[edk2-devel] Is DEBUG output function safe for MP case?

[edk2-devel] Is DEBUG output function safe for MP case?

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 587 bytes --]

Hello
I ask for help for debug output in multi-core case.


Is the current DEBUG marco safe for AP Cores to output message?
If the output message is too long, it takes too much time, will it cause some abnormal exception?


Thanks

-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119406): https://edk2.groups.io/g/devel/message/119406
Mute This Topic: https://groups.io/mt/106403315/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 1254 bytes --]

回复: [edk2-devel] Is DEBUG output function safe for MP case?

[Jeff Fan via groups.io]

[-- Attachment #1: Type: text/plain, Size: 918 bytes --]

AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-05-31 14:22
收件人： devel@edk2.groups.io
主题： [edk2-devel] Is DEBUG output function safe for MP case?
Hello
I ask for help for debug output in multi-core case.

Is the current DEBUG marco safe for AP Cores to output message?
If the output message is too long, it takes too much time, will it cause some abnormal exception?

Thanks



-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119407): https://edk2.groups.io/g/devel/message/119407
Mute This Topic: https://groups.io/mt/106403392/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 2871 bytes --]

Re: [edk2-devel] Is DEBUG output function safe for MP case?

[Andrew Fish via groups.io]

[-- Attachment #1: Type: text/plain, Size: 3974 bytes --]

I think the generic answer is no DEBUG() is not guaranteed to be MP safe. 

Jeff is pointing out in the edk2 your platform build can control what libraries are used to implement DEBUG(), so you might be able to make it work in your case. 

Base libs are libraries that don’t depend on other things so they have a much better chance of working but they will have at least 2 issue:
1) The are not MP safe so the output could get intermixed and things could get lost as multiple agents try to to write to the UART in parallel
2) A Base lib that goes after a x86 IO port has a chance of working at runtime, but an ARM UART driver that uses an MMIO address is not going to work as the that MMIO range would need a virtual mapping and a Base lib is not going to request that.

You can search for instances of the DebugLib (that implements DEBUG) by grepping the inf files. 

/Volumes/Case/edk2(master)>git grep DebugLib -- \*.inf | grep LIBRARY_CLASS | grep DXE_RUNTIME_DRIVER
ArmPkg/Library/SemiHostingDebugLib/SemiHostingDebugLib.inf:19:  LIBRARY_CLASS                  = DebugLib|BASE SEC DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER DXE_SMM_DRIVER UEFI_APPLICATION UEFI_DRIVER
ArmVirtPkg/Library/DebugLibFdtPL011Uart/DxeRuntimeDebugLibFdtPL011Uart.inf:23:  LIBRARY_CLASS  = DebugLib|DXE_RUNTIME_DRIVER
MdeModulePkg/Library/PeiDxeDebugLibReportStatusCode/PeiDxeDebugLibReportStatusCode.inf:19:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER DXE_SMM_DRIVER SMM_CORE PEIM SEC PEI_CORE UEFI_APPLICATION UEFI_DRIVER MM_STANDALONE
MdePkg/Library/DxeRuntimeDebugLibSerialPort/DxeRuntimeDebugLibSerialPort.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_RUNTIME_DRIVER
MdePkg/Library/UefiDebugLibConOut/UefiDebugLibConOut.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
MdePkg/Library/UefiDebugLibDebugPortProtocol/UefiDebugLibDebugPortProtocol.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
MdePkg/Library/UefiDebugLibStdErr/UefiDebugLibStdErr.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
OvmfPkg/Library/PlatformDebugLibIoPort/PlatformDebugLibIoPort.inf:19:  LIBRARY_CLASS                  = DebugLib|PEI_CORE PEIM DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER SMM_CORE DXE_SMM_DRIVER UEFI_DRIVER UEFI_APPLICATION

Note: DXE_RUNTIME_DRIVE may not imply the the DEBUG works at running, and it may only support DEBUG print a Boot Services time. But if you look at the list you will see some of the libs are designed to used at runtime. So that should imply they deal with virtual address mapping and locks. 

