From 618a3c03fcfdf1ac4543247c8ddfb0c9d775ff33 Mon Sep 17 00:00:00 2001
From: Hal Rosenstock <halr@voltaire.com>
Date: Tue, 28 Mar 2006 16:40:04 -0800
Subject: IB/mad: RMPP support for additional classes

Add RMPP support for additional management classes that support it.
Also, validate RMPP is consistent with management class specified.

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 include/rdma/ib_mad.h | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

(limited to 'include')

diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h
index 51ab8eddb295..5ff77558013b 100644
--- a/include/rdma/ib_mad.h
+++ b/include/rdma/ib_mad.h
@@ -3,7 +3,7 @@
  * Copyright (c) 2004 Infinicon Corporation.  All rights reserved.
  * Copyright (c) 2004 Intel Corporation.  All rights reserved.
  * Copyright (c) 2004 Topspin Corporation.  All rights reserved.
- * Copyright (c) 2004 Voltaire Corporation.  All rights reserved.
+ * Copyright (c) 2004-2006 Voltaire Corporation.  All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU
@@ -55,6 +55,10 @@
 #define IB_MGMT_CLASS_DEVICE_MGMT		0x06
 #define IB_MGMT_CLASS_CM			0x07
 #define IB_MGMT_CLASS_SNMP			0x08
+#define IB_MGMT_CLASS_DEVICE_ADM		0x10
+#define IB_MGMT_CLASS_BOOT_MGMT			0x11
+#define IB_MGMT_CLASS_BIS			0x12
+#define IB_MGMT_CLASS_CONG_MGMT			0x21
 #define IB_MGMT_CLASS_VENDOR_RANGE2_START	0x30
 #define IB_MGMT_CLASS_VENDOR_RANGE2_END		0x4F
 
@@ -117,6 +121,8 @@ enum {
 	IB_MGMT_VENDOR_DATA = 216,
 	IB_MGMT_SA_HDR = 56,
 	IB_MGMT_SA_DATA = 200,
+	IB_MGMT_DEVICE_HDR = 64,
+	IB_MGMT_DEVICE_DATA = 192,
 };
 
 struct ib_mad_hdr {
@@ -602,6 +608,25 @@ struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent,
 					    int hdr_len, int data_len,
 					    gfp_t gfp_mask);
 
+/**
+ * ib_is_mad_class_rmpp - returns whether given management class
+ * supports RMPP.
+ * @mgmt_class: management class
+ *
+ * This routine returns whether the management class supports RMPP.
+ */
+int ib_is_mad_class_rmpp(u8 mgmt_class);
+
+/**
+ * ib_get_mad_data_offset - returns the data offset for a given
+ * management class.
+ * @mgmt_class: management class
+ *
+ * This routine returns the data offset in the MAD for the management
+ * class requested.
+ */
+int ib_get_mad_data_offset(u8 mgmt_class);
+
 /**
  * ib_get_rmpp_segment - returns the data buffer for a given RMPP segment.
  * @send_buf: Previously allocated send data buffer.
-- 
cgit v1.2.3


From 3283a67d8618c9a292eced23e8753ab64adc6dba Mon Sep 17 00:00:00 2001
From: Jes Sorensen <jes@sgi.com>
Date: Thu, 30 Mar 2006 10:13:22 -0500
Subject: [IA64] Add __mca_table to the DISCARD list in gate.lds

Add __mca_table to the DISCARD list for the gate.lds linker script to
avoid broken linker references when linking the final vmlinux file.

Also add comment to include/asm-ia64/asmmacros.h to avoid anyone else
hitting this problem in the future.

Credits to James Bottomley <James.Bottomley@SteelEye.com> for spotting
the DISCARD list in gate.lds.S

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 arch/ia64/kernel/gate.lds.S | 1 +
 include/asm-ia64/asmmacro.h | 4 ++++
 2 files changed, 5 insertions(+)

(limited to 'include')

diff --git a/arch/ia64/kernel/gate.lds.S b/arch/ia64/kernel/gate.lds.S
index e1e4aba9ecd0..7c99e6ec3daf 100644
--- a/arch/ia64/kernel/gate.lds.S
+++ b/arch/ia64/kernel/gate.lds.S
@@ -59,6 +59,7 @@ SECTIONS
 	*(.dynbss)
 	*(.bss .bss.* .gnu.linkonce.b.*)
 	*(__ex_table)
+	*(__mca_table)
   }
 }
 
diff --git a/include/asm-ia64/asmmacro.h b/include/asm-ia64/asmmacro.h
index d4cec32083d8..edf2cebb2969 100644
--- a/include/asm-ia64/asmmacro.h
+++ b/include/asm-ia64/asmmacro.h
@@ -38,6 +38,10 @@ name:
 
 /*
  * Helper macros for accessing user memory.
+ *
+ * When adding any new .section/.previous entries here, make sure to
+ * also add it to the DISCARD section in arch/ia64/kernel/gate.lds.S or
+ * unpleasant things will happen.
  */
 
 	.section "__ex_table", "a"		// declare section & section attributes
-- 
cgit v1.2.3


From a81dd18eb974cc34634c53a6447b2799ec0c3158 Mon Sep 17 00:00:00 2001
From: Thibaut VARENE <varenet@parisc-linux.org>
Date: Fri, 3 Feb 2006 18:06:30 -0700
Subject: [PARISC] Clarify pdc_stable license terms

pdc_stable.c is explicitly licensed under GPL version 2.

Signed-off-by: Thibaut VARENE <varenet@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/kernel/pdc_chassis.c | 5 ++---
 drivers/parisc/pdc_stable.c      | 5 ++---
 include/asm-parisc/pdc_chassis.h | 5 ++---
 3 files changed, 6 insertions(+), 9 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/kernel/pdc_chassis.c b/arch/parisc/kernel/pdc_chassis.c
index 0cea6958f427..a45e2e2ffd9f 100644
--- a/arch/parisc/kernel/pdc_chassis.c
+++ b/arch/parisc/kernel/pdc_chassis.c
@@ -5,9 +5,8 @@
  *    Copyright (C) 2002-2004 Thibaut VARENE <varenet@parisc-linux.org>
  *
  *    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
- *    the Free Software Foundation; either version 2 of the License, or
- *    (at your option) any later version.
+ *    it under the terms of the GNU General Public License, version 2, as
+ *    published by the Free Software Foundation.
  *
  *    This program is distributed in the hope that it will be useful,
  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
diff --git a/drivers/parisc/pdc_stable.c b/drivers/parisc/pdc_stable.c
index a28e17898fbd..4e53be9c03ab 100644
--- a/drivers/parisc/pdc_stable.c
+++ b/drivers/parisc/pdc_stable.c
@@ -4,9 +4,8 @@
  *    Copyright (C) 2005-2006 Thibaut VARENE <varenet@parisc-linux.org>
  *
  *    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
- *    the Free Software Foundation; either version 2 of the License, or
- *    (at your option) any later version.
+ *    it under the terms of the GNU General Public License, version 2, as
+ *    published by the Free Software Foundation.
  *
  *    This program is distributed in the hope that it will be useful,
  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
diff --git a/include/asm-parisc/pdc_chassis.h b/include/asm-parisc/pdc_chassis.h
index adac9ac2743f..a609273dc6bf 100644
--- a/include/asm-parisc/pdc_chassis.h
+++ b/include/asm-parisc/pdc_chassis.h
@@ -6,9 +6,8 @@
  *
  *
  *      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
- *      the Free Software Foundation; either version 2, or (at your option)
- *      any later version.
+ *      it under the terms of the GNU General Public License, version 2, as
+ *      published by the Free Software Foundation.
  *      
  *      This program is distributed in the hope that it will be useful,
  *      but WITHOUT ANY WARRANTY; without even the implied warranty of
-- 
cgit v1.2.3


From ab43227c8a568119a3aebc952a95ac3023e1730d Mon Sep 17 00:00:00 2001
From: James Bottomley <jejb@parisc-linux.org>
Date: Wed, 22 Mar 2006 08:28:59 -0700
Subject: [PARISC] Add parisc implementation of flush_anon_page()

This should now allow SG_IO and fuse to function correctly on our
platform.

Signed-off-by: James Bottomley <jejb@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 include/asm-parisc/cacheflush.h | 8 ++++++++
 1 file changed, 8 insertions(+)

(limited to 'include')

diff --git a/include/asm-parisc/cacheflush.h b/include/asm-parisc/cacheflush.h
index c53af9ff41b5..482be77551e3 100644
--- a/include/asm-parisc/cacheflush.h
+++ b/include/asm-parisc/cacheflush.h
@@ -184,6 +184,14 @@ flush_cache_page(struct vm_area_struct *vma, unsigned long vmaddr, unsigned long
 
 }
 
+static inline void
+flush_anon_page(struct page *page, unsigned long vmaddr)
+{
+	if (PageAnon(page))
+		flush_user_dcache_page(vmaddr);
+}
+#define ARCH_HAS_FLUSH_ANON_PAGE
+
 #ifdef CONFIG_DEBUG_RODATA
 void mark_rodata_ro(void);
 #endif
-- 
cgit v1.2.3


From ba57583396585a1ca509e2a84d970a2ff3f9bbfb Mon Sep 17 00:00:00 2001
From: James Bottomley <jejb@parisc-linux.org>
Date: Wed, 22 Mar 2006 09:42:04 -0700
Subject: [PARISC] Add parisc implementation of flush_kernel_dcache_page()

We need to do a little renaming of our original syntax because
of the difference in arguments.

