[patch 0/6] s390 bug fixes for 2.6.29

[patch 0/6] s390 bug fixes for 2.6.29

[Martin Schwidefsky]

Greetings,
some of the bug fixes in my feature patch pile for the next merge
window are severe enough to warrent inclusion in 2.6.29.
There are 6 relevant patches:

Gerald Schaefer (1):
      [S390] Dont check for pfn_valid() in uaccess_pt.c

Heiko Carstens (2):
      [S390] topology: define SD_MC_INIT to fix performance regression
      [S390] ftrace/mcount: fix kernel stack backchain

Martin Schwidefsky (3):
      [S390] __div64_31 broken for CONFIG_MARCH_G5
      [S390] make page table walking more robust
      [S390] make page table upgrade work again

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 1/6] __div64_31 broken for CONFIG_MARCH_G5

[Martin Schwidefsky]

[-- Attachment #1: 001-div64.diff --]
[-- Type: text/plain, Size: 1211 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

The implementation of __div64_31 for G5 machines is broken. The comments
in __div64_31 are correct, only the code does not do what the comments
say. The part "If the remainder has overflown subtract base and increase
the quotient" is only partially realized, the base is subtracted correctly
but the quotient is only increased if the dividend had the last bit set.
Using the correct instruction fixes the problem.

Cc: stable@kernel.org
Reported-by: Frans Pop <elendil@planet.nl>
Tested-by: Frans Pop <elendil@planet.nl>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/lib/div64.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: quilt-2.6/arch/s390/lib/div64.c
===================================================================
--- quilt-2.6.orig/arch/s390/lib/div64.c
+++ quilt-2.6/arch/s390/lib/div64.c
@@ -61,7 +61,7 @@ static uint32_t __div64_31(uint64_t *n, 
 		"	clr	%0,%3\n"
 		"	jl	0f\n"
 		"	slr	%0,%3\n"
-		"	alr	%1,%2\n"
+		"	ahi	%1,1\n"
 		"0:\n"
 		: "+d" (reg2), "+d" (reg3), "=d" (tmp)
 		: "d" (base), "2" (1UL) : "cc" );

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 2/6] topology: define SD_MC_INIT to fix performance regression

[Martin Schwidefsky]

[-- Attachment #1: 002-sd-mc-init.diff --]
[-- Type: text/plain, Size: 942 bytes --]

From: Heiko Carstens <heiko.carstens@de.ibm.com>

The default values for SD_MC_INIT cause an additional cpu usage of up
to 40% on some network benchmarks compared to the plain SD_CPU_INIT
values. So just define SD_MC_INIT to SD_CPU_INIT.
More tuning needs to be done.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/include/asm/topology.h |    2 ++
 1 file changed, 2 insertions(+)

Index: quilt-2.6/arch/s390/include/asm/topology.h
===================================================================
--- quilt-2.6.orig/arch/s390/include/asm/topology.h
+++ quilt-2.6/arch/s390/include/asm/topology.h
@@ -30,6 +30,8 @@ static inline void s390_init_cpu_topolog
 };
 #endif
 
+#define SD_MC_INIT SD_CPU_INIT
+
 #include <asm-generic/topology.h>
 
 #endif /* _ASM_S390_TOPOLOGY_H */

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 3/6] ftrace/mcount: fix kernel stack backchain

[Martin Schwidefsky]

[-- Attachment #1: 003-ftrace-mcount.diff --]
[-- Type: text/plain, Size: 1147 bytes --]

From: Heiko Carstens <heiko.carstens@de.ibm.com>

With packed stack the backchain is at a different location.
Just use __SF_BACKCHAIN as an offset to store the backchain.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/kernel/mcount.S |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

Index: quilt-2.6/arch/s390/kernel/mcount.S
===================================================================
--- quilt-2.6.orig/arch/s390/kernel/mcount.S
+++ quilt-2.6/arch/s390/kernel/mcount.S
@@ -5,6 +5,8 @@
  *
  */
 
