From: Alexander Graf <agraf@suse.de>
To: kvm-ppc@vger.kernel.org
Cc: kvm list <kvm@vger.kernel.org>, Avi Kivity <avi@redhat.com>,
Marcelo Tosatti <mtosatti@redhat.com>,
Paul Mackerras <paulus@samba.org>
Subject: [PATCH 31/52] KVM: PPC: Make the H_ENTER hcall more reliable
Date: Fri, 13 Jan 2012 15:31:34 +0100 [thread overview]
Message-ID: <1326465115-5976-32-git-send-email-agraf@suse.de> (raw)
In-Reply-To: <1326465115-5976-1-git-send-email-agraf@suse.de>
From: Paul Mackerras <paulus@samba.org>
At present, our implementation of H_ENTER only makes one try at locking
each slot that it looks at, and doesn't even retry the ldarx/stdcx.
atomic update sequence that it uses to attempt to lock the slot. Thus
it can return the H_PTEG_FULL error unnecessarily, particularly when
the H_EXACT flag is set, meaning that the caller wants a specific PTEG
slot.
This improves the situation by making a second pass when no free HPTE
slot is found, where we spin until we succeed in locking each slot in
turn and then check whether it is full while we hold the lock. If the
second pass fails, then we return H_PTEG_FULL.
This also moves lock_hpte to a header file (since later commits in this
series will need to use it from other source files) and renames it to
try_lock_hpte, which is a somewhat less misleading name.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
---
arch/powerpc/include/asm/kvm_book3s_64.h | 25 ++++++++++++
arch/powerpc/kvm/book3s_hv_rm_mmu.c | 63 ++++++++++++++++--------------
2 files changed, 59 insertions(+), 29 deletions(-)
diff --git a/arch/powerpc/include/asm/kvm_book3s_64.h b/arch/powerpc/include/asm/kvm_book3s_64.h
index fa3dc79..300ec04 100644
--- a/arch/powerpc/include/asm/kvm_book3s_64.h
+++ b/arch/powerpc/include/asm/kvm_book3s_64.h
@@ -43,6 +43,31 @@ static inline void svcpu_put(struct kvmppc_book3s_shadow_vcpu *svcpu)
#define HPT_HASH_MASK (HPT_NPTEG - 1)
#endif
+/*
+ * We use a lock bit in HPTE dword 0 to synchronize updates and
+ * accesses to each HPTE, and another bit to indicate non-present
+ * HPTEs.
+ */
+#define HPTE_V_HVLOCK 0x40UL
+
+static inline long try_lock_hpte(unsigned long *hpte, unsigned long bits)
+{
+ unsigned long tmp, old;
+
+ asm volatile(" ldarx %0,0,%2\n"
+ " and. %1,%0,%3\n"
+ " bne 2f\n"
+ " ori %0,%0,%4\n"
+ " stdcx. %0,0,%2\n"
+ " beq+ 2f\n"
+ " li %1,%3\n"
+ "2: isync"
+ : "=&r" (tmp), "=&r" (old)
+ : "r" (hpte), "r" (bits), "i" (HPTE_V_HVLOCK)
+ : "cc", "memory");
+ return old == 0;
+}
+
static inline unsigned long compute_tlbie_rb(unsigned long v, unsigned long r,
unsigned long pte_index)
{
diff --git a/arch/powerpc/kvm/book3s_hv_rm_mmu.c b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
index 84dae82..a28a603 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_mmu.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
@@ -53,26 +53,6 @@ static void *real_vmalloc_addr(void *x)
return __va(addr);
}
-#define HPTE_V_HVLOCK 0x40UL
-
-static inline long lock_hpte(unsigned long *hpte, unsigned long bits)
