nsim failed in tests/subsys/jwt/libraries_encoding

[vonhust]

Describe the bug
arc nsim_em and nsim_sem failed in tests/subsys/jwt/library_encoding.

It's caused by the wrong calculation of _timer_int_handler of arcv2_timer0 driver

The address of _timer_int_handler in zephyr.elf and zephyr_prebuilt.elf are different. (It should be same).
In tests/kernel , there are no problems, ,the address of _timer_int_handler in zephyr.elf and zephyr_prebuilt is the same.

I compared the map files of zephyr.elf and zephyr_prebuilt.elf and found there are difference in linking of some functions in libgcc.a

info of zephyr_prebuilt.elf
text 0x0000000000000000 0x1383c
0x0000000000000000 _image_rom_start = .
0x0000000000000000 _image_text_start = .
*(.exc_vector_table)
.exc_vector_table
0x0000000000000000 0x40 arch/arch/arc/core/libarch__arc__core.a(vector_table.c.obj)
0x0000000000000000 _VectorTable
(.exc_vector_table.)
*(.gnu.linkonce.irq_vector_table)
.gnu.linkonce.irq_vector_table
0x0000000000000040 0x10 libzephyr.a(isr_tables.c.obj)
0x0000000000000040 _irq_vector_table
*(.text)
.text 0x0000000000000050 0x378 /opt/zephyr-sdk/sysroots/arc-zephyr-elf/usr/lib/arc-zephyr-elf/6.2.1/em4_dmips/libgcc.a(_udivdi3.o)
0x0000000000000050 __udivdi3
.text 0x00000000000003c8 0xf8 /opt/zephyr-sdk/sysroots/arc-zephyr-elf/usr/lib/em4_dmips/libc.a(lib_a-memcmp-bs-norm.o)

info of zephyr.elf
text 0x0000000000000000 0x138a8
0x0000000000000000 _image_rom_start = .
0x0000000000000000 _image_text_start = .
*(.exc_vector_table)
.exc_vector_table
0x0000000000000000 0x40 arch/arch/arc/core/libarch__arc__core.a(vector_table.c.obj)
0x0000000000000000 _VectorTable
(.exc_vector_table.)
*(.gnu.linkonce.irq_vector_table)
.gnu.linkonce.irq_vector_table
0x0000000000000040 0x10 CMakeFiles/kernel_elf.dir/isr_tables.c.obj
0x0000000000000040 _irq_vector_table
*(.text)
.text 0x0000000000000050 0x30 /opt/zephyr-sdk/sysroots/arc-zephyr-elf/usr/lib/arc-zephyr-elf/6.2.1/em4_dmips/libgcc.a(_millicodethunk_st.o)
0x0000000000000050 __st_r13_to_r25
0x0000000000000054 __st_r13_to_r24
0x0000000000000058 __st_r13_to_r23
0x000000000000005c __st_r13_to_r22
0x0000000000000060 __st_r13_to_r21
0x0000000000000064 __st_r13_to_r20
0x0000000000000068 __st_r13_to_r19
0x000000000000006c __st_r13_to_r18
0x0000000000000070 __st_r13_to_r17
0x0000000000000074 __st_r13_to_r16
0x0000000000000078 __st_r13_to_r15
.text 0x0000000000000080 0x3c /opt/zephyr-sdk/sysroots/arc-zephyr-elf/usr/lib/arc-zephyr-elf/6.2.1/em4_dmips/libgcc.a(_millicodethunk_ret.o)
0x0000000000000080 __ld_r13_to_r25_ret
0x0000000000000084 __ld_r13_to_r24_ret
0x0000000000000088 __ld_r13_to_r23_ret
0x000000000000008c __ld_r13_to_r22_ret
0x0000000000000090 __ld_r13_to_r21_ret
0x0000000000000094 __ld_r13_to_r20_ret
0x0000000000000098 __ld_r13_to_r19_ret
0x000000000000009c __ld_r13_to_r18_ret
0x00000000000000a0 __ld_r13_to_r17_ret
0x00000000000000a4 __ld_r13_to_r16_ret
0x00000000000000a8 __ld_r13_to_r15_ret
0x00000000000000ac __ld_r13_to_r14_ret
.text 0x00000000000000bc 0x378 /opt/zephyr-sdk/sysroots/arc-zephyr-elf/usr/lib/arc-zephyr-elf/6.2.1/em4_dmips/libgcc.a(_udivdi3.o)
0x00000000000000bc __udivdi3

