Clover General discussion

[Jief_Machak]

8 hours ago, Slice said:

But why it is happen only with new commit?

I guess the size of it was just what it needs to need padding between "__beginning_of_section_ctors" symbol and the actual content of the ctors section.

Honestly, I'm surprised too.

 

 

8 hours ago, Slice said:

14 seconds at the end of Device Properties strings.

 

8 hours ago, Slice said:

27:689 14:328 stringlength = 134284

That's quite some injection !

 

Between "ATI injection not set" message and "stringlength =", there is quite a lot. It's in SetDevices().

I'll have a look. Most likely add some more temporary debug message to discover where exactly it happens.

[Slice]

1 hour ago, Jief_Machak said:

 

That's quite some injection !

 

Between "ATI injection not set" message and "stringlength =", there is quite a lot. It's in SetDevices().

I'll have a look. Most likely add some more temporary debug message to discover where exactly it happens.

He injected VideoBIOS so why the length and the time.

[Slice]

Hm, he said there is no such timeout with release-5132. The pause appeared later.

[Jief_Machak]

Hum, not easy to reproduce. I cannot inject a video bios on a VMWare because it's not a ATI or NVIDIA.

Can I do it with Qemu ?

Maybe he can try some of my compilation from here : https://github.com/jief666/CloverCommits so he can tell exactly which one has the problem ?

 

[Jief_Machak]

I was adding some logs and I got an idea. Could he try this : CLOVERX64.efi and tell me if it solved the delay ?

[Slice]

I asked him to produce new log with additional DBG and see:

6:636 0:003 generate string
23:552 16:916 hex2bin
23:557 0:004 stringlength = 134284

The codes are here

  if (StringDirty) {
    EFI_PHYSICAL_ADDRESS BufferPtr = EFI_SYSTEM_TABLE_MAX_ADDRESS; //0xFE000000;
    DBG("generate string\n");
    XString8 newDeviceProperties = devprop_generate_string(device_inject_string);
    DBG("hex2bin\n");
    size_t binaryPropSize = hex2bin(newDeviceProperties, NULL, 0);
    if ( binaryPropSize > MAX_UINT32 ) {
      MsgLog("devprop_generate_string(device_inject_string) is too big");
      newDeviceProperties.setEmpty();

 

[Jief_Machak]

That’s should be fixed in the last efi I’ve posted. Could he try and confirm ?

[Slice]

23 minutes ago, Jief_Machak said:

That’s should be fixed in the last efi I’ve posted. Could he try and confirm ?

We wait.

[Jief_Machak]

I manage to reproduce. Last efi didn't fix. Will post very soon.

[Jief_Machak]

This efi should fix : CLOVERX64.efi Let me know if I can commit.

[Slice]

I compiled commit c11a90046 and test on real hardware #1 in signature.

Clover works faster. It was

5132 (master, commit 6ff5bab8a):

8:197  0:048  GUI ready

and now

5132 (master, commit c11a90046) 

5:727  0:048  GUI ready

 

And in the system 10.14.6 I got strange CPU behaviour. It go crazy!

 

I am in the process to understand what is happen.

[Jief_Machak]

Just to be sure : The fix for long generation of devprop_generate_string is NOT committed. Because I wasn't sure although I can reproduce something that looks like it. So the guy from https://applelife.ru/ has to test the efi I posted here.

4 minutes ago, Slice said:

Clover works faster. It was

5132 (master, commit 6ff5bab8a):

8:197  0:048  GUI ready

and now

5132 (master, commit c11a90046) 

5:727  0:048  GUI ready

No idea why it would be faster here...

[Slice]

I still wait for those guy.

[Slice]

Main pause was here

0:100  0:000  CloverX64 : Image base = 0x4BCAB000
2:682  2:582  
2:682  0:000  Now is 03.04.2021,  20:46:11 (GMT)
2:682  0:000  Starting Clover revision: 5132 (master, commit 6ff5bab8a) on American Megatrends EFI

and now

0:100  0:000  CloverX64 : Image base = 0x4BCAA000
0:162  0:062  
0:162  0:000  Now is 07.04.2021,  15:29:32 (GMT)
0:162  0:000  Starting Clover revision: 5132 (master, commit c11a90046) on American Megatrends EFI

 

20 minutes ago, Slice said:

I still wait for those guy.