-{
- unsigned long tmp, old;
-
- asm volatile(" ldarx %0,0,%2\n"
- " and. %1,%0,%3\n"
- " bne 2f\n"
- " ori %0,%0,%4\n"
- " stdcx. %0,0,%2\n"
- " beq+ 2f\n"
- " li %1,%3\n"
- "2: isync"
- : "=&r" (tmp), "=&r" (old)
- : "r" (hpte), "r" (bits), "i" (HPTE_V_HVLOCK)
- : "cc", "memory");
- return old == 0;
-}
-
long kvmppc_h_enter(struct kvm_vcpu *vcpu, unsigned long flags,
long pte_index, unsigned long pteh, unsigned long ptel)
{
@@ -126,24 +106,49 @@ long kvmppc_h_enter(struct kvm_vcpu *vcpu, unsigned long flags,
pteh &= ~0x60UL;
ptel &= ~(HPTE_R_PP0 - kvm->arch.ram_psize);
ptel |= pa;
+
if (pte_index >= HPT_NPTE)
return H_PARAMETER;
if (likely((flags & H_EXACT) == 0)) {
pte_index &= ~7UL;
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
- for (i = 0; ; ++i) {
- if (i == 8)
- return H_PTEG_FULL;
+ for (i = 0; i < 8; ++i) {
if ((*hpte & HPTE_V_VALID) == 0 &&
- lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID))
+ try_lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID))
break;
hpte += 2;
}
+ if (i == 8) {
+ /*
+ * Since try_lock_hpte doesn't retry (not even stdcx.
+ * failures), it could be that there is a free slot
+ * but we transiently failed to lock it. Try again,
+ * actually locking each slot and checking it.
+ */
+ hpte -= 16;
+ for (i = 0; i < 8; ++i) {
+ while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
+ cpu_relax();
+ if ((*hpte & HPTE_V_VALID) == 0)
+ break;
+ *hpte &= ~HPTE_V_HVLOCK;
+ hpte += 2;
+ }
+ if (i == 8)
+ return H_PTEG_FULL;
+ }
pte_index += i;
} else {
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
- if (!lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID))
- return H_PTEG_FULL;
+ if (!try_lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID)) {
+ /* Lock the slot and check again */
+ while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
+ cpu_relax();
+ if (*hpte & HPTE_V_VALID) {
+ *hpte &= ~HPTE_V_HVLOCK;
+ return H_PTEG_FULL;
+ }
+ }
}
/* Save away the guest's idea of the second HPTE dword */
@@ -189,7 +194,7 @@ long kvmppc_h_remove(struct kvm_vcpu *vcpu, unsigned long flags,
if (pte_index >= HPT_NPTE)
return H_PARAMETER;
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
- while (!lock_hpte(hpte, HPTE_V_HVLOCK))
+ while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
cpu_relax();
if ((hpte[0] & HPTE_V_VALID) == 0 ||
((flags & H_AVPN) && (hpte[0] & ~0x7fUL) != avpn) ||
@@ -248,7 +253,7 @@ long kvmppc_h_bulk_remove(struct kvm_vcpu *vcpu)
break;
}
hp = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
- while (!lock_hpte(hp, HPTE_V_HVLOCK))
+ while (!try_lock_hpte(hp, HPTE_V_HVLOCK))
cpu_relax();
found = 0;
if (hp[0] & HPTE_V_VALID) {
@@ -310,7 +315,7 @@ long kvmppc_h_protect(struct kvm_vcpu *vcpu, unsigned long flags,
if (pte_index >= HPT_NPTE)
return H_PARAMETER;
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
- while (!lock_hpte(hpte, HPTE_V_HVLOCK))
+ while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
