Skip to contentSkip to conversation

sparc32 sunrpc.o

DISCUSSION Pete Zaitcev 4 participants 10 messages Sep 26, 2002 – Sep 30, 2002 1 message not in the archive SPARC + UltraSPARC (sparc/sparc64)

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 10 lines

Since 2.4.20-pre2 or 3, sunrpc.o has had this problem on sparc32:

depmod: *** Unresolved symbols in /lib/modules/2.4.20-pre8/kernel/net/sunrpc/sunrpc.o
depmod:         ___illegal_use_of_BTFIXUP_SETHI_in_module
depmod:         ___f_set_pte
depmod:         fix_kmap_begin
depmod:         ___f_flush_cache_all
depmod:         ___f_pte_clear
depmod:         ___f_mk_pte
depmod:         ___f_flush_tlb_all

Try these two things:

1. diff -urN -X dontdiff linux-2.4.19 linux-2.4.20-pre2 > x.diff vi x.diff
I've got Tigran's dontdiff updated at
http://people.redhat.com/zaitcev/linux/dontdiff.fix

2. make vmlinux modules > build.out 2>&1 </dev/null grep -i warning build.out

-- Pete

#2 lore raw

From: Pete Zaitcev <[email protected]>
Date: Thu, 26 Sep 2002 17:27:24 -0400

Quoted 10 lines

Since 2.4.20-pre2 or 3, sunrpc.o has had this problem on sparc32:

depmod: *** Unresolved symbols in /lib/modules/2.4.20-pre8/kernel/net/sunrpc/sunrpc.o
depmod:         ___illegal_use_of_BTFIXUP_SETHI_in_module
depmod:         ___f_set_pte
depmod:         fix_kmap_begin
depmod:         ___f_flush_cache_all
depmod:         ___f_pte_clear
depmod:         ___f_mk_pte
depmod:         ___f_flush_tlb_all

Try these two things:

No Peter, it really does use kmap_atomic stuff from modules, and this precludes providing those routines inline in highmem.h, they must live statically in main kernel image so that flush/pte calls can be properly BTFIXUP'd.

See my other email.

#3 lore raw
Quoted 14 lines

From: Pete Zaitcev <[email protected]>
Date: Thu, 26 Sep 2002 17:27:24 -0400

Quoted 10 lines

Since 2.4.20-pre2 or 3, sunrpc.o has had this problem on sparc32:

depmod: *** Unresolved symbols in /lib/modules/2.4.20-pre8/kernel/net/sunrpc/sunrpc.o
depmod:         ___illegal_use_of_BTFIXUP_SETHI_in_module
depmod:         ___f_set_pte
depmod:         fix_kmap_begin
depmod:         ___f_flush_cache_all
depmod:         ___f_pte_clear
depmod:         ___f_mk_pte
depmod:         ___f_flush_tlb_all
Quoted 8 lines

Try these two things:

No Peter, it really does use kmap_atomic stuff from modules, and this precludes providing those routines inline in highmem.h, they must live statically in main kernel image so that flush/pte calls can be properly BTFIXUP'd.

See my other email.

Ok, DaveM, could you have a look at this patch?

T.

arch/sparc/kernel/sparc_ksyms.c +11 -0
diff -urN linux-2.4.20-pre8.vanilla/arch/sparc/kernel/sparc_ksyms.c linux-2.4.20-pre8/arch/sparc/kernel/sparc_ksyms.c
@@ -46,6 +46,9 @@
 #include <asm/sbus.h>
 #include <asm/dma.h>
 #endif
+#ifdef CONFIG_HIGHMEM
+#include <asm/highmem.h>
+#endif
 #include <asm/a.out.h>
 #include <asm/io-unit.h>
 
@@ -204,6 +207,14 @@
 EXPORT_SYMBOL(pci_dma_sync_single);
 #endif
 
+/* Highmem helpers from arch/sparc/mm/highmem.c */
+#ifdef CONFIG_HIGHMEM
+EXPORT_SYMBOL(kmap);
+EXPORT_SYMBOL(kunmap);
+EXPORT_SYMBOL(kmap_atomic);
+EXPORT_SYMBOL(kunmap_atomic);
+#endif
+
 /* Solaris/SunOS binary compatibility */
 EXPORT_SYMBOL(svr4_setcontext);
 EXPORT_SYMBOL(svr4_getcontext);
arch/sparc/mm/Makefile +1 -1
diff -urN linux-2.4.20-pre8.vanilla/arch/sparc/mm/Makefile linux-2.4.20-pre8/arch/sparc/mm/Makefile
@@ -11,7 +11,7 @@
 	$(CC) $(AFLAGS) -ansi -c -o $*.o $<
 
 O_TARGET := mm.o
-obj-y    := fault.o init.o loadmmu.o generic.o extable.o btfixup.o
+obj-y    := fault.o init.o loadmmu.o generic.o extable.o highmem.o btfixup.o
 
 ifeq ($(CONFIG_SUN4),y)
 obj-y	 += nosrmmu.o
arch/sparc/mm/highmem.c +101 -0

101 lines not on this page · raw message

include/asm-sparc/highmem.h +9 -92

113 lines not on this page · raw message

#4 lore raw
Quoted 16 lines
Quoted 16 lines

From: Pete Zaitcev <[email protected]>
Date: Thu, 26 Sep 2002 17:27:24 -0400

Quoted 10 lines

Since 2.4.20-pre2 or 3, sunrpc.o has had this problem on sparc32:

depmod: *** Unresolved symbols in /lib/modules/2.4.20-pre8/kernel/net/sunrpc/sunrpc.o
depmod:         ___illegal_use_of_BTFIXUP_SETHI_in_module
depmod:         ___f_set_pte
depmod:         fix_kmap_begin
depmod:         ___f_flush_cache_all
depmod:         ___f_pte_clear
depmod:         ___f_mk_pte
depmod:         ___f_flush_tlb_all

Try these two things:

Quoted 8 lines
Quoted 6 lines

No Peter, it really does use kmap_atomic stuff from modules, and this precludes providing those routines inline in highmem.h, they must live statically in main kernel image so that flush/pte calls can be properly BTFIXUP'd.

See my other email.

Ok, DaveM, could you have a look at this patch?

Please disregard, highmem.c unreasonably included linux/highmem.h. I'll be looking into this some more.

T.

#5 lore raw

From: Tomas Szepe <[email protected]>
Date: Sat, 28 Sep 2002 18:13:16 +0200

Quoted 1 line

Ok, DaveM, could you have a look at this patch?

Please disregard, highmem.c unreasonably included linux/highmem.h. I'll be looking into this some more.

Let us know when new working patch is available :-)

#6 lore raw
Quoted 9 lines

From: Tomas Szepe <[email protected]>
Date: Sat, 28 Sep 2002 18:13:16 +0200

Quoted 1 line

Ok, DaveM, could you have a look at this patch?

Please disregard, highmem.c unreasonably included linux/highmem.h. I'll be looking into this some more.

Let us know when new working patch is available :-)

Right. :)
This took a bit longer cos I had to leave for the night.

Patch against 2.4.20-pre8.

T.

arch/sparc/kernel/sparc_ksyms.c +9 -0
diff -urN 2.4.20-pre8.vanilla/arch/sparc/kernel/sparc_ksyms.c linux-2.4.20-pre8/arch/sparc/kernel/sparc_ksyms.c
@@ -46,6 +46,9 @@
 #include <asm/sbus.h>
 #include <asm/dma.h>
 #endif
+#ifdef CONFIG_HIGHMEM
+#include <asm/highmem.h>
+#endif
 #include <asm/a.out.h>
 #include <asm/io-unit.h>
 
@@ -204,6 +207,12 @@
 EXPORT_SYMBOL(pci_dma_sync_single);
 #endif
 
+/* in arch/sparc/mm/highmem.c */
+#ifdef CONFIG_HIGHMEM
+EXPORT_SYMBOL(kmap_atomic);
+EXPORT_SYMBOL(kunmap_atomic);
+#endif
+
 /* Solaris/SunOS binary compatibility */
 EXPORT_SYMBOL(svr4_setcontext);
 EXPORT_SYMBOL(svr4_getcontext);
arch/sparc/mm/Makefile +1 -1
diff -urN 2.4.20-pre8.vanilla/arch/sparc/mm/Makefile linux-2.4.20-pre8/arch/sparc/mm/Makefile
@@ -11,7 +11,7 @@
 	$(CC) $(AFLAGS) -ansi -c -o $*.o $<
 
