diff options
| author | Benjamin Herrenschmidt <benh@kernel.crashing.org> | 2018-01-12 05:37:12 +0300 |
|---|---|---|
| committer | Paul Mackerras <paulus@ozlabs.org> | 2018-01-19 04:10:21 +0300 |
| commit | bf4159da4751ab8eea43ca6e7c49193dbce8398c (patch) | |
| tree | f388b8cce6117fd8dd3d592dfa18d7d290ea35a9 /arch/powerpc/kvm/book3s_xive.h | |
| parent | c424c108233dc422a9a29ee833154006a5bdf9fc (diff) | |
| download | linux-bf4159da4751ab8eea43ca6e7c49193dbce8398c.tar.xz | |
KVM: PPC: Book3S HV: Enable use of the new XIVE "single escalation" feature
That feature, provided by Power9 DD2.0 and later, when supported
by newer OPAL versions, allows us to sacrifice a queue (priority 7)
in favor of merging all the escalation interrupts of the queues
of a single VP into a single interrupt.
This reduces the number of host interrupts used up by KVM guests
especially when those guests use multiple priorities.
It will also enable a future change to control the masking of the
escalation interrupts more precisely to avoid spurious ones.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Diffstat (limited to 'arch/powerpc/kvm/book3s_xive.h')
| -rw-r--r-- | arch/powerpc/kvm/book3s_xive.h | 15 |
1 files changed, 6 insertions, 9 deletions
diff --git a/arch/powerpc/kvm/book3s_xive.h b/arch/powerpc/kvm/book3s_xive.h index 6ba63f8e8a61..a08ae6fd4c51 100644 --- a/arch/powerpc/kvm/book3s_xive.h +++ b/arch/powerpc/kvm/book3s_xive.h @@ -120,6 +120,8 @@ struct kvmppc_xive { u32 q_order; u32 q_page_order; + /* Flags */ + u8 single_escalation; }; #define KVMPPC_XIVE_Q_COUNT 8 @@ -201,25 +203,20 @@ static inline struct kvmppc_xive_src_block *kvmppc_xive_find_source(struct kvmpp * is as follow. * * Guest request for 0...6 are honored. Guest request for anything - * higher results in a priority of 7 being applied. - * - * However, when XIRR is returned via H_XIRR, 7 is translated to 0xb - * in order to match AIX expectations + * higher results in a priority of 6 being applied. * * Similar mapping is done for CPPR values */ static inline u8 xive_prio_from_guest(u8 prio) { - if (prio == 0xff || prio < 8) + if (prio == 0xff || prio < 6) return prio; - return 7; + return 6; } static inline u8 xive_prio_to_guest(u8 prio) { - if (prio == 0xff || prio < 7) - return prio; - return 0xb; + return prio; } static inline u32 __xive_read_eq(__be32 *qpage, u32 msk, u32 *idx, u32 *toggle) |