Currently the man page for syscall(2) documents superh as using "trap #0x17" with a note [4] that the low bits are the number of arguments. While it was historically documented this way, that was never actually the case; all traps in the range 0x10-0x1f have always behaved identically on sh3/4, and on sh2, a different, gratuitously incompatible rage, 0x20-0x2f, was chosen because some of the traps in that range overlap with hardware exceptions. I fixed the gratuitous incompatibility in kernel commit 3623d13821: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/sh/kernel/entry-common.S?id=3623d138213aedf0e2c89720f6a3cd0e164cb310 by adding 0x1f (which does not clash with hardware exceptions) as a supported syscall trap on sh2, and the recommendation for application code manually making syscalls is to always use 0x1f, which works on all sh3/4 kernels (going back to original linux sh support) and all recent sh2 kernels. Also the trap instruction is misspelled; it should be "trapa" not "trap". I think the contents of the table should be changed from "trap #0x17" to "trapa #0x1f" and the note should be changed to remove the inaccurate information about number of arguments and to note that while any trap in the range 0x10-0x1f works, 0x1f is preferred for compatibility with sh2, and that old (pre-4.6) sh2 kernels need values in the range 0x20-0x2f. Thoughts? Rich -- To unsubscribe from this list: send the line "unsubscribe linux-man" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
