Workaround the Out-of-order Issue of Loongson-2F

PATCH Applied Wu Zhangjin 5 participants 21 messages Mar 13, 2010 – Apr 12, 2010 MIPS

lore

Share & tools

lore

Wu Zhangjin

author

2010-03-13 04:41 UTC #0 lore raw

v3 cover Workaround the Out-of-order Issue of Loongson-2F

From: Wu Zhangjin <[email protected]>

Hi, Ralf

Is that possible to apply this patchset for both 2.6.33(the stable release) and 2.6.34? then we can get a not broken support for loongson-2f with the options(-mfix-loongson2f-nop, -mfix-loongson2f-jump) provided by the latest binutils (>=2.20.1).

Thanks & Regards,
Wu Zhangjin

-------

Changes:

v2->v3:

o Herein, RAS is short for Return Address Stack not Row Address Strobe, at the same time, the "model" in the translation of Chapter 15 should be "mode". (feedback from Shinya Kuribayashi)

v1->v2:

o Cleanup some comments and align some instructions.

----

As the Chapter 15: "Errata: Issue of Out-of-order in loongson"[1] shows, to workaround the Issue of Loongson-2F，We need to do:

o When switching from user mode to kernel mode, you should flush the branch target history such as BTB and RAS.

o Doing some tricks to the indirect branch target to make sure that the indirect branch target can not be in the I/O region.

This patchset applied the above methods and for the binutils patch[3] have been merged into binutils 2.20.1, so, it's time to upstream this patchset now. without this patch, the machines will hang when the instruction sequence hit the Out-of-order Issue of Loongson-2F, therefore, this patchset is very urgent for both 2.6.33 and 2.6.34.

[1] Chinese Version: http://www.loongson.cn/uploadfile/file/20080821113149.pdf
[2] English Version of Chapter 15:
http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source
[3] http://sourceware.org/ml/binutils/2009-11/msg00387.html

Regards,
Wu Zhangjin

Wu Zhangjin (3):
Loongson-2F: Flush the branch target history such as BTB and RAS
Loongson-2F: Enable fixups of binutils 2.20.1
Loongson-2F: Fixup of problems introduced by -mfix-loongson2f-jump of

    binutils 2.20.1

 arch/mips/Makefile                 |    4 +++-
 arch/mips/include/asm/stackframe.h |   19 +++++++++++++++++++
 arch/mips/loongson/common/reset.c  |   11 ++++++++++-
 3 files changed, 32 insertions(+), 2 deletions(-)

Wu Zhangjin

author

2010-03-13 04:45 UTC #1 lore raw

PATCH v3 1/3 Loongson-2F: Flush the branch target history such as BTB and RAS

From: Wu Zhangjin <[email protected]>

As the Chapter 15: "Errata: Issue of Out-of-order in loongson"[1] shows, to workaround the Issue of Loongson-2F，We need to do:

"When switching from user mode to kernel mode, you should flush the branch target history such as BTB and RAS."

This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.

[1] Chinese Version: http://www.loongson.cn/uploadfile/file/200808211
[2] English Version of Chapter 15:
http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source

Signed-off-by Wu Zhangjin <[email protected]>

---
 arch/mips/include/asm/stackframe.h |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)

arch/mips/include/asm/stackframe.h +19 -0

index 3b6da33..c841912 100644
@@ -121,6 +121,25 @@
 		.endm
 #else
 		.macro	get_saved_sp	/* Uniprocessor variation */
+#ifdef CONFIG_CPU_LOONGSON2F
+		/*
+		 * Clear BTB (branch target buffer), forbid RAS (return address
+		 * stack) to workaround the Out-of-order Issue in Loongson2F
+		 * via its diagnostic register.
+		 */
+		move	k0, ra
+		jal	1f
+		 nop
+1:		jal	1f
+		 nop
+1:		jal	1f
+		 nop
+1:		jal	1f
+		 nop
+1:		move	ra, k0
+		li	k0, 3
+		mtc0	k0, $22
+#endif /* CONFIG_CPU_LOONGSON2F */
 #if defined(CONFIG_32BIT) || defined(KBUILD_64BIT_SYM32)
 		lui	k1, %hi(kernelsp)
 #else