Yes, the pause is absent!

https://www.applelife.ru/threads/clover.42089/page-1370#post-934567

[Jief_Machak]

19 hours ago, Slice said:

Main pause was here

0:100  0:000  CloverX64 : Image base = 0x4BCAB000
2:682  2:582  
2:682  0:000  Now is 03.04.2021,  20:46:11 (GMT)
2:682  0:000  Starting Clover revision: 5132 (master, commit 6ff5bab8a) on American Megatrends EFI

and now

0:100  0:000  CloverX64 : Image base = 0x4BCAA000
0:162  0:062  
0:162  0:000  Now is 07.04.2021,  15:29:32 (GMT)
0:162  0:000  Starting Clover revision: 5132 (master, commit c11a90046) on American Megatrends EFI

 

This has nothing to do with the long time needed to construct device injection string.

This one was because of a stall to give time to gdb to connect:

    gBS->Stall(2500000); // to give time to gdb to connect

Mine is enclosed with #ifdef JIEF_DEBUG, but this line was duplicated outside the #ifdef.

I guess I did (commit 3252bf63e0ce7efb735e38d2fcaaf3b6ef0315e4) that when you tried to launch gdb.

 

By the way, does launching with gdb now works ?

[Slice]

Yes, I see gdb started but I still is not famous with it.

[Jief_Machak]

Can you have a backtrace on a panic ?

[Slice]

1 hour ago, Jief_Machak said:

Can you have a backtrace on a panic ?

But I have no panic at this moment

Find the GDB manual and other documentation resources online at:
    <http://www.gnu.org/software/gdb/documentation/>.

For help, type "help".
Type "apropos word" to search for commands related to "word".
Reading symbols from /Users/sergey/src/CloverBootloader/Qemu/../Build/Clover/DEBUG_GCC53/X64/CLOVERX64.debug...
Remote debugging using localhost:9000
Python Exception <type 'exceptions.NameError'> Installation error: gdb.execute_unwinders function is missing: 
0x000000007f5142d4 in ?? ()
Hardware read watchpoint 1: *0
Breakpoint 2 at 0x7dca2bfc: file /Users/sergey/src/CloverBootloader/rEFIt_UEFI/Platform/Posix/abort.cpp, line 37.
Continuing.

 

[Slice]

New commit

/Users/sergey/src/CloverBootloader/rEFIt_UEFI/cpp_unit_test/XString_test.cpp:1570:28: error: no matching function for call to 'XStringW::takeValueFrom(XString32&, int)'
 1570 |   xsw2.takeValueFrom(xsw, 1);
      |                            ^

 

[Jief_Machak]

Using gdb on command line is very hard but annoying. I almost never use it like that.

But having the backtrace in case of panic turn 1 or 2 hours searching with log into 5 minutes.

From the console you sent, it looks like it works. Put a panic anywhere and try :-)

[Slice]

Resolved by excluding second argument.

I.e.

xsw2.takeValueFrom(xsw);

[Jief_Machak]

oups. done.

[Slice]

Yes, I can see panic

Python Exception <type 'exceptions.NameError'> Installation error: gdb.execute_unwinders function is missing: 
Breakpoint 2, panic_ (Python Exception <type 'exceptions.NameError'> Installation error: gdb.execute_unwinders function is missing: 
format=0x7ddbd006 "ххх вам", 
    va=0x7fcef598 "\030\032\376}")
    at /Users/sergey/src/CloverBootloader/rEFIt_UEFI/PlatformEFI/posix/abort.cpp:37
37	  egSetGraphicsModeEnabled(false);
(gdb) 

But I see no where it happened. I inserted panic("ххх вам") into main.cpp.

[Slice]

O yes! It is the command "bt"

 at /Users/sergey/src/CloverBootloader/rEFIt_UEFI/refit/main.cpp:3480

 

What else you can recommend me to use?

[iCanaro]

T-Ryzo