Re: [Xen-devel] [PATCH v2] arm: Fix asynchronous aborts (SError exceptions) due to bogus PTEs
On Tue, 22 Mar 2016, Shanker Donthineni wrote: > On 03/22/2016 05:21 PM, Julien Grall wrote: > > (CC some ARM folks) > > > > On 21/03/2016 23:18, Shanker Donthineni wrote: > >> Hi Julien, > > > > Hello Shanker, > > > > Sorry for the late answer. > > > >> Do you have any other comments to be addressed? > > > > I have a question regarding the implication for what you wrote in the > > commit. > > > > As far as I understand, any speculative table walk might cause an imprecise > > asynchronous abort. So if a guest is using page tables that contain > > garbage, it would be possible to receive an SError. Am I right? > > > > Yes, you are right (applies to EL1 TTBR0/TTBR1, EL2 TTBR0/TTBR1 and EL3 TTBR0 > tables). > > >> > >> On 03/16/2016 02:08 PM, Shanker Donthineni wrote: > >>> From: Vikram Sethi> >>> > >>> ARMv8 architecture allows performing prefetch data/instructions > >>> from memory locations marked as normal memory. Prefetch does not > >>> mean that the data/instruction has to be used/executed in code > >>> flow. All PTEs that appear to be valid to MMU must contain valid > >>> physical address with proper attributes otherwise MMU table walk > >>> might cause imprecise asynchronous aborts. > >>> > >>> The way current XEN code is preparing page tables for frametable > >>> and xenheap memory can create bogus PTEs. This patch fixes the > >>> issue by clearing page table memory before populating EL2 L0/L1 > >>> PTEs. Without this patch XEN crashes on Qualcomm Technologies > >>> server chips due to asynchronous aborts. > >>> > >>> The speculative/prefetch feature explanation is scattered everywhere > >>> in ARM specification but below two sections have useful information. > >>> > >>> E2.8 Memory types and attributes > >>> G4.12.6 External abort on a translation table walk > > > > As said on an earlier version of this patch, please mention the version of > > the spec when you quote it. > > > > Sure, should I post V3 patch mentioning ARM spec version? Yes, please. > >>> Signed-off-by: Vikram Sethi > >>> Signed-off-by: Shanker Donthineni > >>> --- > >>> Changes since v1: > >>> Replace memset() with clear_page() > >>> Edit commit description > >>> > >>> xen/arch/arm/mm.c | 3 +++ > >>> 1 file changed, 3 insertions(+) > >>> > >>> diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c > >>> index 81f9e2e..3fda8f3 100644 > >>> --- a/xen/arch/arm/mm.c > >>> +++ b/xen/arch/arm/mm.c > >>> @@ -730,6 +730,8 @@ void __init setup_xenheap_mappings(unsigned long > >>> base_mfn, > >>> else > >>> { > >>> unsigned long first_mfn = alloc_boot_pages(1, 1); > >>> + > >>> +clear_page(mfn_to_virt(first_mfn)); > >>> pte = mfn_to_xen_entry(first_mfn, WRITEALLOC); > >>> pte.pt.table = 1; > >>> write_pte(p, pte); > >>> @@ -773,6 +775,7 @@ void __init setup_frametable_mappings(paddr_t ps, > >>> paddr_t pe) > >>> second = mfn_to_virt(second_base); > >>> for ( i = 0; i < nr_second; i++ ) > >>> { > >>> +clear_page(mfn_to_virt(second_base + i)); > >>> pte = mfn_to_xen_entry(second_base + i, WRITEALLOC); > >>> pte.pt.table = 1; > >>> > >>> write_pte(_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte); > >> > > > > Regards, > > > > -- > Shanker Donthineni > Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center, Inc. > Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux > Foundation Collaborative Project > ___ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
Re: [Xen-devel] [PATCH v2] arm: Fix asynchronous aborts (SError exceptions) due to bogus PTEs
On 03/22/2016 05:21 PM, Julien Grall wrote: > (CC some ARM folks) > > On 21/03/2016 23:18, Shanker Donthineni wrote: >> Hi Julien, > > Hello Shanker, > > Sorry for the late answer. > >> Do you have any other comments to be addressed? > > I have a question regarding the implication for what you wrote in the commit. > > As far as I understand, any speculative table walk might cause an imprecise > asynchronous abort. So if a guest is using page tables that contain garbage, > it would be possible to receive an SError. Am I right? > Yes, you are right (applies to EL1 TTBR0/TTBR1, EL2 TTBR0/TTBR1 and EL3 TTBR0 tables). >> >> On 03/16/2016 02:08 PM, Shanker Donthineni wrote: >>> From: Vikram Sethi>>> >>> ARMv8 architecture allows performing prefetch data/instructions >>> from memory locations marked as normal memory. Prefetch does not >>> mean that the data/instruction has to be