapr to reference the associated
> >> sPAPREnvironment,
> >> which is required by spapr_find_pci_dev(). Otherwise, we have to pass
> >> sPAPREnvironment
> >> to sPAPRPHBClass::eeh_set_option().
> >
> >AFAICT spapr_pci.c:find_dev() only needs sPAPREnvironment to look up the phb
> >given a buid, but in your case you already have the phb and pass it on to
> >eeh_set_option(), so within eeh_set_option() you can call pci_find_device()
> >just like spapr_pci.c:find_dev() does to do the validation.
> >
>
> Yeah, it's another option I was thinking about. It would introduce
> duplicate code, but it seems it's the best way to go. I'll update
> accordingly in next revision. Thanks for your comments.
This sounds like the best option to me for now.
> >The validation seems to assume the addr value is a config_addr for the device
> >though, isn't it possible we might recieve a pe_addr of the form returned
> >by rtas_ibm_get_config_addr_info2? That value would happen to correspond to
> >bus:n,device:0,func:0,reg:1, and find_dev in that case would just mask off
> >the reg value and verify there's a device in PCI slot 0, instead of whatever
> >actually needs to be validated in that situation (which isn't clear to me).
> >
>
> Yeah, The address passed to rtas_ibm_set_eeh_option() could be device's
> config_addr or PE_addr depending on the options. For option EEH_ENABLE,
> it's device's config_addr.
Ah.. ok, that complicates matters. But that's definitely another
reason to move the validation into the callback that's already
checking the options.
--
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
pgpIAZ4cxZs2V.pgp
Description: PGP signature