Thanks,

Andrew Fish

> On May 31, 2024, at 1:33 AM, Jeff Fan via groups.io <fanjianfeng=byosoft.com.cn@groups.io> wrote:
> 
> AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)
> 
> 
> fanjianfeng@byosoft.com.cn <mailto:fanjianfeng@byosoft.com.cn>
>  
> 发件人： Yoshinoya <mailto:yoshinoyatoko@163.com>
> 发送时间： 2024-05-31 14:22
> 收件人： devel@edk2.groups.io <mailto:devel@edk2.groups.io>
> 主题： [edk2-devel] Is DEBUG output function safe for MP case?
> Hello
> I ask for help for debug output in multi-core case.
> 
> Is the current DEBUG marco safe for AP Cores to output message?
> If the output message is too long, it takes too much time, will it cause some abnormal exception?
> 
> Thanks
> 



-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119409): https://edk2.groups.io/g/devel/message/119409
Mute This Topic: https://groups.io/mt/106405010/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 13578 bytes --]

Re: [edk2-devel] Is DEBUG output function safe for MP case?

[Andrew Fish via groups.io]

[-- Attachment #1: Type: text/plain, Size: 4688 bytes --]

There is also one extra caveat about UEFI runtime that I forgot to mention. There is no way to share the hardware between UEFI and the OS, so UEFI really needs dedicated hardware.

Thanks,

Andrew Fish

PS This makes me think of a Windows story from the 1990’s when Windows added Plug-and-Play for serial to detect modems. A factory updated Windows on a PC that controlled a CnC machine and the modem auto detect AT commands got sent to the CnC machine, and this told the CnC machine to start drilling holes in the CnC machine.  

> On May 31, 2024, at 4:59 AM, Andrew Fish via groups.io <afish=apple.com@groups.io> wrote:
> 
> I think the generic answer is no DEBUG() is not guaranteed to be MP safe. 
> 
> Jeff is pointing out in the edk2 your platform build can control what libraries are used to implement DEBUG(), so you might be able to make it work in your case. 
> 
> Base libs are libraries that don’t depend on other things so they have a much better chance of working but they will have at least 2 issue:
> 1) The are not MP safe so the output could get intermixed and things could get lost as multiple agents try to to write to the UART in parallel
> 2) A Base lib that goes after a x86 IO port has a chance of working at runtime, but an ARM UART driver that uses an MMIO address is not going to work as the that MMIO range would need a virtual mapping and a Base lib is not going to request that.
> 
> You can search for instances of the DebugLib (that implements DEBUG) by grepping the inf files. 
> 
> /Volumes/Case/edk2(master)>git grep DebugLib -- \*.inf | grep LIBRARY_CLASS | grep DXE_RUNTIME_DRIVER
> ArmPkg/Library/SemiHostingDebugLib/SemiHostingDebugLib.inf:19:  LIBRARY_CLASS                  = DebugLib|BASE SEC DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER DXE_SMM_DRIVER UEFI_APPLICATION UEFI_DRIVER
> ArmVirtPkg/Library/DebugLibFdtPL011Uart/DxeRuntimeDebugLibFdtPL011Uart.inf:23:  LIBRARY_CLASS  = DebugLib|DXE_RUNTIME_DRIVER
> MdeModulePkg/Library/PeiDxeDebugLibReportStatusCode/PeiDxeDebugLibReportStatusCode.inf:19:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER DXE_SMM_DRIVER SMM_CORE PEIM SEC PEI_CORE UEFI_APPLICATION UEFI_DRIVER MM_STANDALONE
> MdePkg/Library/DxeRuntimeDebugLibSerialPort/DxeRuntimeDebugLibSerialPort.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_RUNTIME_DRIVER
> MdePkg/Library/UefiDebugLibConOut/UefiDebugLibConOut.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
> MdePkg/Library/UefiDebugLibDebugPortProtocol/UefiDebugLibDebugPortProtocol.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
> MdePkg/Library/UefiDebugLibStdErr/UefiDebugLibStdErr.inf:22:  LIBRARY_CLASS                  = DebugLib|DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER UEFI_APPLICATION UEFI_DRIVER
> OvmfPkg/Library/PlatformDebugLibIoPort/PlatformDebugLibIoPort.inf:19:  LIBRARY_CLASS                  = DebugLib|PEI_CORE PEIM DXE_CORE DXE_DRIVER DXE_RUNTIME_DRIVER SMM_CORE DXE_SMM_DRIVER UEFI_DRIVER UEFI_APPLICATION
> 
> Note: DXE_RUNTIME_DRIVE may not imply the the DEBUG works at running, and it may only support DEBUG print a Boot Services time. But if you look at the list you will see some of the libs are designed to used at runtime. So that should imply they deal with virtual address mapping and locks. 
> 
> Thanks,
> 
> Andrew Fish
> 
>> On May 31, 2024, at 1:33 AM, Jeff Fan via groups.io <fanjianfeng=byosoft.com.cn@groups.io> wrote:
>> 
>> AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)
>> 
>> 
>> fanjianfeng@byosoft.com.cn <mailto:fanjianfeng@byosoft.com.cn>
>>  
>> 发件人： Yoshinoya <mailto:yoshinoyatoko@163.com>
>> 发送时间： 2024-05-31 14:22
>> 收件人： devel@edk2.groups.io <mailto:devel@edk2.groups.io>
>> 主题： [edk2-devel] Is DEBUG output function safe for MP case?
>> Hello
>> I ask for help for debug output in multi-core case.
>> 
>> Is the current DEBUG marco safe for AP Cores to output message?
>> If the output message is too long, it takes too much time, will it cause some abnormal exception?
>> 
>> Thanks
> 
> 



