Skip to contentSkip to conversation

UNCACHED_SDRAM macro issue

U-Boot-Users Luigi 'Comio' Mantellini 3 participants 9 messages Feb 13, 2008 – Feb 28, 2008

lore
Share & tools
lore
#0 lore raw

Hi list,

I'm working on the porting of u-boot (1.3.1 from git) to a mips based SoC. The port seems to work fine unless the UNCACHED_SDRAM macro (used by do_bootm_linux).

Using the version from commit de9a738faa7c2f47286119c3bfebc3dfbfe7d86d (that maps the addresses on KSEG1 instead on physical addresses) the linux kernel is not able to start, while changing the macro to a PHYSADDR(a) (how was before the commit) the linux kernel starts without any problem.

In order to make my system bootable I changed the macro UNCACHED_SDRAM from:

#define UNCACHED_SDRAM(a) KSEG1ADDR(a)

to

#define UNCACHED_SDRAM(a) PHYSADDR(a) (how defined before the commit de9a738...)

What I need to change on my board specific u-boot in order to use the KSEG1 address space?

thanks a lot

luigi

Signature · 8 lines
     ______       Luigi Mantellini
   .'______'.     R&D - Software
  (.'      '.)    Industrie Dial Face S.p.A.
  ( :=----=: )    Via Canzo, 4
  ('.______.')    20068 Peschiera Borromeo (MI), Italy
   '.______.'     Tel.: +39 02 5167 2813
                  Fax:  +39 02 5167 2459
 Ind. Dial Face   Email: luigi.mantellini at idf-hit.com
#1 lore raw
Quoted 13 lines · Luigi 'Comio' Mantellini

I'm working on the porting of u-boot (1.3.1 from git) to a mips based SoC. The port seems to work fine unless the UNCACHED_SDRAM macro (used by do_bootm_linux).

Using the version from commit de9a738faa7c2f47286119c3bfebc3dfbfe7d86d (that maps the addresses on KSEG1 instead on physical addresses) the linux kernel is not able to start, while changing the macro to a PHYSADDR(a) (how was before the commit) the linux kernel starts without any problem.

In order to make my system bootable I changed the macro UNCACHED_SDRAM from:

Quoted 6 lines

#define UNCACHED_SDRAM(a) KSEG1ADDR(a)

to

#define UNCACHED_SDRAM(a) PHYSADDR(a) (how defined before the commit de9a738...)

So it seems Alchemy-based SoC. if so,

Quoted 2 lines

What I need to change on my board specific u-boot in order to use the KSEG1 address space?

what do you mean "in order to use the KSEG1 address space?" I'm not familiar with Au1XXX, but you might need CONFIG_AU1X00 in your board config. Some URLs added[1][2].

Shinya

[1] http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/32316/focus=32372
[2] http://www.linux-mips.org/wiki/Alchemy

#2 lore raw

Hi Shinya,

see inline.

Quoted 1 line · Shinya Kuribayashi

Luigi 'Comio' Mantellini wrote:

...

Quoted 3 lines

So it seems Alchemy-based SoC. if so,

No. I'm working on a SoC that uses a IP core from MIPS (mips4kec). The mips core should be (formally) a standard mips core.

Quoted 7 lines
Quoted 2 lines

What I need to change on my board specific u-boot in order to use the KSEG1 address space?

what do you mean "in order to use the KSEG1 address space?" I'm not familiar with Au1XXX, but you might need CONFIG_AU1X00 in your board config. Some URLs added[1][2].

The CONFIG_AU1X00 doesn't work for me. from datasheet there aren't differences between my SoC and the standalone mips4kec. This is the memory map how depicted by the datasheet:

0x00000000--0x7fffffff KUser - Cached / Mapped 0x80000000--0x9fffffff KSEG0 - Cached / Unmapped 0xA0000000--0x9fffffff KSEG1 - Uncached / Unmapped
....

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

Thanks a lot,

luigi

Quoted 4 lines

Shinya

[1] http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/32316/focus=32372
[2] http://www.linux-mips.org/wiki/Alchemy

Signature · 8 lines
     ______       Luigi Mantellini
   .'______'.     R&D - Software
  (.'      '.)    Industrie Dial Face S.p.A.
  ( :=----=: )    Via Canzo, 4
  ('.______.')    20068 Peschiera Borromeo (MI), Italy
   '.______.'     Tel.: +39 02 5167 2813
                  Fax:  +39 02 5167 2459
 Ind. Dial Face   Email: luigi.mantellini at idf-hit.com
#3 lore raw
Quoted 13 lines · Luigi 'Comio' Mantellini
Quoted 2 lines

So it seems Alchemy-based SoC. if so,

No. I'm working on a SoC that uses a IP core from MIPS (mips4kec). The mips core should be (formally) a standard mips core.

Quoted 6 lines
Quoted 2 lines

What I need to change on my board specific u-boot in order to use the KSEG1 address space?

what do you mean "in order to use the KSEG1 address space?" I'm not familiar with Au1XXX, but you might need CONFIG_AU1X00 in your board config. Some URLs added[1][2].

