On 01/27/18 17:17, Laszlo Ersek wrote:
> Hello Jian,
>
> On 01/15/18 09:54, Jian J Wang wrote:
>> If PcdDxeNxMemoryProtectionPolicy is set to enable protection for memory
>> of EfiBootServicesCode, EfiConventionalMemory, the BIOS will hang at a page
>> fault exception during MP initialization.
>>
>> The root cause is that the AP wake up buffer, which is below 1MB and used
>> to hold both AP init code and data, is type of EfiConventionalMemory (not
>> really allocated because of potential conflict with legacy code), and is
>> marked as non-executable. During the transition from real address mode
>> to long mode, the AP init code has to enable paging which will then cause
>> itself a page fault exception because it's just running in non-executable
>> memory.
>>
>> The solution is splitting AP wake up buffer into two part: lower part is
>> still below 1MB and shared with legacy system, higher part is really
>> allocated memory of BootServicesCode type. The init code in the memory
>> below 1MB will not enable paging but just switch to protected mode and
>> jump to higher memory, in which the init code will enable paging and
>> switch to long mode.
>>
>> Cc: Jiewen Yao <jiewen....@intel.com>
>> Cc: Ruiyu Ni <ruiyu...@intel.com>
>> Cc: Eric Dong <eric.d...@intel.com>
>> Cc: Laszlo Ersek <ler...@redhat.com>
>> Contributed-under: TianoCore Contribution Agreement 1.1
>> Signed-off-by: Jian J Wang <jian.j.w...@intel.com>
>> ---
>>  UefiCpuPkg/Library/MpInitLib/DxeMpLib.c        | 34 ++++++++++
>>  UefiCpuPkg/Library/MpInitLib/Ia32/MpEqu.inc    |  5 ++
>>  UefiCpuPkg/Library/MpInitLib/Ia32/MpFuncs.nasm | 32 +++++-----
>>  UefiCpuPkg/Library/MpInitLib/MpLib.c           | 45 +++++++++++++
>>  UefiCpuPkg/Library/MpInitLib/MpLib.h           | 22 +++++++
>>  UefiCpuPkg/Library/MpInitLib/PeiMpLib.c        | 23 +++++++
>>  UefiCpuPkg/Library/MpInitLib/X64/MpEqu.inc     |  5 +-
>>  UefiCpuPkg/Library/MpInitLib/X64/MpFuncs.nasm  | 87 
>> ++++++++++++++++----------
>>  8 files changed, 204 insertions(+), 49 deletions(-)
>
> This patch breaks OVMF on KVM. The symptom is that the guest crashes
> and reboots as follows (infinite reboot loop):
>
>> Loading PEIM at 0x0007FEB0000 EntryPoint=0x0007FEB5C96 CpuMpPei.efi
>> AP Loop Mode is 1
>> WakeupBufferStart = 9F000, WakeupBufferSize = 1000
>> -- crash & reboot here --
>> SecCoreStartupWithStack(0xFFFCC000, 0x820000)

The following build options were used for this build:

$ build -a X64 -p OvmfPkg/OvmfPkgX64.dsc -D SECURE_BOOT_ENABLE \
    -t GCC48 -b NOOPT -D HTTP_BOOT_ENABLE

The tree was built at 06c1f423e17f ("BeagleBoardPkg: reroute Firmware
Vendor Pcd to MdeModulePkg", 2018-01-26). (This commit is listed at the
top of the bisection log in my previous email.)

Here's the KVM log up to the triple fault:

>              CPU-32283 [004] 13652.374591: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374594: kvm_exit:             reason 
> CR_ACCESS rip 0x3a info 0 0
>              CPU-32283 [004] 13652.374595: kvm_cr:               cr_write 0 = 
> 0x60000013
>              CPU-32283 [004] 13652.374596: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374597: kvm_exit:             reason 
> CR_ACCESS rip 0x9f06a info 4 0
>              CPU-32283 [004] 13652.374598: kvm_cr:               cr_write 4 = 
> 0x20
>              CPU-32283 [004] 13652.374603: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374604: kvm_exit:             reason 
> CR_ACCESS rip 0x9f075 info 103 0
>              CPU-32283 [004] 13652.374605: kvm_cr:               cr_write 3 = 
> 0x800000
>              CPU-32283 [004] 13652.374606: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374607: kvm_exit:             reason 
> MSR_READ rip 0x9f07d info 0 0
>              CPU-32283 [004] 13652.374608: kvm_msr:              msr_read 
> c0000080 = 0x0
>              CPU-32283 [004] 13652.374608: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374609: kvm_exit:             reason 
> MSR_WRITE rip 0x9f083 info 0 0
>              CPU-32283 [004] 13652.374611: kvm_msr:              msr_write 
> c0000080 = 0x100
>              CPU-32283 [004] 13652.374612: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374613: kvm_exit:             reason 
> CR_ACCESS rip 0x9f08c info 0 0
>              CPU-32283 [004] 13652.374613: kvm_cr:               cr_write 0 = 
> 0xe0000013
>              CPU-32283 [004] 13652.374620: kvm_entry:            vcpu 2
>              CPU-32283 [004] 13652.374622: kvm_exit:             reason 
> TRIPLE_FAULT rip 0x9f096 info 0 0

Offset 0x96 (relative to 0x9F000) is from
"UefiCpuPkg/Library/MpInitLib/X64/MpFuncs.nasm":

> 5c66d125eaae5 (Jeff Fan       2016-07-29 21:13:34 +0800 103) 
> SkipEnableExecuteDisableBit:
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 104)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 105)     ; Enable PAE
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 106)     ;
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 107)     mov        
> eax, cr4
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 108)     bts        
> eax, 5
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 109)     mov        
> cr4, eax
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 110)
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 111)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 112)     ; Load page 
> table
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 113)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 114)     mov        
> esi, Cr3Location             ; Save CR3 in ecx
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 115)     mov        
> ecx, [ebx + esi]
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 116)     mov        
> cr3, ecx                    ; Load CR3
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 117)
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 118)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 119)     ; Enable 
> long mode
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 120)     ;
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 121)     mov        
> ecx, 0c0000080h             ; EFER MSR number
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 122)     rdmsr        
>                           ; Read EFER
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 123)     bts        
> eax, 8                      ; Set LME=1
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 124)     wrmsr        
>                           ; Write EFER
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 125)
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 126)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 127)     ; Enable 
> paging
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 128)     ;
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 129)     mov        
> eax, cr0                    ; Read CR0
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 130)     bts        
> eax, 31                     ; Set PG=1
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 131)     mov        
> cr0, eax                    ; Write CR0
> d94e5f672994f (Jeff Fan       2016-07-20 22:44:39 +0800 132)
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 133)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 134)     ; Far jump 
> to 64-bit code
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 135)     ;
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 136)     mov        
> edi, ModeHighMemoryLocation
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 137)     add        
> edi, ebx
> f32bfe6d06142 (Jian J Wang    2017-12-29 09:12:54 +0800 138)     jmp far    
> [edi]                                          <- here

Thanks
Laszlo
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel

Reply via email to