[RFC,kernel] KVM: PPC: Book3S PR: Fix WIMG handling under pHyp

Message ID 20171122034221.647-1-aik@ozlabs.ru
State New
Headers show
Series
  • [RFC,kernel] KVM: PPC: Book3S PR: Fix WIMG handling under pHyp
Related show

Commit Message

Alexey Kardashevskiy Nov. 22, 2017, 3:42 a.m.
96df226 "KVM: PPC: Book3S PR: Preserve storage control bits" added WIMG
bits preserving but it missed 2 special cases:
- a magic page in kvmppc_mmu_book3s_64_xlate() and
- guest real mode in kvmppc_handle_pagefault().

For these ptes WIMG were 0 and pHyp failed on these causing a guest to
stop in the very beginning at NIP=0x100 (due to bd9166ffe
"KVM: PPC: Book3S PR: Exit KVM on failed mapping").

This initializes WIMG to non-zero value HPTE_R_M. The value is chosen
as (0x192 & HPTE_R_WIMG); 0x192 is a magic value from
kvmppc_mmu_map_page().

Fixes: 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits"
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---

This indeed fixes PR KVM + VFIO under pHyp but selection of HPTE_R_M
is arguable.

---
 arch/powerpc/kvm/book3s_64_mmu.c | 1 +
 arch/powerpc/kvm/book3s_pr.c     | 1 +
 2 files changed, 2 insertions(+)

Comments

Alexander Graf Nov. 22, 2017, 3:02 p.m. | #1
On 22.11.17 04:42, Alexey Kardashevskiy wrote:
> 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits" added WIMG
> bits preserving but it missed 2 special cases:
> - a magic page in kvmppc_mmu_book3s_64_xlate() and
> - guest real mode in kvmppc_handle_pagefault().
> 
> For these ptes WIMG were 0 and pHyp failed on these causing a guest to
> stop in the very beginning at NIP=0x100 (due to bd9166ffe
> "KVM: PPC: Book3S PR: Exit KVM on failed mapping").
> 
> This initializes WIMG to non-zero value HPTE_R_M. The value is chosen
> as (0x192 & HPTE_R_WIMG); 0x192 is a magic value from
> kvmppc_mmu_map_page().
> 
> Fixes: 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits"
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> 
> This indeed fixes PR KVM + VFIO under pHyp but selection of HPTE_R_M
> is arguable.

This does indeed fix the breakage we've seen:

Tested-by: Ruediger Oertel <ro@suse.de>


Alex
Greg Kurz Nov. 23, 2017, 11:33 p.m. | #2
On Wed, 22 Nov 2017 14:42:21 +1100
Alexey Kardashevskiy <aik@ozlabs.ru> wrote:

> 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits" added WIMG
> bits preserving but it missed 2 special cases:
> - a magic page in kvmppc_mmu_book3s_64_xlate() and
> - guest real mode in kvmppc_handle_pagefault().
> 
> For these ptes WIMG were 0 and pHyp failed on these causing a guest to
> stop in the very beginning at NIP=0x100 (due to bd9166ffe
> "KVM: PPC: Book3S PR: Exit KVM on failed mapping").
> 
> This initializes WIMG to non-zero value HPTE_R_M. The value is chosen
> as (0x192 & HPTE_R_WIMG); 0x192 is a magic value from
> kvmppc_mmu_map_page().
> 
> Fixes: 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits"
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> 
> This indeed fixes PR KVM + VFIO under pHyp but selection of HPTE_R_M
> is arguable.
> 

The initial page fault at 0x100 on machine startup causes H_ENTER to fail in
pHyp with H_PARAMETER, as described in PAPR:

 The hypervisor checks that the WIMG bits within the PTE are appropriate for the
 physical page number else H_Parameter return. (For System Memory pages WIMG=0010,
 or, 1110 if the SAO option is enabled, and for IO pages WIMG=01**.)

I'm not aware we care for SAO so HPTE_R_M looks like the only sensible choice,
or I'm missing something ?

Anyway, this patch allows to start a PR guest on PowerVM again.

