I was doing a little looking and I noticed that gd->have_console was not being set (is normally set in console_init_f() ). Further greps of the code for ARM show its only used in common/console.c/ctrlc() . Being that the ARM generic code never sets it (nor is it done anywhere else), the function probably didn't work before. I just changed my code (patch I sent) to do : Console_init_f(); /* set have_console = 1 */ Console_init_r(); /* complete console init */
I won't test this till tomorrow but it is obvious that have_console should be set. I can resubmit the patch if you like, or just know to add this.
* One other comment on this enabling the console patch...it should fix a bug which is in common/boot_cmd.c/read_record() . read_record currently EXPECTS syscalls_tbl to have been allocated, and it dereferences the table looking for a console device. Prior to this patch the ARM code failed to ever allocate this small table at all. So, it was likely a dereference of a uninitilized structure.
Regards,
Richard W.
-----Original Message----- From: Woodruff, Richard Sent: Sunday, June 15, 2003 5:05 PM To: u-boot-users@lists.sourceforge.net Cc: 'Wolfgang Denk' Subject: [U-Boot-Users] [PATCH] Enable device and console for ARM.
Hi,
This update enables the standard device abstraction for ARM and enables the console member. It also initializes bi_baudrate for ARM. The superficial effect of this is coninfo works with out hanging a system. More importantly it enables devices, including the console.
The only little bit stopping some of this from working before is that common/devices.c was written such that it was not relocation agnostic (Current ARM code do not do the selective relocation which the PPC ports appear to do). Everything I've tried so far still works as before (or better), and the console seems to work on my ARM board.
Regards,
Richard Woodruff