cpu_relax();
if ((hpte[0] & HPTE_V_VALID) == 0 ||
((flags & H_AVPN) && (hpte[0] & ~0x7fUL) != avpn)) {
--
1.6.0.2
next prev parent reply other threads:[~2012-01-13 14:58 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-01-13 14:31 [PULL 00/52] ppc patch queue 2012-01-13 Alexander Graf
2012-01-13 14:31 ` [PATCH 01/52] KVM: PPC: e500: don't translate gfn to pfn with preemption disabled Alexander Graf
2012-01-13 14:31 ` [PATCH 02/52] KVM: PPC: e500: Eliminate preempt_disable in local_sid_destroy_all Alexander Graf
2012-01-13 14:31 ` [PATCH 03/52] KVM: PPC: e500: clear up confusion between host and guest entries Alexander Graf
2012-01-13 14:31 ` [PATCH 04/52] KVM: PPC: e500: MMU API Alexander Graf
2012-01-13 14:31 ` [PATCH 05/52] KVM: PPC: e500: tlbsx: fix tlb0 esel Alexander Graf
2012-01-13 14:31 ` [PATCH 06/52] KVM: PPC: e500: Don't hardcode PIR=0 Alexander Graf
2012-01-13 14:31 ` [PATCH 07/52] KVM: PPC: E500: Support hugetlbfs Alexander Graf
2012-01-13 14:31 ` [PATCH 08/52] PPC: Fix race in mtmsr paravirt implementation Alexander Graf
2012-01-13 14:31 ` [PATCH 09/52] Fix DEC truncation for greater than 0xffff_ffff/1000 Alexander Graf
2012-01-13 14:31 ` [PATCH 10/52] KVM: booke: Do Not start decrementer when SPRN_DEC set 0 Alexander Graf
2012-01-13 14:31 ` [PATCH 11/52] KVM: PPC: booke: check for signals in kvmppc_vcpu_run Alexander Graf
2012-01-13 14:31 ` [PATCH 12/52] KVM: PPC: Rename deliver_interrupts to prepare_to_enter Alexander Graf
2012-01-13 14:31 ` [PATCH 13/52] KVM: PPC: Move prepare_to_enter call site into subarch code Alexander Graf
2012-01-13 14:31 ` [PATCH 14/52] KVM: PPC: booke: Check for MSR[WE] in prepare_to_enter Alexander Graf
2012-01-13 14:31 ` [PATCH 15/52] KVM: PPC: booke: Fix int_pending calculation for MSR[EE] paravirt Alexander Graf
2012-01-13 14:31 ` [PATCH 16/52] KVM: PPC: booke: Paravirtualize wrtee Alexander Graf
2012-01-13 14:31 ` [PATCH 17/52] KVM: PPC: Paravirtualize SPRG4-7, ESR, PIR, MASn Alexander Graf
2012-01-13 14:31 ` [PATCH 18/52] KVM: PPC: booke: Improve timer register emulation Alexander Graf
2012-01-13 14:31 ` [PATCH 19/52] KVM: PPC: Book3s: PR: Disable preemption in vcpu_run Alexander Graf
2012-01-13 14:31 ` [PATCH 20/52] KVM: PPC: Book3s: PR: No irq_disable " Alexander Graf
2012-01-13 14:31 ` [PATCH 21/52] KVM: PPC: Use get/set for to_svcpu to help preemption Alexander Graf
2012-01-13 14:31 ` [PATCH 22/52] KVM: PPC: align vcpu_kick with x86 Alexander Graf
2012-01-13 14:31 ` [PATCH 23/52] KVM: PPC: Book3S: PR: Fix signal check race Alexander Graf
2012-01-13 14:31 ` [PATCH 24/52] KVM: PPC: e500: Fix TLBnCFG in KVM_CONFIG_TLB Alexander Graf
2012-01-13 14:31 ` [PATCH 25/52] KVM: PPC: e500: use hardware hint when loading TLB0 entries Alexander Graf
2012-01-13 14:31 ` [PATCH 26/52] KVM: PPC: Avoid patching paravirt template code Alexander Graf
2012-01-13 14:31 ` [PATCH 27/52] KVM: PPC: Make wakeups work again for Book3S HV guests Alexander Graf
