Skip to contentSkip to conversation

ARM: socfpga: Put stack at the end of SRAM

PATCH New Marek Vasut 3 participants 11 messages May 12, 2018 – May 15, 2018

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>
---
 include/configs/socfpga_common.h | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +1 -3
index 4de2aa7929..cb67d539b1 100644
@@ -35,10 +35,8 @@
 #define CONFIG_SYS_INIT_RAM_ADDR	0xFFE00000
 #define CONFIG_SYS_INIT_RAM_SIZE	0x40000 /* 256KB */
 #endif
-#define CONFIG_SYS_INIT_SP_OFFSET		\
-	(CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
 #define CONFIG_SYS_INIT_SP_ADDR			\
-	(CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+	(CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)
 
 #define CONFIG_SYS_SDRAM_BASE		PHYS_SDRAM_1
 
Signature · 1 line
2.16.2
#1 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 16 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>
---
  include/configs/socfpga_common.h | 4 +---
  1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
  #define CONFIG_SYS_INIT_RAM_ADDR	0xFFE00000
  #define CONFIG_SYS_INIT_RAM_SIZE	0x40000 /* 256KB */
  #endif
-#define CONFIG_SYS_INIT_SP_OFFSET		\
-	(CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
  #define CONFIG_SYS_INIT_SP_ADDR			\
-	(CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+	(CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

I'd really like to keep some room here for bootcounter and some related things I store here.

Simon

Quoted 4 lines
include/configs/socfpga_common.h +0 -0
  #define CONFIG_SYS_SDRAM_BASE		PHYS_SDRAM_1
 
 
#2 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
  include/configs/socfpga_common.h | 4 +---

  1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
   #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
   #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
   #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
   #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 2 lines

I'd really like to keep some room here for bootcounter and some related things I store here.

So if I understand it correctly, before this patch, the boards overwrote the global data ?

Anyway, if you need some extra space, look at board_init_f_alloc_reserve() .

Signature · 2 lines
Best regards,
Marek Vasut
#3 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 32 lines · Marek Vasut
Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>
---
   include/configs/socfpga_common.h | 4 +---
   1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
   #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
   #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
   #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
   #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 5 lines
Quoted 2 lines

I'd really like to keep some room here for bootcounter and some related things I store here.

So if I understand it correctly, before this patch, the boards overwrote the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

Chances are high that these bytes just fit into the 16-byte alignment allocated by GENERATED_GBL_DATA_SIZE (and on my board it *is* like that).

Quoted 3 lines

Anyway, if you need some extra space, look at board_init_f_alloc_reserve() .

OK, but that function does not seem to be overridable? How can I reserve my own memory here?

And reading this function, I guess your commit message is a bit misleading. It seems that the purpose of CONFIG_SYS_INIT_SP_OFFSET like it was was to reserve memory for the initial gd (before relocation), which is now done in board_init_f_alloc_reserve().

Simon

#4 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 38 lines · Simon Goldschmidt
Quoted 35 lines · Marek Vasut
Quoted 33 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
   include/configs/socfpga_common.h | 4 +---

   1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
    #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
    #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
    #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
    #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

I'd really like to keep some room here for bootcounter and some related things I store here.

Quoted 7 lines
Quoted 2 lines

So if I understand it correctly, before this patch, the boards overwrote the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

+CC Stefan.

Quoted 10 lines

Chances are high that these bytes just fit into the 16-byte alignment allocated by GENERATED_GBL_DATA_SIZE (and on my board it *is* like that).

Quoted 4 lines

Anyway, if you need some extra space, look at
board_init_f_alloc_reserve() .

OK, but that function does not seem to be overridable? How can I reserve my own memory here?

Add a __weak there I guess ?

Quoted 4 lines

And reading this function, I guess your commit message is a bit misleading. It seems that the purpose of CONFIG_SYS_INIT_SP_OFFSET like it was was to reserve memory for the initial gd (before relocation), which is now done in board_init_f_alloc_reserve().

The code is confusing as hell, I give you that.

Signature · 2 lines
Best regards,
Marek Vasut
#5 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 42 lines · Marek Vasut
Quoted 41 lines · Simon Goldschmidt
Quoted 37 lines · Marek Vasut
Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>
---
    include/configs/socfpga_common.h | 4 +---
    1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
    #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
    #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
    #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
    #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 2 lines

I'd really like to keep some room here for bootcounter and some related things I store here.

So if I understand it correctly, before this patch, the boards overwrote the global data ?