-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119410): https://edk2.groups.io/g/devel/message/119410
Mute This Topic: https://groups.io/mt/106405152/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 18419 bytes --]

Re: 回复: [edk2-devel] Is DEBUG output function safe for MP case?

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 1089 bytes --]

Hi, Jeff:

If AP code use base-typed debuglib, then it's mp safe?




Thanks
















At 2024-05-31 14:33:40, "Jeff Fan via groups.io" <fanjianfeng=byosoft.com.cn@groups.io> wrote:

AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-05-31 14:22
收件人： devel@edk2.groups.io
主题： [edk2-devel] Is DEBUG output function safe for MP case?
Hello
I ask for help for debug output in multi-core case.


Is the current DEBUG marco safe for AP Cores to output message?
If the output message is too long, it takes too much time, will it cause some abnormal exception?


Thanks


-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119424): https://edk2.groups.io/g/devel/message/119424
Mute This Topic: https://groups.io/mt/106403392/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 3477 bytes --]

回复: Re: [edk2-devel] Is DEBUG output function safe for MP case?

[Jeff Fan via groups.io]

[-- Attachment #1: Type: text/plain, Size: 1869 bytes --]

As Andrew mentioned, DEBUG() is NOT safe for MP. 

Generally, kernel modules are not suggested to use DEBUG() in AP code becasue they are not the decision maker on the final DebuLib instance.
If you could make sure the DebugLib instance platform selected only based on I/O or MMIO operations, you could use DEBUG() in AP for debuging or testing purpose.
To avoid mess output message due to multiple APs, you may use spinlock to protect DEBUG() invoking.

For runtime (for example SMM AP code),  the upper method may not work becaseu OS may take control on Serial Port.  




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-06-03 16:24
收件人： devel; fanjianfeng
主题： Re:回复: [edk2-devel] Is DEBUG output function safe for MP case?
Hi, Jeff:
If AP code use base-typed debuglib, then it's mp safe?

