[parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

Hi all,

This boot fine on the b180 (32bit kernel), the b2k (32bit kernel only even
with CONFIG_PA20 is Ok but failed with 64bit kernel?) and N4000-55 (same
64bit kernel  build with my own kgcc-3.3.3 :)).

Any way this last test set aside any pb with _my_ gcc compiler :).

Congratulation to all,
    Joel

>-- Original Message --
>To: parisc-linux-cvs@lists.parisc-linux.org
>Date: Tue, 27 Jan 2004 23:05:16 -0700 (MST)
>From: carlos@parisc-linux.org (Carlos O'Donell)
>Reply-To: parisc-linux@lists.parisc-linux.org
>Subject: [parisc-linux-cvs] linux-2.6 carlos
>
>
>CVSROOT:	/var/cvs
Module name:	linux-2.6
Changes by:	carlos	04/01/27 23:05:15

Modified files:
	.              : Makefile 

Log message:
Builds and boots on my A500 (64-bit), builds in 32-bit mode aswell (e.g.
no CONFIG_COMPAT)

_____________
>_________________________________
parisc-linux-cvs mailing list
parisc-linux-cvs@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux-cvs



-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Wed, Jan 28, 2004 at 03:00:48PM +0100, Joel Soete wrote:
> Hi all,
> 
> This boot fine on the b180 (32bit kernel), the b2k (32bit kernel only even
> with CONFIG_PA20 is Ok but failed with 64bit kernel?) and N4000-55 (same
> 64bit kernel  build with my own kgcc-3.3.3 :)).
> 
> Any way this last test set aside any pb with _my_ gcc compiler :).

Undo the patch, rebuild, and see if it still fails?
Could you explain more about the failure?

c.

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

>> Any way this last test set aside any pb with _my_ gcc compiler :).
>
>Undo the patch, rebuild, and see if it still fails?

I don't think it has any relation with any patch: since my first test of
2.6 I always reach to boot a 32bit kernel but never a 64bit kernel on this
b2k (that is why I hesitate a long time before trying this on the N because
suspecting a compiler pb)

>Could you explain more about the failure?

Well the pb is easy to easy:
the last messages the palo banner advising: "... If this is the last message
you see ..." then the kernel do very early a hpmc. I so try to get piminfo
but everything seems to point out of kernel :(. I so try to add some printk()
just at begining of init() (in init/main.c just before lock_kernel()) but
it seems to not even reach this point (but not quiet sure that it is a relevant
place for printk() ). 
So I suspect a pb in very early boot into head64.S or vmlinux.lds.S but here
I could insert printk() ?

Any idea?

Thanks for attention,
    Joel



-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

>>Could you explain more about the failure?
>
>Well the pb is easy to decribe:
>the last messages the palo banner advising: "... If this is the last message
>you see ..." then the kernel do very early a hpmc. I so try to get piminfo
>but everything seems to point out of kernel :(. I so try to add some printk()
>just at begining of init() (in init/main.c just before lock_kernel()) but
>it seems to not even reach this point (but not quiet sure that it is a relevant
>place for printk() ). 

seems Ok :)

>So I suspect a pb in very early boot into head64.S or vmlinux.lds.S but
here
>I could insert printk() ?

Sorry for auto-reply but my suspicious was right: I just revert this patch:
===================================================================
RCS file: /var/lib/cvs/linux-2.5/arch/parisc/kernel/head64.S,v
retrieving revision 1.1.2.1
retrieving revision 1.1.2.2
diff -u -r1.1.2.1 -r1.1.2.2
--- linux-2.5/arch/parisc/kernel/head64.S	2002/12/24 22:15:01	1.1.2.1
+++ linux-2.5/arch/parisc/kernel/head64.S	2003/01/14 13:08:03	1.1.2.2
@@ -176,7 +176,9 @@
 #endif /* CONFIG_SMP */
 
 	/* Save the rfi target address */
-	std		%r11,  TASK_PT_GR11-TASK_SZ_ALGN(%sp)
+	ldo		-THREAD_SZ_ALGN(%sp), %r1
+	ldd		TI_TASK(%r1), %r1
+	std		%r11,  TASK_PT_GR11(%r1)
 
 #ifndef CONFIG_PDC_NARROW
 	/* Switch to wide mode; Superdome doesn't support narrow PDC
@@ -206,7 +208,9 @@
 
 stext_pdc_ret:
 	/* restore rfi target address*/
-	ldd		TASK_PT_GR11-TASK_SZ_ALGN(%sp), %r11
+	ldo		-THREAD_SZ_ALGN(%sp), %r1
+	ldd		TI_TASK(%r1), %r1
+	ldd		TASK_PT_GR11(%r1), %r11
 
 	/* PARANOID: clear user scratch/user space SR's */
 	mtsp	%r0,%sr0
@@ -310,7 +314,7 @@
 	load32		PA(smp_init_current_idle_task),%sp
 	ldd		0(%sp),%sp	/* load task address */
 	mtctl           %sp,%cr30       /* store in cr30 */
-	ldo             TASK_SZ_ALGN(%sp),%sp
+	ldo             THREAD_SZ_ALGN(%sp),%sp
 	tophys_r1       %sp
 
 	/* point CPU to kernel page tables */
=========><=========

and now it boots well but I am very not able to fix it correctly :(.

hth.

Thanks again,
    Joel


-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

[parisc-linux] About 2.6 64bit on b2k?

[Joel Soete]

Hi all,

About this :
> 
> Sorry for auto-reply but my suspicious was right: I just revert this patch:
> ===================================================================
> RCS file: /var/lib/cvs/linux-2.5/arch/parisc/kernel/head64.S,v
> retrieving revision 1.1.2.1
> retrieving revision 1.1.2.2
> diff -u -r1.1.2.1 -r1.1.2.2
> --- linux-2.5/arch/parisc/kernel/head64.S	2002/12/24 22:15:01	1.1.2.1
> +++ linux-2.5/arch/parisc/kernel/head64.S	2003/01/14 13:08:03	1.1.2.2
> @@ -176,7 +176,9 @@
>  #endif /* CONFIG_SMP */
>  
>  	/* Save the rfi target address */
> -	std		%r11,  TASK_PT_GR11-TASK_SZ_ALGN(%sp)
> +	ldo		-THREAD_SZ_ALGN(%sp), %r1
> +	ldd		TI_TASK(%r1), %r1
> +	std		%r11,  TASK_PT_GR11(%r1)
>  
>  #ifndef CONFIG_PDC_NARROW
>  	/* Switch to wide mode; Superdome doesn't support narrow PDC
> @@ -206,7 +208,9 @@
>  
>  stext_pdc_ret:
>  	/* restore rfi target address*/
> -	ldd		TASK_PT_GR11-TASK_SZ_ALGN(%sp), %r11
> +	ldo		-THREAD_SZ_ALGN(%sp), %r1
> +	ldd		TI_TASK(%r1), %r1
> +	ldd		TASK_PT_GR11(%r1), %r11
>  
>  	/* PARANOID: clear user scratch/user space SR's */
>  	mtsp	%r0,%sr0
> @@ -310,7 +314,7 @@
>  	load32		PA(smp_init_current_idle_task),%sp
>  	ldd		0(%sp),%sp	/* load task address */
>  	mtctl           %sp,%cr30       /* store in cr30 */
> -	ldo             TASK_SZ_ALGN(%sp),%sp
> +	ldo             THREAD_SZ_ALGN(%sp),%sp
>  	tophys_r1       %sp
>  
>  	/* point CPU to kernel page tables */
> =========><=========
> 
> and now it boots well but I am very not able to fix it correctly :(.
> 
> hth.
> 
I don't understand why it works on c3k and not on b2k. Could it be because by b2k has only 256k of physical ram?