used/executed in code >>> flow. All PTEs that appear to be valid to MMU must contain valid >>> physical address with proper attributes otherwise MMU table walk >>> might cause imprecise asynchronous aborts. >>> >>> The way current XEN code is preparing page tables for frametable >>> and xenheap memory can create bogus PTEs. This patch fixes the >>> issue by clearing page table memory before populating EL2 L0/L1 >>> PTEs. Without this patch XEN crashes on Qualcomm Technologies >>> server chips due to asynchronous aborts. >>> >>> The speculative/prefetch feature explanation is scattered everywhere >>> in ARM specification but below two sections have useful information. >>> >>> E2.8 Memory types and attributes >>> G4.12.6 External abort on a translation table walk > > As said on an earlier version of this patch, please mention the version of > the spec when you quote it. > Sure, should I post V3 patch mentioning ARM spec version? >>> >>> Signed-off-by: Vikram Sethi >>> Signed-off-by: Shanker Donthineni >>> --- >>> Changes since v1: >>> Replace memset() with clear_page() >>> Edit commit description >>> >>> xen/arch/arm/mm.c | 3 +++ >>> 1 file changed, 3 insertions(+) >>> >>> diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c >>> index 81f9e2e..3fda8f3 100644 >>> --- a/xen/arch/arm/mm.c >>> +++ b/xen/arch/arm/mm.c >>> @@ -730,6 +730,8 @@ void __init setup_xenheap_mappings(unsigned long >>> base_mfn, >>> else >>> { >>> unsigned long first_mfn = alloc_boot_pages(1, 1); >>> + >>> +clear_page(mfn_to_virt(first_mfn)); >>> pte = mfn_to_xen_entry(first_mfn, WRITEALLOC); >>> pte.pt.table = 1; >>> write_pte(p, pte); >>> @@ -773,6 +775,7 @@ void __init setup_frametable_mappings(paddr_t ps, >>> paddr_t pe) >>> second = mfn_to_virt(second_base); >>> for ( i = 0; i < nr_second; i++ ) >>> { >>> +clear_page(mfn_to_virt(second_base + i)); >>> pte = mfn_to_xen_entry(second_base + i, WRITEALLOC); >>> pte.pt.table = 1; >>> >>> write_pte(_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte); >> > > Regards, > -- Shanker Donthineni Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project ___ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
Re: [Xen-devel] [PATCH v2] arm: Fix asynchronous aborts (SError exceptions) due to bogus PTEs
(CC some ARM folks) On 21/03/2016 23:18, Shanker Donthineni wrote: Hi Julien, Hello Shanker, Sorry for the late answer. Do you have any other comments to be addressed? I have a question regarding the implication for what you wrote in the commit. As far as I understand, any speculative table walk might cause an imprecise asynchronous abort. So if a guest is using page tables that contain garbage, it would be possible to receive an SError. Am I right? On 03/16/2016 02:08 PM, Shanker Donthineni wrote: From: Vikram SethiARMv8 architecture allows performing prefetch data/instructions from memory locations marked as normal memory. Prefetch does not mean that the data/instruction has to be used/executed in code flow. All PTEs that appear to be valid to MMU must contain valid physical address with proper attributes otherwise MMU table walk might cause imprecise asynchronous aborts. The way current XEN code is preparing page tables for frametable and xenheap memory can create bogus PTEs. This patch fixes the issue by clearing page table memory before populating EL2 L0/L1 PTEs. Without this patch XEN crashes on Qualcomm Technologies server chips due to asynchronous aborts. The speculative/prefetch feature explanation is scattered everywhere in ARM specification but below two sections have useful information. E2.8 Memory types and attributes G4.12.6 External abort on a translation table walk As said on an earlier version of this patch, please mention the version of the spec when you quote it. Signed-off-by: Vikram Sethi Signed-off-by: Shanker Donthineni --- Changes since v1: Replace memset() with clear_page() Edit commit description xen/arch/arm/mm.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c index 81f9e2e..3fda8f3 100644 --- a/xen/arch/arm/mm.c +++ b/xen/arch/arm/mm.c @@ -730,6 +730,8 @@ void __init setup_xenheap_mappings(unsigned long base_mfn, else { unsigned long first_mfn = alloc_boot_pages(1, 1); + +clear_page(mfn_to_virt(first_mfn)); pte = mfn_to_xen_entry(first_mfn, WRITEALLOC); pte.pt.table = 1; write_pte(p, pte); @@ -773,6 +775,7 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t pe) second = mfn_to_virt(second_base); for ( i = 0; i < nr_second; i++ ) { +clear_page(mfn_to_virt(second_base + i)); pte = mfn_to_xen_entry(second_base + i, WRITEALLOC); pte.pt.table = 1; write_pte(_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte); Regards, -- Julien Grall ___ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
