Re: [PATCH v5 07/19] arm64: kvm: Enable access to TRBE support for host

[Mathieu Poirier]

On Tue, 30 Mar 2021 at 09:35, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Mar 30, 2021 at 09:23:14AM -0600, Mathieu Poirier wrote:
> > On Tue, Mar 30, 2021 at 11:38:18AM +0100, Suzuki K Poulose wrote:
> > > On 26/03/2021 16:55, Mathieu Poirier wrote:
> > > > On Tue, Mar 23, 2021 at 12:06:35PM +0000, Suzuki K Poulose wrote:
> > > > > For a nvhe host, the EL2 must allow the EL1&0 translation
> > > > > regime for TraceBuffer (MDCR_EL2.E2TB == 0b11). This must
> > > > > be saved/restored over a trip to the guest. Also, before
> > > > > entering the guest, we must flush any trace data if the
> > > > > TRBE was enabled. And we must prohibit the generation
> > > > > of trace while we are in EL1 by clearing the TRFCR_EL1.
> > > > >
> > > > > For vhe, the EL2 must prevent the EL1 access to the Trace
> > > > > Buffer.
> > > > >
> > > > > Cc: Will Deacon <will@xxxxxxxxxx>
> > > > > Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
> > > > > Cc: Marc Zyngier <maz@xxxxxxxxxx>
> > > > > Cc: Mark Rutland <mark.rutland@xxxxxxx>
> > > > > Cc: Anshuman Khandual <anshuman.khandual@xxxxxxx>
> > > > > Acked-by: Mathieu Poirier <mathieu.poirier@xxxxxxxxxx>
> > > > > Signed-off-by: Suzuki K Poulose <suzuki.poulose@xxxxxxx>
> > > > > ---
> > > > >   arch/arm64/include/asm/el2_setup.h | 13 +++++++++
> > > > >   arch/arm64/include/asm/kvm_arm.h   |  2 ++
> > > > >   arch/arm64/include/asm/kvm_host.h  |  2 ++
> > > > >   arch/arm64/kernel/hyp-stub.S       |  3 ++-
> > > > >   arch/arm64/kvm/debug.c             |  6 ++---
> > > > >   arch/arm64/kvm/hyp/nvhe/debug-sr.c | 42 ++++++++++++++++++++++++++++++
> > > > >   arch/arm64/kvm/hyp/nvhe/switch.c   |  1 +
> > > > >   7 files changed, 65 insertions(+), 4 deletions(-)
> > > > >
> > > >
> > > > Marc - do you want me to pick up this one?
> > >
> > > I think the kvmarm tree is the best route for this patch, given the amount
> > > of changes the tree is going through, in the areas this patch
> > > touches. Or else there would be conflicts with merging. And this patch
> > > depends on the patches from this series that were queued.
> > >
> > > Here is the depency tree :
> > >
> > > a) kvm-arm fixes for debug (Patch 1, 2) & SPE save-restore fix (queued in
> > > v5.12-rc3)
> > >
> > > b) TRBE defintions and Trace synchronization barrier (Patches 5 & 6)
> > >
> > > c) kvm-arm TRBE host support (Patch 7)
> > >
> > > d) TRBE driver support (and the ETE changes)
> > >
> > >
> > > (c) code merge depends on -> (a) + (b)
> > > (d) build (no conflicts) depends on -> (b)
> > >
> > >
> > > Now (d) has an indirect dependency on (c) for operational correctness at
> > > runtime.
> > > So, if :
> > >
> > > kvmarm tree picks up : b + c
> > > coresight tree picksup : b + d
> > >
> > > and if we could ensure the merge order of the trees are in
> > > kvmarm
> > > greg-kh (device-misc tree) (coresight goes via this tree)
> > >
> >
> > Greg's char-misc tree is based on the rc releases rather than next.  As such it
> > is a while before other branches like kvmarm get merged, causing all sort of
> > compilation breakage.
>
> My tree can not be based on -next, and neither can any other
> maintainer's tree, as next is composed of maintainer trees :)
>

Exactly

> > > we should be fine.
> > >
> > > Additionally, we could rip out the Kconfig changes from the TRBE patch
> > > and add it only at the rc1, once we verify both the trees are in to make
> > > sure the runtime operation dependency is not triggered.
> > >
> >
> > We could also do that but Greg might frown at the tactic, and rightly so.  The
> > usual way to work with complex merge dependencies is to proceed in steps, which
> > would mean that all KVM related patches go in the v5.13 merge window.  When that
> > is done we add the ETE/TRBE for the v5.14 merge window.  I agree that we waste
> > an entire cycle but it guarantees to avoid breaking builds and follows the
> > conventional way to do things.
>
> Or someone creates a single branch with a signed tag and it gets pulled
> into multiple maintainer's trees and never rebased.  We've done that
> lots of time, nothing new there.  Or everything goes through one tree,
> or you wait a release cycle.
>
> You have 3 choices, pick one :)

I'm perfectly happy with getting this entire set merged via Marc's
kvmarm tree, as long as you are fine with it.

>
> thanks,
>
> greg k-h