Also, I notice that in head64.S there aren't .import like:
[snip]
         .import init_thread_union,data
         .import $global$                /* forward declaration */
         .import fault_vector_11,code    /* IVA parisc 1.1 32 bit */
         .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */

[snip]

as in its 32bit brother head.S

I don't yet find the meaning of '.import' but just in case ;)

hth,
	Joel

Re: [parisc-linux] About 2.6 64bit on b2k?

[Grant Grundler]

On Sun, Feb 01, 2004 at 12:40:51AM +0000, Joel Soete wrote:
> I don't understand why it works on c3k and not on b2k.
> Could it be because by b2k has only 256k of physical ram?

No. And you meant 256MB.
But c3k and b2k both supported the same DIMMs originally (128 and 256MB
DIMMS). Minimum mem possible was thus 128MB and I thought 256MB was
the minimum shipped with each box.

Sorry - I don't know why reverting that diff matters.

> Also, I notice that in head64.S there aren't .import like:
> [snip]
>         .import init_thread_union,data
>         .import $global$                /* forward declaration */
>         .import fault_vector_11,code    /* IVA parisc 1.1 32 bit */
>         .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */
> [snip]
> 
> as in its 32bit brother head.S

head64.S only runs wide mode on PA20 HW.
No need for any PA11 or narrow mode code.
Look in the Makefile and see when head64.o is linked in.
Then you'll *SMACK* yourself on the forehead and say "DOH!" :^)


> I don't yet find the meaning of '.import' but just in case ;)

AFAIK, it's just telling the assembler where to expect global symbols.

grant

Re: [parisc-linux] About 2.6 64bit on b2k?

[Joel Soete]

HI Grant,


Grant Grundler wrote:
> On Sun, Feb 01, 2004 at 12:40:51AM +0000, Joel Soete wrote:
> 
>>I don't understand why it works on c3k and not on b2k.
>>Could it be because by b2k has only 256k of physical ram?
> 
> 
> No. And you meant 256MB.
Oops yes :)

> But c3k and b2k both supported the same DIMMs originally (128 and 256MB
> DIMMS). Minimum mem possible was thus 128MB and I thought 256MB was
> the minimum shipped with each box.
> 
> Sorry - I don't know why reverting that diff matters.
> 
Not more, I will try to objdump the two head64.o and see what the diff?

> 
>>Also, I notice that in head64.S there aren't .import like:
>>[snip]
>>        .import init_thread_union,data
>>        .import $global$                /* forward declaration */
>>        .import fault_vector_11,code    /* IVA parisc 1.1 32 bit */
>>        .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */
>>[snip]
>>
>>as in its 32bit brother head.S
> 
> 
> head64.S only runs wide mode on PA20 HW.
> No need for any PA11 or narrow mode code.
I understand that but less '.import init_thread_union,data' when
[snip]
         /* And the stack pointer too */
         load32          PA(init_thread_union+THREAD_SZ_ALGN),%sp

[snip]
(never the less in the working 2.4 it was the same, so I trust that the problem is very well in the computed pointers?
But I don't have material right now to check ;) )
I am also curious to revert patch on the N (on which the patch works)?

> Look in the Makefile and see when head64.o is linked in.
> Then you'll *SMACK* yourself on the forehead and say "DOH!" :^)
> 
> 
> 
>>I don't yet find the meaning of '.import' but just in case ;)
> 
> 
> AFAIK, it's just telling the assembler where to expect global symbols.
> 
> grant
> 
Many thanks,
	Joel

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Randolph Chung]

> Sorry for auto-reply but my suspicious was right: I just revert this patch:

can you try this one instead?

randolph

Index: arch/parisc/kernel/head64.S
===================================================================
RCS file: /var/cvs/linux-2.6/arch/parisc/kernel/head64.S,v
retrieving revision 1.3
diff -u -p -r1.3 head64.S
--- arch/parisc/kernel/head64.S 17 Jan 2004 15:00:32 -0000      1.3
+++ arch/parisc/kernel/head64.S 1 Feb 2004 18:09:13 -0000
@@ -165,8 +165,7 @@ common_stext:
 #endif /* CONFIG_SMP */

        /* Save the rfi target address */
-       ldo             -THREAD_SZ_ALGN(%sp), %r1
-       ldd             TI_TASK(%r1), %r1
+       ldd             TI_TASK-THREAD_SZ_ALGN(%sp), %r1
        std             %r11,  TASK_PT_GR11(%r1)

 #ifndef CONFIG_PDC_NARROW
@@ -301,7 +300,8 @@ smp_slave_stext:

        /*  Initialize the SP - monarch sets up smp_init_current_idle_task */
        load32          PA(smp_init_current_idle_task),%sp
-       ldd             0(%sp),%sp      /* load task address */
+       ldd             0(%sp),%sp      /* load thread_info address */
+       ldd             TASK_THREAD_INFO(%sp), %sp
        mtctl           %sp,%cr30       /* store in cr30 */
        ldo             THREAD_SZ_ALGN(%sp),%sp
        tophys_r1       %sp

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

Hi Randolph,

Very sorry but it doesn't works :( 
(Sorry also if I couldn't help more: my knowledge of kernel internals is
still too limited :( )

Thanks for your attention,
    Joel


>-- Original Message --
>Date: Sun, 1 Feb 2004 12:05:40 -0800
>From: Randolph Chung <randolph@tausq.org>
>To: Joel Soete <soete.joel@tiscali.be>
>Cc: Carlos O'Donell <carlos@baldric.uwo.ca>,
>	James Bottomley <James.Bottomley@steeleye.com>,
>	parisc-linux@lists.parisc-linux.org
>Reply-To: Randolph Chung <randolph@tausq.org>
>Subject: Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos
>
>
>> Sorry for auto-reply but my suspicious was right: I just revert this patch:

can you try this one instead?

randolph

Index: arch/parisc/kernel/head64.S
===================================================================
RCS file: /var/cvs/li
>ux-2.6/arch/parisc/kernel/head64.S,v
retrieving revision 1.3
diff -u -p -r1.3 head64.S
--- arch/parisc/kernel/head64.S 17 Jan 2004 15:00:32 -0000      1.3
+++ arch/parisc/kernel/head64.S 1 Feb 2004 18:09:13 -0000
@@ -165,8 +165,7 @@ common_stext:

> #endif /* CONFIG_SMP */

        /* Save the rfi target address */
-       ldo             -THREAD_SZ_ALGN(%sp), %r1
-       ldd             TI_TASK(%r1), %r1
+       ldd             TI_TASK-THREAD_SZ_ALGN(%sp), %r1
        std             %r11,
> TASK_PT_GR11(%r1)

 #ifndef CONFIG_PDC_NARROW
@@ -301,7 +300,8 @@ smp_slave_stext:

        /*  Initialize the SP - monarch sets up smp_init_current_idle_task
*/
        load32          PA(smp_init_current_idle_task),%sp
-       ldd            
>0(%sp),%sp      /* load task address */
+       ldd             0(%sp),%sp      /* load thread_info address */
+       ldd             TASK_THREAD_INFO(%sp), %sp
        mtctl           %sp,%cr30       /* store in cr30 */
        ldo             TH
>EAD_SZ_ALGN(%sp),%sp
        tophys_r1       %sp



-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] RE: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

Hi Randolph,

Very sorry but it doesn't works :( 
(Sorry also if I couldn't help more: my knowledge of kernel internals is
still too limited :( )