Re: [Xen-devel] [PATCH v2] arm: Fix asynchronous aborts (SError exceptions) due to bogus PTEs
Hi Julien, Do you have any other comments to be addressed? On 03/16/2016 02:08 PM, Shanker Donthineni wrote: > From: Vikram Sethi> > ARMv8 architecture allows performing prefetch data/instructions > from memory locations marked as normal memory. Prefetch does not > mean that the data/instruction has to be used/executed in code > flow. All PTEs that appear to be valid to MMU must contain valid > physical address with proper attributes otherwise MMU table walk > might cause imprecise asynchronous aborts. > > The way current XEN code is preparing page tables for frametable > and xenheap memory can create bogus PTEs. This patch fixes the > issue by clearing page table memory before populating EL2 L0/L1 > PTEs. Without this patch XEN crashes on Qualcomm Technologies > server chips due to asynchronous aborts. > > The speculative/prefetch feature explanation is scattered everywhere > in ARM specification but below two sections have useful information. > > E2.8 Memory types and attributes > G4.12.6 External abort on a translation table walk > > Signed-off-by: Vikram Sethi > Signed-off-by: Shanker Donthineni > --- > Changes since v1: > Replace memset() with clear_page() > Edit commit description > > xen/arch/arm/mm.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c > index 81f9e2e..3fda8f3 100644 > --- a/xen/arch/arm/mm.c > +++ b/xen/arch/arm/mm.c > @@ -730,6 +730,8 @@ void __init setup_xenheap_mappings(unsigned long base_mfn, > else > { > unsigned long first_mfn = alloc_boot_pages(1, 1); > + > +clear_page(mfn_to_virt(first_mfn)); > pte = mfn_to_xen_entry(first_mfn, WRITEALLOC); > pte.pt.table = 1; > write_pte(p, pte); > @@ -773,6 +775,7 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t > pe) > second = mfn_to_virt(second_base); > for ( i = 0; i < nr_second; i++ ) > { > +clear_page(mfn_to_virt(second_base + i)); > pte = mfn_to_xen_entry(second_base + i, WRITEALLOC); > pte.pt.table = 1; > write_pte(_first[first_table_offset(FRAMETABLE_VIRT_START)+i], > pte); -- Shanker Donthineni Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project ___ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
[Xen-devel] [PATCH v2] arm: Fix asynchronous aborts (SError exceptions) due to bogus PTEs
From: Vikram SethiARMv8 architecture allows performing prefetch data/instructions from memory locations marked as normal memory. Prefetch does not mean that the data/instruction has to be used/executed in code flow. All PTEs that appear to be valid to MMU must contain valid physical address with proper attributes otherwise MMU table walk might cause imprecise asynchronous aborts. The way current XEN code is preparing page tables for frametable and xenheap memory can create bogus PTEs. This patch fixes the issue by clearing page table memory before populating EL2 L0/L1 PTEs. Without this patch XEN crashes on Qualcomm Technologies server chips due to asynchronous aborts. The speculative/prefetch feature explanation is scattered everywhere in ARM specification but below two sections have useful information. E2.8 Memory types and attributes G4.12.6 External abort on a translation table walk Signed-off-by: Vikram Sethi Signed-off-by: Shanker Donthineni --- Changes since v1: Replace memset() with clear_page() Edit commit description xen/arch/arm/mm.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c index 81f9e2e..3fda8f3 100644 --- a/xen/arch/arm/mm.c +++ b/xen/arch/arm/mm.c @@ -730,6 +730,8 @@ void __init setup_xenheap_mappings(unsigned long base_mfn, else { unsigned long first_mfn = alloc_boot_pages(1, 1); + +clear_page(mfn_to_virt(first_mfn)); pte = mfn_to_xen_entry(first_mfn, WRITEALLOC); pte.pt.table = 1; write_pte(p, pte); @@ -773,6 +775,7 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t pe) second = mfn_to_virt(second_base); for ( i = 0; i < nr_second; i++ ) { +clear_page(mfn_to_virt(second_base + i)); pte = mfn_to_xen_entry(second_base + i, WRITEALLOC); pte.pt.table = 1; write_pte(_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte); -- Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project ___ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel