x86: skip check for spurious faults for non-present faults
PATCH 1/1 x86: skip check for spurious faults for non-present faults
If a fault on a kernel address is due to a non-present page, then it cannot be the result of stale TLB entry from a protection change (RO to RW or NX to X). Thus the pagetable walk in spurious_fault() can be skipped.
This avoids spurious_fault() oopsing in some cases if the pagetables it attempts to walk are not accessible. This obscures the location of the original fault.
This also fixes a crash with Xen PV guests when they access entries in the M2P corresponding to device MMIO regions. The M2P is mapped (read-only) by Xen into the kernel address space of the guest and this mapping may contains holes for non-RAM regions. Read faults will result in calls to spurious_fault(), but because the page tables for the M2P mappings are not accessible by the guest the pagetable walk would fault.
This was not normally a problem as MMIO mappings would not normally result in a M2P lookup because of the use of the _PAGE_IOMAP bit the PTE. However, removing the _PAGE_IOMAP bit requires M2P lookups for MMIO mappings as well.
---
x86 maintainers, this is a prerequisite for removing Xen's usage of _PAGE_IOMAP so I think this is best merged via the Xen tree.
See the full series at https://lkml.org/lkml/2014/4/15/390
--- arch/x86/mm/fault.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
arch/x86/mm/fault.c +4 -2
@@ -936,8 +936,10 @@ spurious_fault(unsigned long error_code, unsigned long address)pte_t *pte;int ret; -/* Reserved-bit violation or user access to kernel space? */ -if (error_code & (PF_USER |PF_RSVD )) +/* Only check for spurious faults on supervisor write or +instruction faults. */ +if (error_code != (PF_WRITE |PF_PROT ) + && error_code != (PF_INSTR |PF_PROT ))return 0 ; pgd = init_mm.pgd +pgd_index (address);
Signature · 1 line
1.7.10.4
arch/x86/mm/fault.c spurious_fault(unsigned long error_code, unsigned long address) · line 943
Quoted 7 lines · David Vrabel
arch/x86/mm/fault.c +4
-2-/* Reserved-bit violation or user access to kernel space? */ -if (error_code & (PF_USER |PF_RSVD )) +/* Only check for spurious faults on supervisor write or +instruction faults. */ +if (error_code != (PF_WRITE |PF_PROT ) + && error_code != (PF_INSTR |PF_PROT ))return 0 ;
This changes the semantics a bit too much for me to feel happy about it.
This is at best missing quite a bit of detail from the changelog.
1. 'return 0' means "this was not a spurious fault"
2. We used to check for the presence of PF_USER|PF_RSVD
3. This patch checks now for two _explicit_ conditions, which implicitly check for the _absence_ of the two bits we checked for before.
I do believe your patch is correct, but it took me a bit to convince myself that it was the right thing. Please be explicit (in the comment) about the exact PTE transitions that you expect to get you here.
Also, I have to wonder if you can just leave the original if() in there.
You're making this _more_ restrictive than it was before, and I wonder
if it might just be more clear if you have both checks. The compiler might even compile it down to the same code, just changing the immediate that was generated for the mask that you're checking.
H. Peter Anvin
reviewerI'm trying to wrap my head around any forward compatibility concerns... if we misidentify a fault as spurious that would be bad.
Quoted 31 lines · Dave Hansen
arch/x86/mm/fault.c spurious_fault(unsigned long error_code, unsigned long address) · line 943
Quoted 7 lines · David Vrabel
arch/x86/mm/fault.c +4
-2-/* Reserved-bit violation or user access to kernel space? */ -if (error_code & (PF_USER |PF_RSVD )) +/* Only check for spurious faults on supervisor write or +instruction faults. */ +if (error_code != (PF_WRITE |PF_PROT ) + && error_code != (PF_INSTR |PF_PROT ))return 0 ;This changes the semantics a bit too much for me to feel happy about it.
This is at best missing quite a bit of detail from the changelog.1. 'return 0' means "this was not a spurious fault"
2. We used to check for the presence of PF_USER|PF_RSVD
3. This patch checks now for two _explicit_ conditions, which implicitly check for the _absence_ of the two bits we checked for before.I do believe your patch is correct, but it took me a bit to convince myself that it was the right thing. Please be explicit (in the comment)
about the exact PTE transitions that you expect to get you here.Also, I have to wonder if you can just leave the original
if()in there.
You're making this _more_ restrictive than it was before, and I wonder if it might just be more clear if you have both checks. The compiler might even compile it down to the same code, just changing the immediate
that was generated for the mask that you're checking.
Signature · 1 line
Sent from my mobile phone. Please pardon brevity and lack of formatting.
Quoted 2 lines · H. Peter Anvin
I'm trying to wrap my head around any forward compatibility concerns... if we misidentify a fault as spurious that would be bad.
FWIW, I did go through the bullets in the SDM's "4.10.4.3 Optional Invalidation". The only condition not covered by this patch is the one that talks about "CR4.SMEP = 0...".
But, I don't know of any case where we take a userspace page and make it supervisor-only, so I don't think this is possible now.