 O_TARGET := mm.o
-obj-y    := fault.o init.o loadmmu.o generic.o extable.o btfixup.o
+obj-y    := fault.o init.o loadmmu.o generic.o extable.o highmem.o btfixup.o
 
 ifeq ($(CONFIG_SUN4),y)
 obj-y	 += nosrmmu.o
arch/sparc/mm/highmem.c +84 -0

84 lines not on this page · raw message

include/asm-sparc/highmem.h +14 -78

110 lines not on this page · raw message

#7 lore raw

arch/sparc/kernel/sparc_ksyms.c line 51

Quoted 7 lines

Date: Sun, 29 Sep 2002 12:22:39 +0200
From: Tomas Szepe <[email protected]>

+++ linux-2.4.20-pre8/arch/sparc/kernel/sparc_ksyms.c 2002-09-29 11:45:33.000000000 +0200

arch/sparc/kernel/sparc_ksyms.c +3 -0
+#ifdef CONFIG_HIGHMEM
+#include <asm/highmem.h>
+#endif

OK, this is actually correct, I think. Looks funny. :)

arch/sparc/mm/Makefile line 14

Quoted 7 lines

+++ linux-2.4.20-pre8/arch/sparc/mm/Makefile 2002-09-29 11:45:33.000000000 +0200

arch/sparc/mm/Makefile +1 -1
@@ -11,7 +11,7 @@
 	$(CC) $(AFLAGS) -ansi -c -o $*.o $<
 
 O_TARGET := mm.o
-obj-y    := fault.o init.o loadmmu.o generic.o extable.o btfixup.o
+obj-y    := fault.o init.o loadmmu.o generic.o extable.o highmem.o btfixup.o

Why is this not obj-$(CONFIG_HIGHMEM) ?

include/asm-sparc/highmem.h line 38

Quoted 3 lines
include/asm-sparc/highmem.h +3 -0
+/* in mm/memory.c */
+extern struct page *highmem_start_page;
+

I would not do this. I would try to include <linux/highmem.h> into arch/sparc/mm/highmem.c (instead of <asm/highmem.h> as you did).

Also, now that you moved a bunch of implementation out of inlines, try to trim the #include list at the top of <asm-sparc/highmem.h>. For instance, asm/vaddrs.h is a suspect.

The rest of the patch is sane, as far as I can tell. I do not have a ready to run 2.4 sparc box, sorry. Please ask sparclinux@vger people to test, especially Uzi.

-- Pete

#8 lore raw

From: Pete Zaitcev <[email protected]>
Date: Sun, 29 Sep 2002 19:56:12 -0400

The rest of the patch is sane, as far as I can tell. I do not have a ready to run 2.4 sparc box, sorry. Please ask sparclinux@vger people to test, especially Uzi.

Let me know when new acceptable and tested patch is available.

#9 lore raw
In reply to Pete Zaitcev
Quoted 10 lines · Pete Zaitcev
Quoted 4 lines

Date: Sun, 29 Sep 2002 12:22:39 +0200
From: Tomas Szepe <[email protected]>

+++ linux-2.4.20-pre8/arch/sparc/kernel/sparc_ksyms.c 2002-09-29 11:45:33.000

000000 +0200

arch/sparc/kernel/sparc_ksyms.c line 51

Quoted 3 lines
arch/sparc/kernel/sparc_ksyms.c +3 -0
+#ifdef CONFIG_HIGHMEM
+#include <asm/highmem.h>
+#endif

OK, this is actually correct, I think. Looks funny. :)

Do it inside the included file then?

Signature · 4 lines
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513
#10 lore raw
In reply to Pete Zaitcev

Hi,

Quoted 9 lines

Review line 14

Quoted 7 lines

+++ linux-2.4.20-pre8/arch/sparc/mm/Makefile 2002-09-29 11:45:33.000000000 +0200

(quoted diff) +1 -1
@@ -11,7 +11,7 @@
 	$(CC) $(AFLAGS) -ansi -c -o $*.o $<
 
 O_TARGET := mm.o
-obj-y    := fault.o init.o loadmmu.o generic.o extable.o btfixup.o
+obj-y    := fault.o init.o loadmmu.o generic.o extable.o highmem.o btfixup.o

Why is this not obj-$(CONFIG_HIGHMEM) ?

My bad. I took the liberty of fixing up the formatting of the makefile when I was at changing this.

Quoted 6 lines

Review line 17

Quoted 3 lines
(quoted diff) +3 -0
+/* in mm/memory.c */
+extern struct page *highmem_start_page;
+

I would not do this. I would try to include <linux/highmem.h> into arch/sparc/mm/highmem.c (instead of <asm/highmem.h> as you did).

Ok. linux/highmem.h initially looked suspicious to me. Also if this extern isn't put in asm-sparc/highmem.h, linux/highmem.h has to be included everywhere instead (which is not really a problem I think).

Quoted 3 lines

Also, now that you moved a bunch of implementation out of inlines, try to trim the #include list at the top of <asm-sparc/highmem.h>. For instance, asm/vaddrs.h is a suspect.

Indeed! And linux/init.h and asm/pgtable.h aren't needed either.

Quoted 3 lines

The rest of the patch is sane, as far as I can tell. I do not have a ready to run 2.4 sparc box, sorry. Please ask sparclinux@vger people to test, especially Uzi.

Will do. Hopefully aurora people will have a go with the patch too. v3 as per your comments follows.

T.

arch/sparc/kernel/sparc_ksyms.c +9 -0
diff -urN 2.4.20-pre8.vanilla/arch/sparc/kernel/sparc_ksyms.c linux-2.4.20-pre8/arch/sparc/kernel/sparc_ksyms.c
@@ -46,6 +46,9 @@
 #include <asm/sbus.h>
 #include <asm/dma.h>
 #endif
+#ifdef CONFIG_HIGHMEM
+#include <linux/highmem.h>
+#endif
 #include <asm/a.out.h>
 #include <asm/io-unit.h>
 
@@ -204,6 +207,12 @@
 EXPORT_SYMBOL(pci_dma_sync_single);
 #endif
 
+/* in arch/sparc/mm/highmem.c */
+#ifdef CONFIG_HIGHMEM
+EXPORT_SYMBOL(kmap_atomic);
+EXPORT_SYMBOL(kunmap_atomic);
+#endif
+
 /* Solaris/SunOS binary compatibility */
 EXPORT_SYMBOL(svr4_setcontext);
 EXPORT_SYMBOL(svr4_getcontext);
arch/sparc/mm/Makefile +11 -6
diff -urN 2.4.20-pre8.vanilla/arch/sparc/mm/Makefile linux-2.4.20-pre8/arch/sparc/mm/Makefile
@@ -10,19 +10,24 @@
 .S.o:
 	$(CC) $(AFLAGS) -ansi -c -o $*.o $<
 
-O_TARGET := mm.o
-obj-y    := fault.o init.o loadmmu.o generic.o extable.o btfixup.o
+O_TARGET	:= mm.o
+obj-y		:= fault.o init.o loadmmu.o generic.o extable.o btfixup.o
 
 ifeq ($(CONFIG_SUN4),y)
-obj-y	 += nosrmmu.o
+obj-y		+= nosrmmu.o
 else
-obj-y	 += srmmu.o iommu.o io-unit.o hypersparc.o viking.o tsunami.o swift.o
+obj-y		+= srmmu.o iommu.o io-unit.o hypersparc.o viking.o \
+			tsunami.o swift.o
+endif
+
+ifdef CONFIG_HIGHMEM
+obj-y		+= highmem.o
 endif
 
 ifdef CONFIG_SMP
-obj-y   += nosun4c.o
+obj-y		+= nosun4c.o
 else
-obj-y   += sun4c.o
+obj-y		+= sun4c.o
 endif
 
 include $(TOPDIR)/Rules.make
arch/sparc/mm/highmem.c +84 -0

84 lines not on this page · raw message

include/asm-sparc/highmem.h +11 -81

116 lines not on this page · raw message