+#include <asm/asm-offsets.h>
+
 #ifndef CONFIG_64BIT
 .globl _mcount
 _mcount:
@@ -14,7 +16,7 @@ _mcount:
 	ahi	%r15,-96
 	l	%r3,100(%r15)
 	la	%r2,0(%r14)
-	st	%r1,0(%r15)
+	st	%r1,__SF_BACKCHAIN(%r15)
 	la	%r3,0(%r3)
 	bras	%r14,0f
 	.long	ftrace_trace_function
@@ -38,7 +40,7 @@ _mcount:
 	stg	%r14,112(%r15)
 	lgr	%r1,%r15
 	aghi	%r15,-160
-	stg	%r1,0(%r15)
+	stg	%r1,__SF_BACKCHAIN(%r15)
 	lgr	%r2,%r14
 	lg	%r3,168(%r15)
 	larl	%r14,ftrace_trace_function

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 4/6] Dont check for pfn_valid() in uaccess_pt.c

[Martin Schwidefsky]

[-- Attachment #1: 004-valid.diff --]
[-- Type: text/plain, Size: 2332 bytes --]

From: Gerald Schaefer <gerald.schaefer@de.ibm.com>

pfn_valid() actually checks for a valid struct page and not for a
valid pfn. Using xip mappings w/o struct pages, this will result in
-EFAULT returned by the (page table walk) user copy functions,
even though there is valid memory. Those user copy functions don't
need a struct page, so this patch just removes the pfn_valid() check.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/lib/uaccess_pt.c |   18 ------------------
 1 file changed, 18 deletions(-)

Index: quilt-2.6/arch/s390/lib/uaccess_pt.c
===================================================================
--- quilt-2.6.orig/arch/s390/lib/uaccess_pt.c
+++ quilt-2.6/arch/s390/lib/uaccess_pt.c
@@ -119,8 +119,6 @@ retry:
 			goto fault;
 
 		pfn = pte_pfn(*pte);
-		if (!pfn_valid(pfn))
-			goto out;
 
 		offset = uaddr & (PAGE_SIZE - 1);
 		size = min(n - done, PAGE_SIZE - offset);
@@ -135,7 +133,6 @@ retry:
 		done += size;
 		uaddr += size;
 	} while (done < n);
-out:
 	spin_unlock(&mm->page_table_lock);
 	return n - done;
 fault:
@@ -163,9 +160,6 @@ retry:
 		goto fault;
 
 	pfn = pte_pfn(*pte);
-	if (!pfn_valid(pfn))
-		goto out;
-
 	ret = (pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE - 1));
 out:
 	return ret;
@@ -244,11 +238,6 @@ retry:
 			goto fault;
 
 		pfn = pte_pfn(*pte);
-		if (!pfn_valid(pfn)) {
-			done = -1;
-			goto out;
-		}
-
 		offset = uaddr & (PAGE_SIZE-1);
 		addr = (char *)(pfn << PAGE_SHIFT) + offset;
 		len = min(count - done, PAGE_SIZE - offset);
@@ -256,7 +245,6 @@ retry:
 		done += len_str;
 		uaddr += len_str;
 	} while ((len_str == len) && (done < count));
-out:
 	spin_unlock(&mm->page_table_lock);
 	return done + 1;
 fault:
@@ -325,12 +313,7 @@ retry:
 		}
 
 		pfn_from = pte_pfn(*pte_from);
-		if (!pfn_valid(pfn_from))
-			goto out;
 		pfn_to = pte_pfn(*pte_to);
-		if (!pfn_valid(pfn_to))
-			goto out;
-
 		offset_from = uaddr_from & (PAGE_SIZE-1);
 		offset_to = uaddr_from & (PAGE_SIZE-1);
 		offset_max = max(offset_from, offset_to);
@@ -342,7 +325,6 @@ retry:
 		uaddr_from += size;
 		uaddr_to += size;
 	} while (done < n);
-out:
 	spin_unlock(&mm->page_table_lock);
 	return n - done;
 fault:

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 5/6] make page table walking more robust

[Martin Schwidefsky]