Signed-off-by: James Bottomley <jejb@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/kernel/cache.c      | 6 +++---
 arch/parisc/kernel/pacache.S    | 4 ++--
 include/asm-parisc/cache.h      | 2 +-
 include/asm-parisc/cacheflush.h | 9 ++++++++-
 include/asm-parisc/page.h       | 2 +-
 5 files changed, 15 insertions(+), 8 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/kernel/cache.c b/arch/parisc/kernel/cache.c
index d8a4ca021aac..360b7391cb8c 100644
--- a/arch/parisc/kernel/cache.c
+++ b/arch/parisc/kernel/cache.c
@@ -89,7 +89,7 @@ update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t pte)
 	if (pfn_valid(page_to_pfn(page)) && page_mapping(page) &&
 	    test_bit(PG_dcache_dirty, &page->flags)) {
 
-		flush_kernel_dcache_page(page_address(page));
+		flush_kernel_dcache_page(page);
 		clear_bit(PG_dcache_dirty, &page->flags);
 	}
 }
@@ -278,7 +278,7 @@ void flush_dcache_page(struct page *page)
 		return;
 	}
 
-	flush_kernel_dcache_page(page_address(page));
+	flush_kernel_dcache_page(page);
 
 	if (!mapping)
 		return;
@@ -317,7 +317,7 @@ EXPORT_SYMBOL(flush_dcache_page);
 
 /* Defined in arch/parisc/kernel/pacache.S */
 EXPORT_SYMBOL(flush_kernel_dcache_range_asm);
-EXPORT_SYMBOL(flush_kernel_dcache_page);
+EXPORT_SYMBOL(flush_kernel_dcache_page_asm);
 EXPORT_SYMBOL(flush_data_cache_local);
 EXPORT_SYMBOL(flush_kernel_icache_range_asm);
 
diff --git a/arch/parisc/kernel/pacache.S b/arch/parisc/kernel/pacache.S
index 9534ee17b9be..7a4f07e8d3c3 100644
--- a/arch/parisc/kernel/pacache.S
+++ b/arch/parisc/kernel/pacache.S
@@ -621,9 +621,9 @@ __clear_user_page_asm:
 
 	.procend
 
-	.export flush_kernel_dcache_page
+	.export flush_kernel_dcache_page_asm
 
-flush_kernel_dcache_page:
+flush_kernel_dcache_page_asm:
 	.proc
 	.callinfo NO_CALLS
 	.entry
diff --git a/include/asm-parisc/cache.h b/include/asm-parisc/cache.h
index ae50f8e12eed..c831665473cb 100644
--- a/include/asm-parisc/cache.h
+++ b/include/asm-parisc/cache.h
@@ -48,7 +48,7 @@ extern void flush_user_icache_range_asm(unsigned long, unsigned long);
 extern void flush_kernel_icache_range_asm(unsigned long, unsigned long);
 extern void flush_user_dcache_range_asm(unsigned long, unsigned long);
 extern void flush_kernel_dcache_range_asm(unsigned long, unsigned long);
-extern void flush_kernel_dcache_page(void *);
+extern void flush_kernel_dcache_page_asm(void *);
 extern void flush_kernel_icache_page(void *);
 extern void disable_sr_hashing(void);   /* turns off space register hashing */
 extern void disable_sr_hashing_asm(int); /* low level support for above */
diff --git a/include/asm-parisc/cacheflush.h b/include/asm-parisc/cacheflush.h
index 482be77551e3..76b6b7d6046a 100644
--- a/include/asm-parisc/cacheflush.h
+++ b/include/asm-parisc/cacheflush.h
@@ -62,7 +62,7 @@ extern void flush_dcache_page(struct page *page);
 #define flush_dcache_mmap_unlock(mapping) \
 	write_unlock_irq(&(mapping)->tree_lock)
 
-#define flush_icache_page(vma,page)	do { flush_kernel_dcache_page(page_address(page)); flush_kernel_icache_page(page_address(page)); } while (0)
+#define flush_icache_page(vma,page)	do { flush_kernel_dcache_page(page); flush_kernel_icache_page(page_address(page)); } while (0)
 
 #define flush_icache_range(s,e)		do { flush_kernel_dcache_range_asm(s,e); flush_kernel_icache_range_asm(s,e); } while (0)
 
@@ -192,6 +192,13 @@ flush_anon_page(struct page *page, unsigned long vmaddr)
 }
 #define ARCH_HAS_FLUSH_ANON_PAGE
 
+static inline void
+flush_kernel_dcache_page(struct page *page)
+{
+	flush_kernel_dcache_page_asm(page_address(page));
+}
+#define ARCH_HAS_FLUSH_KERNEL_DCACHE_PAGE
+
 #ifdef CONFIG_DEBUG_RODATA
 void mark_rodata_ro(void);
 #endif
diff --git a/include/asm-parisc/page.h b/include/asm-parisc/page.h
index 9f303c0c3cd7..c6c9d5793070 100644
--- a/include/asm-parisc/page.h
+++ b/include/asm-parisc/page.h
@@ -26,7 +26,7 @@ static inline void
 copy_user_page(void *vto, void *vfrom, unsigned long vaddr, struct page *pg)
 {
 	copy_user_page_asm(vto, vfrom);
-	flush_kernel_dcache_page(vto);
+	flush_kernel_dcache_page_asm(vto);
 	/* XXX: ppc flushes icache too, should we? */
 }
 
-- 
cgit v1.2.3


From b31059f7634931a06f6811247ae9217d1a833a46 Mon Sep 17 00:00:00 2001
From: Helge Deller <deller@parisc-linux.org>
Date: Thu, 23 Mar 2006 15:48:15 -0700
Subject: [PARISC] Add STRICT_MM_TYPECHECKS

Add STRICT_MM_TYPECHECKS to page.h as other architectures do.

Signed-off-by: Helge Deller <deller@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 include/asm-parisc/page.h | 56 +++++++++++++++++++++++++++++++++--------------
 1 file changed, 40 insertions(+), 16 deletions(-)

(limited to 'include')

diff --git a/include/asm-parisc/page.h b/include/asm-parisc/page.h
index c6c9d5793070..45e02aa5bf4b 100644
--- a/include/asm-parisc/page.h
+++ b/include/asm-parisc/page.h
@@ -40,14 +40,19 @@ clear_user_page(void *page, unsigned long vaddr, struct page *pg)
 /*
  * These are used to make use of C type-checking..
  */
-#ifdef __LP64__
-typedef struct { unsigned long pte; } pte_t;
-#else
-typedef struct {
-	unsigned long pte;
-	unsigned long flags;
-} pte_t;
+#define STRICT_MM_TYPECHECKS
+#ifdef STRICT_MM_TYPECHECKS
+typedef struct { unsigned long pte;
+#if !defined(CONFIG_64BIT)
+                 unsigned long future_flags;
+ /* XXX: it's possible to remove future_flags and change BITS_PER_PTE_ENTRY
+	 to 2, but then strangely the identical 32bit kernel boots on a
+	 c3000(pa20), but not any longer on a 715(pa11).
+	 Still investigating... HelgeD.
+  */
 #endif
+} pte_t; /* either 32 or 64bit */
+
 /* NOTE: even on 64 bits, these entries are __u32 because we allocate
  * the pmd and pgd in ZONE_DMA (i.e. under 4GB) */
 typedef struct { __u32 pmd; } pmd_t;
@@ -55,25 +60,44 @@ typedef struct { __u32 pgd; } pgd_t;
 typedef struct { unsigned long pgprot; } pgprot_t;
 
 #define pte_val(x)	((x).pte)
-#ifdef __LP64__
-#define pte_flags(x)	(*(__u32 *)&((x).pte))
-#else
-#define pte_flags(x)	((x).flags)
-#endif
-
 /* These do not work lvalues, so make sure we don't use them as such. */
 #define pmd_val(x)	((x).pmd + 0)
 #define pgd_val(x)	((x).pgd + 0)
 #define pgprot_val(x)	((x).pgprot)
 
-#define __pmd_val_set(x,n) (x).pmd = (n)
-#define __pgd_val_set(x,n) (x).pgd = (n)
-
 #define __pte(x)	((pte_t) { (x) } )
 #define __pmd(x)	((pmd_t) { (x) } )
 #define __pgd(x)	((pgd_t) { (x) } )
 #define __pgprot(x)	((pgprot_t) { (x) } )
 