Signature · 1 line

1.7.0.1

Ralf Baechle

Applied

2010-03-17 15:11 UTC #2 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 8 lines · Wu Zhangjin

From: Wu Zhangjin <[email protected]>
As the Chapter 15: "Errata: Issue of Out-of-order in loongson"[1] shows, to workaround the Issue of Loongson-2F，We need to do:
"When switching from user mode to kernel mode, you should flush the branch target history such as BTB and RAS."

Quoted 6 lines

This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.
[1] Chinese Version: http://www.loongson.cn/uploadfile/file/200808211
[2] English Version of Chapter 15:
http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source

Thanks, applied.

Ralf

Andreas Barth

2010-04-02 15:04 UTC #3 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

In reply to Wu Zhangjin

Quoted 2 lines · * Wu Zhangjin ([email protected]) [100313 05:45]:

This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.

Unfortunatly the Loongson 2F here still fails with this patch, compiled with the new binutils and both options enabled.

Testcase: plain debian unstable chroot, build binutils in that chroot.

More ideas, codes, whatever welcome.

Andi

Wu Zhangjin

author

2010-04-03 01:50 UTC #4 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 10 lines · Andreas Barth

Quoted 2 lines · * Wu Zhangjin ([email protected]) [100313 05:45]:
This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.
Unfortunatly the Loongson 2F here still fails with this patch, compiled with the new binutils and both options enabled.
Testcase: plain debian unstable chroot, build binutils in that chroot.
More ideas, codes, whatever welcome.

Hi,

This patch is not enough, please use the kernel(>=2.6.32) from
http://dev.lemote.com/code/rt4ls or
http://dev.lemote.com/code/linux-loongson-community

Regards,
Wu Zhangjin

Andreas Barth

2010-04-06 19:10 UTC #5 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 8 lines · * Wu Zhangjin ([email protected]) [100403 03:50]:

Quoted 7 lines · Andreas Barth
Quoted 2 lines · * Wu Zhangjin ([email protected]) [100313 05:45]:
This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.
Unfortunatly the Loongson 2F here still fails with this patch, compiled with the new binutils and both options enabled.

Quoted 8 lines

Quoted 3 lines
Testcase: plain debian unstable chroot, build binutils in that chroot.
More ideas, codes, whatever welcome.
This patch is not enough, please use the kernel(>=2.6.32) from
http://dev.lemote.com/code/rt4ls or
http://dev.lemote.com/code/linux-loongson-community

The kernel vmlinuz-2.6.33-lemote2f-bfs inside of
http://www.anheng.com.cn/loongson/install/loongson2_debian6_20100328.tar.lzma
(linked via linux-loongson-community) fails at the same place:

touch stamp-picdir
if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/regex.c -o pic/regex.o; \
else true; fi
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic ../../libiberty/regex.c -o regex.o if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/cplus-dem.c -o pic/cplus-dem.o; \
else true; fi

Any other kernel I should try?

Andi

Wu Zhangjin

author

2010-04-07 07:38 UTC #6 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

On Tue, 2010-04-06 at 21:10 +0200, Andreas Barth wrote:
[...]

Quoted 5 lines

The kernel vmlinuz-2.6.33-lemote2f-bfs inside of
http://www.anheng.com.cn/loongson/install/loongson2_debian6_20100328.tar.lzma
(linked via linux-loongson-community) fails at the same place:

Quoted 9 lines

touch stamp-picdir
if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/regex.c -o pic/regex.o; \
else true; fi
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic ../../libiberty/regex.c -o regex.o if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/cplus-dem.c -o pic/cplus-dem.o; \
else true; fi

When & where did you get the above information?

do you mean the kernel can not boot or there are some other problems after the kernel booting?