[-- Attachment #1: 005-page-walk.diff --]
[-- Type: text/plain, Size: 3112 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Make page table walking on s390 more robust. The current code requires
that the pgd/pud/pmd/pte loop is only done for address ranges that are
below the end address of the last vma of the address space. But this
is not always true, e.g. the generic page table walker does not guarantee
this. Change TASK_SIZE/TASK_SIZE_OF to reflect the current size of the
address space. This makes the generic page table walker happy but it
breaks the upgrade of a 3 level page table to a 4 level page table.
To make the upgrade work again another fix is required.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/include/asm/processor.h |    5 ++---
 arch/s390/mm/mmap.c               |    4 ++--
 arch/s390/mm/pgtable.c            |    2 ++
 3 files changed, 6 insertions(+), 5 deletions(-)

Index: quilt-2.6/arch/s390/include/asm/processor.h
===================================================================
--- quilt-2.6.orig/arch/s390/include/asm/processor.h
+++ quilt-2.6/arch/s390/include/asm/processor.h
@@ -61,7 +61,7 @@ extern void print_cpu_info(struct cpuinf
 extern int get_cpu_capability(unsigned int *);
 
 /*
- * User space process size: 2GB for 31 bit, 4TB for 64 bit.
+ * User space process size: 2GB for 31 bit, 4TB or 8PT for 64 bit.
  */
 #ifndef __s390x__
 
@@ -70,8 +70,7 @@ extern int get_cpu_capability(unsigned i
 
 #else /* __s390x__ */
 
-#define TASK_SIZE_OF(tsk)	(test_tsk_thread_flag(tsk,TIF_31BIT) ? \
-					(1UL << 31) : (1UL << 53))
+#define TASK_SIZE_OF(tsk)	((tsk)->mm->context.asce_limit)
 #define TASK_UNMAPPED_BASE	(test_thread_flag(TIF_31BIT) ? \
 					(1UL << 30) : (1UL << 41))
 #define TASK_SIZE		TASK_SIZE_OF(current)
Index: quilt-2.6/arch/s390/mm/mmap.c
===================================================================
--- quilt-2.6.orig/arch/s390/mm/mmap.c
+++ quilt-2.6/arch/s390/mm/mmap.c
@@ -35,7 +35,7 @@
  * Leave an at least ~128 MB hole.
  */
 #define MIN_GAP (128*1024*1024)
-#define MAX_GAP (TASK_SIZE/6*5)
+#define MAX_GAP (STACK_TOP/6*5)
 
 static inline unsigned long mmap_base(void)
 {
@@ -46,7 +46,7 @@ static inline unsigned long mmap_base(vo
 	else if (gap > MAX_GAP)
 		gap = MAX_GAP;
 
-	return TASK_SIZE - (gap & PAGE_MASK);
+	return STACK_TOP - (gap & PAGE_MASK);
 }
 
 static inline int mmap_is_legacy(void)
Index: quilt-2.6/arch/s390/mm/pgtable.c
===================================================================
--- quilt-2.6.orig/arch/s390/mm/pgtable.c
+++ quilt-2.6/arch/s390/mm/pgtable.c
@@ -117,6 +117,7 @@ repeat:
 		crst_table_init(table, entry);
 		pgd_populate(mm, (pgd_t *) table, (pud_t *) pgd);
 		mm->pgd = (pgd_t *) table;
+		mm->task_size = mm->context.asce_limit;
 		table = NULL;
 	}
 	spin_unlock(&mm->page_table_lock);
@@ -154,6 +155,7 @@ void crst_table_downgrade(struct mm_stru
 			BUG();
 		}
 		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
+		mm->task_size = mm->context.asce_limit;
 		crst_table_free(mm, (unsigned long *) pgd);
 	}
 	update_mm(mm, current);

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.

[patch 6/6] make page table upgrade work again

[Martin Schwidefsky]

[-- Attachment #1: 006-large-map.diff --]
[-- Type: text/plain, Size: 3739 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

After TASK_SIZE now gives the current size of the address space the
upgrade of a 64 bit process from 3 to 4 levels of page table  needs
to use the arch_mmap_check hook to catch large mmap lengths. The
get_unmapped_area* functions need to check for -ENOMEM from the
arch_get_unmapped_area*, upgrade the page table and retry.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

 arch/s390/include/asm/mman.h |    5 ++++
 arch/s390/mm/mmap.c          |   44 +++++++++++++++++++++++++++++--------------
 2 files changed, 35 insertions(+), 14 deletions(-)

Index: quilt-2.6/arch/s390/include/asm/mman.h
===================================================================
--- quilt-2.6.orig/arch/s390/include/asm/mman.h
+++ quilt-2.6/arch/s390/include/asm/mman.h
@@ -22,4 +22,9 @@
 #define MCL_CURRENT	1		/* lock all current mappings */
 #define MCL_FUTURE	2		/* lock all future mappings */
 
+#if defined(__KERNEL__) && !defined(__ASSEMBLY__) && defined(CONFIG_64BIT)
+int s390_mmap_check(unsigned long addr, unsigned long len);
+#define arch_mmap_check(addr,len,flags)	s390_mmap_check(addr,len)
+#endif
+
 #endif /* __S390_MMAN_H__ */
Index: quilt-2.6/arch/s390/mm/mmap.c
===================================================================
--- quilt-2.6.orig/arch/s390/mm/mmap.c
+++ quilt-2.6/arch/s390/mm/mmap.c
@@ -89,42 +89,58 @@ EXPORT_SYMBOL_GPL(arch_pick_mmap_layout)
 
 #else
 
+int s390_mmap_check(unsigned long addr, unsigned long len)
+{
+	if (!test_thread_flag(TIF_31BIT) &&
+	    len >= TASK_SIZE && TASK_SIZE < (1UL << 53))
+		return crst_table_upgrade(current->mm, 1UL << 53);
+	return 0;
+}
+
 static unsigned long
 s390_get_unmapped_area(struct file *filp, unsigned long addr,
 		unsigned long len, unsigned long pgoff, unsigned long flags)
 {
 	struct mm_struct *mm = current->mm;
+	unsigned long area;
 	int rc;
 
-	addr = arch_get_unmapped_area(filp, addr, len, pgoff, flags);
-	if (addr & ~PAGE_MASK)
-		return addr;
-	if (unlikely(mm->context.asce_limit < addr + len)) {
-		rc = crst_table_upgrade(mm, addr + len);
+	area = arch_get_unmapped_area(filp, addr, len, pgoff, flags);
+	if (!(area & ~PAGE_MASK))
+		return area;
+	if (area == -ENOMEM &&
+	    !test_thread_flag(TIF_31BIT) && TASK_SIZE < (1UL << 53)) {
+		/* Upgrade the page table to 4 levels and retry. */
+		rc = crst_table_upgrade(mm, 1UL << 53);
 		if (rc)
 			return (unsigned long) rc;
+		area = arch_get_unmapped_area(filp, addr, len, pgoff, flags);
 	}
-	return addr;
+	return area;
 }
 
 static unsigned long
-s390_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
+s390_get_unmapped_area_topdown(struct file *filp, const unsigned long addr,
 			  const unsigned long len, const unsigned long pgoff,
 			  const unsigned long flags)
 {
 	struct mm_struct *mm = current->mm;
-	unsigned long addr = addr0;
+	unsigned long area;
 	int rc;
 
-	addr = arch_get_unmapped_area_topdown(filp, addr, len, pgoff, flags);
-	if (addr & ~PAGE_MASK)
-		return addr;
-	if (unlikely(mm->context.asce_limit < addr + len)) {
-		rc = crst_table_upgrade(mm, addr + len);
+	area = arch_get_unmapped_area_topdown(filp, addr, len, pgoff, flags);
+	if (!(area & ~PAGE_MASK))
+		return area;
+	if (area == -ENOMEM &&
+	    !test_thread_flag(TIF_31BIT) && TASK_SIZE < (1UL << 53)) {
+		/* Upgrade the page table to 4 levels and retry. */
+		rc = crst_table_upgrade(mm, 1UL << 53);
 		if (rc)
 			return (unsigned long) rc;
+		area = arch_get_unmapped_area_topdown(filp, addr, len,
+						      pgoff, flags);
 	}
-	return addr;
+	return area;
 }
 /*
  * This function, called very early during the creation of a new

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.