Quoted 6 lines
Quoted 4 lines

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

Quoted 11 lines
Quoted 10 lines

Chances are high that these bytes just fit into the 16-byte alignment allocated by GENERATED_GBL_DATA_SIZE (and on my board it *is* like that).

Quoted 4 lines

Anyway, if you need some extra space, look at
board_init_f_alloc_reserve() .

OK, but that function does not seem to be overridable? How can I reserve my own memory here?

Quoted 8 lines

Add a __weak there I guess ?

Quoted 4 lines

And reading this function, I guess your commit message is a bit misleading. It seems that the purpose of CONFIG_SYS_INIT_SP_OFFSET like it was was to reserve memory for the initial gd (before relocation), which is now done in board_init_f_alloc_reserve().

The code is confusing as hell, I give you that.

Thanks,
Stefan

#6 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 50 lines · Stefan Roese
Quoted 49 lines · Marek Vasut
Quoted 47 lines · Simon Goldschmidt
Quoted 32 lines · Marek Vasut
Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
    include/configs/socfpga_common.h | 4 +---

    1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
     #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
     #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
     #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
     #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 7 lines
Quoted 3 lines

I'd really like to keep some room here for bootcounter and some related
things I store here.

So if I understand it correctly, before this patch, the boards overwrote
the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

Quoted 7 lines
Quoted 1 line

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

I just checked out ae996c80 and from what I can tell, the RAM layout was the same at that time: include/generated/generated-asm-offsets.h contains:
- #define GENERATED_GBL_DATA_SIZE 208
- #define GD_SIZE 200

That leaves us with an 8 byte alignment gap at the end of the SRAM and that's where the boot counter is located. So the global data is not overwritten, but this is more or less by chance, any change to the struct definition might change this.

Looking at just another example: the x600 board defines its SRAM_SIZE 8 bytes smaller than what it really is. That seems like a good idea for socfpga, too. I'll see if I can provide a patch.

Slightly off-topic: I have a constellation where the boot-counter is not reset although the power has been pulled. This device has big capacitors and the SRAM content seems to survive although the processor is reset. To me it seems as if looking at the boot reason might be safer than checking for a magic word... Any ideas?

Thanks,
Simon

Quoted 25 lines
Quoted 13 lines
Quoted 6 lines

Chances are high that these bytes just fit into the 16-byte alignment allocated by GENERATED_GBL_DATA_SIZE (and on my board it *is* like that).

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

Anyway, if you need some extra space, look at
board_init_f_alloc_reserve() .

OK, but that function does not seem to be overridable? How can I reserve my own memory here?

Quoted 8 lines

Add a __weak there I guess ?

Quoted 4 lines

And reading this function, I guess your commit message is a bit misleading. It seems that the purpose of CONFIG_SYS_INIT_SP_OFFSET like it was was to reserve memory for the initial gd (before relocation), which is now done in board_init_f_alloc_reserve().

The code is confusing as hell, I give you that.

Thanks,
Stefan

#7 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 21 lines · Simon Goldschmidt
Quoted 18 lines · Stefan Roese
Quoted 17 lines · Marek Vasut
Quoted 16 lines · Simon Goldschmidt
Quoted 13 lines · Marek Vasut
Quoted 12 lines · Simon Goldschmidt
Quoted 11 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
    include/configs/socfpga_common.h | 4 +---

    1 file changed, 1 insertion(+), 3 deletions(-)
Quoted 14 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644
@@ -35,10 +35,8 @@
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
     #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
     #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
     #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
     #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

Anyway, after reading up on this in the code and the resulting binaries, I do think this patch is correct, so:

Sorry for the confusion, but this really helped to clarify my understanding of the initial memory allocation ;-)

Thanks,
Simon

Quoted 79 lines
Quoted 18 lines
Quoted 18 lines
Quoted 17 lines
Quoted 5 lines
Quoted 5 lines

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 7 lines
Quoted 3 lines

I'd really like to keep some room here for bootcounter and some related
things I store here.

So if I understand it correctly, before this patch, the boards overwrote
the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

Quoted 7 lines
Quoted 1 line

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

I just checked out ae996c80 and from what I can tell, the RAM layout was the same at that time: include/generated/generated-asm-offsets.h contains:
- #define GENERATED_GBL_DATA_SIZE 208
- #define GD_SIZE 200

That leaves us with an 8 byte alignment gap at the end of the SRAM and that's where the boot counter is located. So the global data is not overwritten, but this is more or less by chance, any change to the struct definition might change this.

Looking at just another example: the x600 board defines its SRAM_SIZE 8 bytes smaller than what it really is. That seems like a good idea for socfpga, too. I'll see if I can provide a patch.

Slightly off-topic: I have a constellation where the boot-counter is not reset although the power has been pulled. This device has big capacitors and the SRAM content seems to survive although the processor is reset. To me it seems as if looking at the boot reason might be safer than checking for a magic word... Any ideas?

Thanks,
Simon

Quoted 21 lines
Quoted 21 lines
Quoted 6 lines

Chances are high that these bytes just fit into the 16-byte alignment allocated by GENERATED_GBL_DATA_SIZE (and on my board it *is* like that).

Quoted 1 line
Quoted 7 lines
Quoted 3 lines

Anyway, if you need some extra space, look at
board_init_f_alloc_reserve() .

OK, but that function does not seem to be overridable? How can I reserve
my own memory here?

Add a __weak there I guess ?

Quoted 4 lines

And reading this function, I guess your commit message is a bit misleading. It seems that the purpose of CONFIG_SYS_INIT_SP_OFFSET like it was was to reserve memory for the initial gd (before relocation), which is now done in board_init_f_alloc_reserve().

Quoted 5 lines
Quoted 1 line

The code is confusing as hell, I give you that.

Thanks,
Stefan

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot

#8 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

In reply to Simon Goldschmidt
Quoted 66 lines · Simon Goldschmidt
Quoted 50 lines · Stefan Roese
Quoted 49 lines · Marek Vasut
Quoted 47 lines · Simon Goldschmidt
Quoted 32 lines · Marek Vasut
Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
    include/configs/socfpga_common.h | 4 +---

    1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
     #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
     #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
     #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
     #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 7 lines
Quoted 3 lines

I'd really like to keep some room here for bootcounter and some related
things I store here.

So if I understand it correctly, before this patch, the boards overwrote
the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

Quoted 7 lines
Quoted 1 line

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

I just checked out ae996c80 and from what I can tell, the RAM layout was the same at that time: include/generated/generated-asm-offsets.h contains:
- #define GENERATED_GBL_DATA_SIZE 208
- #define GD_SIZE 200

Quoted 8 lines

That leaves us with an 8 byte alignment gap at the end of the SRAM and that's where the boot counter is located. So the global data is not overwritten, but this is more or less by chance, any change to the struct definition might change this.

Looking at just another example: the x600 board defines its SRAM_SIZE 8 bytes smaller than what it really is. That seems like a good idea for socfpga, too. I'll see if I can provide a patch.

This looks like a hack and a bogus one. Lying about RAM size is never a good idea.

Quoted 5 lines

Slightly off-topic: I have a constellation where the boot-counter is not reset although the power has been pulled. This device has big capacitors and the SRAM content seems to survive although the processor is reset. To me it seems as if looking at the boot reason might be safer than checking for a magic word... Any ideas?

What boot reason options do you get ?

[...]

Signature · 2 lines
Best regards,
Marek Vasut
#9 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 79 lines · Marek Vasut
Quoted 66 lines · Simon Goldschmidt
Quoted 50 lines · Stefan Roese
Quoted 49 lines · Marek Vasut
Quoted 47 lines · Simon Goldschmidt
Quoted 32 lines · Marek Vasut
Quoted 31 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>
---
     include/configs/socfpga_common.h | 4 +---
     1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
     #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
     #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
     #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
     #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 7 lines
Quoted 3 lines

I'd really like to keep some room here for bootcounter and some related
things I store here.

So if I understand it correctly, before this patch, the boards overwrote
the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained by denx, so maybe you could clarify this?

Quoted 7 lines
Quoted 1 line

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

I just checked out ae996c80 and from what I can tell, the RAM layout was the same at that time: include/generated/generated-asm-offsets.h contains:
- #define GENERATED_GBL_DATA_SIZE 208
- #define GD_SIZE 200

Quoted 8 lines

That leaves us with an 8 byte alignment gap at the end of the SRAM and that's where the boot counter is located. So the global data is not overwritten, but this is more or less by chance, any change to the struct definition might change this.

Looking at just another example: the x600 board defines its SRAM_SIZE 8 bytes smaller than what it really is. That seems like a good idea for socfpga, too. I'll see if I can provide a patch.

This looks like a hack and a bogus one. Lying about RAM size is never a good idea.

Quoted 7 lines
Quoted 5 lines

Slightly off-topic: I have a constellation where the boot-counter is not reset although the power has been pulled. This device has big capacitors and the SRAM content seems to survive although the processor is reset. To me it seems as if looking at the boot reason might be safer than checking for a magic word... Any ideas?

What boot reason options do you get ?

The reset manager has a status register [1] that has various bits for different reasons of warm or cold reset. I'm just not sure if it's worth going through the hassle or if I'm alone with that problem of the bootcounter surviving a pull-and-replug.

[1] https://www.altera.com/hps/cyclone-v/topic/sfo1410067763568.html

Simon

#10 lore raw

PATCH 1/1 ARM: socfpga: Put stack at the end of SRAM

Quoted 91 lines · Simon Goldschmidt
Quoted 88 lines · Marek Vasut
Quoted 69 lines · Simon Goldschmidt
Quoted 52 lines · Stefan Roese
Quoted 51 lines · Marek Vasut
Quoted 49 lines · Simon Goldschmidt
Quoted 33 lines · Marek Vasut
Quoted 32 lines · Simon Goldschmidt
Quoted 17 lines · Marek Vasut

The global data are in the .data section, so there's no point in reserving any space for it above stack. Put stack at the end of SRAM.

Cc Chin Liang See <[email protected]>
Cc Dinh Nguyen <[email protected]>

---
     include/configs/socfpga_common.h | 4 +---

     1 file changed, 1 insertion(+), 3 deletions(-)
include/configs/socfpga_common.h +0 -0
index 4de2aa7929..cb67d539b1 100644

include/configs/socfpga_common.h line 39

Quoted 8 lines
include/configs/socfpga_common.h +1 -3
@@ -35,10 +35,8 @@
      #define CONFIG_SYS_INIT_RAM_ADDR    0xFFE00000
      #define CONFIG_SYS_INIT_RAM_SIZE    0x40000 /* 256KB */
      #endif
-#define CONFIG_SYS_INIT_SP_OFFSET        \
-    (CONFIG_SYS_INIT_RAM_SIZE - GENERATED_GBL_DATA_SIZE)
      #define CONFIG_SYS_INIT_SP_ADDR            \
-    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_SP_OFFSET)
+    (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_INIT_RAM_SIZE)

I think this interferes with the bootcount being stored at the end of
SRAM. See the defconfigs of 'is1' and 'sr1500' boards - or my local defconfig here :-)

Quoted 7 lines
Quoted 3 lines

I'd really like to keep some room here for bootcounter and some related
things I store here.

So if I understand it correctly, before this patch, the boards overwrote
the global data ?

Ehrm, reading up on the init code path in the S and C files, it seems like that, yes. I just used the two boards as an example. Are they actively maintained? get_maintainer.pl shows the boards are maintained
by denx, so maybe you could clarify this?

Quoted 7 lines
Quoted 1 line

+CC Stefan.

Thanks. At the time I added the bootcounter support on this platform, it used to work. I might have missed something though. Right now, I don't have access to such a platform, so its hard for me to test something. If you have some patches to fix any potential overlapping memory usages, I'll gladly review them.

I just checked out ae996c80 and from what I can tell, the RAM layout was the same at that time: include/generated/generated-asm-offsets.h contains:
- #define GENERATED_GBL_DATA_SIZE 208
- #define GD_SIZE 200

Quoted 8 lines

That leaves us with an 8 byte alignment gap at the end of the SRAM and that's where the boot counter is located. So the global data is not overwritten, but this is more or less by chance, any change to the struct definition might change this.

Looking at just another example: the x600 board defines its SRAM_SIZE 8 bytes smaller than what it really is. That seems like a good idea for socfpga, too. I'll see if I can provide a patch.

This looks like a hack and a bogus one. Lying about RAM size is never a good idea.

Quoted 5 lines

Slightly off-topic: I have a constellation where the boot-counter is not reset although the power has been pulled. This device has big capacitors and the SRAM content seems to survive although the processor is reset. To me it seems as if looking at the boot reason might be safer than checking for a magic word... Any ideas?

Quoted 8 lines
Quoted 1 line

What boot reason options do you get ?

The reset manager has a status register [1] that has various bits for different reasons of warm or cold reset. I'm just not sure if it's worth going through the hassle or if I'm alone with that problem of the bootcounter surviving a pull-and-replug.

[1] https://www.altera.com/hps/cyclone-v/topic/sfo1410067763568.html

Do you get warm or cold reset indication during plug-and-replug ?

Signature · 2 lines
Best regards,
Marek Vasut