RE: [PATCH v8 2/8] igb: handle PF/VF reset properly

[Sriram Yagnaraman]

> -----Original Message-----
> From: Philippe Mathieu-Daudé <philmd@linaro.org>
> Sent: Wednesday, 22 March 2023 11:29
> To: Sriram Yagnaraman <sriram.yagnaraman@est.tech>
> Cc: qemu-devel@nongnu.org; Akihiko Odaki <akihiko.odaki@daynix.com>;
> Jason Wang <jasowang@redhat.com>; Dmitry Fleytman
> <dmitry.fleytman@gmail.com>; Michael S . Tsirkin <mst@redhat.com>; Marcel
> Apfelbaum <marcel.apfelbaum@gmail.com>
> Subject: Re: [PATCH v8 2/8] igb: handle PF/VF reset properly
> 
> On 22/3/23 10:26, Sriram Yagnaraman wrote:
> > Use PFRSTD to reset RSTI bit for VFs, and raise VFLRE interrupt when
> > VF is reset.
> >
> > Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech>
> > ---
> >   hw/net/igb_core.c   | 33 +++++++++++++++++++++------------
> >   hw/net/igb_regs.h   |  3 +++
> >   hw/net/trace-events |  2 ++
> >   3 files changed, 26 insertions(+), 12 deletions(-)
> >
> > diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index
> > 596039aab8..fe6c7518e9 100644
> > --- a/hw/net/igb_core.c
> > +++ b/hw/net/igb_core.c
> > @@ -1895,14 +1895,6 @@ static void igb_set_eims(IGBCore *core, int
> index, uint32_t val)
> >       igb_update_interrupt_state(core);
> >   }
> >
> > -static void igb_vf_reset(IGBCore *core, uint16_t vfn) -{
> > -    /* TODO: Reset of the queue enable and the interrupt registers of the 
> > VF.
> */
> > -
> > -    core->mac[V2PMAILBOX0 + vfn] &= ~E1000_V2PMAILBOX_RSTI;
> > -    core->mac[V2PMAILBOX0 + vfn] = E1000_V2PMAILBOX_RSTD;
> > -}
> > -
> >   static void mailbox_interrupt_to_vf(IGBCore *core, uint16_t vfn)
> >   {
> >       uint32_t ent = core->mac[VTIVAR_MISC + vfn]; @@ -1980,6 +1972,17
> > @@ static void igb_set_vfmailbox(IGBCore *core, int index, uint32_t val)
> >       }
> >   }
> >
> > +static void igb_vf_reset(IGBCore *core, uint16_t vfn) {
> > +    /* disable Rx and Tx for the VF*/
> > +    core->mac[VFTE] &= ~BIT(vfn);
> > +    core->mac[VFRE] &= ~BIT(vfn);
> > +    /* indicate VF reset to PF */
> > +    core->mac[VFLRE] |= BIT(vfn);
> > +    /* VFLRE and mailbox use the same interrupt cause */
> > +    mailbox_interrupt_to_pf(core);
> > +}
> 
> Orthogonal to this patch, I'm surprised to see a function named
> igb_vf_reset() which is not called by igb_reset().

Thanks for the pertinent comment, will fix it. On PF reset, the hardware will 
assert RSTI bit on all VF mailboxes, which should in turn trigger a VF reset 
after the PF reset is complete.