The root cause is unknown,

To Reproduce
Steps to reproduce the behavior:
0. cd tests\subsys\jwt\library_encoding

1. mkdir build; cd build
2. cmake -DBOARD=nsim_em ..
3. make run
4. see error
5. comapre the value (the first entry)in isr_table.c and the value of _timer_int_handler in zephyr.map.

Expected behavior
If the address caculation is ok. The test will pass as shown below

***** Booting Zephyr OS v1.7.99-15745-g00c0cb066c5f *****
Running test suite lib_jwt_test
===================================================================
starting test - test_jwt
JWT:
eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCJ9.eyJhdWQiOiJpb3Qtd29yay0xOTk0MTkiLCJleHAiOjE1MzAzMTIwMjYsImlhdCI6MTUzMDMwODQyNn0.VHKDT-dvWW3UDBGokkttP7P6DNDnd4jjz_wLsOvn9RVBoDAAVG1mQnlROPtOHyRTotWzFYf2g02-VyZoh326yDFXM-66U7ycqyUsJdWBZWYqub6iZ17l0hjodoSpZBJqPB6F9c8Q5u9x1gWVw2l2XNASQyTDIZ3db_GiQbx1SYnNtyxsospoKClIebee-EQkcut9jBPNyiKrzZRPuX0xCiPJYTumHdHSp0UPIDIij2hJl_tacXs-hz_ckqT5HZvJPsxgJCeXz3tP7IlgbeGbPFMFmaYNX4UxYD4XS10MPXUK4_qp8wK4EqjT2YBNF7xXi_sjWMj7isc0i0YSIl6_nQ
len: 459
PASS - test_jwt
===================================================================
Test suite lib_jwt_test succeeded
===================================================================
PROJECT EXECUTION SUCCESSFUL

Impact
What impact does this issue have on your progress (e.g., annoyance, showstopper)

Screenshots or console output
If applicable, add a screenshot (drag-and-drop an image), or console logs
(cut-and-paste text and put a code fence (```) before and after, to help
explain the issue.

Environment (please complete the following information):

• OS: (e.g. Linux, MacOS, Windows)
• Toolchain (e.g Zephyr SDK, ...)
• Commit SHA or Version used

Additional context
Add any other context about the problem here.

[andrewboie]

Might be a symptom of something very bad going on -- there's supposed to be no difference in symbol offsets between zephyr.elf and zephyr_prebuilt.elf. I'll take a look.

[andrewboie]

As part of working on this I'm going to try to implement some test that proves memory addresses are not shifting between zephyr_prebuilt.elf and zephyr.elf, this is something we have seen before (with various root causes)

[andrewboie]

I've reproduced this and indeed, the zephyr final kernel binary is pulling in all this __st_XXX_to_YYY and __ld_XXX_to_YYY_ret functions. They were not present in zephyr_prebuilt.elf. Every function after them is getting shifted, starting with __udivdi3.

I don't comprehend yet what these "millicodethunk" functions do.

Presumably, these are getting pulled in by functions that are only present in zephyr.elf and not zephyr_prebuilt.elf. I'm still investigating but I believe this is just the functions generated by gperf. What I don't understand yet is why it's just this test that is having problems.

[andrewboie]

From the GCC manual:

       -mno-millicode
           When optimizing for size (using -Os), prologues and epilogues
           that have to save or restore a large number of registers are
           often shortened by using call to a special function in libgcc;
           this is referred to as a millicode call.  As these calls can pose
           performance issues, and/or cause linking issues when linking in a
           nonstandard way, this option is provided to turn off millicode
           call generation.

Now I could just add this to ARC's cflags...but it doesn't make sense that these are there in zephyr.elf and not zephyr_prebuilt.elf