Reviewed-by: Greg Kurz <groug@kaod.org>

and

Tested-by: Greg Kurz <groug@kaod.org>

> ---
>  arch/powerpc/kvm/book3s_64_mmu.c | 1 +
>  arch/powerpc/kvm/book3s_pr.c     | 1 +
>  2 files changed, 2 insertions(+)
> 
> diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
> index 29ebe2f..a93d719 100644
> --- a/arch/powerpc/kvm/book3s_64_mmu.c
> +++ b/arch/powerpc/kvm/book3s_64_mmu.c
> @@ -235,6 +235,7 @@ static int kvmppc_mmu_book3s_64_xlate(struct kvm_vcpu *vcpu, gva_t eaddr,
>  		gpte->may_read = true;
>  		gpte->may_write = true;
>  		gpte->page_size = MMU_PAGE_4K;
> +		gpte->wimg = HPTE_R_M;
>  
>  		return 0;
>  	}
> diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
> index 69a0944..0152efe 100644
> --- a/arch/powerpc/kvm/book3s_pr.c
> +++ b/arch/powerpc/kvm/book3s_pr.c
> @@ -557,6 +557,7 @@ int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
>  		pte.eaddr = eaddr;
>  		pte.vpage = eaddr >> 12;
>  		pte.page_size = MMU_PAGE_64K;
> +		pte.wimg = HPTE_R_M;
>  	}
>  
>  	switch (kvmppc_get_msr(vcpu) & (MSR_DR|MSR_IR)) {
Alexander Graf Dec. 11, 2017, 10:51 a.m. | #3
On 24.11.17 00:33, Greg Kurz wrote:
> On Wed, 22 Nov 2017 14:42:21 +1100
> Alexey Kardashevskiy <aik@ozlabs.ru> wrote:
> 
>> 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits" added WIMG
>> bits preserving but it missed 2 special cases:
>> - a magic page in kvmppc_mmu_book3s_64_xlate() and
>> - guest real mode in kvmppc_handle_pagefault().
>>
>> For these ptes WIMG were 0 and pHyp failed on these causing a guest to
>> stop in the very beginning at NIP=0x100 (due to bd9166ffe
>> "KVM: PPC: Book3S PR: Exit KVM on failed mapping").
>>
>> This initializes WIMG to non-zero value HPTE_R_M. The value is chosen
>> as (0x192 & HPTE_R_WIMG); 0x192 is a magic value from
>> kvmppc_mmu_map_page().
>>
>> Fixes: 96df226 "KVM: PPC: Book3S PR: Preserve storage control bits"
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>>
>> This indeed fixes PR KVM + VFIO under pHyp but selection of HPTE_R_M
>> is arguable.
>>
> 
> The initial page fault at 0x100 on machine startup causes H_ENTER to fail in
> pHyp with H_PARAMETER, as described in PAPR:
> 
>  The hypervisor checks that the WIMG bits within the PTE are appropriate for the
>  physical page number else H_Parameter return. (For System Memory pages WIMG=0010,
>  or, 1110 if the SAO option is enabled, and for IO pages WIMG=01**.)
> 
> I'm not aware we care for SAO so HPTE_R_M looks like the only sensible choice,
> or I'm missing something ?
> 
> Anyway, this patch allows to start a PR guest on PowerVM again.
> 
> Reviewed-by: Greg Kurz <groug@kaod.org>
> 
> and
> 
> Tested-by: Greg Kurz <groug@kaod.org>

So doesn't that mean we should pick it up, mark it CC stable and push it
upstream?


Alex

Patch

diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
index 29ebe2f..a93d719 100644
--- a/arch/powerpc/kvm/book3s_64_mmu.c
+++ b/arch/powerpc/kvm/book3s_64_mmu.c
@@ -235,6 +235,7 @@  static int kvmppc_mmu_book3s_64_xlate(struct kvm_vcpu *vcpu, gva_t eaddr,
 		gpte->may_read = true;
 		gpte->may_write = true;
 		gpte->page_size = MMU_PAGE_4K;
+		gpte->wimg = HPTE_R_M;
 
 		return 0;
 	}
diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
index 69a0944..0152efe 100644
--- a/arch/powerpc/kvm/book3s_pr.c
+++ b/arch/powerpc/kvm/book3s_pr.c
@@ -557,6 +557,7 @@  int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 		pte.eaddr = eaddr;
 		pte.vpage = eaddr >> 12;
 		pte.page_size = MMU_PAGE_64K;
+		pte.wimg = HPTE_R_M;
 	}
 
 	switch (kvmppc_get_msr(vcpu) & (MSR_DR|MSR_IR)) {