Questionable address adjustment in SB buffers may break alignment requirements

[skliper]

Describe the bug
Software bus message buffers should handle most restrictive alignment requirements for a message. Address arithmetic used in manipulating buffers may break this alignment on some systems. Even if not broken, a maintenance issue since an update to CFE_SB_BufferD_t could break things unexpectedly.

cFE/fsw/cfe-core/src/sb/cfe_sb_api.c

Lines 1637 to 1641 in 32f3dee

	/* first set ptr to actual msg buffer the same as ptr to descriptor */
	address = (cpuaddr)bd;
	/* increment actual msg buffer ptr beyond the descriptor */
	address += sizeof(CFE_SB_BufferD_t);

cFE/fsw/cfe-core/src/sb/cfe_sb_buf.c

Lines 84 to 88 in 32f3dee

	/* first set ptr to actual msg buffer the same as ptr to descriptor */
	address = (uint8 *)bd;
	/* increment actual msg buffer ptr beyond the descriptor */
	address += sizeof(CFE_SB_BufferD_t);

cFE/fsw/cfe-core/src/sb/cfe_sb_buf.c

Line 119 in 32f3dee

CFE_SB_BufferD_t *bd = (CFE_SB_BufferD_t *)((cpuaddr)Address - sizeof(CFE_SB_BufferD_t));

To Reproduce
Not confirmed, but likely won't meet alignment requirement for a message with long double.

Expected behavior
Safer to use the real buffer type (instead of void *) in the descriptor along with offsetof to size the buffer correctly (CFE_SB_Msg_t for now, or maybe CFE_SB_Buffer_t from #1019)

Code snips
See above

System observed on:
NA - inspection

Additional context
#1019, #1009

Reporter Info
Jacob Hageman - NASA/GSFC

[skliper]

Cleaning this up could also simplify the ZeroCopy logic for passing around the buffer descriptor, and it's relation to the zero copy descriptor (could eliminate CFE_SB_GetBufferFromCaller).