+#define __pmd_val_set(x,n) (x).pmd = (n)
+#define __pgd_val_set(x,n) (x).pgd = (n)
+
+#else
+/*
+ * .. while these make it easier on the compiler
+ */
+typedef unsigned long pte_t;
+typedef         __u32 pmd_t;
+typedef         __u32 pgd_t;
+typedef unsigned long pgprot_t;
+
+#define pte_val(x)      (x)
+#define pmd_val(x)      (x)
+#define pgd_val(x)      (x)
+#define pgprot_val(x)   (x)
+
+#define __pte(x)        (x)
+#define __pmd(x)	(x)
+#define __pgd(x)        (x)
+#define __pgprot(x)     (x)
+
+#define __pmd_val_set(x,n) (x) = (n)
+#define __pgd_val_set(x,n) (x) = (n)
+
+#endif /* STRICT_MM_TYPECHECKS */
+
+
 typedef struct __physmem_range {
 	unsigned long start_pfn;
 	unsigned long pages;       /* PAGE_SIZE pages */
-- 
cgit v1.2.3


From b8ce0aadcdebbaf5ec013c57e2a0803060817bcc Mon Sep 17 00:00:00 2001
From: Helge Deller <deller@parisc-linux.org>
Date: Tue, 7 Mar 2006 14:12:13 -0700
Subject: [PARISC] Add CONFIG_HPPA_IOREMAP to conditionally enable ioremap

Instead of making it a #define in asm/io.h, allow user to select
to turn on IOREMAP from the config menu.

Signed-off-by: Helge Deller <deller@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/Kconfig      | 11 +++++++++++
 arch/parisc/mm/ioremap.c | 13 ++++++-------
 include/asm-parisc/io.h  | 16 +++++-----------
 3 files changed, 22 insertions(+), 18 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index 6b3c50964ca9..fc5755d1db57 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -138,6 +138,17 @@ config 64BIT
 	  enable this option otherwise. The 64bit kernel is significantly bigger
 	  and slower than the 32bit one.
 
+config HPPA_IOREMAP
+	bool "Enable IOREMAP functionality (EXPERIMENTAL)"
+	depends on EXPERIMENTAL
+	default n
+	help
+	  Enable this option if you want to enable real IOREMAPPING on PA-RISC.
+	  Currently we just "simulate" I/O remapping, and enabling this option
+	  will just crash your machine.
+
+	  Say N here, unless you are a real PA-RISC Linux kernel hacker.
+
 config SMP
 	bool "Symmetric multi-processing support"
 	---help---
diff --git a/arch/parisc/mm/ioremap.c b/arch/parisc/mm/ioremap.c
index edd9a9559cba..028772144191 100644
--- a/arch/parisc/mm/ioremap.c
+++ b/arch/parisc/mm/ioremap.c
@@ -72,7 +72,7 @@ remap_area_pmd(pmd_t *pmd, unsigned long address, unsigned long size,
 	return 0;
 }
 
-#if USE_HPPA_IOREMAP
+#ifdef CONFIG_HPPA_IOREMAP
 static int 
 remap_area_pages(unsigned long address, unsigned long phys_addr,
 		 unsigned long size, unsigned long flags)
@@ -114,7 +114,7 @@ remap_area_pages(unsigned long address, unsigned long phys_addr,
 
 	return error;
 }
-#endif /* USE_HPPA_IOREMAP */
+#endif /* CONFIG_HPPA_IOREMAP */
 
 #ifdef CONFIG_DEBUG_IOREMAP
 static unsigned long last = 0;
@@ -154,8 +154,7 @@ EXPORT_SYMBOL(__raw_bad_addr);
  */
 void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned long flags)
 {
-#if !(USE_HPPA_IOREMAP)
-
+#if !defined(CONFIG_HPPA_IOREMAP)
 	unsigned long end = phys_addr + size - 1;
 	/* Support EISA addresses */
 	if ((phys_addr >= 0x00080000 && end < 0x000fffff)
@@ -222,10 +221,10 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
 
 void iounmap(void __iomem *addr)
 {
-#if !(USE_HPPA_IOREMAP)
-	return;
-#else
+#ifdef CONFIG_HPPA_IOREMAP
 	if (addr > high_memory)
 		return vfree((void *) (PAGE_MASK & (unsigned long __force) addr));
+#else
+	return;
 #endif
 }
diff --git a/include/asm-parisc/io.h b/include/asm-parisc/io.h
index be0c7234a6da..2fb253255102 100644
--- a/include/asm-parisc/io.h
+++ b/include/asm-parisc/io.h
@@ -26,11 +26,7 @@ extern unsigned long parisc_vmerge_max_size;
  */
 
 #ifdef CONFIG_DEBUG_IOREMAP
-#ifdef CONFIG_64BIT
-#define NYBBLE_SHIFT 60
-#else
-#define NYBBLE_SHIFT 28
-#endif
+#define NYBBLE_SHIFT (BITS_PER_LONG - 4)
 extern void gsc_bad_addr(unsigned long addr);
 extern void __raw_bad_addr(const volatile void __iomem *addr);
 #define gsc_check_addr(addr)					\
@@ -181,13 +177,11 @@ extern inline void * ioremap_nocache(unsigned long offset, unsigned long size)
 extern void iounmap(void __iomem *addr);
 
 /*
- * USE_HPPA_IOREMAP is the magic flag to enable or disable real ioremap()
+ * CONFIG_HPPA_IOREMAP is the magic flag to enable or disable real ioremap()
  * functionality.  It's currently disabled because it may not work on some
  * machines.
  */
-#define USE_HPPA_IOREMAP 0
-
-#if USE_HPPA_IOREMAP
+#ifdef CONFIG_HPPA_IOREMAP
 static inline unsigned char __raw_readb(const volatile void __iomem *addr)
 {
 	return (*(volatile unsigned char __force *) (addr));
@@ -221,7 +215,7 @@ static inline void __raw_writeq(unsigned long long b, volatile void __iomem *add
 {
 	*(volatile unsigned long long __force *) addr = b;
 }
-#else /* !USE_HPPA_IOREMAP */
+#else /* !CONFIG_HPPA_IOREMAP */
 static inline unsigned char __raw_readb(const volatile void __iomem *addr)
 {
 	__raw_check_addr(addr);
@@ -271,7 +265,7 @@ static inline void __raw_writeq(unsigned long long b, volatile void __iomem *add
 
 	gsc_writeq(b, (unsigned long) addr);
 }
-#endif /* !USE_HPPA_IOREMAP */
+#endif /* !CONFIG_HPPA_IOREMAP */
 
 /* readb can never be const, so use __fswab instead of le*_to_cpu */
 #define readb(addr) __raw_readb(addr)
-- 
cgit v1.2.3


From 29ef8295327653ff09a56285c35213cd31fa54b3 Mon Sep 17 00:00:00 2001
From: Helge Deller <deller@parisc-linux.org>
Date: Thu, 23 Mar 2006 00:32:46 -0700
Subject: [PARISC] Enable ioremap functionality unconditionally

Enable CONFIG_HPPA_IOREMAP by default and remove all now unnecessary code.

Signed-off-by: Helge Deller <deller@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/Kconfig      | 11 ---------
 arch/parisc/mm/ioremap.c | 16 ++-----------
 include/asm-parisc/io.h  | 58 +-----------------------------------------------
 3 files changed, 3 insertions(+), 82 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index fc5755d1db57..6b3c50964ca9 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -138,17 +138,6 @@ config 64BIT
 	  enable this option otherwise. The 64bit kernel is significantly bigger
 	  and slower than the 32bit one.
 
