Intel E6320 Specification Update - Page 54

VEX.L is Not Ignored with VCVT*2SI Instructions, MCI_ADDR May be Incorrect For Cache Parity Errors

Page 54 highlights

Workaround: Before performing a VM entry to the shutdown state, software should check whether the "virtual NMIs" and "NMI-window exiting" VM-execution controls are both 1. If they are, software should clear "NMI-window exiting" and inject an NMI as part of VM entry. Status: For the steppings affected, see the Summary Tables of Changes. BJ112. Execution of INVVPID Outside 64-Bit Mode Cannot Invalidate Translations For 64-Bit Linear Addresses Problem: Executions of the INVVPID instruction outside 64-bit mode with the INVVPID type "individual-address invalidation" ignore bits 63:32 of the linear address in the INVVPID descriptor and invalidate translations for bits 31:0 of the linear address. Implication: The INVVPID instruction may fail to invalidate translations for linear addresses that set bits in the range 63:32. Because this erratum applies only to executions outside 64-bit mode, it applies only to attempts by a 32-bit virtual-machine monitor (VMM) to invalidate translations for a 64-bit guest. Intel has not observed this erratum with any commercially available software. Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Changes. BJ113. VEX.L is Not Ignored with VCVT*2SI Instructions Problem: The VEX.L bit should be ignored for the VCVTSS2SI, VCVTSD2SI, VCVTTSS2SI, and VCVTTSD2SI instructions, however due to this erratum the VEX.L bit is not ignored and will cause a #UD. Implication: Unexpected #UDs will be seen when the VEX.L bit is set to 1 with VCVTSS2SI, VCVTSD2SI, VCVTTSS2SI, and VCVTTSD2SI instructions. Workaround: Software should ensure that the VEX.L bit is set to 0 for all scalar instructions. Status: For the steppings affected, see the Summary Tables of Changes. BJ114. MCI_ADDR May be Incorrect For Cache Parity Errors Problem: In cases when a WBINVD instruction evicts a line containing an address or data parity error (MCACOD of 0x124, and MSCOD of 0x10), the address of this error should be logged in the MCi_ADDR register. Due to this erratum, the logged address may be incorrect, even though MCi_Status.ADDRV (bit 63) is set. Implication: The address reported in MCi_ADDR may not be correct for cases of a parity error found during WBINVD execution. Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Changes. BJ115. Instruction Fetches Page-Table Walks May be Made Speculatively to Uncacheable Memory Problem: Page-table walks on behalf of instruction fetches may be made speculatively to uncacheable (UC) memory. Implication: If any paging structures are located at addresses in uncacheable memory that are used for memory-mapped I/O, such I/O operations may be invoked as a result of speculative execution that would never actually occur in the executed code path. Intel has not observed this erratum with any commercially available software. Workaround: Software should avoid locating paging structures at addresses in uncacheable memory that are used for memory-mapped I/O Status: For the steppings affected, see the Summary Tables of Changes. 54 Specification Update

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51
  • 52
  • 53
  • 54
  • 55
  • 56
  • 57
  • 58
  • 59
  • 60
  • 61
  • 62
  • 63
  • 64

54
Specification Update
Workaround:
Before performing a VM entry to the shutdown state, software should check whether
the “virtual NMIs” and “NMI-window exiting” VM-execution controls are both 1. If they
are, software should clear “NMI-window exiting” and inject an NMI as part of VM entry.
Status:
For the steppings affected, see the Summary Tables of Changes.
BJ112.
Execution of INVVPID Outside 64-Bit Mode Cannot Invalidate
Translations For 64-Bit Linear Addresses
Problem:
Executions of the INVVPID instruction outside 64-bit mode with the INVVPID type
“individual-address invalidation” ignore bits 63:32 of the linear address in the INVVPID
descriptor and invalidate translations for bits 31:0 of the linear address.
Implication:
The INVVPID instruction may fail to invalidate translations for linear addresses that set
bits in the range 63:32. Because this erratum applies only to executions outside 64-bit
mode, it applies only to attempts by a 32-bit virtual-machine monitor (VMM) to
invalidate translations for a 64-bit guest. Intel has not observed this erratum with any
commercially available software.
Workaround:
None identified.
Status:
For the steppings affected, see the Summary Tables of Changes.
BJ113.
VEX.L is Not Ignored with VCVT*2SI Instructions
Problem:
The VEX.L bit should be ignored for the VCVTSS2SI, VCVTSD2SI, VCVTTSS2SI, and
VCVTTSD2SI instructions, however due to this erratum the VEX.L bit is not ignored and
will cause a #UD.
Implication:
Unexpected #UDs will be seen when the VEX.L bit is set to 1 with VCVTSS2SI,
VCVTSD2SI, VCVTTSS2SI, and VCVTTSD2SI instructions.
Workaround:
Software should ensure that the VEX.L bit is set to 0 for all scalar instructions.
Status:
For the steppings affected, see the Summary Tables of Changes.
BJ114.
MCI_ADDR May be Incorrect For Cache Parity Errors
Problem:
In cases when a WBINVD instruction evicts a line containing an address or data parity
error (MCACOD of 0x124, and MSCOD of 0x10), the address of this error should be
logged in the MCi_ADDR register. Due to this erratum, the logged address may be
incorrect, even though MCi_Status.ADDRV (bit 63) is set.
Implication:
The address reported in MCi_ADDR may not be correct for cases of a parity error found
during WBINVD execution.
Workaround:
None identified.
Status:
For the steppings affected, see the Summary Tables of Changes.
BJ115.
Instruction Fetches Page-Table Walks May be Made Speculatively to
Uncacheable Memory
Problem:
Page-table walks on behalf of instruction fetches may be made speculatively to
uncacheable (UC) memory.
Implication:
If any paging structures are located at addresses in uncacheable memory that are used
for memory-mapped I/O, such I/O operations may be invoked as a result of speculative
execution that would never actually occur in the executed code path. Intel has not
observed this erratum with any commercially available software.
Workaround:
Software should avoid locating paging structures at addresses in uncacheable memory
that are used for memory-mapped I/O
Status:
For the steppings affected, see the Summary Tables of Changes.