Before this was reworked, in the file for mt7621 subtarget
(target/linux/ramips/image/lzma-loader/src/board-mt7621.c)
the "Transmitter shift register empty" bit TEMT was used instead of
the "Transmitter holding register empty" bit THRE,
but after the rework, this value was labeled as the THRE bit instead.
Functionally there is no difference, but this is confusing to read,
as it suggests that the subtargets have different bits for the same
register in UART when in reality they are exactly the same.
One can use either bit, or both, at user's descretion
in order to determine whether the UART TX buffer is ready.
The generic kernel early-printk uses both,
(arch/mips/kernel/early_printk_8250.c)
while the ralink-specific early-printk uses only THRE,
(arch/mips/ralink/early_printk.c).
Define both bits and rewrite macros for readability,
keep the same values, as changing which to use should be tested first.
Ref: c31319b66 ("ramips: lzma-loader: Refactor loader")
Signed-off-by: Michael Pratt <mcpratt@pm.me>
(cherry picked from commit 2e47913c64)
Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
The native bus address for UART was entered for rt305x UART_BASE,
but the bootloaders have memory space remapped with the same
virtual memory map the kernel uses for program addressing at boot time.
In UBoot, the remapped address is often defined as TEXT_BASE.
In the kernel, for rt305x this remapped address is RT305X_SYSC_BASE.
(arch/mips/include/asm/mach-ralink/rt305x.h)
Because the ralink I/O busses begin at a low address of 0x10000000,
they are remapped using KSEG0 or KSEG1, which for all 32-bit MIPS SOCs
(arch/mips/include/asm/addrspace.h)
are offsets of 0x80000000 and 0xa0000000 respectively.
This is consistent with the other UART_BASE macros here
and with MIPS memory map documentation.
Before the recent rework of the lzma-loader for ramips,
the original board-$(PLATFORM).c files also did not
use KSEG1ADDR for UART_BASE despite being defined,
which made this mistake easier to occur.
Fix this by defining KSEG1ADDR again and actually use it.
Copy and paste from the kernel's macros for consistency.
Link: https://training.mips.com/basic_mips/PDF/Memory_Map.pdf
Fixes: c31319b66 ("ramips: lzma-loader: Refactor loader")
Reported-by: Lech Perczak <lech.perczak@gmail.com>
Signed-off-by: Michael Pratt <mcpratt@pm.me>
(cherry picked from commit 4c1e9bd858)
Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
* Delete unused lantiq makefile
* Delete redundant makefiles and unify them into the main makefile
* Refactor and unify board code into a single file
* Add support and review subtarget specific board support
Signed-off-by: Antonio Vázquez <antoniovazquezblanco@gmail.com>
FLASH_START is supposed to point at the memory area where NOR flash are
mapped. We currently have an incorrect FLASH_START copied from ar71xx
back then and the loader doesn't work under OKLI mode.
On ramips, mt7621 has it's flash mapped to 0x1fc00000 and other SoCs
uses 0x1c000000. This commit makes FLASH_START a configurable value to
handle both cases.
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
According to the calling convention of the o32 ABI the
caller function must reserve stack space for $a0-$a3
registers in case the callee needs to save its arguments.
The assembly startup code does not reserve stack space
for these registers thus when the main C function needs
to save its arguments, that will cause a stack overflow.
Fix the assembly code to reserve stack space for the
registers to avoid that.
Untested. It seems that the lzma-loader is not used at all?
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 49062