-config HPPA_IOREMAP
-	bool "Enable IOREMAP functionality (EXPERIMENTAL)"
-	depends on EXPERIMENTAL
-	default n
-	help
-	  Enable this option if you want to enable real IOREMAPPING on PA-RISC.
-	  Currently we just "simulate" I/O remapping, and enabling this option
-	  will just crash your machine.
-
-	  Say N here, unless you are a real PA-RISC Linux kernel hacker.
-
 config SMP
 	bool "Symmetric multi-processing support"
 	---help---
diff --git a/arch/parisc/mm/ioremap.c b/arch/parisc/mm/ioremap.c
index 028772144191..5a22e1cab217 100644
--- a/arch/parisc/mm/ioremap.c
+++ b/arch/parisc/mm/ioremap.c
@@ -72,7 +72,6 @@ remap_area_pmd(pmd_t *pmd, unsigned long address, unsigned long size,
 	return 0;
 }
 
-#ifdef CONFIG_HPPA_IOREMAP
 static int 
 remap_area_pages(unsigned long address, unsigned long phys_addr,
 		 unsigned long size, unsigned long flags)
@@ -114,7 +113,6 @@ remap_area_pages(unsigned long address, unsigned long phys_addr,
 
 	return error;
 }
-#endif /* CONFIG_HPPA_IOREMAP */
 
 #ifdef CONFIG_DEBUG_IOREMAP
 static unsigned long last = 0;
@@ -154,21 +152,16 @@ EXPORT_SYMBOL(__raw_bad_addr);
  */
 void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned long flags)
 {
-#if !defined(CONFIG_HPPA_IOREMAP)
+#ifdef CONFIG_EISA
+	#error FIXME.
 	unsigned long end = phys_addr + size - 1;
 	/* Support EISA addresses */
 	if ((phys_addr >= 0x00080000 && end < 0x000fffff)
 			|| (phys_addr >= 0x00500000 && end < 0x03bfffff)) {
 		phys_addr |= 0xfc000000;
 	}
-
-#ifdef CONFIG_DEBUG_IOREMAP
-	return (void __iomem *)(phys_addr - (0x1UL << NYBBLE_SHIFT));
-#else
-	return (void __iomem *)phys_addr;
 #endif
 
-#else
 	void *addr;
 	struct vm_struct *area;
 	unsigned long offset, last_addr;
@@ -216,15 +209,10 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
 	}
 
 	return (void __iomem *) (offset + (char *)addr);
-#endif
 }
 
 void iounmap(void __iomem *addr)
 {
-#ifdef CONFIG_HPPA_IOREMAP
 	if (addr > high_memory)
 		return vfree((void *) (PAGE_MASK & (unsigned long __force) addr));
-#else
-	return;
-#endif
 }
diff --git a/include/asm-parisc/io.h b/include/asm-parisc/io.h
index 2fb253255102..bcfe0b09c93f 100644
--- a/include/asm-parisc/io.h
+++ b/include/asm-parisc/io.h
@@ -176,12 +176,7 @@ extern inline void * ioremap_nocache(unsigned long offset, unsigned long size)
 
 extern void iounmap(void __iomem *addr);
 
-/*
- * CONFIG_HPPA_IOREMAP is the magic flag to enable or disable real ioremap()
- * functionality.  It's currently disabled because it may not work on some
- * machines.
- */
-#ifdef CONFIG_HPPA_IOREMAP
+
 static inline unsigned char __raw_readb(const volatile void __iomem *addr)
 {
 	return (*(volatile unsigned char __force *) (addr));
@@ -215,57 +210,6 @@ static inline void __raw_writeq(unsigned long long b, volatile void __iomem *add
 {
 	*(volatile unsigned long long __force *) addr = b;
 }
-#else /* !CONFIG_HPPA_IOREMAP */
-static inline unsigned char __raw_readb(const volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	return gsc_readb((unsigned long) addr);
-}
-static inline unsigned short __raw_readw(const volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	return gsc_readw((unsigned long) addr);
-}
-static inline unsigned int __raw_readl(const volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	return gsc_readl((unsigned long) addr);
-}
-static inline unsigned long long __raw_readq(const volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	return gsc_readq((unsigned long) addr);
-}
-
-static inline void __raw_writeb(unsigned char b, volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	gsc_writeb(b, (unsigned long) addr);
-}
-static inline void __raw_writew(unsigned short b, volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	gsc_writew(b, (unsigned long) addr);
-}
-static inline void __raw_writel(unsigned int b, volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	gsc_writel(b, (unsigned long) addr);
-}
-static inline void __raw_writeq(unsigned long long b, volatile void __iomem *addr)
-{
-	__raw_check_addr(addr);
-
-	gsc_writeq(b, (unsigned long) addr);
-}
-#endif /* !CONFIG_HPPA_IOREMAP */
 
 /* readb can never be const, so use __fswab instead of le*_to_cpu */
 #define readb(addr) __raw_readb(addr)
-- 
cgit v1.2.3


From a41d3862dfd44a1b09a0f6243bb34773061fd9a2 Mon Sep 17 00:00:00 2001
From: Helge Deller <deller@parisc-linux.org>
Date: Thu, 23 Mar 2006 01:07:00 -0700
Subject: [PARISC] Remove obsolete CONFIG_DEBUG_IOREMAP

Remove CONFIG_DEBUG_IOREMAP, it's now obsolete and won't work anyway.
Remove it from lib/KConfig since it was only available on parisc.

Signed-off-by: Helge Deller <deller@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/lib/iomap.c  |  4 ----
 arch/parisc/mm/ioremap.c | 24 ------------------------
 include/asm-parisc/io.h  | 33 ---------------------------------
 lib/Kconfig.debug        | 13 -------------
 4 files changed, 74 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/lib/iomap.c b/arch/parisc/lib/iomap.c
index 01bec8fcbd0d..f4a811690ab3 100644
--- a/arch/parisc/lib/iomap.c
+++ b/arch/parisc/lib/iomap.c
@@ -263,11 +263,7 @@ static const struct iomap_ops iomem_ops = {
 
 const struct iomap_ops *iomap_ops[8] = {
 	[0] = &ioport_ops,
-#ifdef CONFIG_DEBUG_IOREMAP
-	[6] = &iomem_ops,
-#else
 	[7] = &iomem_ops
-#endif
 };
 
 
diff --git a/arch/parisc/mm/ioremap.c b/arch/parisc/mm/ioremap.c
index 21a16747e56e..a5967b7d3726 100644
--- a/arch/parisc/mm/ioremap.c
+++ b/arch/parisc/mm/ioremap.c
@@ -114,30 +114,6 @@ remap_area_pages(unsigned long address, unsigned long phys_addr,
 	return error;
 }
 
-#ifdef CONFIG_DEBUG_IOREMAP
-static unsigned long last = 0;
-
-void gsc_bad_addr(unsigned long addr)
-{
-	if (time_after(jiffies, last + HZ*10)) {
-		printk("gsc_foo() called with bad address 0x%lx\n", addr);
-		dump_stack();
-		last = jiffies;
-	}
-}
-EXPORT_SYMBOL(gsc_bad_addr);
-
-void __raw_bad_addr(const volatile void __iomem *addr)
-{
-	if (time_after(jiffies, last + HZ*10)) {
-		printk("__raw_foo() called with bad address 0x%p\n", addr);
-		dump_stack();
-		last = jiffies;
-	}
-}
-EXPORT_SYMBOL(__raw_bad_addr);
-#endif
-
 /*
  * Generic mapping function (not visible outside):
  */
diff --git a/include/asm-parisc/io.h b/include/asm-parisc/io.h
index bcfe0b09c93f..29da31194b91 100644
--- a/include/asm-parisc/io.h
+++ b/include/asm-parisc/io.h
@@ -25,31 +25,11 @@ extern unsigned long parisc_vmerge_max_size;
  *   eg dev->hpa or 0xfee00000.
  */
 
-#ifdef CONFIG_DEBUG_IOREMAP
-#define NYBBLE_SHIFT (BITS_PER_LONG - 4)
-extern void gsc_bad_addr(unsigned long addr);
-extern void __raw_bad_addr(const volatile void __iomem *addr);
-#define gsc_check_addr(addr)					\
-	if ((addr >> NYBBLE_SHIFT) != 0xf) {			\
-		gsc_bad_addr(addr);				\
-		addr |= 0xfUL << NYBBLE_SHIFT;			\
-	}
-#define __raw_check_addr(addr)					\
-	if (((unsigned long)addr >> NYBBLE_SHIFT) != 0xe)	\
-		__raw_bad_addr(addr);			\
-	addr = (void __iomem *)((unsigned long)addr | (0xfUL << NYBBLE_SHIFT));
-#else
-#define gsc_check_addr(addr)
-#define __raw_check_addr(addr)
-#endif
-
 static inline unsigned char gsc_readb(unsigned long addr)
 {
 	long flags;
 	unsigned char ret;
 
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	rsm	2,%0\n"
 	"	ldbx	0(%2),%1\n"
@@ -64,8 +44,6 @@ static inline unsigned short gsc_readw(unsigned long addr)
 	long flags;
 	unsigned short ret;
 
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	rsm	2,%0\n"
 	"	ldhx	0(%2),%1\n"
@@ -79,8 +57,6 @@ static inline unsigned int gsc_readl(unsigned long addr)
 {
 	u32 ret;
 
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	ldwax	0(%1),%0\n"
 	: "=r" (ret) : "r" (addr) );
@@ -91,7 +67,6 @@ static inline unsigned int gsc_readl(unsigned long addr)
 static inline unsigned long long gsc_readq(unsigned long addr)
 {
 	unsigned long long ret;
-	gsc_check_addr(addr);
 
 #ifdef __LP64__
 	__asm__ __volatile__(
@@ -108,8 +83,6 @@ static inline unsigned long long gsc_readq(unsigned long addr)
 static inline void gsc_writeb(unsigned char val, unsigned long addr)
 {
 	long flags;
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	rsm	2,%0\n"
 	"	stbs	%1,0(%2)\n"
@@ -120,8 +93,6 @@ static inline void gsc_writeb(unsigned char val, unsigned long addr)
 static inline void gsc_writew(unsigned short val, unsigned long addr)
 {
 	long flags;
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	rsm	2,%0\n"
 	"	sths	%1,0(%2)\n"
@@ -131,8 +102,6 @@ static inline void gsc_writew(unsigned short val, unsigned long addr)
 
 static inline void gsc_writel(unsigned int val, unsigned long addr)
 {
-	gsc_check_addr(addr);
-
 	__asm__ __volatile__(
 	"	stwas	%0,0(%1)\n"
 	: :  "r" (val), "r" (addr) );
@@ -140,8 +109,6 @@ static inline void gsc_writel(unsigned int val, unsigned long addr)
 
 static inline void gsc_writeq(unsigned long long val, unsigned long addr)
 {
-	gsc_check_addr(addr);
-
 #ifdef __LP64__
 	__asm__ __volatile__(
 	"	stda	%0,0(%1)\n"
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 6e8a60f67c7a..d57fd9181b18 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -157,19 +157,6 @@ config DEBUG_INFO
 
 	  If unsure, say N.
 
-config DEBUG_IOREMAP
-	bool "Enable ioremap() debugging"
-	depends on DEBUG_KERNEL && PARISC
-	help
-	  Enabling this option will cause the kernel to distinguish between
-	  ioremapped and physical addresses.  It will print a backtrace (at
-	  most one every 10 seconds), hopefully allowing you to see which
-	  drivers need work.  Fixing all these problems is a prerequisite
-	  for turning on USE_HPPA_IOREMAP.  The warnings are harmless;
-	  the kernel has enough information to fix the broken drivers
-	  automatically, but we'd like to make it more efficient by not
-	  having to do that.
-
 config DEBUG_FS
 	bool "Debug Filesystem"
 	depends on SYSFS
-- 
cgit v1.2.3


From 10267cdd0c2dee46a3f59d93fbfac7229d416dba Mon Sep 17 00:00:00 2001
From: Helge Deller <deller@parisc-linux.org>
Date: Sun, 26 Mar 2006 01:54:16 -0700
Subject: [PARISC] Fixup CONFIG_EISA a bit

Fix up some ISA/EISA stuff.

(Note: isa_ accessors have been removed from asm/io.h)

Signed-off-by: Helge Deller <deller@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/mm/ioremap.c | 8 +++-----
 include/asm-parisc/pci.h | 5 +++++
 2 files changed, 8 insertions(+), 5 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/mm/ioremap.c b/arch/parisc/mm/ioremap.c
index a5967b7d3726..5067ea90d91a 100644
--- a/arch/parisc/mm/ioremap.c
+++ b/arch/parisc/mm/ioremap.c
@@ -135,11 +135,9 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
 #ifdef CONFIG_EISA
 	unsigned long end = phys_addr + size - 1;
 	/* Support EISA addresses */
-	if ((phys_addr >= 0x00080000 && end < 0x000fffff)
-			|| (phys_addr >= 0x00500000 && end < 0x03bfffff)) {
-		phys_addr |= 0xfc000000;
-#warning "FIXME: EISA regions do not work yet..."
-		return NULL;  /* XXX */
+	if ((phys_addr >= 0x00080000 && end < 0x000fffff) ||
+	    (phys_addr >= 0x00500000 && end < 0x03bfffff)) {
+		phys_addr |= F_EXTEND(0xfc000000);
 	}
 #endif
 
diff --git a/include/asm-parisc/pci.h b/include/asm-parisc/pci.h
index fe7f6a2f5aa7..77bbafb7f73e 100644
--- a/include/asm-parisc/pci.h
+++ b/include/asm-parisc/pci.h
@@ -289,4 +289,9 @@ static inline void pcibios_add_platform_entries(struct pci_dev *dev)
 {
 }
 
+static inline void pcibios_penalize_isa_irq(int irq, int active)
+{
+	/* We don't need to penalize isa irq's */
+}
+
 #endif /* __ASM_PARISC_PCI_H */
-- 
cgit v1.2.3


From bc8846c522264d2522b0082321ec8c2051a4536f Mon Sep 17 00:00:00 2001
From: Kyle McMartin <kyle@parisc-linux.org>
Date: Fri, 24 Mar 2006 21:22:02 -0700
Subject: [PARISC] More useful readwrite lock helpers

spinlock.c needs _can_lock helpers. Rewrite _is_locked helpers to be
_can_lock helpers.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 include/asm-parisc/spinlock.h | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

(limited to 'include')

diff --git a/include/asm-parisc/spinlock.h b/include/asm-parisc/spinlock.h
index 16c2ac075fc5..a93960e232cf 100644
--- a/include/asm-parisc/spinlock.h
+++ b/include/asm-parisc/spinlock.h
@@ -134,14 +134,22 @@ static  __inline__ int __raw_write_trylock(raw_rwlock_t *rw)
 	return 1;
 }
 
-static __inline__ int __raw_is_read_locked(raw_rwlock_t *rw)
+/*
+ * read_can_lock - would read_trylock() succeed?
+ * @lock: the rwlock in question.
+ */
+static __inline__ int __raw_read_can_lock(raw_rwlock_t *rw)
 {
-	return rw->counter > 0;
+	return rw->counter >= 0;
 }
 
-static __inline__ int __raw_is_write_locked(raw_rwlock_t *rw)
+/*
+ * write_can_lock - would write_trylock() succeed?
+ * @lock: the rwlock in question.
+ */
+static __inline__ int __raw_write_can_lock(raw_rwlock_t *rw)
 {
-	return rw->counter < 0;
+	return !rw->counter;
 }
 
 #endif /* __ASM_SPINLOCK_H */
-- 
cgit v1.2.3


From 50a34dbd612925f2ec55b1781632835ef36b97d5 Mon Sep 17 00:00:00 2001
From: Kyle McMartin <kyle@parisc-linux.org>
Date: Fri, 24 Mar 2006 21:24:21 -0700
Subject: [PARISC] Add PREEMPT support

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 arch/parisc/Kconfig              |  6 +-----
 arch/parisc/kernel/entry.S       | 39 ++++++++++++++++++++++++++++++++++++---
 include/asm-parisc/thread_info.h |  3 ++-
 3 files changed, 39 insertions(+), 9 deletions(-)

(limited to 'include')

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index 6b3c50964ca9..2fdf21989dc2 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -177,14 +177,10 @@ config ARCH_DISCONTIGMEM_DEFAULT
 	def_bool y
 	depends on ARCH_DISCONTIGMEM_ENABLE
 
+source "kernel/Kconfig.preempt"
 source "kernel/Kconfig.hz"
 source "mm/Kconfig"
 
-config PREEMPT
-	bool
-#	bool "Preemptible Kernel"
-	default n
-
 config COMPAT
 	def_bool y
 	depends on 64BIT
diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S
index 6d17c0a3431b..7c95d7663c29 100644
--- a/arch/parisc/kernel/entry.S
+++ b/arch/parisc/kernel/entry.S
@@ -1014,14 +1014,21 @@ intr_restore:
 	nop
 	nop
 
+#ifndef CONFIG_PREEMPT
+# define intr_do_preempt	intr_restore
+#endif /* !CONFIG_PREEMPT */
+
 	.import schedule,code
 intr_do_resched:
-	/* Only do reschedule if we are returning to user space */
+	/* Only call schedule on return to userspace. If we're returning
+	 * to kernel space, we may schedule if CONFIG_PREEMPT, otherwise
+	 * we jump back to intr_restore.
+	 */
 	LDREG	PT_IASQ0(%r16), %r20
-	CMPIB= 0,%r20,intr_restore /* backward */
+	CMPIB=	0, %r20, intr_do_preempt
 	nop
 	LDREG	PT_IASQ1(%r16), %r20
-	CMPIB= 0,%r20,intr_restore /* backward */
+	CMPIB=	0, %r20, intr_do_preempt
 	nop
 
 #ifdef CONFIG_64BIT
@@ -1037,6 +1044,32 @@ intr_do_resched:
 #endif
 	ldo	R%intr_check_sig(%r2), %r2
 
+	/* preempt the current task on returning to kernel
+	 * mode from an interrupt, iff need_resched is set,
+	 * and preempt_count is 0. otherwise, we continue on
+	 * our merry way back to the current running task.
+	 */
+#ifdef CONFIG_PREEMPT
+	.import preempt_schedule_irq,code
+intr_do_preempt:
+	rsm	PSW_SM_I, %r0		/* disable interrupts */
+
+	/* current_thread_info()->preempt_count */
+	mfctl	%cr30, %r1
+	LDREG	TI_PRE_COUNT(%r1), %r19
+	CMPIB<>	0, %r19, intr_restore	/* if preempt_count > 0 */
+	nop				/* prev insn branched backwards */
+
+	/* check if we interrupted a critical path */
+	LDREG	PT_PSW(%r16), %r20
+	bb,<,n	%r20, 31 - PSW_SM_I, intr_restore
+	nop
+
+	BL	preempt_schedule_irq, %r2
+	nop
+
+	b	intr_restore		/* ssm PSW_SM_I done by intr_restore */
+#endif /* CONFIG_PREEMPT */
 
 	.import do_signal,code
 intr_do_signal:
diff --git a/include/asm-parisc/thread_info.h b/include/asm-parisc/thread_info.h
index ac32f140b83a..f2f83b04cd8b 100644
--- a/include/asm-parisc/thread_info.h
+++ b/include/asm-parisc/thread_info.h
@@ -49,7 +49,8 @@ struct thread_info {
 
 #endif /* !__ASSEMBLY */
 
-#define PREEMPT_ACTIVE          0x10000000
+#define PREEMPT_ACTIVE_BIT	28
+#define PREEMPT_ACTIVE		(1 << PREEMPT_ACTIVE_BIT)
 
 /*
  * thread information flags
-- 
cgit v1.2.3


From 2746ae14fe55f9483ae94ef89d4495d0eb8ee03d Mon Sep 17 00:00:00 2001
From: Kyle McMartin <kyle@parisc-linux.org>
Date: Wed, 29 Mar 2006 07:04:49 -0700
Subject: [PARISC] Make local_t use atomic_long_t

As done in asm-generic/local.h in mainline. Otherwise local_t was 32-bit
even on a 64-bit kernel.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 include/asm-parisc/local.h | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

(limited to 'include')

diff --git a/include/asm-parisc/local.h b/include/asm-parisc/local.h
index 892b3b2c4962..d0f550912755 100644
--- a/include/asm-parisc/local.h
+++ b/include/asm-parisc/local.h
@@ -4,16 +4,16 @@
 #include <linux/percpu.h>
 #include <asm/atomic.h>
 
-typedef atomic_t local_t;
+typedef atomic_long_t local_t;
 
-#define LOCAL_INIT(i)	ATOMIC_INIT(i)
-#define local_read(v)	atomic_read(v)
-#define local_set(v,i)	atomic_set(v,i)
+#define LOCAL_INIT(i)	ATOMIC_LONG_INIT(i)
+#define local_read(v)	atomic_long_read(v)
+#define local_set(v,i)	atomic_long_set(v,i)
 
-#define local_inc(v)	atomic_inc(v)
-#define local_dec(v)	atomic_dec(v)
-#define local_add(i, v)	atomic_add(i, v)
-#define local_sub(i, v)	atomic_sub(i, v)
+#define local_inc(v)	atomic_long_inc(v)
+#define local_dec(v)	atomic_long_dec(v)
+#define local_add(i, v)	atomic_long_add(i, v)
+#define local_sub(i, v)	atomic_long_sub(i, v)
 
 #define __local_inc(v)		((v)->counter++)
 #define __local_dec(v)		((v)->counter--)
-- 
cgit v1.2.3


From 4da9f131a74d12de56c44da6d522a9116da06805 Mon Sep 17 00:00:00 2001
From: Kyle McMartin <kyle@parisc-linux.org>
Date: Wed, 29 Mar 2006 19:47:32 -0500
Subject: [PARISC] Add atomic_sub_and_test

Define atomic_sub_and_test to fix build failures.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
---
 include/asm-parisc/atomic.h | 3 +++
 1 file changed, 3 insertions(+)

(limited to 'include')

diff --git a/include/asm-parisc/atomic.h b/include/asm-parisc/atomic.h
index 4dc7253ff5d0..403ea97316cf 100644
--- a/include/asm-parisc/atomic.h
+++ b/include/asm-parisc/atomic.h
@@ -210,6 +210,8 @@ static __inline__ int atomic_read(const atomic_t *v)
 
 #define atomic_dec_and_test(v)	(atomic_dec_return(v) == 0)
 
+#define atomic_sub_and_test(i,v)	(atomic_sub_return((i),(v)) == 0)
+
 #define ATOMIC_INIT(i)	((atomic_t) { (i) })
 
 #define smp_mb__before_atomic_dec()	smp_mb()
@@ -267,6 +269,7 @@ atomic64_read(const atomic64_t *v)
 
 #define atomic64_inc_and_test(v) 	(atomic64_inc_return(v) == 0)
 #define atomic64_dec_and_test(v)	(atomic64_dec_return(v) == 0)
+#define atomic64_sub_and_test(i,v)	(atomic64_sub_return((i),(v)) == 0)
 
 #endif /* __LP64__ */
 
-- 
cgit v1.2.3


From 5274f052e7b3dbd81935772eb551dfd0325dfa9d Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@suse.de>
Date: Thu, 30 Mar 2006 15:15:30 +0200
Subject: [PATCH] Introduce sys_splice() system call

This adds support for the sys_splice system call. Using a pipe as a
transport, it can connect to files or sockets (latter as output only).

From the splice.c comments:

   "splice": joining two ropes together by interweaving their strands.

   This is the "extended pipe" functionality, where a pipe is used as
   an arbitrary in-memory buffer. Think of a pipe as a small kernel
   buffer that you can use to transfer data from one end to the other.

   The traditional unix read/write is extended with a "splice()" operation
   that transfers data buffers to or from a pipe buffer.

   Named by Larry McVoy, original implementation from Linus, extended by
   Jens to support splicing to files and fixing the initial implementation
   bugs.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
---
 arch/i386/kernel/syscall_table.S |   1 +
 arch/ia64/kernel/entry.S         |   1 +
 fs/Makefile                      |   2 +-
 fs/ext2/file.c                   |   2 +
 fs/ext3/file.c                   |   2 +
 fs/pipe.c                        |  33 ++-
 fs/reiserfs/file.c               |   2 +
 fs/splice.c                      | 612 +++++++++++++++++++++++++++++++++++++++
 include/asm-i386/unistd.h        |   3 +-
 include/asm-ia64/unistd.h        |   3 +-
 include/asm-powerpc/unistd.h     |   3 +-
 include/asm-x86_64/unistd.h      |   4 +-
 include/linux/fs.h               |   4 +
 include/linux/syscalls.h         |   2 +
 net/socket.c                     |   6 +-
 15 files changed, 669 insertions(+), 11 deletions(-)
 create mode 100644 fs/splice.c

(limited to 'include')

diff --git a/arch/i386/kernel/syscall_table.S b/arch/i386/kernel/syscall_table.S
index 326595f3fa4d..ce3ef4fa0551 100644
--- a/arch/i386/kernel/syscall_table.S
+++ b/arch/i386/kernel/syscall_table.S
@@ -312,3 +312,4 @@ ENTRY(sys_call_table)
 	.long sys_unshare		/* 310 */
 	.long sys_set_robust_list
 	.long sys_get_robust_list
+	.long sys_splice
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index 0e3eda99e549..750e8e7fbdc3 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -1605,5 +1605,6 @@ sys_call_table:
 	data8 sys_ni_syscall			// reserved for pselect
 	data8 sys_ni_syscall			// 1295 reserved for ppoll
 	data8 sys_unshare
+	data8 sys_splice
 
 	.org sys_call_table + 8*NR_syscalls	// guard against failures to increase NR_syscalls
diff --git a/fs/Makefile b/fs/Makefile
index 080b3867be4d..f3a4f7077175 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -10,7 +10,7 @@ obj-y :=	open.o read_write.o file_table.o buffer.o  bio.o super.o \
 		ioctl.o readdir.o select.o fifo.o locks.o dcache.o inode.o \
 		attr.o bad_inode.o file.o filesystems.o namespace.o aio.o \
 		seq_file.o xattr.o libfs.o fs-writeback.o mpage.o direct-io.o \
-		ioprio.o pnode.o drop_caches.o
+		ioprio.o pnode.o drop_caches.o splice.o
 
 obj-$(CONFIG_INOTIFY)		+= inotify.o
 obj-$(CONFIG_EPOLL)		+= eventpoll.o
diff --git a/fs/ext2/file.c b/fs/ext2/file.c
index 509cceca04db..23e2c7ccec1d 100644
--- a/fs/ext2/file.c
+++ b/fs/ext2/file.c
@@ -53,6 +53,8 @@ const struct file_operations ext2_file_operations = {
 	.readv		= generic_file_readv,
 	.writev		= generic_file_writev,
 	.sendfile	= generic_file_sendfile,
+	.splice_read	= generic_file_splice_read,
+	.splice_write	= generic_file_splice_write,
 };
 