Thanks





At 2024-05-31 14:33:40, "Jeff Fan via groups.io" <fanjianfeng=byosoft.com.cn@groups.io> wrote:
AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-05-31 14:22
收件人： devel@edk2.groups.io
主题： [edk2-devel] Is DEBUG output function safe for MP case?
Hello
I ask for help for debug output in multi-core case.

Is the current DEBUG marco safe for AP Cores to output message?
If the output message is too long, it takes too much time, will it cause some abnormal exception?

Thanks



-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119441): https://edk2.groups.io/g/devel/message/119441
Mute This Topic: https://groups.io/mt/106475080/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 5496 bytes --]

Re: 回复: Re: [edk2-devel] Is DEBUG output function safe for MP case?

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 2029 bytes --]

Hi, Jeff/Andrew;

Thanks for your help!




Appreciates it.




best wishes,













At 2024-06-04 09:11:50, "Jeff Fan via groups.io" <fanjianfeng=byosoft.com.cn@groups.io> wrote:

As Andrew mentioned, DEBUG() is NOT safe for MP. 


Generally, kernel modules are not suggested to use DEBUG() in AP code becasue they are not the decision maker on the final DebuLib instance.
If you could make sure the DebugLib instance platform selected only based on I/O or MMIO operations, you could use DEBUG() in AP for debuging or testing purpose.
To avoid mess output message due to multiple APs, you may use spinlock to protect DEBUG() invoking.


For runtime (for example SMM AP code),  the upper method may not work becaseu OS may take control on Serial Port.  




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-06-03 16:24
收件人： devel; fanjianfeng
主题： Re:回复: [edk2-devel] Is DEBUG output function safe for MP case?

Hi, Jeff:

If AP code use base-typed debuglib, then it's mp safe?




Thanks
















At 2024-05-31 14:33:40, "Jeff Fan via groups.io" <fanjianfeng=byosoft.com.cn@groups.io> wrote:

AP cannot run any UEFI Service. Please make sure AP code use base-typed DebugLib instance (Such as BaseDebugLibSerialport.inf from MdePkg)




fanjianfeng@byosoft.com.cn
 
发件人： Yoshinoya
发送时间： 2024-05-31 14:22
收件人： devel@edk2.groups.io
主题： [edk2-devel] Is DEBUG output function safe for MP case?
Hello
I ask for help for debug output in multi-core case.


Is the current DEBUG marco safe for AP Cores to output message?
If the output message is too long, it takes too much time, will it cause some abnormal exception?


Thanks


-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119506): https://edk2.groups.io/g/devel/message/119506
Mute This Topic: https://groups.io/mt/106475080/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 6095 bytes --]

[edk2-devel] processing compressed section during CAR phase

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

Hello

A question about compressed peims in uefi bios.




for example:

Current x86 cache size is large enough.

Is there possible to process compressed peim during cache as ram phase?




Is there any limitations?




Thanks

-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119557): https://edk2.groups.io/g/devel/message/119557
Mute This Topic: https://groups.io/mt/106627102/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 1747 bytes --]

[edk2-devel] uncompress peims during car phase

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

Hello

A question about compressed peims in uefi bios.




for example:

Current x86 cache size is large enough.

Is there possible to process compressed peim during cache as ram phase?




Is there any limitations?




Thanks

-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119650): https://edk2.groups.io/g/devel/message/119650
Mute This Topic: https://groups.io/mt/106776870/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 1956 bytes --]

[edk2-devel] pci command supports CXL device type

[Yoshinoya]

[-- Attachment #1: Type: text/plain, Size: 486 bytes --]

Hello

Maybe current shell pci command doesn't support parse cxl device type.

Do we have plan to update this command?




Thanks

-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#119505): https://edk2.groups.io/g/devel/message/119505
Mute This Topic: https://groups.io/mt/106517896/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-



[-- Attachment #2: Type: text/html, Size: 1108 bytes --]