Quoted 8 lines

The CONFIG_AU1X00 doesn't work for me. from datasheet there aren't differences between my SoC and the standalone mips4kec. This is the memory map how depicted by the datasheet:

0x00000000--0x7fffffff KUser - Cached / Mapped 0x80000000--0x9fffffff KSEG0 - Cached / Unmapped 0xA0000000--0x9fffffff KSEG1 - Uncached / Unmapped
....

Then I'm confused. Physical address map of your SoC will be helpful. Where is the SDRAM mapped to?

Quoted 5 lines

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

IMHO it's not related to cache.

Shinya

#4 lore raw
Quoted 7 lines · Shinya Kuribayashi
Quoted 5 lines

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

IMHO it's not related to cache.

How do you set ERL and EXL bits? Please try to clear them at the STATUS register initialization like:

reset:

        <snip>

        /* STATUS register */
        mfc0    k0, CP0_STATUS
-       li      k1, ~ST0_IE
+       li      k1, ~(ST0_ERL | ST0_EXL | ST0_IE)
        and     k0, k1
        mtc0    k0, CP0_STATUS

ERL and EXL disable exceptions. Due to this spec, we are in danger of overlooking something critical. If this change brings in new exception(s), please fix the causes of them first. Hope this helps.

Shinya

#5 lore raw
Quoted 21 lines · Shinya Kuribayashi
Quoted 6 lines · Shinya Kuribayashi
Quoted 5 lines

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

IMHO it's not related to cache.

How do you set ERL and EXL bits? Please try to clear them at the STATUS register initialization like:

reset:

        <snip>

        /* STATUS register */
        mfc0    k0, CP0_STATUS
(quoted diff) +1 -1
-       li      k1, ~ST0_IE
+       li      k1, ~(ST0_ERL | ST0_EXL | ST0_IE)
        and     k0, k1
        mtc0    k0, CP0_STATUS
Quoted 3 lines

ERL and EXL disable exceptions. Due to this spec, we are in danger of overlooking something critical. If this change brings in new exception(s), please fix the causes of them first. Hope this helps.

Err, sorry for confusing example. Here's the right one:

cpu/mips/start.S +8 -2
index c92b162..02797f7 100644
@@ -211,14 +211,20 @@ reset:
 	mtc0	zero, CP0_WATCHLO
 	mtc0	zero, CP0_WATCHHI
 
+	/* Inhibit deffered WATCH exception */
+	mfc0	k0, CP0_CAUSE
+	li	k1, ~(1UL << 22)	# CP0.Cause.WP = 0
+	and     k0, k0, k1
+	mtc0	k0, CP0_CAUSE
+
 	/* STATUS register */
 #ifdef  CONFIG_TB0229
 	li	k0, ST0_CU0
 #else
 	mfc0	k0, CP0_STATUS
 #endif
-	li	k1, ~ST0_IE
-	and	k0, k1
+	ori	k0, (ST0_ERL | ST0_EXL | ST0_IE)
+	xori	k0, (ST0_ERL | ST0_EXL)
 	mtc0	k0, CP0_STATUS
 
 	/* CAUSE register */

Again, it's highy recommended to make sure U-Boot works fine under interrupts enabled (CP0.Status.IE=1), before digging into UNCACHED_ SDRAM problem.

Shinya

#6 lore raw

Hi Shinya,

Thanks for your support and your code patches.
I'm waiting for a Lauterbach trace32 to debug deeply the code and understand better the behaviour.
I'll send you my feedback next week.

Best regards,

luigi

Quoted 56 lines · Shinya Kuribayashi
Quoted 21 lines · Shinya Kuribayashi
Quoted 6 lines · Shinya Kuribayashi
Quoted 5 lines

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

IMHO it's not related to cache.

How do you set ERL and EXL bits? Please try to clear them at the STATUS register initialization like:

reset:

        <snip>

        /* STATUS register */
        mfc0    k0, CP0_STATUS
cpu/mips/start.S +1 -1
-       li      k1, ~ST0_IE
+       li      k1, ~(ST0_ERL | ST0_EXL | ST0_IE)
        and     k0, k1
        mtc0    k0, CP0_STATUS
Quoted 3 lines

ERL and EXL disable exceptions. Due to this spec, we are in danger of overlooking something critical. If this change brings in new exception(s), please fix the causes of them first. Hope this helps.

Err, sorry for confusing example. Here's the right one:

cpu/mips/start.S +8 -2
index c92b162..02797f7 100644
@@ -211,14 +211,20 @@ reset:
 	mtc0	zero, CP0_WATCHLO
 	mtc0	zero, CP0_WATCHHI
 
+	/* Inhibit deffered WATCH exception */
+	mfc0	k0, CP0_CAUSE
+	li	k1, ~(1UL << 22)	# CP0.Cause.WP = 0
+	and     k0, k0, k1
+	mtc0	k0, CP0_CAUSE
+
 	/* STATUS register */
 #ifdef  CONFIG_TB0229
 	li	k0, ST0_CU0
 #else
 	mfc0	k0, CP0_STATUS
 #endif
-	li	k1, ~ST0_IE
-	and	k0, k1
+	ori	k0, (ST0_ERL | ST0_EXL | ST0_IE)
+	xori	k0, (ST0_ERL | ST0_EXL)
 	mtc0	k0, CP0_STATUS
 
 	/* CAUSE register */
Quoted 6 lines

Again, it's highy recommended to make sure U-Boot works fine under interrupts enabled (CP0.Status.IE=1), before digging into UNCACHED_ SDRAM problem.

Shinya

Industrie Dial Face S.p.A.
Luigi Mantellini
R&D - Software
Industrie Dial Face S.p.A.
Via Canzo, 4
20068 Peschiera Borromeo (MI), Italy

Tel.:   +39 02 5167 2813
Fax:    +39 02 5167 2459
E-mail: luigi.mantellini at idf-hit.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.denx.de/pipermail/u-boot/attachments/20080216/0165b4f9/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: idf_logo.gif
Type: image/gif
Size: 4122 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080216/0165b4f9/attachment.gif

#7 lore raw
In reply to Shinya Kuribayashi

Hi Shinya,

I found the error... but it's not an u-boot issue. I'm using a (old) customized linux 2.6.10 (from my vendor) that also adds the 0xa0000000 offset to the pointers passed by u-boot, assuming that address is expressed in termos of physical address. I corrected the linux code masking and (after) addind the segment.

Thanks a lot for your help.

luigi

Quoted 56 lines · Shinya Kuribayashi
Quoted 21 lines · Shinya Kuribayashi
Quoted 6 lines · Shinya Kuribayashi
Quoted 5 lines

I think that on my application the UNCACHED_SDRAM should map the address on KSEG1 (how it is now) but this simply doesn't work. Instead, using the PHYSADDR(a) macro... the kernel is able to start.

I suspect that there are issues on cache management. Can be?

IMHO it's not related to cache.

How do you set ERL and EXL bits? Please try to clear them at the STATUS register initialization like:

reset:

        <snip>

        /* STATUS register */
        mfc0    k0, CP0_STATUS
cpu/mips/start.S +1 -1
-       li      k1, ~ST0_IE
+       li      k1, ~(ST0_ERL | ST0_EXL | ST0_IE)
        and     k0, k1
        mtc0    k0, CP0_STATUS
Quoted 3 lines

ERL and EXL disable exceptions. Due to this spec, we are in danger of overlooking something critical. If this change brings in new exception(s), please fix the causes of them first. Hope this helps.

Err, sorry for confusing example. Here's the right one:

cpu/mips/start.S +8 -2
index c92b162..02797f7 100644
@@ -211,14 +211,20 @@ reset:
 	mtc0	zero, CP0_WATCHLO
 	mtc0	zero, CP0_WATCHHI
 
+	/* Inhibit deffered WATCH exception */
+	mfc0	k0, CP0_CAUSE
+	li	k1, ~(1UL << 22)	# CP0.Cause.WP = 0
+	and     k0, k0, k1
+	mtc0	k0, CP0_CAUSE
+
 	/* STATUS register */
 #ifdef  CONFIG_TB0229
 	li	k0, ST0_CU0
 #else
 	mfc0	k0, CP0_STATUS
 #endif
-	li	k1, ~ST0_IE
-	and	k0, k1
+	ori	k0, (ST0_ERL | ST0_EXL | ST0_IE)
+	xori	k0, (ST0_ERL | ST0_EXL)
 	mtc0	k0, CP0_STATUS
 
 	/* CAUSE register */
Quoted 6 lines

Again, it's highy recommended to make sure U-Boot works fine under interrupts enabled (CP0.Status.IE=1), before digging into UNCACHED_ SDRAM problem.

Shinya

Industrie Dial Face S.p.A.
Luigi Mantellini
R&D - Software
Industrie Dial Face S.p.A.
Via Canzo, 4
20068 Peschiera Borromeo (MI), Italy

Tel.:   +39 02 5167 2813
Fax:    +39 02 5167 2459
E-mail: luigi.mantellini at idf-hit.com
GPG fingerprint: 3DD1 7B71 FBDF 6376
1B4A
                 B003 175F E979 907E

1650
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.denx.de/pipermail/u-boot/attachments/20080228/0b332f66/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: idf_logo.gif
Type: image/gif
Size: 4122 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080228/0b332f66/attachment.gif

#8 lore raw

Hi Luigi,

Quoted 7 lines · Luigi 'Comio' Mantellini

Hi Shinya,

I found the error... but it's not an u-boot issue. I'm using a (old) customized linux 2.6.10 (from my vendor) that also adds the 0xa0000000 offset to the pointers passed by u-boot, assuming that address is expressed in termos of physical address. I corrected the linux code masking and (after) addind the segment.

I appricate your letting me know about that.

Thanks,

Shinya