 #ifdef CONFIG_EXT2_FS_XIP
diff --git a/fs/ext3/file.c b/fs/ext3/file.c
index 783a796220bb..1efefb630ea9 100644
--- a/fs/ext3/file.c
+++ b/fs/ext3/file.c
@@ -119,6 +119,8 @@ const struct file_operations ext3_file_operations = {
 	.release	= ext3_release_file,
 	.fsync		= ext3_sync_file,
 	.sendfile	= generic_file_sendfile,
+	.splice_read	= generic_file_splice_read,
+	.splice_write	= generic_file_splice_write,
 };
 
 struct inode_operations ext3_file_inode_operations = {
diff --git a/fs/pipe.c b/fs/pipe.c
index e2f4f1d9ffc2..2414bf270db6 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -15,6 +15,7 @@
 #include <linux/pipe_fs_i.h>
 #include <linux/uio.h>
 #include <linux/highmem.h>
+#include <linux/pagemap.h>
 
 #include <asm/uaccess.h>
 #include <asm/ioctls.h>
@@ -94,11 +95,20 @@ static void anon_pipe_buf_release(struct pipe_inode_info *info, struct pipe_buff
 {
 	struct page *page = buf->page;
 
-	if (info->tmp_page) {
-		__free_page(page);
+	/*
+	 * If nobody else uses this page, and we don't already have a
+	 * temporary page, let's keep track of it as a one-deep
+	 * allocation cache
+	 */
+	if (page_count(page) == 1 && !info->tmp_page) {
+		info->tmp_page = page;
 		return;
 	}
-	info->tmp_page = page;
+
+	/*
+	 * Otherwise just release our reference to it
+	 */
+	page_cache_release(page);
 }
 
 static void *anon_pipe_buf_map(struct file *file, struct pipe_inode_info *info, struct pipe_buffer *buf)
@@ -152,6 +162,11 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 				chars = total_len;
 
 			addr = ops->map(filp, info, buf);
+			if (IS_ERR(addr)) {
+				if (!ret)
+					ret = PTR_ERR(addr);
+				break;
+			}
 			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars);
 			ops->unmap(info, buf);
 			if (unlikely(error)) {
@@ -254,8 +269,16 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 		struct pipe_buf_operations *ops = buf->ops;
 		int offset = buf->offset + buf->len;
 		if (ops->can_merge && offset + chars <= PAGE_SIZE) {
-			void *addr = ops->map(filp, info, buf);
-			int error = pipe_iov_copy_from_user(offset + addr, iov, chars);
+			void *addr;
+			int error;
+
+			addr = ops->map(filp, info, buf);
+			if (IS_ERR(addr)) {
+				error = PTR_ERR(addr);
+				goto out;
+			}
+			error = pipe_iov_copy_from_user(offset + addr, iov,
+							chars);
 			ops->unmap(info, buf);
 			ret = error;
 			do_wakeup = 1;
diff --git a/fs/reiserfs/file.c b/fs/reiserfs/file.c
index 010094d14da6..cf6e1cf40351 100644
--- a/fs/reiserfs/file.c
+++ b/fs/reiserfs/file.c
@@ -1576,6 +1576,8 @@ const struct file_operations reiserfs_file_operations = {
 	.sendfile = generic_file_sendfile,
 	.aio_read = generic_file_aio_read,
 	.aio_write = reiserfs_aio_write,
+	.splice_read = generic_file_splice_read,
+	.splice_write = generic_file_splice_write,
 };
 
