[U-Boot] [PATCH 31/31] 83xx, kmeter1: added bootcount feature
add CONFIG_BOUTCOUNT_LIMIT feature for 83xx CPU. activate this feature for the kmeter1 board.
Signed-off-by: Heiko Schocher hs@denx.de --- cpu/mpc83xx/cpu.c | 27 +++++++++++++++++++++++++++ include/configs/keymile-common.h | 2 -- 2 files changed, 27 insertions(+), 2 deletions(-)
diff --git a/cpu/mpc83xx/cpu.c b/cpu/mpc83xx/cpu.c index 587fca3..7c12672 100644 --- a/cpu/mpc83xx/cpu.c +++ b/cpu/mpc83xx/cpu.c @@ -34,6 +34,9 @@ #include <libfdt.h> #include <tsec.h> #include <netdev.h> +#ifdef CONFIG_BOOTCOUNT_LIMIT +#include <asm/io.h> +#endif
DECLARE_GLOBAL_DATA_PTR;
@@ -385,3 +388,27 @@ int cpu_eth_init(bd_t *bis) #endif return 0; } + +#ifdef CONFIG_BOOTCOUNT_LIMIT + +#if !defined(CONFIG_BOOTCOUNT_ADDR) +#define CONFIG_BOOTCOUNT_ADDR 0x11bff8 +#endif + +void bootcount_store (ulong a) +{ + void *reg = (unsigned __iomem *)(CONFIG_SYS_IMMR + CONFIG_BOOTCOUNT_ADDR); + out_be32 (reg, a); + out_be32 (reg + 4, BOOTCOUNT_MAGIC); +} + +ulong bootcount_load (void) +{ + void *reg = (unsigned __iomem *)(CONFIG_SYS_IMMR + CONFIG_BOOTCOUNT_ADDR); + + if (in_be32 (reg + 4) != BOOTCOUNT_MAGIC) + return 0; + else + return in_be32 (reg); +} +#endif /* CONFIG_BOOTCOUNT_LIMIT */ diff --git a/include/configs/keymile-common.h b/include/configs/keymile-common.h index 3736cd8..6a25262 100644 --- a/include/configs/keymile-common.h +++ b/include/configs/keymile-common.h @@ -27,9 +27,7 @@ /* Do boardspecific init for all boards */ #define CONFIG_BOARD_EARLY_INIT_R 1
-#if defined(CONFIG_MGCOGE) || defined(CONFIG_MGSUVD) || defined(CONFIG_KMSUPX4) #define CONFIG_BOOTCOUNT_LIMIT -#endif
/* * Command line configuration.
On Wed, 28 Jan 2009 10:41:02 +0100 Heiko Schocher hs@denx.de wrote:
diff --git a/cpu/mpc83xx/cpu.c b/cpu/mpc83xx/cpu.c
+#ifdef CONFIG_BOOTCOUNT_LIMIT
+#if !defined(CONFIG_BOOTCOUNT_ADDR) +#define CONFIG_BOOTCOUNT_ADDR 0x11bff8
it's CONFIG_SYS and magic number, but now that I've found it, sorry, I can't accept this - not all 83xx have a QE and thus this MURAM. Even if they did this would require changing the device tree muram node property, to indicate its size has shrunk 8 bytes, right? In fact, not all QE's have 48Kbytes of MURAM either - the 8323 only has 16Kbytes. Problem is, I don't know of a better place to put the bootcount. Ideas?
Kim
Hello Kim,
Kim Phillips wrote:
On Wed, 28 Jan 2009 10:41:02 +0100 Heiko Schocher hs@denx.de wrote:
diff --git a/cpu/mpc83xx/cpu.c b/cpu/mpc83xx/cpu.c
+#ifdef CONFIG_BOOTCOUNT_LIMIT
+#if !defined(CONFIG_BOOTCOUNT_ADDR) +#define CONFIG_BOOTCOUNT_ADDR 0x11bff8
it's CONFIG_SYS and magic number, but now that I've found it, sorry, I can't accept this - not all 83xx have a QE and thus this MURAM. Even
Hmm.. maybe we make this dependent on CONFIG_QE ?
if they did this would require changing the device tree muram node property, to indicate its size has shrunk 8 bytes, right? In fact, not
Yes, we should do this.
all QE's have 48Kbytes of MURAM either - the 8323 only has 16Kbytes.
Thats why you could define it with CONFIG_BOOTCOUNT_ADDR, where exactly this 8 bytes are.
Problem is, I don't know of a better place to put the bootcount. Ideas?
I thought to make it as on 82xx (using parameter Ram of SCC1), because UCC5 should be compatible to the SCC1 see 8360ERM.pdf Table 19-11 on page 19-20. But on the 8360, after reset, the complete parameter RAM is initialized with 0 ... so that didn't work. Other places I couldn't found for this feature :-(
bye Heiko
On Thu, 29 Jan 2009 10:22:06 +0100 Heiko Schocher hs@denx.de wrote:
Hello Kim,
Kim Phillips wrote:
On Wed, 28 Jan 2009 10:41:02 +0100 Heiko Schocher hs@denx.de wrote:
diff --git a/cpu/mpc83xx/cpu.c b/cpu/mpc83xx/cpu.c
+#ifdef CONFIG_BOOTCOUNT_LIMIT
+#if !defined(CONFIG_BOOTCOUNT_ADDR) +#define CONFIG_BOOTCOUNT_ADDR 0x11bff8
it's CONFIG_SYS and magic number, but now that I've found it, sorry, I can't accept this - not all 83xx have a QE and thus this MURAM. Even
Hmm.. maybe we make this dependent on CONFIG_QE ?
still, the 8323 and 8360 would need different address values...so not only CONFIG_QE.
if they did this would require changing the device tree muram node property, to indicate its size has shrunk 8 bytes, right? In fact, not
Yes, we should do this.
I hope the QE drivers don't significantly drop their performance given a muram size is not a multiple of a power of 2.
all QE's have 48Kbytes of MURAM either - the 8323 only has 16Kbytes.
Thats why you could define it with CONFIG_BOOTCOUNT_ADDR, where exactly this 8 bytes are.
it would be nice if all 83xx had the same address though.
Problem is, I don't know of a better place to put the bootcount. Ideas?
I thought to make it as on 82xx (using parameter Ram of SCC1), because UCC5 should be compatible to the SCC1 see 8360ERM.pdf Table 19-11 on page 19-20. But on the 8360, after reset, the complete parameter RAM is initialized with 0 ... so that didn't work. Other places I couldn't found for this feature :-(
you thinking is right, I just can't think of another place to put it - and I suspect this isn't something that can be put in memory that lies outside the IMMR.
if you really want this, make it only for the QE based chips, and include the device tree modification. I still don't know if this should be enabled by default in the configs, however - debugging QE performance down to this change would be relatively hard to find.
Kim
Dear Kim Phillips,
In message 20090129102814.6e267b96.kim.phillips@freescale.com you wrote:
you thinking is right, I just can't think of another place to put it - and I suspect this isn't something that can be put in memory that lies outside the IMMR.
We could use any registers that are guaranteed to maintaion their current values through a core reset. Documentation usually gives little or no guarantees, though.
Best regards,
Wolfgang Denk
participants (3)
-
Heiko Schocher -
Kim Phillips -
Wolfgang Denk