Fix bug in patching args in multiple runs

[rbramand-xilinx]

Problem solved by the commit

When kernel is run multiple times and in some of the runs if arg is updated using set_arg the present code patches the arg address on top of existing already patched value this will result in test failures. This PR fixes this issue

Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered

How problem was solved, alternative solutions (if any) and why they were rejected

Storing BD data ptr entries in patch info and marking it dirty after first patching.
When second or subsequent run is made with set_arg the original BD data ptr values stored are restored and we patch on top of it which is proper.

Risks (if any) associated the changes in the commit

Moderate as this code changes xrt code critical path

What has been tested and how, request additional testing if necessary

Tested xrt elf flow test on aie2p by calling set_arg in multiple kernel runs and test passes
Also tested vadd application on aie2ps by modifying host application accordingly.
Thorough testing is needed to make sure this change doesn't break all the existing flows

Documentation impact (if any)

NA

[chvamshi-xilinx]

I know, this file is not following other XRT user space style of coding, Please update name of variable to lower case. "max_bd_words"

[larry9523]

This fix introduces a new dependency between xrt_core and AIE architecture. Ideally, all patching knowledges should come from ELF.

But given the plan that we will move all the patching service out of XRT to aiebu, I think this temporal solution is OK.

[larry9523]

@rbramand-xilinx please provide a list of tests we have done before we are confident enough to merge this PR

[rbramand-xilinx]

@rbramand-xilinx please provide a list of tests we have done before we are confident enough to merge this PR

Hi @larry9523 , I have tested xrt flow test on aie2p and vadd test on aie2ps by modifying host application and test passes on both of them.

[rbramand-xilinx]

@rbramand-xilinx please provide a list of tests we have done before we are confident enough to merge this PR

Hi @larry9523 , I have tested xrt flow test on aie2p and vadd test on aie2ps by modifying host application and test passes on both of them.

@larry9523 I have verified non control packet , control packet elf flow tests and also full elf (xclbin2elf) flows and patching args in multiple runs tests passes only after applying this change.
@sayyanna is verifying preemption test, if that test passes I guess we are good to go.

[sayyanna]

@rbramand-xilinx please provide a list of tests we have done before we are confident enough to merge this PR

Hi @larry9523 , I have tested xrt flow test on aie2p and vadd test on aie2ps by modifying host application and test passes on both of them.

@larry9523 I have verified non control packet , control packet elf flow tests and also full elf (xclbin2elf) flows and patching args in multiple runs tests passes only after applying this change. @sayyanna is verifying preemption test, if that test passes I guess we are good to go.

Verified preemption test with patching args multiple times and tests passes only after applying this change.