Thanks for your attention,
    Joel

PS: btw I try to compare the dump of two .o files:
palx2000:/usr/src/linux-2.6.2-rc2-pa3/arch/parisc/kernel# diff -Nau head64.working.dmp
head64.orig.dmp 
--- head64.working.dmp	2004-02-02 19:39:05.295417088 +0100
+++ head64.orig.dmp	2004-02-02 19:34:56.138294728 +0100
@@ -62,39 +62,43 @@
   cc:	f0 66 0c 00 	depd r6,31,32,r3
   d0:	68 00 00 20 	stw r0,10(r0)
   d4:	68 00 00 50 	stw r0,28(r0)
-  d8:	73 cb 38 31 	std r11,-3e8(sp)
-  dc:	00 00 14 a2 	mfia rp
-  e0:	34 42 00 28 	ldo 14(rp),rp
-  e4:	f4 40 0c 00 	depdi 0,31,32,rp
-  e8:	e8 40 c0 00 	bv r0(rp)
-  ec:	02 00 0d 60 	ssm 200,r0
-  f0:	34 1a 00 2a 	ldi 15,r26
-  f4:	34 19 00 04 	ldi 2,r25
-  f8:	34 18 00 04 	ldi 2,r24
-  fc:	20 40 00 00 	ldil 0,rp
- 100:	34 42 00 00 	ldo 0(rp),rp
- 104:	e8 60 c0 00 	bv r0(r3)
- 108:	08 00 02 57 	copy r0,r23
+  d8:	37 c1 3e 01 	ldo -100(sp),r1
+  dc:	0c 20 10 c1 	ldd 0(,r1),r1
+  e0:	70 2b 0b 30 	std r11,598(r1)
+  e4:	00 00 14 a2 	mfia rp
+  e8:	34 42 00 28 	ldo 14(rp),rp
+  ec:	f4 40 0c 00 	depdi 0,31,32,rp
+  f0:	e8 40 c0 00 	bv r0(rp)
+  f4:	02 00 0d 60 	ssm 200,r0
+  f8:	34 1a 00 2a 	ldi 15,r26
+  fc:	34 19 00 04 	ldi 2,r25
+ 100:	34 18 00 04 	ldi 2,r24
+ 104:	20 40 00 00 	ldil 0,rp
+ 108:	34 42 00 00 	ldo 0(rp),rp
+ 10c:	e8 60 c0 00 	bv r0(r3)
+ 110:	08 00 02 57 	copy r0,r23
 
-000000000000010c <stext_pdc_ret>:
- 10c:	53 cb 38 31 	ldd -3e8(sp),r11
- 110:	00 00 18 20 	mtsp r0,sr0
- 114:	00 00 58 20 	mtsp r0,sr1
- 118:	00 00 98 20 	mtsp r0,sr2
- 11c:	00 00 d8 20 	mtsp r0,sr3
- 120:	01 00 18 40 	mtctl r0,pidr1
- 124:	01 20 18 40 	mtctl r0,pidr2
- 128:	01 80 18 40 	mtctl r0,pidr3
- 12c:	01 a0 18 40 	mtctl r0,pidr4
- 130:	20 20 02 00 	ldil 10000000,r1
- 134:	08 3e 06 3e 	add sp,r1,sp
- 138:	23 60 00 00 	ldil 0,dp
- 13c:	37 7b 00 00 	ldo 0(dp),dp
- 140:	21 40 00 00 	ldil 0,r10
- 144:	35 4a 00 00 	ldo 0(r10),r10
- 148:	01 ca 18 40 	mtctl r10,iva
- 14c:	e8 00 01 58 	b,l 200 <aligned_rfi>,r0
- 150:	08 00 02 40 	nop
+0000000000000114 <stext_pdc_ret>:
+ 114:	37 c1 3e 01 	ldo -100(sp),r1
+ 118:	0c 20 10 c1 	ldd 0(,r1),r1
+ 11c:	50 2b 0b 30 	ldd 598(r1),r11
+ 120:	00 00 18 20 	mtsp r0,sr0
+ 124:	00 00 58 20 	mtsp r0,sr1
+ 128:	00 00 98 20 	mtsp r0,sr2
+ 12c:	00 00 d8 20 	mtsp r0,sr3
+ 130:	01 00 18 40 	mtctl r0,pidr1
+ 134:	01 20 18 40 	mtctl r0,pidr2
+ 138:	01 80 18 40 	mtctl r0,pidr3
+ 13c:	01 a0 18 40 	mtctl r0,pidr4
+ 140:	20 20 02 00 	ldil 10000000,r1
+ 144:	08 3e 06 3e 	add sp,r1,sp
+ 148:	23 60 00 00 	ldil 0,dp
+ 14c:	37 7b 00 00 	ldo 0(dp),dp
+ 150:	21 40 00 00 	ldil 0,r10
+ 154:	35 4a 00 00 	ldo 0(r10),r10
+ 158:	01 ca 18 40 	mtctl r10,iva
+ 15c:	e8 00 01 38 	b,l 200 <aligned_rfi>,r0
+ 160:	08 00 02 40 	nop
 	...
 
 0000000000000200 <aligned_rfi>:
=========><=========

head64.working.dmp being the one which working (ie with mentioned reverted
patch)

head64.orig.dmp being the cvs one.

Do you noticed compiler error?

hth,
    Joel


-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Fri, Apr 30, 2004 at 08:15:48AM -0600, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/04/30 08:15:48
> 
> Modified files:
> 	include/asm-parisc: unistd.h 
> 
> Log message:
> Update K_INLINE_SYSCALL so userspace apps using the macro work properly.

Lifted from the version I wrote for glibc. I get to clean this up yet
again when we remove the register saves across syscalls.

Index: include/asm-parisc/unistd.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-parisc/unistd.h,v
retrieving revision 1.5
diff -u -p -r1.5 unistd.h
--- a/include/asm-parisc/unistd.h	28 Apr 2004 23:13:37 -0000	1.5
+++ b/include/asm-parisc/unistd.h	30 Apr 2004 14:15:17 -0000
@@ -735,94 +735,141 @@
 
 #ifndef __ASSEMBLY__
 
-/* The old syscall code here didn't work, and it looks like it's only used
- * by applications such as fdisk which for some reason need to produce
- * their own syscall instead of using same from libc.  The code below
- * is leveraged from glibc/sysdeps/unix/sysv/linux/hppa/sysdep.h where
- * it is essentially duplicated -- which sucks.  -PB
- */
-
 #define SYS_ify(syscall_name)   __NR_##syscall_name
 
-/* The system call number MUST ALWAYS be loaded in the delay slot of
-   the ble instruction, or restarting system calls WILL NOT WORK.  See
-   arch/parisc/kernel/signal.c - dhd, 2000-07-26 */
-#define K_INLINE_SYSCALL(name, nr, args...)       ({              \
-        unsigned long __sys_res;                                \
-        {                                                       \
-                register unsigned long __res asm("r28");        \
-                K_LOAD_ARGS_##nr(args)                            \
-                asm volatile(                                   \
-			"ble  0x100(%%sr2, %%r0)\n\t"           \
-                        " ldi %1, %%r20"                        \
-                        : "=r" (__res)                          \
-                        : "i" (SYS_ify(name)) K_ASM_ARGS_##nr   \
-			  );                                    \
-                __sys_res = __res;                              \
-        }                                                       \
-        if (__sys_res >= (unsigned long)-4095) {                \
-		errno = -__sys_res;				\
-                __sys_res = (unsigned long)-1;                 \
-        }                                                       \
-        __sys_res;                                              \
+/* Assume all syscalls are done from PIC code just to be
+ * safe. The worst case scenario is that you lose a register
+ * and save/restore r19 across the syscall. */
+#define PIC
+
+#ifndef ASM_LINE_SEP
+# define ASM_LINE_SEP ;
+#endif
+
+/* Definition taken from glibc 2.3.3
+ * sysdeps/unix/sysv/linux/hppa/sysdep.h
+ */
+
+#ifdef PIC
+/* WARNING: CANNOT BE USED IN A NOP! */
+# define K_STW_PIC stw %r19, -32(%sr0, %sp) ASM_LINE_SEP
+# define K_LDW_PIC ldw -32(%sr0, %sp), %r19 ASM_LINE_SEP
+# define K_STW_ASM_PIC	"       copy %%r19, %%r4\n"
+# define K_LDW_ASM_PIC	"       copy %%r4, %%r19\n"
+# define K_USING_GR4	"%r4",
+#else
+# define K_STW_PIC ASM_LINE_SEP
+# define K_LDW_PIC ASM_LINE_SEP
+# define K_STW_ASM_PIC	" \n"
+# define K_LDW_ASM_PIC	" \n"
+# define K_USING_GR4
+#endif
+
+/* GCC has to be warned that a syscall may clobber all the ABI
+   registers listed as "caller-saves", see page 8, Table 2
+   in section 2.2.6 of the PA-RISC RUN-TIME architecture
+   document. However! r28 is the result and will conflict with
+   the clobber list so it is left out. Also the input arguments
+   registers r20 -> r26 will conflict with the list so they
+   are treated specially. Although r19 is clobbered by the syscall
+   we cannot say this because it would violate ABI, thus we say
+   r4 is clobbered and use that register to save/restore r19
+   across the syscall. */
+
+#define K_CALL_CLOB_REGS "%r1", "%r2", K_USING_GR4 \
+	        	 "%r20", "%r29", "%r31"
+
+#undef K_INLINE_SYSCALL
+#define K_INLINE_SYSCALL(name, nr, args...)	({			\
+	long __sys_res;							\
+	{								\
+		register unsigned long __res asm("r28");		\
+		K_LOAD_ARGS_##nr(args)					\
+		/* FIXME: HACK stw/ldw r19 around syscall */		\
+		asm volatile(						\
+			K_STW_ASM_PIC					\
+			"	ble  0x100(%%sr2, %%r0)\n"		\
+			"	ldi %1, %%r20\n"			\
+			K_LDW_ASM_PIC					\
+			: "=r" (__res)					\
+			: "i" (SYS_ify(name)) K_ASM_ARGS_##nr   	\
+			: K_CALL_CLOB_REGS K_CLOB_ARGS_##nr		\
+		);							\
+		__sys_res = (long)__res;				\
+	}								\
+	if ( (unsigned long)__sys_res >= (unsigned long)-4095 ){	\
+		errno = -__sys_res);		        		\
+		__sys_res = -1;						\
+	}								\
+	__sys_res;							\
 })
 
 #define K_LOAD_ARGS_0()
-#define K_LOAD_ARGS_1(r26)                                        \
-        register unsigned long __r26 __asm__("r26") = (unsigned long)r26;       \
-        K_LOAD_ARGS_0()
-#define K_LOAD_ARGS_2(r26,r25)                                    \
-        register unsigned long __r25 __asm__("r25") = (unsigned long)r25;       \
-        K_LOAD_ARGS_1(r26)
-#define K_LOAD_ARGS_3(r26,r25,r24)                                \
-        register unsigned long __r24 __asm__("r24") = (unsigned long)r24;       \
-        K_LOAD_ARGS_2(r26,r25)
-#define K_LOAD_ARGS_4(r26,r25,r24,r23)                            \
-        register unsigned long __r23 __asm__("r23") = (unsigned long)r23;       \
-        K_LOAD_ARGS_3(r26,r25,r24)
-#define K_LOAD_ARGS_5(r26,r25,r24,r23,r22)                        \
-        register unsigned long __r22 __asm__("r22") = (unsigned long)r22;       \
-        K_LOAD_ARGS_4(r26,r25,r24,r23)
-#define K_LOAD_ARGS_6(r26,r25,r24,r23,r22,r21)                    \
-        register unsigned long __r21 __asm__("r21") = (unsigned long)r21;       \
-        K_LOAD_ARGS_5(r26,r25,r24,r23,r22)
+#define K_LOAD_ARGS_1(r26)					\
+	register unsigned long __r26 __asm__("r26") = (unsigned long)(r26);   \
+	K_LOAD_ARGS_0()
+#define K_LOAD_ARGS_2(r26,r25)					\
+	register unsigned long __r25 __asm__("r25") = (unsigned long)(r25);   \
+	K_LOAD_ARGS_1(r26)
+#define K_LOAD_ARGS_3(r26,r25,r24)				\
+	register unsigned long __r24 __asm__("r24") = (unsigned long)(r24);   \
+	K_LOAD_ARGS_2(r26,r25)
+#define K_LOAD_ARGS_4(r26,r25,r24,r23)				\
+	register unsigned long __r23 __asm__("r23") = (unsigned long)(r23);   \
+	K_LOAD_ARGS_3(r26,r25,r24)
+#define K_LOAD_ARGS_5(r26,r25,r24,r23,r22)			\
+	register unsigned long __r22 __asm__("r22") = (unsigned long)(r22);   \
+	K_LOAD_ARGS_4(r26,r25,r24,r23)
+#define K_LOAD_ARGS_6(r26,r25,r24,r23,r22,r21)			\
+	register unsigned long __r21 __asm__("r21") = (unsigned long)(r21);   \
+	K_LOAD_ARGS_5(r26,r25,r24,r23,r22)
 
+/* Even with zero args we use r20 for the syscall number */
 #define K_ASM_ARGS_0
-#define K_ASM_ARGS_1 , "r" (__r26)
-#define K_ASM_ARGS_2 , "r" (__r26), "r" (__r25)
-#define K_ASM_ARGS_3 , "r" (__r26), "r" (__r25), "r" (__r24)
-#define K_ASM_ARGS_4 , "r" (__r26), "r" (__r25), "r" (__r24), "r" (__r23)
-#define K_ASM_ARGS_5 , "r" (__r26), "r" (__r25), "r" (__r24), "r" (__r23), "r" (__r22)
-#define K_ASM_ARGS_6 , "r" (__r26), "r" (__r25), "r" (__r24), "r" (__r23), "r" (__r22), "r" (__r21)
+#define K_ASM_ARGS_1 K_ASM_ARGS_0, "r" (__r26)
+#define K_ASM_ARGS_2 K_ASM_ARGS_1, "r" (__r25)
+#define K_ASM_ARGS_3 K_ASM_ARGS_2, "r" (__r24)
+#define K_ASM_ARGS_4 K_ASM_ARGS_3, "r" (__r23)
+#define K_ASM_ARGS_5 K_ASM_ARGS_4, "r" (__r22)
+#define K_ASM_ARGS_6 K_ASM_ARGS_5, "r" (__r21)
+
+/* The registers not listed as inputs but clobbered */
+#define K_CLOB_ARGS_6
+#define K_CLOB_ARGS_5 K_CLOB_ARGS_6, "%r21"
+#define K_CLOB_ARGS_4 K_CLOB_ARGS_5, "%r22"
+#define K_CLOB_ARGS_3 K_CLOB_ARGS_4, "%r23"
+#define K_CLOB_ARGS_2 K_CLOB_ARGS_3, "%r24"
+#define K_CLOB_ARGS_1 K_CLOB_ARGS_2, "%r25"
+#define K_CLOB_ARGS_0 K_CLOB_ARGS_1, "%r26"
 
-#define _syscall0(type,name)						      \
-type name(void)								      \
-{									      \
-    return K_INLINE_SYSCALL(name, 0);	\
+#define _syscall0(type,name)						\
+type name(void)								\
+{									\
+    return K_INLINE_SYSCALL(name, 0);	                                \
 }
 
-#define _syscall1(type,name,type1,arg1)					      \
-type name(type1 arg1)							      \
-{									      \
-    return K_INLINE_SYSCALL(name, 1, arg1);	\
+#define _syscall1(type,name,type1,arg1)					\
+type name(type1 arg1)							\
+{									\
+    return K_INLINE_SYSCALL(name, 1, arg1);	                        \
 }
 
-#define _syscall2(type,name,type1,arg1,type2,arg2)			      \
-type name(type1 arg1, type2 arg2)					      \
-{									      \
-    return K_INLINE_SYSCALL(name, 2, arg1, arg2);	\
+#define _syscall2(type,name,type1,arg1,type2,arg2)			\
+type name(type1 arg1, type2 arg2)					\
+{									\
+    return K_INLINE_SYSCALL(name, 2, arg1, arg2);	                \
 }
 
-#define _syscall3(type,name,type1,arg1,type2,arg2,type3,arg3)		      \
-type name(type1 arg1, type2 arg2, type3 arg3)				      \
-{									      \
-    return K_INLINE_SYSCALL(name, 3, arg1, arg2, arg3);	\
+#define _syscall3(type,name,type1,arg1,type2,arg2,type3,arg3)		\
+type name(type1 arg1, type2 arg2, type3 arg3)				\
+{									\
+    return K_INLINE_SYSCALL(name, 3, arg1, arg2, arg3);	                \
 }
 
-#define _syscall4(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4)      \
-type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4)		      \
-{									      \
-    return K_INLINE_SYSCALL(name, 4, arg1, arg2, arg3, arg4);	\
+#define _syscall4(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4)		\
+{									\
+    return K_INLINE_SYSCALL(name, 4, arg1, arg2, arg3, arg4);	        \
 }
 
 /* select takes 5 arguments */
@@ -834,13 +881,11 @@ type name(type1 arg1, type2 arg2, type3 
 
 
 /* mmap & mmap2 take 6 arguments */
-
 #define _syscall6(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5,type6,arg6) \
 type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5, type6 arg6) \
-{									\
+{									        \
     return K_INLINE_SYSCALL(name, 6, arg1, arg2, arg3, arg4, arg5, arg6);	\
 }
-
 
 #ifdef __KERNEL_SYSCALLS__
 

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Wed, Apr 28, 2004 at 05:13:38PM -0600, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/04/28 17:13:38
> 
> Modified files:
> 	arch/parisc/kernel: syscall_table.S 
> 	include/asm-parisc: unistd.h 
> 	ipc            : compat_mq.c 
>
> Log message:
> First cut to get mq_* functions working in userspace. 
> Fix CONFIG_POSIX_MQUEUE build. Connect mq_* syscalls. 
> Make our syscall_table match unistd.h definitions. 
> Boots on my A500, testing new glibc build as we speak.

Not too much glue required.

Index: arch/parisc/kernel/syscall_table.S
===================================================================
RCS file: /var/cvs/linux-2.6/arch/parisc/kernel/syscall_table.S,v
retrieving revision 1.6
diff -u -p -r1.6 syscall_table.S
--- a/arch/parisc/kernel/syscall_table.S	29 Mar 2004 12:51:52 -0000	1.6
+++ b/arch/parisc/kernel/syscall_table.S	28 Apr 2004 23:09:07 -0000
@@ -334,3 +334,12 @@
 	ENTRY_SAME(epoll_ctl)		/* 225 */
 	ENTRY_SAME(epoll_wait)
  	ENTRY_SAME(remap_file_pages)
+  ENTRY_SAME(semtimedop)
+  ENTRY_SAME(mq_open)
+  ENTRY_SAME(mq_unlink)   /* 230 */
+  ENTRY_SAME(mq_timedsend)
+  ENTRY_SAME(mq_timedreceive)
+  ENTRY_SAME(mq_notify)
+  ENTRY_SAME(mq_getsetattr)
+  /* Nothing yet */       /* 235 */
+
Index: include/asm-parisc/unistd.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-parisc/unistd.h,v
retrieving revision 1.4
diff -u -p -r1.4 unistd.h
--- a/include/asm-parisc/unistd.h	16 Mar 2004 15:40:38 -0000	1.4
+++ b/include/asm-parisc/unistd.h	28 Apr 2004 23:09:07 -0000
@@ -700,30 +700,35 @@
 #define __NR_acl_set            (__NR_Linux + 205)
 #define __NR_gettid             (__NR_Linux + 206)
 #define __NR_readahead          (__NR_Linux + 207)
-#define __NR_tkill		(__NR_Linux + 208)
-#define __NR_sendfile64		(__NR_Linux + 209)
-#define __NR_futex		(__NR_Linux + 210)
-#define __NR_sched_setaffinity	(__NR_Linux + 211)
-#define __NR_sched_getaffinity	(__NR_Linux + 212)
-#define __NR_set_thread_area	(__NR_Linux + 213)
-#define __NR_get_thread_area	(__NR_Linux + 214)
-#define __NR_io_setup		(__NR_Linux + 215)
-#define __NR_io_destroy		(__NR_Linux + 216)
-#define __NR_io_getevents	(__NR_Linux + 217)
-#define __NR_io_submit		(__NR_Linux + 218)
-#define __NR_io_cancel		(__NR_Linux + 219)
-#define __NR_alloc_hugepages	(__NR_Linux + 220)
-#define __NR_free_hugepages	(__NR_Linux + 221)
-#define __NR_exit_group		(__NR_Linux + 222)
-#define __NR_lookup_dcookie	(__NR_Linux + 223)
-#define __NR_epoll_create	(__NR_Linux + 224)
-#define __NR_epoll_ctl		(__NR_Linux + 225)
-#define __NR_epoll_wait		(__NR_Linux + 226)
-#define __NR_remap_file_pages	(__NR_Linux + 227)
-#define __NR_semtimedop		(__NR_Linux + 228)
+#define __NR_tkill              (__NR_Linux + 208)
+#define __NR_sendfile64         (__NR_Linux + 209)
+#define __NR_futex              (__NR_Linux + 210)
+#define __NR_sched_setaffinity  (__NR_Linux + 211)
+#define __NR_sched_getaffinity  (__NR_Linux + 212)
+#define __NR_set_thread_area    (__NR_Linux + 213)
+#define __NR_get_thread_area    (__NR_Linux + 214)
+#define __NR_io_setup           (__NR_Linux + 215)
+#define __NR_io_destroy         (__NR_Linux + 216)
+#define __NR_io_getevents       (__NR_Linux + 217)
+#define __NR_io_submit          (__NR_Linux + 218)
+#define __NR_io_cancel          (__NR_Linux + 219)
+#define __NR_alloc_hugepages    (__NR_Linux + 220)
+#define __NR_free_hugepages     (__NR_Linux + 221)
+#define __NR_exit_group         (__NR_Linux + 222)
+#define __NR_lookup_dcookie     (__NR_Linux + 223)
+#define __NR_epoll_create       (__NR_Linux + 224)
+#define __NR_epoll_ctl          (__NR_Linux + 225)
+#define __NR_epoll_wait         (__NR_Linux + 226)
+#define __NR_remap_file_pages   (__NR_Linux + 227)
+#define __NR_semtimedop         (__NR_Linux + 228)
+#define __NR_mq_open            (__NR_Linux + 229) /* Keep the mq_* syscalls together */
+#define __NR_mq_unlink          (__NR_Linux + 230)
+#define __NR_mq_timedsend       (__NR_Linux + 231)
+#define __NR_mq_timedreceive    (__NR_Linux + 232)
+#define __NR_mq_notify          (__NR_Linux + 233)
+#define __NR_mq_getsetattr      (__NR_Linux + 234)
 
-
-#define __NR_Linux_syscalls     228
+#define __NR_Linux_syscalls     235
 
 #define HPUX_GATEWAY_ADDR       0xC0000004
 #define LINUX_GATEWAY_ADDR      0x100
Index: ipc/compat_mq.c
===================================================================
RCS file: /var/cvs/linux-2.6/ipc/compat_mq.c,v
retrieving revision 1.2
diff -u -p -r1.2 compat_mq.c
--- a/ipc/compat_mq.c	15 Apr 2004 18:05:29 -0000	1.2
+++ b/ipc/compat_mq.c	28 Apr 2004 23:09:07 -0000
@@ -7,6 +7,7 @@
  */
 
 #include <linux/compat.h>
+#include <linux/compat_siginfo.h>
 #include <linux/fs.h>
 #include <linux/kernel.h>
 #include <linux/mqueue.h>

Re: [parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Wed, Apr 28, 2004 at 07:15:07PM -0400, Carlos O'Donell wrote:
> On Wed, Apr 28, 2004 at 05:13:38PM -0600, Carlos O'Donell wrote:
> > CVSROOT:	/var/cvs
> > Module name:	linux-2.6
> > Changes by:	carlos	04/04/28 17:13:38
> > 
> > Modified files:
> > 	arch/parisc/kernel: syscall_table.S 
> > 	include/asm-parisc: unistd.h 
> > 	ipc            : compat_mq.c 
> >
> > Log message:
> > First cut to get mq_* functions working in userspace. 
> > Fix CONFIG_POSIX_MQUEUE build. Connect mq_* syscalls. 
> > Make our syscall_table match unistd.h definitions. 
> > Boots on my A500, testing new glibc build as we speak.
> 
> Not too much glue required.
> 

Okay, glibc build completed, tests pass. We have functioning posix mq_*
support for hppa.

c.

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Sun, Apr 25, 2004 at 10:46:48PM -0600, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/04/25 22:46:48
> 
> Modified files:
> 	kernel         : compat_signal.c 
> 
> Log message:
> Add __SI_MESGQ case to compat (simple one liner)

Add the __SI_MESGQ case so that siginfo delivers the right data.

Index: kernel/compat_signal.c
===================================================================
RCS file: /var/cvs/linux-2.6/kernel/compat_signal.c,v
retrieving revision 1.2
retrieving revision 1.3
diff -u -p -r1.2 -r1.3
--- a/kernel/compat_signal.c	28 Jan 2004 05:12:46 -0000	1.2
+++ b/kernel/compat_signal.c	26 Apr 2004 04:46:48 -0000	1.3
@@ -102,6 +102,7 @@ int compat_copy_siginfo_to_user(compat_s
 		err |= __put_user(compat_from.si_stime, &to->si_stime);
 		break;
 	case __SI_RT: /* This is not generated by the kernel as of now. */
+	case __SI_MESGQ: /* But this is */
 		compat_from.si_pid = (compat_pid_t)(from->si_pid);
 		compat_from.si_uid = (__ARCH_SI_COMPAT_UID_T)(from->si_uid);
 		compat_from.si_int = (compat_int_t)(from->si_int);

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Tue, Feb 03, 2004 at 04:04:10PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/02/03 16:04:10
> 
> Modified files:
> 	include/linux  : compat.h 
> 
> Log message:
> Personality is a bitmask, so we use personality() to extract just the base personality before checking for PER_LINUX32 e.g. A compat task
> 

This booboo would have bitten eventually.

c.

Index: include/linux/compat.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/linux/compat.h,v
retrieving revision 1.7
retrieving revision 1.8
diff -u -p -r1.7 -r1.8
--- include/linux/compat.h      28 Jan 2004 05:12:46 -0000      1.7
+++ include/linux/compat.h      3 Feb 2004 23:04:09 -0000       1.8
@@ -23,7 +23,7 @@
 
 /* Non-native task requiring compat */
 #ifndef HAVE_ARCH_IS_COMPAT_TASK
-#define is_compat_task(x) (x->personality == PER_LINUX32)
+#define is_compat_task(x) (personality(x->personality) == PER_LINUX32)
 #endif
 
 struct compat_itimerspec { 

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Tue, Feb 03, 2004 at 02:51:13PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/02/03 14:51:13
> 
> Modified files:
> 	include/linux  : compat_siginfo.h 
> 
> Log message:
> Change compat_copy_siginfo_to_user to take the true argument values of the function.
> 
> _______________________________________________
> parisc-linux-cvs mailing list
> parisc-linux-cvs@lists.parisc-linux.org
> http://lists.parisc-linux.org/mailman/listinfo/parisc-linux-cvs

This matches what I've sent upstream to akpm.

Index: include/linux/compat_siginfo.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/linux/compat_siginfo.h,v
retrieving revision 1.2
diff -u -p -r1.2 compat_siginfo.h
--- include/linux/compat_siginfo.h      28 Jan 2004 05:12:46 -0000      1.2
+++ include/linux/compat_siginfo.h      3 Feb 2004 21:50:30 -0000
@@ -11,7 +11,8 @@
 typedef struct compat_siginfo{
 } compat_siginfo_t;
 
-static inline int compat_copy_siginfo_to_user(void *x, unsigned long y)
+static inline int compat_copy_siginfo_to_user(compat_siginfo_t __user *to, 
+                                               struct siginfo *from)
 {
        return -1;
 }

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Tue, Jan 27, 2004 at 10:14:15PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/01/27 22:14:15
> 
> Modified files:
> 	include/asm-parisc: posix_types.h 
> 
> Log message:
> __kernel_time_t should be of type long on 32-bit systems, it used to be int
> 

The generic code expects this to be of type long on all builds, even
32-bit. This might cause other warnings in parisc drivers that have
gotten used to having time_t as int.

Index: include/asm-parisc/posix_types.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-parisc/posix_types.h,v
retrieving revision 1.2
retrieving revision 1.3
diff -u -p -r1.2 -r1.3
--- include/asm-parisc/posix_types.h	28 Sep 2003 04:06:17 -0000	1.2
+++ include/asm-parisc/posix_types.h	28 Jan 2004 05:14:14 -0000	1.3
@@ -29,7 +29,7 @@ typedef long			__kernel_time_t;
 typedef unsigned int		__kernel_size_t;
 typedef int			__kernel_ssize_t;
 typedef int			__kernel_ptrdiff_t;
-typedef int			__kernel_time_t;
+typedef long			__kernel_time_t;
 #endif
 typedef char *			__kernel_caddr_t;
 

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Tue, Jan 27, 2004 at 10:12:47PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	04/01/27 22:12:47
> 
> Modified files:
> 	kernel         : signal.c compat_signal.c 
> 	include/linux  : compat_siginfo.h compat.h 
> 	include/asm-parisc: compat_ucontext.h compat_rt_sigframe.h 
> 	include/asm-generic: compat_signal.h 
> 
> Log message:
> Update the compat_siginfo_t framework.
> 

This patch is an update based on feedback from the linux-arch community.
Notably we add is_compat_task() and let the compiler optimize out the
code when CONFIG_COMPAT is not set. It also completes the changeover to
using include/linux/compat_siginfo.h instead of the asm-generic version,
which required that all arches add a compat_siginfo.h to their asm
directory.


Index: kernel/signal.c
===================================================================
RCS file: /var/cvs/linux-2.6/kernel/signal.c,v
retrieving revision 1.10
retrieving revision 1.11
diff -u -p -r1.10 -r1.11
--- kernel/signal.c	11 Dec 2003 00:48:55 -0000	1.10
+++ kernel/signal.c	28 Jan 2004 05:12:46 -0000	1.11
@@ -24,15 +24,11 @@
 #include <linux/binfmts.h>
 #include <linux/security.h>
 #include <linux/ptrace.h>
+#include <linux/compat_siginfo.h>
 #include <asm/param.h>
 #include <asm/uaccess.h>
 #include <asm/siginfo.h>
 
-#ifdef CONFIG_COMPAT
-#include <linux/personality.h>
-#include <asm/compat_siginfo.h>
-#endif
-
 /*
  * SLAB caches for signal bits.
  */
@@ -2011,12 +2007,11 @@ int copy_siginfo_to_user(siginfo_t __use
 	if (from->si_code < 0)
 		return __copy_to_user(to, from, sizeof(siginfo_t))
 			? -EFAULT : 0;
-# ifdef CONFIG_COMPAT
+	
 	/* Use compat_siginfo_t with 32-bit signals */
-	if(personality(current->personality) == PER_LINUX32){
+	if(is_compat_task(current)){
 		return compat_copy_siginfo_to_user((compat_siginfo_t __user *)to,from);
 	}
-# endif
 	
