This code gets us to a working linux boot on the alix1c. I have not tested
Ethernet yet. The fixes are a board-specific fake spd_read_byte, cleaning up
comments, and just in general customizing for the 1c.
The lxraminit
change fixes a bug (&& used instead of ||), adds some debug prints which were
VERY useful debugging the alix1c, changes fatal error messages from print_debug
to print_emerg, and adds two functions:
banner, which just prints out a string with a banner, and
hcf, which print an emergency message and then pushes null bytes
into the uart forever, just to make sure that no bytes get lost
for any reason.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@2899 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
diff --git a/src/mainboard/pcengines/alix1c/cache_as_ram_auto.c b/src/mainboard/pcengines/alix1c/cache_as_ram_auto.c
index 16aefc0..087502f 100644
--- a/src/mainboard/pcengines/alix1c/cache_as_ram_auto.c
+++ b/src/mainboard/pcengines/alix1c/cache_as_ram_auto.c
@@ -20,6 +20,7 @@
#define ASSEMBLY 1
#include <stdint.h>
+#include <spd.h>
#include <device/pci_def.h>
#include <arch/io.h>
#include <device/pnp_def.h>
@@ -37,13 +38,70 @@
#define POST_CODE(x) outb(x, 0x80)
#define SERIAL_DEV PNP_DEV(0x2e, W83627HF_SP1)
-#include "southbridge/amd/cs5536/cs5536_early_smbus.c"
+/* The alix1c has no SMBUS; the setup is hard-wired. */
+void cs5536_enable_smbus(void)
+{
+}
+
#include "southbridge/amd/cs5536/cs5536_early_setup.c"
#include "superio/winbond/w83627hf/w83627hf_early_serial.c"
-static inline int spd_read_byte(unsigned device, unsigned address)
+/* the part is a hynix hy5du121622ctp-d43
+ * HY 5D U 12 16 2 2 C <blank> T <blank> P D43
+ * Hynix
+ * DDR SDRAM (5D)
+ * VDD 2.5 VDDQ 2.5 (U)
+ * 512M 8K REFRESH (12)
+ * x16 (16)
+ * 4banks (2)
+ * SSTL_2 (2)
+ * 4th GEN die (C)
+ * Normal Power Consumption (<blank> )
+ * TSOP (T)
+ * Single Die (<blank>)
+ * Lead Free (P)
+ * DDR400 3-3-3 (D43)
+ */
+/* spd array */
+static u8 spdbytes[] = {
+ [SPD_ACCEPTABLE_CAS_LATENCIES] = 0x10,
+ [SPD_BANK_DENSITY] = 0x40,
+ [SPD_DEVICE_ATTRIBUTES_GENERAL] = 0xff,
+ [SPD_MEMORY_TYPE] = 7,
+ [SPD_MIN_CYCLE_TIME_AT_CAS_MAX] = 10, /* This is a guess for tRAC value */
+ [SPD_MODULE_ATTRIBUTES] = 0xff, /* fix me later when we figure out */
+ [SPD_NUM_BANKS_PER_SDRAM] = 4,
+ [SPD_PRIMARY_SDRAM_WIDTH] = 8,
+ /* alix1c is 1 bank. */
+ [SPD_NUM_DIMM_BANKS] = 1,
+ [SPD_NUM_COLUMNS] = 0xa,
+ [SPD_NUM_ROWS] = 3,
+ [SPD_REFRESH] = 0x3a,
+ [SPD_SDRAM_CYCLE_TIME_2ND] = 60,
+ [SPD_SDRAM_CYCLE_TIME_3RD] = 75,
+ [SPD_tRAS] = 40,
+ [SPD_tRCD] = 15,
+ [SPD_tRFC] = 70,
+ [SPD_tRP] = 15,
+ [SPD_tRRD] = 10,
+};
+
+static u8 spd_read_byte(unsigned device, unsigned address)
{
- return smbus_read_byte(device, address);
+ print_debug("spd_read_byte dev ");
+ print_debug_hex8(device);
+
+ if (device != (0x50<<1)){
+ print_debug(" returns 0xff\n");
+ return 0xff;
+ }
+
+ print_debug(" addr ");
+ print_debug_hex8(address);
+ print_debug(" returns ");
+ print_debug_hex8(spdbytes[address]);
+ print_debug("\r\n");
+ return spdbytes[address];
}
#define ManualConf 0 /* Do automatic strapped PLL config */
@@ -83,13 +141,12 @@
void cache_as_ram_main(void)
{
+ static const struct mem_controller memctrl[] = {
+ {.channel0 = {0x50}},
+ };
extern void RestartCAR();
POST_CODE(0x01);
- static const struct mem_controller memctrl [] = {
- {.channel0 = {(0xa<<3)|0, (0xa<<3)|1}}
- };
-
SystemPreInit();
msr_init();
@@ -114,24 +171,34 @@
/* Check memory */
ram_check(0x00000000, 640 * 1024);
- /* Switch from Cache as RAM to real RAM */
- /* There are two ways we could think about this.
- 1. If we are using the auto.inc ROMCC way, the stack is going to be re-setup in the code following this code.
- Just wbinvd the stack to clear the cache tags. We don't care where the stack used to be.
- 2. This file is built as a normal .c -> .o and linked in etc. The stack might be used to return etc.
- That means we care about what is in the stack. If we are smart we set the CAR stack to the same location
- as the rest of LinuxBIOS. If that is the case we can just do a wbinvd. The stack will be written into real
- RAM that is now setup and we continue like nothing happened. If the stack is located somewhere other than
- where LB would like it, you need to write some code to do a copy from cache to RAM
-
- We use method 1 on Norwich and on this board too.
- */
+ /* Switch from Cache as RAM to real RAM
+ * There are two ways we could think about this.
+ *
+ * 1. If we are using the auto.inc ROMCC way, the stack is
+ * going to be re-setup in the code following this code. Just
+ * wbinvd the stack to clear the cache tags. We don't care
+ * where the stack used to be.
+ *
+ * 2. This file is built as a normal .c -> .o and linked in
+ * etc. The stack might be used to return etc. That means we
+ * care about what is in the stack. If we are smart we set
+ * the CAR stack to the same location as the rest of
+ * LinuxBIOS. If that is the case we can just do a wbinvd.
+ * The stack will be written into real RAM that is now setup
+ * and we continue like nothing happened. If the stack is
+ * located somewhere other than where LB would like it, you
+ * need to write some code to do a copy from cache to RAM
+ *
+ * We use method 1 on Norwich and on this board too.
+ */
POST_CODE(0x02);
print_err("POST 02\n");
__asm__("wbinvd\n");
print_err("Past wbinvd\n");
- /* we are finding the return does not work on this board. Explicitly call the label that is
- * after the call to us. This is gross, but sometimes at this level it is the only way out
+ /* we are finding the return does not work on this
+ * board. Explicitly call the label that is after the call to
+ * us. This is gross, but sometimes at this level it is the
+ * only way out
*/
done_cache_as_ram_main();
}