I guess: the whole system crashed when you was compiling something? then please ensure the as & ld is ok via fixing the NOPS with the tool (fix-nop.c) from http://dev.lemote.com/code/linux-loongson-community :

$ ./fix-nop `which as`
$ ./fix-nop `which ld`

Quoted 3 lines

Any other kernel I should try?

The kernel in the above link should boot, or you can compile one yourself.

Regards,
Wu Zhangjin

Andreas Barth

2010-04-10 08:10 UTC #7 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 17 lines · * Wu Zhangjin ([email protected]) [100407 09:38]:

On Tue, 2010-04-06 at 21:10 +0200, Andreas Barth wrote:
[...]
Quoted 5 lines
The kernel vmlinuz-2.6.33-lemote2f-bfs inside of
http://www.anheng.com.cn/loongson/install/loongson2_debian6_20100328.tar.lzma
(linked via linux-loongson-community) fails at the same place:
Quoted 9 lines
touch stamp-picdir
if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/regex.c -o pic/regex.o; \
else true; fi
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic ../../libiberty/regex.c -o regex.o if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/cplus-dem.c -o pic/cplus-dem.o; \
else true; fi

Quoted 8 lines

When & where did you get the above information?
do you mean the kernel can not boot or there are some other problems after the kernel booting?
I guess: the whole system crashed when you was compiling something? then please ensure the as & ld is ok via fixing the NOPS with the tool (fix-nop.c) from http://dev.lemote.com/code/linux-loongson-community :

The kernel does boot, but the whole machines crashes.

I know the fixups (I have adjusted binutils), but I need an kernel that userland cannot crash (otherwise it gets a bit hard to run that as debian buildd).

Andi

Wu Zhangjin

author

2010-04-12 03:35 UTC #8 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 27 lines · Andreas Barth

Quoted 17 lines · * Wu Zhangjin ([email protected]) [100407 09:38]:
On Tue, 2010-04-06 at 21:10 +0200, Andreas Barth wrote:
[...]
Quoted 5 lines
The kernel vmlinuz-2.6.33-lemote2f-bfs inside of
http://www.anheng.com.cn/loongson/install/loongson2_debian6_20100328.tar.lzma
(linked via linux-loongson-community) fails at the same place:
Quoted 9 lines
touch stamp-picdir
if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/regex.c -o pic/regex.o; \
else true; fi
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic ../../libiberty/regex.c -o regex.o if [ x"-fPIC" != x ]; then \
gcc -c -DHAVE_CONFIG_H -g -O2 -I. -I../../libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fPIC ../../libiberty/cplus-dem.c -o pic/cplus-dem.o; \
else true; fi
Quoted 8 lines
When & where did you get the above information?
do you mean the kernel can not boot or there are some other problems after the kernel booting?
I guess: the whole system crashed when you was compiling something? then please ensure the as & ld is ok via fixing the NOPS with the tool (fix-nop.c) from http://dev.lemote.com/code/linux-loongson-community :

Quoted 7 lines

The kernel does boot, but the whole machines crashes.
I know the fixups (I have adjusted binutils), but I need an kernel that userland cannot crash (otherwise it gets a bit hard to run that as debian buildd).

The userland canot be prevented from crash without user-space fixups, that's why we need the fix-nop.c to fix the NOPs in the binaries of user-land applications or using the -mfix-loongson-nop to compile the user-land applications.

If you just need to rebuild debian, you just need to fix the NOPs in the as and ld with fix-nop.c as I have mentioned before:

$ ./fix-nop $(which as)
$ ./fix-nop $(which ld)

Regards,
Wu Zhangjin

Zhang Le

2010-04-03 01:54 UTC #9 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

In reply to Andreas Barth

Quoted 6 lines · Andreas Barth

Quoted 2 lines · * Wu Zhangjin ([email protected]) [100313 05:45]:
This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.
Unfortunatly the Loongson 2F here still fails with this patch, compiled with the new binutils and both options enabled.

Which version of binutils exactly?
The patch is in cvs, but latest binutils-2.20.1 still didn't include it. You need to patch it.

