[parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[James Bottomley]

On Thu, 2003-03-27 at 23:47, James Bottomley wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.5
> Changes by:	jejb	03/03/27 22:47:02
> 
> Modified files:
> 	include/asm-parisc: pgtable.h 
> 
> Log message:
> Add support for file-offsets-in-pte's
> 
> See http://mail.nl.linux.org/linux-mm/2003-03/msg00174.html

===== include/asm-parisc/pgtable.h 1.11 vs edited =====
--- 1.11/include/asm-parisc/pgtable.h	Thu Mar  6 10:19:09 2003
+++ edited/include/asm-parisc/pgtable.h	Thu Mar 27 21:42:26 2003
@@ -122,6 +122,7 @@
 #define _PAGE_GATEWAY_BIT  28   /* (0x008) privilege promotion allowed */
 #define _PAGE_DMB_BIT      27   /* (0x010) Data Memory Break enable (B bit) */
 #define _PAGE_DIRTY_BIT    26   /* (0x020) Page Dirty (D bit) */
+#define _PAGE_FILE_BIT	_PAGE_DIRTY_BIT	/* overload this bit */
 #define _PAGE_REFTRAP_BIT  25   /* (0x040) Page Ref. Trap enable (T bit) */
 #define _PAGE_NO_CACHE_BIT 24   /* (0x080) Uncached Page (U bit) */
 #define _PAGE_ACCESSED_BIT 23   /* (0x100) Software: Page Accessed */
@@ -135,6 +136,17 @@
 
 #define xlate_pabit(x) (31 - x)
 
+/* this defines the shift to the usable bits in the PTE it is set so
+ * that the valid bits _PAGE_PRESENT_BIT and _PAGE_USER_BIT are set
+ * to zero */
+#define PTE_SHIFT	   	xlate_pabit(_PAGE_USER_BIT)
+
+/* this is how many bits may be used by the file functions */
+#define PTE_FILE_MAX_BITS	(BITS_PER_LONG - PTE_SHIFT)
+
+#define pte_to_pgoff(pte) (pte_val(pte) >> PTE_SHIFT)
+#define pgoff_to_pte(off) ((pte_t) { ((off) << PTE_SHIFT) | _PAGE_FILE })
+
 #define _PAGE_READ     (1 << xlate_pabit(_PAGE_READ_BIT))
 #define _PAGE_WRITE    (1 << xlate_pabit(_PAGE_WRITE_BIT))
 #define _PAGE_RW       (_PAGE_READ | _PAGE_WRITE)
@@ -148,6 +160,7 @@
 #define _PAGE_PRESENT  (1 << xlate_pabit(_PAGE_PRESENT_BIT))
 #define _PAGE_FLUSH    (1 << xlate_pabit(_PAGE_FLUSH_BIT))
 #define _PAGE_USER     (1 << xlate_pabit(_PAGE_USER_BIT))
+#define _PAGE_FILE     (1 << xlate_pabit(_PAGE_FILE_BIT))
 
 #define _PAGE_TABLE	(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE |  _PAGE_DIRTY | _PAGE_ACCESSED)
 #define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
@@ -256,6 +269,7 @@
 extern inline int pte_dirty(pte_t pte)		{ return pte_val(pte) & _PAGE_DIRTY; }
 extern inline int pte_young(pte_t pte)		{ return pte_val(pte) & _PAGE_ACCESSED; }
 extern inline int pte_write(pte_t pte)		{ return pte_val(pte) & _PAGE_WRITE; }
+extern inline int pte_file(pte_t pte)		{ return pte_val(pte) & _PAGE_FILE; }
 
 extern inline pte_t pte_rdprotect(pte_t pte)	{ pte_val(pte) &= ~_PAGE_READ; return pte; }
 extern inline pte_t pte_mkclean(pte_t pte)	{ pte_val(pte) &= ~_PAGE_DIRTY; return pte; }

[parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[Carlos O'Donell]

> CVSROOT:	/var/cvs
> Module name:	linux-2.5
> Changes by:	jejb	03/02/16 11:10:21
> 
> Modified files:
> 	arch/parisc/kernel: signal.c sys32.h 
> 	include/asm-parisc: signal.h 
> 
> Log message:
> More signal change fixes

Thanks for all the great work James!

Just to comment on this, we have incorrect sigcontext's from 64-bit kernels.
If you look at the usage of "HACK" you'll see that the 32-bit userspace 
gets some rather unusable 64-bit values in the sigcontext. I'm not sure how 
other arches fix this? Cut off the top 32 bits?

This item is on my personal todo list, but I haven't talked to anyone who has 
suggested a solution. I'm just a lowly GNU/Libc hacker, but if you suggest 
something I _will_ write it. 

c.

Re: [parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[James Bottomley]

On Mon, 2003-02-17 at 10:57, Carlos O'Donell wrote:
> > CVSROOT:	/var/cvs
> > Module name:	linux-2.5
> > Changes by:	jejb	03/02/16 11:10:21
> > 
> > Modified files:
> > 	arch/parisc/kernel: signal.c sys32.h 
> > 	include/asm-parisc: signal.h 
> > 
> > Log message:
> > More signal change fixes
> 
> Thanks for all the great work James!

You're welcome, but these were really just compile fixes.

> Just to comment on this, we have incorrect sigcontext's from 64-bit kernels.
> If you look at the usage of "HACK" you'll see that the 32-bit userspace 
> gets some rather unusable 64-bit values in the sigcontext. I'm not sure how 
> other arches fix this? Cut off the top 32 bits?
> 
> This item is on my personal todo list, but I haven't talked to anyone who has 
> suggested a solution. I'm just a lowly GNU/Libc hacker, but if you suggest 
> something I _will_ write it. 

In general, we should probably be relying more on the generic 32/64
compatibility layer recently introduced.  I suspect the best role model
for all of this is sparc64, since that's habitually run as a 64 bit
kernel with (mostly) 32 bit user space.

The HACK thing should continue to work OK as long as PA-RISC has no 64
bit user binary support, though.

James

Re: [parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[Carlos O'Donell]

> > This item is on my personal todo list, but I haven't talked to anyone who has 
> > suggested a solution. I'm just a lowly GNU/Libc hacker, but if you suggest 
> > something I _will_ write it. 
> 
> In general, we should probably be relying more on the generic 32/64
> compatibility layer recently introduced.  I suspect the best role model
> for all of this is sparc64, since that's habitually run as a 64 bit
> kernel with (mostly) 32 bit user space.

The sparc64 code is quite frightening, but I will look to it for a role
model.
 
> The HACK thing should continue to work OK as long as PA-RISC has no 64
> bit user binary support, though.

Yeah, it's OK until I attempt to write make/get/set/swap/context support
in libc :)

c.

[parisc-linux] Re: [parisc-linux-cvs] linux-2.5 jejb

[James Bottomley]

On Mon, 2003-02-10 at 18:49, James Bottomley wrote:
> CVSROOT:	/var/cvs
> Module name:	linux-2.5
> Changes by:	jejb	03/02/10 17:49:17
> 
> Modified files:
> 	arch/parisc/kernel: init_task.c signal.c 
> 	include/asm-parisc: ptrace.h 
> 
> Log message:
> Update for new kernel signal/sigaction split to replace sig

I should add a warning here that doing this has eliminated the pa-risc
private PTRACE_SETSIGINFO and PTRACE_GETSIGINFO definitions in favour of
the newly minted global kernel ones in linux/ptrace.h

Unfortunately, the values have changed, so some user space tracing tools
that relied on these may have to change.

I think they were unused previously, so all may be well, but someone who
knows this area should check.

James