hello_uart.S: Fix symbol label to point to string

[mablinov]

Tested on Nexys4 DDR.

We want to make sure that symbol str actually sits directly above the .string directive. If the str label is defined directly before the .section .data directive, then the symbol won't point to the actual string bytes but somewhere above it, and will consequently print garbage:

00010074 <_start>:
   10098:       00000597                auipc   a1,0x0
   1009c:       03258593                addi    a1,a1,50 # 100ca <str>
   100a0:       00058283                lb      t0,0(a1)
...

Disassembly of section .data:

000110ca <__DATA_BEGIN__>:
   110ca:       52657753                .4byte  0x52657753
   110ce:       2b56                    fld     fs6,336(sp)
...

Note the discrepancy between 100ca <str> (the address the assembler thinks its at) and the data actually sitting at 0x110ca.

I think this is because if you look in the default linker script with riscv32-unknown-elf-ld -verbose ..., the .data section is preceeded by a page alignment:

* Adjust the address for the data segment.  We want to adjust up to
     the same address within the page on the next page up.  */
  . = DATA_SEGMENT_ALIGN (CONSTANT (MAXPAGESIZE), CONSTANT (COMMONPAGESIZE));

Hence the discrepancy of about 0x1000 (4kb). After this change, the symbols correspond to eachother:

0010074 <_start>:
...
   10098:       00001597                auipc   a1,0x1
   1009c:       03258593                addi    a1,a1,50 # 110ca <__DATA_BEGIN__>
   100a0:       00058283                lb      t0,0(a1)
...

Disassembly of section .data:

000110ca <__DATA_BEGIN__>:
   110ca:       52657753                .4byte  0x52657753
   110ce:       2b56                    fld     fs6,336(sp)
...

[olofk]

Many thanks for the fix and the great explanation. I'm terrible with linkers and stuff like this so there's a good chance there are more things to fix. Picked and pushed!