On Wed, Apr 15, 2015 at 12:33:43PM +0200, Kamil Lulko wrote:
2015-04-14 20:07 GMT+02:00 Matt Porter mporter@konsulko.com:
On ARM v7M, the processor will return to ARM mode when executing a blx instruction with bit 0 of the address == 0. Always set it to 1 to stay in thumb mode.
Signed-off-by: Matt Porter mporter@konsulko.com
common/cmd_boot.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/common/cmd_boot.c b/common/cmd_boot.c index 8f2e070..20ce652 100644 --- a/common/cmd_boot.c +++ b/common/cmd_boot.c @@ -38,6 +38,10 @@ static int do_go(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[]) * pass address parameter as argv[0] (aka command name), * and all remaining args */ +#ifdef CONFIG_CPU_V7M
/* For ARM V7M, set bit zero to stay in Thumb mode */addr++;+#endif rc = do_go_exec ((void *)addr, argc - 1, argv + 1); if (rc != 0) rcode = 1;
I think addr |= 1 would be better - there is always a possibility that kernel image has the zero bit already set (this is the case in my own Buildroot build setup I am using for STM32F4 builds). Anyways - keeping this bit set should be the responsibility of kernel image build process so such machine specific quirk can be kept out of the common code.
I'd agree about |='ing in 1. But it's not a machine quirk, it's a requirement of the Cortex-M family that you not exit Thumb-mode (since it's Thumb-only) and given how we end up trying to jump to the address (here or in 'go' which Matt didn't post the patch for, but same logic) we can / will have problems if we don't do this. You can work around it for images we throw a header into but 'go' is where this gets really annoying.