2012-01-13 14:31 ` [PATCH 28/52] KVM: PPC: Keep a record of HV guest view of hashed page table entries Alexander Graf
2012-01-13 14:31 ` [PATCH 29/52] KVM: PPC: Keep page physical addresses in per-slot arrays Alexander Graf
2012-01-13 14:31 ` [PATCH 30/52] KVM: PPC: Add an interface for pinning guest pages in Book3s HV guests Alexander Graf
2012-01-13 14:31 ` Alexander Graf [this message]
2012-01-13 14:31 ` [PATCH 32/52] KVM: PPC: Only get pages when actually needed, not in prepare_memory_region() Alexander Graf
2012-01-13 14:31 ` [PATCH 33/52] KVM: PPC: Allow use of small pages to back Book3S HV guests Alexander Graf
2012-01-13 14:31 ` [PATCH 34/52] KVM: PPC: Allow I/O mappings in memory slots Alexander Graf
2012-01-13 14:31 ` [PATCH 35/52] KVM: PPC: Maintain a doubly-linked list of guest HPTEs for each gfn Alexander Graf
2012-01-13 14:31 ` [PATCH 36/52] KVM: PPC: Implement MMIO emulation support for Book3S HV guests Alexander Graf
2012-01-13 14:31 ` [PATCH 37/52] KVM: PPC: Implement MMU notifiers " Alexander Graf
2012-01-13 14:31 ` [PATCH 38/52] KVM: Add barriers to allow mmu_notifier_retry to be used locklessly Alexander Graf
2012-01-13 14:31 ` [PATCH 39/52] KVM: PPC: Allow for read-only pages backing a Book3S HV guest Alexander Graf
2012-01-13 14:31 ` [PATCH 40/52] KVM: PPC: Fix vcpu_create dereference before validity check Alexander Graf
2012-01-13 14:31 ` [PATCH 41/52] KVM: PPC: Add KVM_CAP_NR_VCPUS and KVM_CAP_MAX_VCPUS Alexander Graf
2012-01-13 14:31 ` [PATCH 42/52] KVM: PPC: Book3S HV: Keep HPTE locked when invalidating Alexander Graf
2012-01-13 14:31 ` [PATCH 43/52] KVM: PPC: Book3s HV: Maintain separate guest and host views of R and C bits Alexander Graf
2012-01-13 14:31 ` [PATCH 44/52] KVM: PPC: Book3S HV: Use the hardware referenced bit for kvm_age_hva Alexander Graf
2012-01-13 14:31 ` [PATCH 45/52] KVM: PPC: Book3s HV: Implement get_dirty_log using hardware changed bit Alexander Graf
2012-01-13 14:31 ` [PATCH 46/52] KVM: PPC: booke: Add booke206 TLB trace Alexander Graf
2012-01-13 14:31 ` [PATCH 47/52] KVM: PPC: Use the vcpu kmem_cache when allocating new VCPUs Alexander Graf
2012-01-13 14:31 ` [PATCH 48/52] KVM: PPC: Add generic single register ioctls Alexander Graf
2012-01-13 14:31 ` [PATCH 49/52] KVM: PPC: Add support for explicit HIOR setting Alexander Graf
2012-01-13 14:31 ` [PATCH 50/52] KVM: PPC: Move kvm_vcpu_ioctl_[gs]et_one_reg down to platform-specific code Alexander Graf
2012-01-13 14:31 ` [PATCH 51/52] KVM: PPC: Rename MMIO register identifiers Alexander Graf
2012-01-13 14:31 ` [PATCH 52/52] KVM: PPC: refer to paravirt docs in header file Alexander Graf
2012-01-16 10:55 ` [PULL 00/52] ppc patch queue 2012-01-13 Marcelo Tosatti
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1326465115-5976-32-git-send-email-agraf@suse.de \
--to=agraf@suse.de \
--cc=avi@redhat.com \
--cc=kvm-ppc@vger.kernel.org \
--cc=kvm@vger.kernel.org \
--cc=mtosatti@redhat.com \
--cc=paulus@samba.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox