Some fixes and improvements for PR KVM
Subsystems (3)
cover Some fixes and improvements for PR KVM
This series of 5 patches is against the KVM next branch. It fixes some bugs in PR-style KVM on Book 3S PPC and adds support for the guest using 1TB segments as well as 256MB segments. My ultimate goal is to make it possible to configure both HV and PR KVM into the same kernel binary, and this is just the first few steps.
Paul.
PATCH 1/5 KVM: PPC: Book3S PR: Fix proto-VSID calculations
This makes sure the calculation of the proto-VSIDs used by PR KVM is done with 64-bit arithmetic. Since vcpu3s->context_id[] is int, when we do vcpu3s->context_id[0] << ESID_BITS the shift will be done with 32-bit instructions, possibly leading to significant bits getting lost, as the context id can be up to 524283 and ESID_BITS is
18. To fix this we cast the context id to u64 before shifting.
--- arch/powerpc/kvm/book3s_64_mmu_host.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/powerpc/kvm/book3s_64_mmu_host.c +2 -2
@@ -325,9 +325,9 @@ int kvmppc_mmu_init(struct kvm_vcpu *vcpu)return -1 ; vcpu3s->context_id[0 ] = err; - vcpu3s->proto_vsid_max = ((vcpu3s->context_id[0 ] +1 ) + vcpu3s->proto_vsid_max = ((u64)(vcpu3s->context_id[0 ] +1 ) <<ESID_BITS ) -1 ; - vcpu3s->proto_vsid_first = vcpu3s->context_id[0 ] <<ESID_BITS ; + vcpu3s->proto_vsid_first = (u64 )vcpu3s->context_id[0 ] <<ESID_BITS ; vcpu3s->proto_vsid_next = vcpu3s->proto_vsid_first;kvmppc_mmu_hpte_init (vcpu);
Signature · 1 line
1.8.3.1
PATCH 2/5 KVM: PPC: Book3S PR: Fix invalidation of SLB entry 0 on guest entry
In reply to Paul MackerrasOn entering a PR KVM guest, we invalidate the whole SLB before loading up the guest entries. We do this using an slbia instruction, which invalidates all entries except entry 0, followed by an slbie to invalidate entry 0. However, the slbie turns out to be ineffective in some circumstances (specifically when the host linear mapping uses 64k pages) because of errors in computing the parameter to the slbie. The result is that the guest kernel hangs very early in boot because it takes a DSI the first time it tries to access kernel data using a linear mapping address in real mode.
Currently we construct bits 36 - 43 (big-endian numbering) of the slbie parameter by taking bits 56 - 63 of the SLB VSID doubleword. These bits for the tlbie are C (class, 1 bit), B (segment size, 2 bits) and 5 reserved bits. For the SLB VSID doubleword these are C (class, 1 bit), reserved (1 bit), LP (large page size, 2 bits), and 4 reserved bits. Thus we are not setting the B field correctly, and when LP = 01 as it is for 64k pages, we are setting a reserved bit.
Rather than add more instructions to calculate the slbie parameter correctly, this takes a simpler approach, which is to set entry 0 to zeroes explicitly. Normally slbmte should not be used to invalidate an entry, since it doesn't invalidate the ERATs, but it is OK to use it to invalidate an entry if it is immediately followed by slbia, which does invalidate the ERATs. (This has been confirmed with the Power architects.) This approach takes fewer instructions and will work whatever the contents of entry 0.
--- arch/powerpc/kvm/book3s_64_slb.S | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)
arch/powerpc/kvm/book3s_64_slb.S +2 -11
@@ -66,10 +66,6 @@ slb_exit_skip_ ## num: ld r12, PACA_SLBSHADOWPTR(r13) - /* Save off the first entry so we can slbie it later */ - ld r10, SHADOW_SLB_ESID(0)(r12) - ld r11, SHADOW_SLB_VSID(0)(r12) - /* Remove bolted entries */ UNBOLT_SLB_ENTRY(0) UNBOLT_SLB_ENTRY(1)
@@ -81,15 +77,10 @@ slb_exit_skip_ ## num: /* Flush SLB */ + li r10, 0 + slbmte r10, r10 slbia - /* r0 = esid & ESID_MASK */ - rldicr r10, r10, 0, 35 - /* r0 |= CLASS_BIT(VSID) */ - rldic r12, r11, 56 - 36, 36 - or r10, r10, r12 - slbie r10 - /* Fill SLB with our shadow */ lbz r12, SVCPU_SLB_MAX(r3)
Signature · 1 line
1.8.3.1
PATCH 3/5 KVM: PPC: Book3S PR: Don't keep scanning HPTEG after we find a match
In reply to Paul MackerrasThe loop in kvmppc_mmu_book3s_64_xlate() that looks up a translation in the guest hashed page table (HPT) keeps going if it finds an HPTE that matches but doesn't allow access. This is incorrect; it is different from what the hardware does, and there should never be more than one matching HPTE anyway. This fixes it to stop when any matching HPTE is found.
--- arch/powerpc/kvm/book3s_64_mmu.c | 14 ++------------ 1 file changed, 2 insertions(+), 12 deletions(-)
arch/powerpc/kvm/book3s_64_mmu.c +2 -12
@@ -167,7 +167,6 @@ static int kvmppc_mmu_book3s_64_xlate(struct kvm_vcpu *vcpu, gva_t eaddr,int i;u8 key =0 ;bool found = false; -bool perm_err = false;int second =0 ;ulong mp_ea = vcpu->arch.magic_page_ea;
@@ -248,11 +247,6 @@ do_second:break ; } -if (!gpte->may_read) { - perm_err = true; -continue ; - } -dprintk ("KVM MMU: Translated 0x%lx [0x%llx] -> 0x%llx " "-> 0x%lx\n" , eaddr, avpn, gpte->vpage, gpte->raddr);
@@ -281,6 +275,8 @@ do_second:if (pteg[i+1 ] != oldr)copy_to_user ((void __user *)ptegp, pteg,sizeof (pteg)); +if (!gpte->may_read) +return -EPERM ;return 0 ; }else { dprintk("KVM MMU: No PTE found (ea=0x%lx sdr1=0x%llx "
@@ -296,13 +292,7 @@ do_second: } } - no_page_found: - - -if (perm_err) -return -EPERM ; -return -ENOENT ; no_seg_found:
Signature · 1 line
1.8.3.1
Re: [PATCH 3/5] KVM: PPC: Book3S PR: Don't keep scanning HPTEG after we find a match
Quoted 6 lines · Paul Mackerras
The loop in
kvmppc_mmu_book3s_64_xlate()that looks up a translation in the guest hashed page table (HPT) keeps going if it finds an HPTE that matches but doesn't allow access. This is incorrect; it is different from what the hardware does, and there should never be more than one matching HPTE anyway. This fixes it to stop when any matching HPTE is found.
IIRC I put in that logic to make it compatible with how QEMU handled HTAB lookups. Since then this has changed in QEMU though, and doing it this way is architecturally as correct as the other (the spec just says undefined behavior for duplicate entries), so I'm fine with the change.
However, does book3s_32 differ here? I would like to keep them at least behave identically for undefined behavior. So in this case, unless you know of a difference for 32bit, please also provide a patch that does this change in book3s_32_mmu.c.
Alex
Quoted 60 lines
Signed-off-by Paul Mackerras <[email protected]>---
arch/powerpc/kvm/book3s_64_mmu.c| 14 ++------------1 file changed, 2 insertions(+), 12 deletions(-)arch/powerpc/kvm/book3s_64_mmu.c +2
-7@@ -167,7 +167,6 @@ static int kvmppc_mmu_book3s_64_xlate(struct kvm_vcpu *vcpu, gva_t eaddr,int i;u8 key =0 ;bool found = false; -bool perm_err = false;int second =0 ;ulong mp_ea = vcpu->arch.magic_page_ea;@@ -248,11 +247,6 @@ do_second:break ; } -if (!gpte->may_read) { - perm_err = true; -continue ; - } -dprintk ("KVM MMU: Translated 0x%lx [0x%llx] -> 0x%llx " "-> 0x%lx\n" , eaddr, avpn, gpte->vpage, gpte->raddr);@@ -281,6 +275,8 @@ do_second:if (pteg[i+1 ] != oldr)copy_to_user ((void __user *)ptegp, pteg,sizeof (pteg)); +if (!gpte->may_read) +return -EPERM ;return 0 ; }else { dprintk("KVM MMU: No PTE found (ea=0x%lx sdr1=0x%llx "@@ -296,13 +292,7 @@ do_second: } } -
no_page_found:arch/powerpc/kvm/book3s_64_mmu.c +0
-5- - -if (perm_err) -return -EPERM ; -return -ENOENT;
no_seg_found:Signature · 6 lines
1.8.3.1 -- To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
PATCH 4/5 KVM: PPC: Book3S PR: Invalidate SLB entries properly
In reply to Paul MackerrasAt present, if the guest creates a valid SLB (segment lookaside buffer) entry with the slbmte instruction, then invalidates it with the slbie instruction, then reads the entry with the slbmfee/slbmfev instructions, the result of the slbmfee will have the valid bit set, even though the entry is not actually considered valid by the host. This is confusing, if not worse. This fixes it by zeroing out the orige and origv fields of the SLB entry structure when the entry is invalidated.
--- arch/powerpc/kvm/book3s_64_mmu.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
arch/powerpc/kvm/book3s_64_mmu.c +6 -1
@@ -376,6 +376,8 @@ static void kvmppc_mmu_book3s_64_slbie(struct kvm_vcpu *vcpu, u64 ea)dprintk ("KVM MMU: slbie(0x%llx, 0x%llx)\n" , ea, slbe->esid); slbe->valid = false; + slbe->orige =0 ; + slbe->origv =0 ;kvmppc_mmu_map_segment (vcpu, ea); }
@@ -386,8 +388,11 @@ static void kvmppc_mmu_book3s_64_slbia(struct kvm_vcpu *vcpu)dprintk ("KVM MMU: slbia()\n" ); -for (i =1 ; i < vcpu->arch.slb_nr; i++) +for (i =1 ; i < vcpu->arch.slb_nr; i++) { vcpu->arch.slb[i].valid = false; + vcpu->arch.slb[i].orige =0 ; + vcpu->arch.slb[i].origv =0 ; + }if (vcpu->arch.shared->msr &MSR_IR ) {kvmppc_mmu_flush_segments (vcpu);
Signature · 1 line
1.8.3.1
Re: [PATCH 4/5] KVM: PPC: Book3S PR: Invalidate SLB entries properly
Quoted 9 lines · Paul Mackerras
At present, if the guest creates a valid SLB (segment lookaside buffer) entry with the slbmte instruction, then invalidates it with the slbie instruction, then reads the entry with the slbmfee/slbmfev instructions, the result of the slbmfee will have the valid bit set, even though the entry is not actually considered valid by the host. This is confusing, if not worse. This fixes it by zeroing out the orige and origv fields of the SLB entry structure when the entry is invalidated.
Signed-off-by Paul Mackerras <[email protected]>
Could you please change this to only remove the V bit from orige? I've found it very useful for debugging to see old, invalidated entries in the SLB when dumping it. The spec declares anything but the toggle of the V bit as undefined.
Alex
Quoted 37 lines
---
arch/powerpc/kvm/book3s_64_mmu.c| 7 ++++++-1 file changed, 6 insertions(+), 1 deletion(-)arch/powerpc/kvm/book3s_64_mmu.c +6
-1@@ -376,6 +376,8 @@ static void kvmppc_mmu_book3s_64_slbie(struct kvm_vcpu *vcpu, u64 ea)dprintk ("KVM MMU: slbie(0x%llx, 0x%llx)\n" , ea, slbe->esid); slbe->valid = false; + slbe->orige =0 ; + slbe->origv =0 ;kvmppc_mmu_map_segment (vcpu, ea); }@@ -386,8 +388,11 @@ static void kvmppc_mmu_book3s_64_slbia(struct kvm_vcpu *vcpu)dprintk ("KVM MMU: slbia()\n" ); -for (i =1 ; i < vcpu->arch.slb_nr; i++) +for (i =1 ; i < vcpu->arch.slb_nr; i++) { vcpu->arch.slb[i].valid = false; + vcpu->arch.slb[i].orige =0 ; + vcpu->arch.slb[i].origv =0 ; + }if (vcpu->arch.shared->msr &MSR_IR ) {kvmppc_mmu_flush_segments (vcpu);Signature · 6 lines
1.8.3.1 -- To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 4/5] KVM: PPC: Book3S PR: Invalidate SLB entries properly
Quoted 11 lines · Alexander Graf
Quoted 8 lines · Paul Mackerras
At present, if the guest creates a valid SLB (segment lookaside buffer) entry with the slbmte instruction, then invalidates it with the slbie instruction, then reads the entry with the slbmfee/slbmfev instructions, the result of the slbmfee will have the valid bit set, even though the entry is not actually considered valid by the host. This is confusing, if not worse. This fixes it by zeroing out the orige and origv fields of the SLB entry structure when the entry is invalidated.
Quoted 3 lines
Quoted 1 line
Signed-off-by Paul Mackerras <[email protected]>Could you please change this to only remove the V bit from orige? I've found it very useful for debugging to see old, invalidated entries in the SLB when dumping it. The spec declares anything but the toggle of the V bit as undefined.
I did it like this since the architecture (since version 2.03) specifies that slbmfee and slbmfev both return all zeroes for invalid entries. I'm not sure what you mean by your last sentence there.
Paul.
Alexander Graf
AppliedRe: [PATCH 4/5] KVM: PPC: Book3S PR: Invalidate SLB entries properly
Quoted 12 lines · Paul Mackerras
Quoted 11 lines · Alexander Graf
Quoted 8 lines · Paul Mackerras
At present, if the guest creates a valid SLB (segment lookaside buffer) entry with the slbmte instruction, then invalidates it with the slbie instruction, then reads the entry with the slbmfee/slbmfev instructions, the result of the slbmfee will have the valid bit set, even though the entry is not actually considered valid by the host. This is confusing, if not worse. This fixes it by zeroing out the orige and origv fields of the SLB entry structure when the entry is invalidated.
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
Signed-off-by Paul Mackerras <[email protected]>Could you please change this to only remove the V bit from orige? I've found it very useful for debugging to see old, invalidated entries in the SLB when dumping it. The spec declares anything but the toggle of the V bit as undefined.
I did it like this since the architecture (since version 2.03) specifies that slbmfee and slbmfev both return all zeroes for invalid entries. I'm not sure what you mean by your last sentence there.
Oh, really? I based all of the work back then on 2.01, so maybe that changed passed by me unnoticed. But you're right, it's certainly explicitly mentioned in 2.06. Guess this patch is perfectly valid then :).
Thanks, applied to kvm-ppc-queue.
Alex
PATCH 5/5 KVM: PPC: Book3S PR: Allow guest to use 1TB segments
In reply to Paul MackerrasWith this, the guest can use 1TB segments as well as 256MB segments. Since we now have the situation where a single emulated guest segment could correspond to multiple shadow segments (as the shadow segments are still 256MB segments), this adds a new kvmppc_mmu_flush_segment() to scan for all shadow segments that need to be removed.
This restructures the guest HPT (hashed page table) lookup code to use the correct hashing and matching functions for HPTEs within a 1TB segment. We use the standard hpt_hash() function instead of open-coding the hash calculation, and we use HPTE_V_COMPARE() with an AVPN value that has the B (segment size) field included. The calculation of avpn is done a little earlier since it doesn't change in the loop starting at the do_second label.
The computation in kvmppc_mmu_book3s_64_esid_to_vsid() changes so that it returns a 256MB VSID even if the guest SLB entry is a 1TB entry. This is because the users of this function are creating 256MB SLB entries. We set a new VSID_1T flag so that entries created from 1T segments don't collide with entries from 256MB segments.
--- arch/powerpc/include/asm/kvm_book3s.h | 6 ++-- arch/powerpc/kvm/book3s_64_mmu.c | 60 +++++++++++++++++++++++++---------- arch/powerpc/kvm/book3s_64_mmu_host.c | 17 ++++++++++ arch/powerpc/kvm/book3s_pr.c | 3 +- 4 files changed, 66 insertions(+), 20 deletions(-)
arch/powerpc/include/asm/kvm_book3s.h +4 -2
@@ -107,8 +107,9 @@ struct kvmppc_vcpu_book3s {#define CONTEXT_GUEST 1#define CONTEXT_GUEST_END 2 -#define VSID_REAL 0x1fffffffffc00000ULL -#define VSID_BAT 0x1fffffffffb00000ULL +#define VSID_REAL 0x0fffffffffc00000ULL +#define VSID_BAT 0x0fffffffffb00000ULL +#define VSID_1T 0x1000000000000000ULL#define VSID_REAL_DR 0x2000000000000000ULL#define VSID_REAL_IR 0x4000000000000000ULL#define VSID_PR 0x8000000000000000ULL
@@ -123,6 +124,7 @@ extern void kvmppc_mmu_book3s_32_init(struct kvm_vcpu *vcpu);extern void kvmppc_mmu_book3s_hv_init (struct kvm_vcpu *vcpu);extern int kvmppc_mmu_map_page (struct kvm_vcpu *vcpu,struct kvmppc_pte *pte);extern int kvmppc_mmu_map_segment (struct kvm_vcpu *vcpu,ulong eaddr); +extern void kvmppc_mmu_flush_segment (struct kvm_vcpu *vcpu,ulong eaddr,ulong seg_size);extern void kvmppc_mmu_flush_segments (struct kvm_vcpu *vcpu); extern int kvmppc_book3s_hv_page_fault(struct kvm_run *run, struct kvm_vcpu *vcpu, unsigned long addr,
arch/powerpc/kvm/book3s_64_mmu.c +44 -16
142 lines not on this page · raw message
arch/powerpc/kvm/book3s_64_mmu_host.c +17 -0
@@ -301,6 +301,23 @@ out:return r; } +void kvmppc_mmu_flush_segment (struct kvm_vcpu *vcpu,ulong ea,ulong seg_size) +{ +struct kvmppc_book3s_shadow_vcpu *svcpu =svcpu_get (vcpu); +ulong seg_mask = -seg_size; +int i; + +for (i =1 ; i < svcpu->slb_max; i++) { +if ((svcpu->slb[i].esid &SLB_ESID_V ) && + (svcpu->slb[i].esid & seg_mask) = ea) { +/* Invalidate this entry */ + svcpu->slb[i].esid =0 ; + } + } + +svcpu_put (svcpu); +} +void kvmppc_mmu_flush_segments (struct kvm_vcpu *vcpu) {struct kvmppc_book3s_shadow_vcpu *svcpu =svcpu_get (vcpu);
arch/powerpc/kvm/book3s_pr.c +1 -2
@@ -1239,8 +1239,7 @@ out:#ifdef CONFIG_PPC64 int kvm_vm_ioctl_get_smmu_info (struct kvm *kvm,struct kvm_ppc_smmu_info *info) { -/* No flags */ - info->flags =0 ; + info->flags =KVM_PPC_1T_SEGMENTS ;/* SLB is always 64 entries */ info->slb_size =64 ;
Signature · 1 line
1.8.3.1
Alexander Graf
AppliedQuoted 5 lines · Paul Mackerras
This series of 5 patches is against the KVM next branch. It fixes some bugs in PR-style KVM on Book 3S PPC and adds support for the guest using 1TB segments as well as 256MB segments. My ultimate goal is to make it possible to configure both HV and PR KVM into the same kernel binary, and this is just the first few steps.
Thanks, applied all except 4/5 to kvm-ppc-queue.
Alex