Zhang, Le

Andreas Barth

2010-04-03 03:11 UTC #10 lore raw

Re: [PATCH v3 1/3] Loongson-2F: Flush the branch target history such as BTB and RAS

Quoted 11 lines · * Zhang Le ([email protected]) [100403 04:20]:

Quoted 6 lines · Andreas Barth
Quoted 2 lines · * Wu Zhangjin ([email protected]) [100313 05:45]:
This patch did clear BTB(branch target buffer), forbid RAS(return address stack) via Loongson-2F's 64bit diagnostic register.
Unfortunatly the Loongson 2F here still fails with this patch, compiled with the new binutils and both options enabled.
Which version of binutils exactly?
The patch is in cvs, but latest binutils-2.20.1 still didn't include it. You need to patch it.

Yes, binutils are patched with
http://sourceware.org/ml/binutils/2009-11/msg00387.html

Andi

Wu Zhangjin

author

2010-03-13 04:46 UTC #11 lore raw

PATCH v3 2/3 Loongson-2F: Enable fixups of binutils 2.20.1

In reply to Wu Zhangjin

From: Wu Zhangjin <[email protected]>

As the "Fixups of Loongson2F" patch[1] to binutils have been applied into binutils 2.20.1. It's time to enable the options provided by the patch to compile the kernel.

Without these fixups, the system will hang unexpectedly for the bug of processor.

To learn more about these fixups, please refer to the following references.

[1] "Fixups of Loongson2F" patch for binutils(actually for gas)
http://sourceware.org/ml/binutils/2009-11/msg00387.html
[2] Chapter 15 of "Loongson2F User Manual"(Chinese Version)
http://www.loongson.cn/uploadfile/file/200808211
[3] English Version of the above chapter 15
http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source

Signed-off-by Zhang Le <[email protected]>

Signed-off-by Wu Zhangjin <[email protected]>

---
 arch/mips/Makefile |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

arch/mips/Makefile +3 -1

index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)
 
 cflags-$(CONFIG_CPU_MIPS32_R1)	+= $(call cc-option,-march=mips32,-mips32 -U_MIPS_ISA -D_MIPS_ISA=_MIPS_ISA_MIPS32) \
 			-Wa,-mips32 -Wa,--trap

Signature · 1 line

1.7.0.1

Ralf Baechle

2010-03-17 13:52 UTC #12 lore raw

Re: [PATCH v3 2/3] Loongson-2F: Enable fixups of binutils 2.20.1

arch/mips/Makefile cflags-$(CONFIG_CPU_LOONGSON2) += -Wa,--trap · line 140

Quoted 12 lines · Wu Zhangjin

arch/mips/Makefile +3 -1

index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)

Shouldn't these options be used unconditionally? It seems a kernel build should rather fail than a possibly unreliable kernel be built - possibly even without the user noticing the problem.

Ralf

Zhang Le

2010-03-27 16:29 UTC #13 lore raw

Re: [PATCH v3 2/3] Loongson-2F: Enable fixups of binutils 2.20.1

Quoted 18 lines · Ralf Baechle

arch/mips/Makefile cflags-$(CONFIG_CPU_LOONGSON2) += -Wa,--trap · line 140
Quoted 12 lines · Wu Zhangjin
arch/mips/Makefile +3 -1
index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)
Shouldn't these options be used unconditionally? It seems a kernel build should rather fail than a possibly unreliable kernel be built - possibly even without the user noticing the problem.

Zhangjin has been busy preparing for his graduation paper.
I just talked to him. He said later batches of 2F processor is not affected by these two problems, according to Zhang Fuxin, manager of Lemote.

I am not sure on which model of Fuloong and Yeeloong these "good" 2F processors have been used. I think Fuxin should know this.

If Fuxin could told us now, we can make a new patch. In this patch, we decide whether to add these options or not base on the model number.

Otherwise, for now, I think we should enable these options unconditionally.

Zhang, Le

Zhang Le

2010-03-27 17:21 UTC #14 lore raw

Re: [PATCH v3 2/3] Loongson-2F: Enable fixups of binutils 2.20.1

Quoted 24 lines · Zhang Le

Quoted 18 lines · Ralf Baechle
arch/mips/Makefile cflags-$(CONFIG_CPU_LOONGSON2) += -Wa,--trap · line 140
Quoted 12 lines · Wu Zhangjin
arch/mips/Makefile +3 -1
index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)
Shouldn't these options be used unconditionally? It seems a kernel build should rather fail than a possibly unreliable kernel be built - possibly even without the user noticing the problem.
Zhangjin has been busy preparing for his graduation paper.
I just talked to him. He said later batches of 2F processor is not affected by these two problems, according to Zhang Fuxin, manager of Lemote.

Quoted 7 lines

I am not sure on which model of Fuloong and Yeeloong these "good" 2F processors have been used. I think Fuxin should know this.
If Fuxin could told us now, we can make a new patch. In this patch, we decide whether to add these options or not base on the model number.
Otherwise, for now, I think we should enable these options unconditionally.

Sorry, I got Zhang Fuxin's email wrong. Now fixed.

Zhang, Le

zhangfx

2010-03-27 23:18 UTC #15 lore raw

Re: [PATCH v3 2/3] Loongson-2F: Enable fixups of binutils 2.20.1

The revision with these bugs fixed is 2F03. Up to now all processors used in current products are 2F01/02. 2F03 is in production and expected in this summer.

Quoted 42 lines · Zhang Le

Quoted 26 lines · Zhang Le
Quoted 20 lines · Ralf Baechle
arch/mips/Makefile cflags-$(CONFIG_CPU_LOONGSON2) += -Wa,--trap · line 141
Quoted 13 lines · Wu Zhangjin
arch/mips/Makefile +3 -1
index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)
 
Shouldn't these options be used unconditionally? It seems a kernel build should rather fail than a possibly unreliable kernel be built - possibly even without the user noticing the problem.
Zhangjin has been busy preparing for his graduation paper.
I just talked to him. He said later batches of 2F processor is not affected by these two problems, according to Zhang Fuxin, manager of Lemote.
Quoted 8 lines
I am not sure on which model of Fuloong and Yeeloong these "good" 2F processors have been used. I think Fuxin should know this.
If Fuxin could told us now, we can make a new patch. In this patch, we decide whether to add these options or not base on the model number.
Otherwise, for now, I think we should enable these options unconditionally.
Sorry, I got Zhang Fuxin's email wrong. Now fixed.
Zhang, Le

Signature · 2 lines

张福新 Zhang Fuxin
龙芯梦兰 管理部 总经理 Lemote General Manager [email protected]

Wu Zhangjin

author

2010-04-07 13:09 UTC #16 lore raw

Re: [PATCH v3 2/3] Loongson-2F: Enable fixups of binutils 2.20.1

In reply to Ralf Baechle

Quoted 18 lines · Ralf Baechle

arch/mips/Makefile cflags-$(CONFIG_CPU_LOONGSON2) += -Wa,--trap · line 140
Quoted 12 lines · Wu Zhangjin
arch/mips/Makefile +3 -1
index 2f2eac2..5ae342e 100644
@@ -135,7 +135,9 @@ cflags-$(CONFIG_CPU_LOONGSON2)	+= -Wa,--trap
 cflags-$(CONFIG_CPU_LOONGSON2E) += \
 	$(call cc-option,-march=loongson2e,-march=r4600)
 cflags-$(CONFIG_CPU_LOONGSON2F) += \
-	$(call cc-option,-march=loongson2f,-march=r4600)
+	$(call cc-option,-march=loongson2f,-march=r4600) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-nop,) \
+	$(call as-option,-Wa$(comma)-mfix-loongson2f-jump,)
Shouldn't these options be used unconditionally? It seems a kernel build should rather fail than a possibly unreliable kernel be built - possibly even without the user noticing the problem.

Thanks for your good suggestion!

Just added a new kernel config option: CPU_LOONGSON2F_WORKAROUNDS to allow the users to enable the workarounds for the necessary loongson2f batches and it is enabled by default. And to force the users to use the right binutils with the workarounds, errors will be printed on the standard output with the following stuff:

ifeq ($(call as-option,-Wa$(comma)-mfix-loongson2f-nop,),)

    $(error gcc does not support needed option -mfix-loongson2f-nop)
  else
    cflags-$(CONFIG_CPU_NOP_WORKAROUNDS) += -Wa

$(comma)-mfix-loongson2f-nop
endif

Will resend the new patchset with your feedbacks asap.

Regards,
Wu Zhangjin

Wu Zhangjin

author

2010-03-13 04:46 UTC #17 lore raw

PATCH v3 3/3 Loongson-2F: Fixup of problems introduced by -mfix-loongson2f-jump of binutils 2.20.1

In reply to Wu Zhangjin

From: Wu Zhangjin <[email protected]>

The -mfix-loongson2f-jump option provided by the binutils 2.20.1 have fixed the Out-of-order Issue of Loongson-2F described in Chapter 15 of "Loongson2F User Manual"[1,2], but introduced some problems.

The option changes all of the jumping target to "addr & 0xcfffffff" through the at($1) register, but for the REBOOT address of loongson-2F: 0xbfc00000, this is totally wrong, so, this patch try to avoid the problem via telling the assembler not to use at($1) register.

[1] Loongson2F User Manual(Chinese Version)
http://www.loongson.cn/uploadfile/file/200808211
[2] English Version of Chapter 15:
http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source

Reported-and-tested-by Liu Shiwei <[email protected]>

Signed-off-by Wu Zhangjin <[email protected]>

---
 arch/mips/loongson/common/reset.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

arch/mips/loongson/common/reset.c +10 -1

index 4bd9c18..dc311f2 100644
@@ -21,8 +21,17 @@ static void loongson_restart(char *command)
 	/* do preparation for reboot */
 	mach_prepare_reboot();
 
-	/* reboot via jumping to boot base address */
+	/* reboot via jumping to boot base address
+	 *
+	 * ".set noat" and ".set at" are used to ensure the address not break
+	 * by the -mfix-loongson2f-jump option provided by binutils 2.20.1 (or
+	 * higher version) which try to change the jumping address to "addr &
+	 * 0xcfffffff" via the at($1) register, this is totally wrong for
+	 * 0xbfc00000 (LOONGSON_BOOT_BASE).
+	 */
+	__asm__ __volatile__(".set noat\n");
 	((void (*)(void))ioremap_nocache(LOONGSON_BOOT_BASE, 4)) ();
+	__asm__ __volatile__(".set at\n");
 }
 
 static void loongson_poweroff(void)

Signature · 1 line

1.7.0.1

Ralf Baechle

2010-03-17 15:02 UTC #18 lore raw

Re: [PATCH v3 3/3] Loongson-2F: Fixup of problems introduced by -mfix-loongson2f-jump of binutils 2.20.1

arch/mips/loongson/common/reset.c static void loongson_restart(char *command) · line 34

Quoted 12 lines · Wu Zhangjin

arch/mips/loongson/common/reset.c +10 -1

-	/* reboot via jumping to boot base address */
+	/* reboot via jumping to boot base address
+	 *
+	 * ".set noat" and ".set at" are used to ensure the address not break
+	 * by the -mfix-loongson2f-jump option provided by binutils 2.20.1 (or
+	 * higher version) which try to change the jumping address to "addr &
+	 * 0xcfffffff" via the at($1) register, this is totally wrong for
+	 * 0xbfc00000 (LOONGSON_BOOT_BASE).
+	 */
+	__asm__ __volatile__(".set noat\n");
 	((void (*)(void))ioremap_nocache(LOONGSON_BOOT_BASE, 4)) ();
+	__asm__ __volatile__(".set at\n");

Ouch. This is fragile and totally puts the kernels fate at the mercy of gcc and the ioremap_nocache() implementation. GCC might emit a .set noat at any time. Something like

void loongson_restart(char *command)

{
	void (*func)(void);

	/* do preparation for reboot */
	mach_prepare_reboot();

	/* reboot via jumping to boot base address */
	func = (void *) ioremap_nocache(LOONGSON_BOOT_BASE, 4);

	__asm__ __volatile__(
	"	.set	noat						\n"
	"	jr	%[func]						\n"
	"	.set	at						\n"
	: /* No outputs */
	: [func] "r" (func));
}

should be safe against -mfix-loongson2f-jump I think.

