on-chip random numbers

[Raistlin]

From what I've read, RdRand is pretty fast (requests in the nanosecond ranges) and reasonably "secure" for cryptographic use.
The compatibility issue however is of more concern for generic implementations. I will be adding its detection to my hardware enumerator/system profiler in any case.

@hutch : https://www.nytimes.com/2013/09/06/us/nsa-foils-much-internet-encryption.html   
<-This is where the NSA tried to alter RdRand to make it less random

Just for interest, anyone worked with hle/rtm (Transactional Synchronization Extensions) CPUID functions (Multi threading) ?
CPUID function 7 -> EBX bits : 4 and 11

Instructions: XACQUIRE,LOCK,XRELEASE,XBEGIN,XEND,XABORT   

[Raistlin]

Never-mind - found some more information. Let me share, this looks promising to say the least.

https://www.felixcloutier.com/x86/XACQUIRE:XRELEASE.html  (Hardware Lock Elision) HLE / TSX

https://www.scss.tcd.ie/Jeremy.Jones/CS4021/transactional%20memory.pdf (Restricted Transactional Memory) RTM

[aw27]

Just for interest, anyone worked with hle/rtm (Transactional Synchronization Extensions) CPUID functions (Multi threading) ?

Probably not, given the amount of zero byte locking files databases continue to spill over.

[Raistlin]

@aw27 - I was thinking of using TSX/HLE for our multi-threaded IOCP servers.
We could potentially eliminate the Critical sections (my case) / Semaphore (your case)
mechanisms that currently steal valuable clock cycles. Seems less clunky as well.

[aw27]

You may try, but listen to the silence of Intel....
What I mean is that according to the grapevine and the errata, TSX is not yet free of bugs.

Errata:
https://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/desktop-6th-gen-core-family-spec-update.pdf

[aw27]

In addition, I could not find any processor able to support either RTM or HLE.  :(
I run also the Intel emulator without success.  :(

Code Select Expand


includelib \masm32\lib64\msvcrt.lib
printf PROTO :PTR, :VARARG
includelib \masm32\lib64\kernel32.lib
ExitProcess PROTO :DWORD

.data
nortm db "RTM not supported",10,0
nohle db "HLE not supported",10,0

.code


main proc
sub rsp, 28h
; Are RTM & HLE supported?
mov eax, 7
mov ecx, 0
cpuid
bt ebx, 11
jc hle
mov rcx, offset nortm
call printf
hle:
bt ebx, 4
jc exit; supported
mov rcx, offset nohle
call printf

exit:
mov rcx, 0
call ExitProcess

main endp

end


And MASM does not recognize any of the instructions XACQUIRE,LOCK,XRELEASE,XBEGIN,XEND,XABORT.

So, it is a whole fantasy.

[Raistlin]

D@mn!t  - oh well. That is strange though, TSX/HLE has been around since 2014 (Haswell).

[hutch--]

Rudi,

Thanks for the link, its an interesting read and gives you every good reason to avoid anything that is well known in the field of encryption as it is probably compromised before it hits the market. Its the normal response, if the good guys (sic) can use it, the bad guys can as well. Any of the big players already know this and I imagine that any of them already have very big computer grunt for breaking anything that relies on the key bit size.

The stupid part is that anyone who seriously needs to encrypt data is not using crap like this, they are using their own techniques and this goes for fruitcakes like the Islamic State, spies of any of the government agencies, truly big business and folks doing secret research on some massive range of subjects. Security agencies grabbing at information are not doing it for security reasons, they are doing it for control and clandestine theft of technology.

[aw27]

D@mn!t  - oh well. That is strange though, TSX/HLE has been around since 2014 (Haswell).

Oh yeah, d@mn!t:
https://www.youtube.com/watch?v=AIXUgtNC4Kc

[Raistlin]

 

[K_F]

Oh yeah, d@mn!t:
https://www.youtube.com/watch?v=AIXUgtNC4Kc

So sad.. but that's representative of the state that SA has descended to..  :lol:

[aw27]

It is difficult for Raistlin to do TSX/HLE assembly language in there.  :(

[zedd151]

....Try this one out on your computer, it test if RDRAND is available and if it is, it runs it.

Code Select Expand

Entropy = 7.999998 bits per byte.

Optimum compression would reduce the size
of this 80000000 byte file by 0 percent.

Chi square distribution for 80000000 samples is 232.44, and randomly
would exceed this value 75.00 percent of the times.

Arithmetic mean value of data bytes is 127.5112 (127.5 = random).
Monte Carlo value for Pi is 3.140695879 (error 0.03 percent).
Serial correlation coefficient is 0.000011 (totally uncorrelated = 0.0).

That's all folks
Press any key to exit ....


Processor   AMD A6-9220e RADEON R4, 5 COMPUTE CORES 2C+3G, 1600 Mhz, 2 Core(s), 2 Logical Processor(s)   

[aw27]

After scratching my head looking for the reasons the Intel emulator does not properly identify TSX instructions I found that the answer lies in a mysterious switch that we only know about in the extended help.
So, if I do: "sde -help" in the command line, nothing surfaces about TSX. But if I do "sde -long-help" it is all there.

Now, if I have a file called tsx.exe with TSX instructions, I can make the emulator run as if it had an Haswell CPU (before TSX being disabled in the Haswell CPU  :lol: ). Well, it appears that in real life, TSX instructions are currently disabled in all recent CPUs, except some server models, but someday all will be back to normal and we all will be more happy with those TSX instructions.

This is my reviewed TSX identification snippet, tested on the emulator.

Code Select Expand



includelib \masm32\lib64\msvcrt.lib
printf PROTO :PTR, :VARARG
includelib \masm32\lib64\kernel32.lib
ExitProcess PROTO :DWORD

.data
nortm db "RTM not supported",10,0
rtm db "RTM supported",10,0
nohle db "HLE not supported",10,0
hle db "HLE supported",10,0

.code

main proc
sub rsp, 28h
; Are RTM & HLE supported?
mov eax, 7
mov ecx, 0
cpuid
bt ebx, 11
jc rtmsup
mov rcx, offset nortm
call printf
jmp short skiprtmsup
rtmsup:
mov rcx, offset rtm
call printf
skiprtmsup:
bt ebx, 4
jc hlesup
mov rcx, offset nohle
call printf
jmp short exit
hlesup:
mov rcx, offset hle
call printf

exit:
mov rcx, 0
call ExitProcess

main endp

end


I was also badmouthing MASM for not recognizing any of the instructions XACQUIRE,LOCK,XRELEASE,XBEGIN,XEND,XABORT. Actually, it does recognize, I was not using the correct syntax.

[felipe]

Nice program aw27, i like your style, thanks for sharing:

Code Select Expand


C:\masm32\xxx>xxx
RTM not supported
HLE not supported

C:\masm32\xxx>


:icon14: