f06da27eb8
The point is to allow userspace to probe for CPUCFG without possibly triggering invalid instructions. In addition to that, future Loongson feature bits could all be stuffed into CPUCFG bit fields (or "leaves" in x86-speak) if Loongson does not make mistakes, so ELF HWCAP bits are conserved. Userspace can determine native CPUCFG availability by checking the LCSRP (Loongson CSR Present) bit in CPUCFG output after seeing CPUCFG bit in HWCAP. Native CPUCFG always sets the LCSRP bit, as CPUCFG is part of the Loongson CSR ASE, while the emulation intentionally leaves this bit clear. The other existing Loongson-specific HWCAP bits are, to my best knowledge, unused, as (1) they are fairly recent additions, (2) Loongson never back-ported the patch into their kernel fork, and (3) Loongson's existing installed base rarely upgrade, if ever; However, they are still considered userspace ABI, hence unfortunately unremovable. But hopefully at least we could stop adding new Loongson HWCAP bits in the future. Cc: Paul Burton <paulburton@kernel.org> Cc: Jiaxun Yang <jiaxun.yang@flygoat.com> Cc: Huacai Chen <chenhc@lemote.com> Signed-off-by: WANG Xuerui <git@xen0n.name> Reviewed-by: Huacai Chen <chenhc@lemote.com> Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> |
||
---|---|---|
.. | ||
auxvec.h | ||
bitfield.h | ||
bitsperlong.h | ||
break.h | ||
byteorder.h | ||
cachectl.h | ||
errno.h | ||
fcntl.h | ||
hwcap.h | ||
inst.h | ||
ioctl.h | ||
ioctls.h | ||
Kbuild | ||
kvm_para.h | ||
kvm.h | ||
mman.h | ||
msgbuf.h | ||
param.h | ||
poll.h | ||
posix_types.h | ||
ptrace.h | ||
reg.h | ||
resource.h | ||
sembuf.h | ||
setup.h | ||
sgidefs.h | ||
shmbuf.h | ||
sigcontext.h | ||
siginfo.h | ||
signal.h | ||
socket.h | ||
sockios.h | ||
stat.h | ||
statfs.h | ||
swab.h | ||
sysmips.h | ||
termbits.h | ||
termios.h | ||
types.h | ||
ucontext.h | ||
unistd.h |