 	/*
 	 * If you change siginfo_t structure, please be sure
Index: kernel/compat_signal.c
===================================================================
RCS file: /var/cvs/linux-2.6/kernel/compat_signal.c,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -p -r1.1 -r1.2
--- kernel/compat_signal.c	10 Dec 2003 22:37:50 -0000	1.1
+++ kernel/compat_signal.c	28 Jan 2004 05:12:46 -0000	1.2
@@ -4,7 +4,7 @@
  *  2003-12-20  Carlos O'Donell
  *              Copied linux/kernel/compat_signal.c (copy_siginfo_to_user)
  *              and modified to use compat_siginfo_t for thunking down to
- *              32-bit userspace with an ELF64 kernel.
+ *              32-bit userspace from a 64-bit kernel.
  *              
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -19,14 +19,12 @@
  *
  * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
  *
  */
 
-/*#define __KERNEL_SYSCALLS__*/
-
+#include <linux/compat_siginfo.h>
 #include <asm/errno.h>
-#include <asm/compat_siginfo.h>
 #include <asm/uaccess.h>
 #include <asm/siginfo.h>
 
@@ -41,8 +39,8 @@ int compat_copy_siginfo_to_user(compat_s
 		return -EFAULT;
 	
 	/*
-	 * If you change siginfo_t structure, please be sure
-	 * this code is fixed accordingly.
+	 * If you change compat_siginfo_t structure *or* siginfo_t, 
+	 * please be sure this code is fixed accordingly.
 	 * It should never copy any pad contained in the structure
 	 * to avoid security leaks, but must copy the generic
 	 * 3 ints plus the relevant union member.
@@ -86,7 +84,7 @@ int compat_copy_siginfo_to_user(compat_s
 	case __SI_FAULT:
 		compat_from.si_addr = (compat_uptr_t)((u64)(from->si_addr) & 0xffffffffUL);
 		err |= __put_user(compat_from.si_addr, &to->si_addr);
-#ifdef __ARCH_SI_TRAPNO
+#ifdef __ARCH_SI_COMPAT_TRAPNO
 		compat_from.si_trapno = (compat_int_t)(from->si_addr);
 		err |= __put_user(compat_from.si_trapno, &to->si_trapno);
 #endif
Index: include/linux/compat_siginfo.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/linux/compat_siginfo.h,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -p -r1.1 -r1.2
--- include/linux/compat_siginfo.h	26 Jan 2004 21:09:14 -0000	1.1
+++ include/linux/compat_siginfo.h	28 Jan 2004 05:12:46 -0000	1.2
@@ -6,18 +6,18 @@
 
 #ifndef CONFIG_COMPAT
 
-/* No compatibility layer required, add empty definitions
-   for the compiler */
+/* No compatibility layer required, add empty definitions for the compiler */
 
 typedef struct compat_siginfo{
 } compat_siginfo_t;
 
 static inline int compat_copy_siginfo_to_user(void *x, unsigned long y)
 {
-	return 0;
+	return -1;
 }
 
 #else
+
 #include <linux/compiler.h>
 #include <asm/siginfo.h>
 
@@ -166,3 +166,4 @@ extern int compat_copy_siginfo_to_user(c
 
 #endif /* CONFIG_COMPAT */
 #endif /* _ASM_GENERIC_COMPAT_SIGINFO_H */
+
Index: include/linux/compat.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/linux/compat.h,v
retrieving revision 1.6
retrieving revision 1.7
diff -u -p -r1.6 -r1.7
--- include/linux/compat.h	27 Jan 2004 22:02:36 -0000	1.6
+++ include/linux/compat.h	28 Jan 2004 05:12:46 -0000	1.7
@@ -6,15 +6,26 @@
  */
 #include <linux/config.h>
 
-#ifdef CONFIG_COMPAT
+#ifndef CONFIG_COMPAT
+
+/* Non-native task requiring compat... doesn't exist */
+#define is_compat_task(x) 0
+
+#else
 
 #include <linux/stat.h>
 #include <linux/param.h>	/* for HZ */
+#include <linux/personality.h>  /* Conditional process compat */
 #include <asm/compat.h>
 
 #define compat_jiffies_to_clock_t(x)	\
 		(((unsigned long)(x) * COMPAT_USER_HZ) / HZ)
 
+/* Non-native task requiring compat */
+#ifndef HAVE_ARCH_IS_COMPAT_TASK
+#define is_compat_task(x) (x->personality == PER_LINUX32)
+#endif
+	
 struct compat_itimerspec { 
 	struct compat_timespec it_interval;
 	struct compat_timespec it_value;
Index: include/asm-parisc/compat_ucontext.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-parisc/compat_ucontext.h,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -p -r1.1 -r1.2
--- include/asm-parisc/compat_ucontext.h	10 Dec 2003 22:37:10 -0000	1.1
+++ include/asm-parisc/compat_ucontext.h	28 Jan 2004 05:12:47 -0000	1.2
@@ -4,7 +4,7 @@
 #include<linux/compat.h>
 #include<asm/compat_signal.h>
 
-/* ELF32 ucontext as seen from an ELF64 kernel */
+/* 32-bit ucontext as seen from an 64-bit kernel */
 struct compat_ucontext {
 	compat_uint_t uc_flags;
 	compat_uptr_t uc_link;
Index: include/asm-parisc/compat_rt_sigframe.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-parisc/compat_rt_sigframe.h,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -p -r1.1 -r1.2
--- include/asm-parisc/compat_rt_sigframe.h	10 Dec 2003 22:36:09 -0000	1.1
+++ include/asm-parisc/compat_rt_sigframe.h	28 Jan 2004 05:12:47 -0000	1.2
@@ -1,5 +1,5 @@
 #include<linux/compat.h>
-#include<asm/compat_siginfo.h>
+#include<linux/compat_siginfo.h>
 #include<asm/compat_ucontext.h>
 
 #ifndef _ASM_PARISC_COMPAT_RT_SIGFRAME_H
Index: include/asm-generic/compat_signal.h
===================================================================
RCS file: /var/cvs/linux-2.6/include/asm-generic/compat_signal.h,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -p -r1.1 -r1.2
--- include/asm-generic/compat_signal.h	10 Dec 2003 22:35:48 -0000	1.1
+++ include/asm-generic/compat_signal.h	28 Jan 2004 05:12:47 -0000	1.2
@@ -1,8 +1,8 @@
 #ifndef _ASM_GENERIC_COMPAT_SIGNAL_H
 #define _ASM_GENERIC_COMPAT_SIGNAL_H
 
-# ifndef __ASSEMBLY__
-#  include <linux/types.h>
+#ifndef __ASSEMBLY__
+#include <linux/compat.h>
 
 typedef compat_uptr_t compat_sighandler_t;
 
@@ -12,8 +12,6 @@ typedef struct compat_sigaltstack {
 	compat_size_t ss_size;
 } compat_stack_t;
 
-#  ifdef __KERNEL__
-
 /* Most things should be clean enough to redefine this at will, if care
    is taken to make libc match.  */
 
@@ -23,6 +21,5 @@ struct compat_sigaction {
 	compat_sigset_t sa_mask;		/* mask last for extensibility */
 };
 
-#  endif /* __KERNEL__ */
-# endif /* !__ASSEMBLY */
-#endif /* _ASM_GENERIC_COMPAT_SIGNAL_H */
+#endif /* !__ASSEMBLY__ */
+#endif /* !_ASM_GENERIC_COMPAT_SIGNAL_H */

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Wed, Dec 10, 2003 at 05:48:56PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	03/12/10 17:48:56
> 
> Modified files:
> 	include/asm-parisc: rt_sigframe.h 
> 	kernel         : Makefile signal.c 
> 
> Log message:
> Wrap the signal compat in CONFIG_COMPAT so 32-bit builds work properly.

I assumed CONFIG_COMPAT=Y for all the scenarios, and when Helge said his
32-bit build didn't work I had to slap my forehead.

So this patch removes all the compat signal code when doing a 32-bit
compile.

c.

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Wed, Dec 10, 2003 at 03:40:26PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	03/12/10 15:40:25
> 
> Modified files:
> 	include/asm-parisc: compat.h rt_sigframe.h siginfo.h ucontext.h 
> 
> Log message:
> Cleanup for non-native thread signals, and ensure that during 64-bit builds everything works too.

You may have noticed the following:

  8 #define SIGRETURN_TRAMP 4
  9 #define SIGRESTARTBLOCK_TRAMP 5 
 10 #define TRAMP_SIZE (SIGRETURN_TRAMP + SIGRESTARTBLOCK_TRAMP)
 11 
 12 struct rt_sigframe {
 13         /* XXX: Must match trampoline size in arch/parisc/kernel/signal.c 
 14                 Secondary to that it must protect the ERESTART_RESTARTBLOCK
 15                 trampoline we left on the stack (we were bad and didn't 
 16                 change sp so we could run really fast.) */
 17         unsigned int tramp[TRAMP_SIZE];
 18         struct siginfo info;
 19         struct ucontext uc;
 20 };

And wondered "Why do people write XXX: followed by some scary sounding
stuff?" Here is the part where we clarify what this means.

When Randolph wrote the ERESTART_RESTARTBLOCK code he used a stack
trampoline to rewrite the syscall entry into the kernel with a different
syscall number. This stack trampoline is *unprotected* and the stack
pointer is not adjusted.

If one were to take a signal while executing the ERESTART_RESTARTBLOCK
stack trampoline the signal's frame would destroy the trampoline. Instead
we *notch* 5 more words out of the front of the signal frame to protect a
possible ERESTART_RESTARTBLOCK trampoline. The signal code used to do
exactly the same thing before, it had it's own notch for the sigreturn
trampoline on the stack. Thus now we protect both trampolines.

What if you took another ERESTART_RESTARTBLOCK signal while walking the
last ERESTART_RESTARTBLOCK trampoline? Well that scenario isn't allowed
to happen, and according to Linux we have to disable
ERESTART_RESTARTBLOCK until we are back in the kernel handling the
restart. We still haven't fixed this last bit.

c.

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Joel Soete]

Just because it still failed to boot on my b2k (Suckio pb), may I ask 
you your .config file (if it is not the defconfig) and also which 64bit 
compiler did you used.

Thanks again for additional help,
	Joel

Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	03/11/08 14:00:57
> 
> Modified files:
> 	.              : Makefile 
> 
> Log message:
> Trap handling changes made. Boots on A500, 64-bit kernel. Bumped to -pa5
> 
> _______________________________________________
> parisc-linux-cvs mailing list
> parisc-linux-cvs@lists.parisc-linux.org
> http://lists.parisc-linux.org/mailman/listinfo/parisc-linux-cvs
> 

Re: [parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Sat, Nov 08, 2003 at 09:16:38PM +0000, Joel Soete wrote:
> Just because it still failed to boot on my b2k (Suckio pb), may I ask 
> you your .config file (if it is not the defconfig) and also which 64bit 
> compiler did you used.

Custom built 64-bit cross based on 3.3.1 release and upstream binutils.
Adding your fix for the suckio pb right now.

c.

[parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Sat, Nov 08, 2003 at 01:59:55PM -0700, Carlos O'Donell wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.6
> Changes by:	carlos	03/11/08 13:59:55
> 
> Modified files:
> 	arch/parisc/kernel: traps.c 
> 
> Log message:
> Cleanup trap handling for codes 13,15,16 and 17. In particular forward port 13, and make 17 non-fatal to the kernel (only the process should die).

tausq,

Here's the trap handler cleanup I promised. Please comment.

c.

Index: Makefile
===================================================================
RCS file: /var/cvs/linux-2.6/Makefile,v
retrieving revision 1.83
diff -u -p -r1.83 Makefile
--- Makefile	2 Nov 2003 04:40:49 -0000	1.83
+++ Makefile	8 Nov 2003 20:53:32 -0000
@@ -1,7 +1,7 @@
 VERSION = 2
 PATCHLEVEL = 6
 SUBLEVEL = 0
-EXTRAVERSION = -test9-pa4
+EXTRAVERSION = -test9-pa5
 
 # *DOCUMENTATION*
 # To see a list of typical targets execute "make help"
Index: arch/parisc/kernel/traps.c
===================================================================
RCS file: /var/cvs/linux-2.6/arch/parisc/kernel/traps.c,v
retrieving revision 1.10
diff -u -p -r1.10 traps.c
--- arch/parisc/kernel/traps.c	8 Oct 2003 20:52:07 -0000	1.10
+++ arch/parisc/kernel/traps.c	8 Nov 2003 20:53:35 -0000
@@ -557,20 +557,41 @@ void handle_interruption(int code, struc
 		si.si_addr = (void *) regs->iaoq[0];
 		force_sig_info(SIGFPE, &si, current);
 		return;
-
+		
+	case 13:
+		/* Conditional Trap
+		   The condition succees in an instruction which traps 
+		   on condition  */
+		si.si_signo = SIGFPE;
+		/* Set to zero, and let the userspace app figure it out from
+		   the insn pointed to by si_addr */
+		si.si_code = 0;
+		si.si_addr = (void *) regs->iaoq[0];
+		force_sig_info(SIGFPE, &si, current);
+		return;
+		
 	case 14:
 		/* Assist Exception Trap, i.e. floating point exception. */
 		die_if_kernel("Floating point exception", regs, 0); /* quiet */
 		handle_fpe(regs);
 		return;
-
+		
+	case 15:
+		/* Data TLB miss fault/Data page fault */
+		/* Fall through */
+	case 16:
+		/* Non-access instruction TLB miss fault */
+		/* The instruction TLB entry needed for the target address of the FIC
+		   is absent, and hardware can't find it, so we get to cleanup */
+		/* Fall through */
 	case 17:
 		/* Non-access data TLB miss fault/Non-access data page fault */
 		/* TODO: Still need to add slow path emulation code here */
-		pdc_chassis_send_status(PDC_CHASSIS_DIRECT_PANIC);
-		
+		/* TODO: Understand what is meant by the TODO listed
+		         above this one. (Carlos) */
 		fault_address = regs->ior;
-		parisc_terminate("Non access data tlb fault!",regs,code,fault_address);
+		fault_space = regs->isr;
+		break;
 
 	case 18:
 		/* PCXS only -- later cpu's split this into types 26,27 & 28 */
@@ -580,9 +601,8 @@ void handle_interruption(int code, struc
 			return;
 		}
 		/* Fall Through */
-
-	case 15: /* Data TLB miss fault/Data page fault */
-	case 26: /* PCXL: Data memory access rights trap */
+	case 26: 
+		/* PCXL: Data memory access rights trap */
 		fault_address = regs->ior;
 		fault_space   = regs->isr;
 		break;
@@ -638,7 +658,6 @@ void handle_interruption(int code, struc
 			up_read(&current->mm->mmap_sem);
 		}
 		/* Fall Through */
-
 	case 27: 
 		/* Data memory protection ID trap */
 		die_if_kernel("Protection id trap", regs, code);

Re: [parisc-linux] Re: [parisc-linux-cvs] linux-2.6 carlos

[Carlos O'Donell]

On Sat, Nov 08, 2003 at 09:44:23PM +0000, Joel Soete wrote:
> +       case 13:
> +               if (user_mode(regs)) {
> +#ifdef PRINT_USER_FAULTS
> +                       printk(KERN_DEBUG "\nhandle_interruption() id=%d 
> command='%s'\n",
> +                           current->pid, current->comm);
> +                       show_regs(regs);
> +#endif
> +                       /* Conditional Trap
> +                          The condition succees in an instruction which 
> traps on condition  */
> +                       si.si_signo = SIGFPE;
> [...]
> +                       return;
> +               } else break;
> 
> (sorry for bad wrap but I also join the patch I suggested)
> 
> But you can prefer to ignore as it is just a work around and that I 
> don't yet see what other solution could be bring to such pb?

Adding the check for user_mode. Thanks Joel!
It seems that the kernel tries to use 'addi,tc' and a few other whichs
might generate a trap.

c.