The workaround in http://sourceware.org/ml/binutils/2009-11/msg00387.html
will also cause problems calling functions with bits 28..29 set, that is in the ranges 0x81000000..0xbffffff and 0xd0000000..0xffffffff. The first range is not much of a problem as only the kernel proper resides there and the kernel load address is manually selected in the Makefile. The 2nd range might be used for under certain circumstances.

Ralf

Zhang Le

2010-03-27 17:13 UTC #19 lore raw

Re: [PATCH v3 3/3] Loongson-2F: Fixup of problems introduced by -mfix-loongson2f-jump of binutils 2.20.1

Quoted 2 lines · Ralf Baechle

The workaround in http://sourceware.org/ml/binutils/2009-11/msg00387.html
will also cause problems calling functions with bits 28..29 set, that is

Yes. If all functions address doesn't have 28..29 set, this patch will not have any side effect. That means all functions should reside in two ranges:
0x80000000..0x8fffffff
0xc0000000..0xcfffffff

Quoted 1 line

in the ranges 0x81000000..0xbffffff and 0xd0000000..0xffffffff. The

                ^^^^^^^^^^

I think you mean 0x90000000, right?

Quoted 3 lines

first range is not much of a problem as only the kernel proper resides there and the kernel load address is manually selected in the Makefile. The 2nd range might be used for under certain circumstances.

It seems the Loongson kernel I compiled satisfies the requirement I listed above:

ffffffff80200000 A _text
...
ffffffff8089bbe0 A _end

So, if I understand it correctly, as long as the kernel size is smaller than 254MB, the patch should be ok. Is there anything I overlooked?

Zhang, Le

Wu Zhangjin

author

2010-04-07 13:04 UTC #20 lore raw

Re: [PATCH v3 3/3] Loongson-2F: Fixup of problems introduced by -mfix-loongson2f-jump of binutils 2.20.1

In reply to Ralf Baechle

Quoted 37 lines · Ralf Baechle

arch/mips/loongson/common/reset.c static void loongson_restart(char *command) · line 34

Quoted 12 lines · Wu Zhangjin

arch/mips/loongson/common/reset.c +10 -1

-	/* reboot via jumping to boot base address */
+	/* reboot via jumping to boot base address
+	 *
+	 * ".set noat" and ".set at" are used to ensure the address not break
+	 * by the -mfix-loongson2f-jump option provided by binutils 2.20.1 (or
+	 * higher version) which try to change the jumping address to "addr &
+	 * 0xcfffffff" via the at($1) register, this is totally wrong for
+	 * 0xbfc00000 (LOONGSON_BOOT_BASE).
+	 */
+	__asm__ __volatile__(".set noat\n");
 	((void (*)(void))ioremap_nocache(LOONGSON_BOOT_BASE, 4)) ();
+	__asm__ __volatile__(".set at\n");

Ouch. This is fragile and totally puts the kernels fate at the mercy of gcc and the ioremap_nocache() implementation. GCC might emit a .set noat at any time. Something like

void loongson_restart(char *command)

{
	void (*func)(void);

	/* do preparation for reboot */
	mach_prepare_reboot();

	/* reboot via jumping to boot base address */
	func = (void *) ioremap_nocache(LOONGSON_BOOT_BASE, 4);

	__asm__ __volatile__(
	"	.set	noat						\n"
	"	jr	%[func]						\n"
	"	.set	at						\n"
	: /* No outputs */
	: [func] "r" (func));
}

Quoted 1 line

should be safe against -mfix-loongson2f-jump I think.

Thanks very much, will apply it in the revision of this patch.

Regards,
Wu Zhangjin