 struct inode_operations reiserfs_file_inode_operations = {
diff --git a/fs/splice.c b/fs/splice.c
new file mode 100644
index 000000000000..efa47c1c4e13
--- /dev/null
+++ b/fs/splice.c
@@ -0,0 +1,612 @@
+/*
+ * "splice": joining two ropes together by interweaving their strands.
+ *
+ * This is the "extended pipe" functionality, where a pipe is used as
+ * an arbitrary in-memory buffer. Think of a pipe as a small kernel
+ * buffer that you can use to transfer data from one end to the other.
+ *
+ * The traditional unix read/write is extended with a "splice()" operation
+ * that transfers data buffers to or from a pipe buffer.
+ *
+ * Named by Larry McVoy, original implementation from Linus, extended by
+ * Jens to support splicing to files and fixing the initial implementation
+ * bugs.
+ *
+ * Copyright (C) 2005 Jens Axboe <axboe@suse.de>
+ * Copyright (C) 2005 Linus Torvalds <torvalds@osdl.org>
+ *
+ */
+#include <linux/fs.h>
+#include <linux/file.h>
+#include <linux/pagemap.h>
+#include <linux/pipe_fs_i.h>
+#include <linux/mm_inline.h>
+
+/*
+ * Passed to the actors
+ */
+struct splice_desc {
+	unsigned int len, total_len;	/* current and remaining length */
+	unsigned int flags;		/* splice flags */
+	struct file *file;		/* file to read/write */
+	loff_t pos;			/* file position */
+};
+
+static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
+					struct pipe_buffer *buf)
+{
+	page_cache_release(buf->page);
+	buf->page = NULL;
+}
+
+static void *page_cache_pipe_buf_map(struct file *file,
+				     struct pipe_inode_info *info,
+				     struct pipe_buffer *buf)
+{
+	struct page *page = buf->page;
+
+	lock_page(page);
+
+	if (!PageUptodate(page)) {
+		unlock_page(page);
+		return ERR_PTR(-EIO);
+	}
+
+	if (!page->mapping) {
+		unlock_page(page);
+		return ERR_PTR(-ENODATA);
+	}
+
+	return kmap(buf->page);
+}
+
+static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
+				      struct pipe_buffer *buf)
+{
+	unlock_page(buf->page);
+	kunmap(buf->page);
+}
+
+static struct pipe_buf_operations page_cache_pipe_buf_ops = {
+	.can_merge = 0,
+	.map = page_cache_pipe_buf_map,
+	.unmap = page_cache_pipe_buf_unmap,
+	.release = page_cache_pipe_buf_release,
+};
+
+static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
+			    int nr_pages, unsigned long offset,
+			    unsigned long len)
+{
+	struct pipe_inode_info *info;
+	int ret, do_wakeup, i;
+
+	ret = 0;
+	do_wakeup = 0;
+	i = 0;
+
+	mutex_lock(PIPE_MUTEX(*inode));
+
+	info = inode->i_pipe;
+	for (;;) {
+		int bufs;
+
+		if (!PIPE_READERS(*inode)) {
+			send_sig(SIGPIPE, current, 0);
+			if (!ret)
+				ret = -EPIPE;
+			break;
+		}
+
+		bufs = info->nrbufs;
+		if (bufs < PIPE_BUFFERS) {
+			int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS - 1);
+			struct pipe_buffer *buf = info->bufs + newbuf;
+			struct page *page = pages[i++];
+			unsigned long this_len;
+
+			this_len = PAGE_CACHE_SIZE - offset;
+			if (this_len > len)
+				this_len = len;
+
+			buf->page = page;
+			buf->offset = offset;
+			buf->len = this_len;
+			buf->ops = &page_cache_pipe_buf_ops;
+			info->nrbufs = ++bufs;
+			do_wakeup = 1;
+
+			ret += this_len;
+			len -= this_len;
+			offset = 0;
+			if (!--nr_pages)
+				break;
+			if (!len)
+				break;
+			if (bufs < PIPE_BUFFERS)
+				continue;
+
+			break;
+		}
+
+		if (signal_pending(current)) {
+			if (!ret)
+				ret = -ERESTARTSYS;
+			break;
+		}
+
+		if (do_wakeup) {
+			wake_up_interruptible_sync(PIPE_WAIT(*inode));
+			kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO,
+				    POLL_IN);
+			do_wakeup = 0;
+		}
+
+		PIPE_WAITING_WRITERS(*inode)++;
+		pipe_wait(inode);
+		PIPE_WAITING_WRITERS(*inode)--;
+	}
+
+	mutex_unlock(PIPE_MUTEX(*inode));
+
+	if (do_wakeup) {
+		wake_up_interruptible(PIPE_WAIT(*inode));
+		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+	}
+
+	while (i < nr_pages)
+		page_cache_release(pages[i++]);
+
+	return ret;
+}
+
+static int __generic_file_splice_read(struct file *in, struct inode *pipe,
+				      size_t len)
+{
+	struct address_space *mapping = in->f_mapping;
+	unsigned int offset, nr_pages;
+	struct page *pages[PIPE_BUFFERS], *shadow[PIPE_BUFFERS];
+	struct page *page;
+	pgoff_t index, pidx;
+	int i, j;
+
+	index = in->f_pos >> PAGE_CACHE_SHIFT;
+	offset = in->f_pos & ~PAGE_CACHE_MASK;
+	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+
+	if (nr_pages > PIPE_BUFFERS)
+		nr_pages = PIPE_BUFFERS;
+
+	/*
+	 * initiate read-ahead on this page range
+	 */
+	do_page_cache_readahead(mapping, in, index, nr_pages);
+
+	/*
+	 * Get as many pages from the page cache as possible..
+	 * Start IO on the page cache entries we create (we
+	 * can assume that any pre-existing ones we find have
+	 * already had IO started on them).
+	 */
+	i = find_get_pages(mapping, index, nr_pages, pages);
+
+	/*
+	 * common case - we found all pages and they are contiguous,
+	 * kick them off
+	 */
+	if (i && (pages[i - 1]->index == index + i - 1))
+		goto splice_them;
+
+	/*
+	 * fill shadow[] with pages at the right locations, so we only
+	 * have to fill holes
+	 */
+	memset(shadow, 0, i * sizeof(struct page *));
+	for (j = 0, pidx = index; j < i; pidx++, j++)
+		shadow[pages[j]->index - pidx] = pages[j];
+
+	/*
+	 * now fill in the holes
+	 */
+	for (i = 0, pidx = index; i < nr_pages; pidx++, i++) {
+		int error;
+
+		if (shadow[i])
+			continue;
+
+		/*
+		 * no page there, look one up / create it
+		 */
+		page = find_or_create_page(mapping, pidx,
+						   mapping_gfp_mask(mapping));
+		if (!page)
+			break;
+
+		if (PageUptodate(page))
+			unlock_page(page);
+		else {
+			error = mapping->a_ops->readpage(in, page);
+
+			if (unlikely(error)) {
+				page_cache_release(page);
+				break;
+			}
+		}
+		shadow[i] = page;
+	}
+
+	if (!i) {
+		for (i = 0; i < nr_pages; i++) {
+			 if (shadow[i])
+				page_cache_release(shadow[i]);
+		}
+		return 0;
+	}
+
+	memcpy(pages, shadow, i * sizeof(struct page *));
+
+	/*
+	 * Now we splice them into the pipe..
+	 */
+splice_them:
+	return move_to_pipe(pipe, pages, i, offset, len);
+}
+
+ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
+				 size_t len, unsigned int flags)
+{
+	ssize_t spliced;
+	int ret;
+
+	ret = 0;
+	spliced = 0;
+	while (len) {
+		ret = __generic_file_splice_read(in, pipe, len);
+
+		if (ret <= 0)
+			break;
+
+		in->f_pos += ret;
+		len -= ret;
+		spliced += ret;
+	}
+
+	if (spliced)
+		return spliced;
+
+	return ret;
+}
+
+/*
+ * Send 'len' bytes to socket from 'file' at position 'pos' using sendpage().
+ */
+static int pipe_to_sendpage(struct pipe_inode_info *info,
+			    struct pipe_buffer *buf, struct splice_desc *sd)
+{
+	struct file *file = sd->file;
+	loff_t pos = sd->pos;
+	unsigned int offset;
+	ssize_t ret;
+	void *ptr;
+
+	/*
+	 * sub-optimal, but we are limited by the pipe ->map. we don't
+	 * need a kmap'ed buffer here, we just want to make sure we
+	 * have the page pinned if the pipe page originates from the
+	 * page cache
+	 */
+	ptr = buf->ops->map(file, info, buf);
+	if (IS_ERR(ptr))
+		return PTR_ERR(ptr);
+
+	offset = pos & ~PAGE_CACHE_MASK;
+
+	ret = file->f_op->sendpage(file, buf->page, offset, sd->len, &pos,
+					sd->len < sd->total_len);
+
+	buf->ops->unmap(info, buf);
+	if (ret == sd->len)
+		return 0;
+
+	return -EIO;
+}
+
+/*
+ * This is a little more tricky than the file -> pipe splicing. There are
+ * basically three cases:
+ *
+ *	- Destination page already exists in the address space and there
+ *	  are users of it. For that case we have no other option that
+ *	  copying the data. Tough luck.
+ *	- Destination page already exists in the address space, but there
+ *	  are no users of it. Make sure it's uptodate, then drop it. Fall
+ *	  through to last case.
+ *	- Destination page does not exist, we can add the pipe page to
+ *	  the page cache and avoid the copy.
+ *
+ * For now we just do the slower thing and always copy pages over, it's
+ * easier than migrating pages from the pipe to the target file. For the
+ * case of doing file | file splicing, the migrate approach had some LRU
+ * nastiness...
+ */
+static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
+			struct splice_desc *sd)
+{
+	struct file *file = sd->file;
+	struct address_space *mapping = file->f_mapping;
+	unsigned int offset;
+	struct page *page;
+	char *src, *dst;
+	pgoff_t index;
+	int ret;
+
+	/*
+	 * after this, page will be locked and unmapped
+	 */
+	src = buf->ops->map(file, info, buf);
+	if (IS_ERR(src))
+		return PTR_ERR(src);
+
+	index = sd->pos >> PAGE_CACHE_SHIFT;
+	offset = sd->pos & ~PAGE_CACHE_MASK;
+
+find_page:
+	ret = -ENOMEM;
+	page = find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
+	if (!page)
+		goto out;
+
+	/*
+	 * If the page is uptodate, it is also locked. If it isn't
+	 * uptodate, we can mark it uptodate if we are filling the
+	 * full page. Otherwise we need to read it in first...
+	 */
+	if (!PageUptodate(page)) {
+		if (sd->len < PAGE_CACHE_SIZE) {
+			ret = mapping->a_ops->readpage(file, page);
+			if (unlikely(ret))
+				goto out;
+
+			lock_page(page);
+
+			if (!PageUptodate(page)) {
+				/*
+				 * page got invalidated, repeat
+				 */
+				if (!page->mapping) {
+					unlock_page(page);
+					page_cache_release(page);
+					goto find_page;
+				}
+				ret = -EIO;
+				goto out;
+			}
+		} else {
+			WARN_ON(!PageLocked(page));
+			SetPageUptodate(page);
+		}
+	}
+
+	ret = mapping->a_ops->prepare_write(file, page, 0, sd->len);
+	if (ret)
+		goto out;
+
+	dst = kmap_atomic(page, KM_USER0);
+	memcpy(dst + offset, src + buf->offset, sd->len);
+	flush_dcache_page(page);
+	kunmap_atomic(dst, KM_USER0);
+
+	ret = mapping->a_ops->commit_write(file, page, 0, sd->len);
+	if (ret < 0)
+		goto out;
+
+	set_page_dirty(page);
+	ret = write_one_page(page, 0);
+out:
+	if (ret < 0)
+		unlock_page(page);
+	page_cache_release(page);
+	buf->ops->unmap(info, buf);
+	return ret;
+}
+
+typedef int (splice_actor)(struct pipe_inode_info *, struct pipe_buffer *,
+			   struct splice_desc *);
+
+static ssize_t move_from_pipe(struct inode *inode, struct file *out,
+			      size_t len, unsigned int flags,
+			      splice_actor *actor)
+{
+	struct pipe_inode_info *info;
+	int ret, do_wakeup, err;
+	struct splice_desc sd;
+
+	ret = 0;
+	do_wakeup = 0;
+
+	sd.total_len = len;
+	sd.flags = flags;
+	sd.file = out;
+	sd.pos = out->f_pos;
+
+	mutex_lock(PIPE_MUTEX(*inode));
+
+	info = inode->i_pipe;
+	for (;;) {
+		int bufs = info->nrbufs;
+
+		if (bufs) {
+			int curbuf = info->curbuf;
+			struct pipe_buffer *buf = info->bufs + curbuf;
+			struct pipe_buf_operations *ops = buf->ops;
+
+			sd.len = buf->len;
+			if (sd.len > sd.total_len)
+				sd.len = sd.total_len;
+
+			err = actor(info, buf, &sd);
+			if (err) {
+				if (!ret && err != -ENODATA)
+					ret = err;
+
+				break;
+			}
+
+			ret += sd.len;
+			buf->offset += sd.len;
+			buf->len -= sd.len;
+			if (!buf->len) {
+				buf->ops = NULL;
+				ops->release(info, buf);
+				curbuf = (curbuf + 1) & (PIPE_BUFFERS - 1);
+				info->curbuf = curbuf;
+				info->nrbufs = --bufs;
+				do_wakeup = 1;
+			}
+
+			sd.pos += sd.len;
+			sd.total_len -= sd.len;
+			if (!sd.total_len)
+				break;
+		}
+
+		if (bufs)
+			continue;
+		if (!PIPE_WRITERS(*inode))
+			break;
+		if (!PIPE_WAITING_WRITERS(*inode)) {
+			if (ret)
+				break;
+		}
+
+		if (signal_pending(current)) {
+			if (!ret)
+				ret = -ERESTARTSYS;
+			break;
+		}
+
+		if (do_wakeup) {
+			wake_up_interruptible_sync(PIPE_WAIT(*inode));
+			kill_fasync(PIPE_FASYNC_WRITERS(*inode),SIGIO,POLL_OUT);
+			do_wakeup = 0;
+		}
+
+		pipe_wait(inode);
+	}
+
+	mutex_unlock(PIPE_MUTEX(*inode));
+
+	if (do_wakeup) {
+		wake_up_interruptible(PIPE_WAIT(*inode));
+		kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+	}
+
+	mutex_lock(&out->f_mapping->host->i_mutex);
+	out->f_pos = sd.pos;
+	mutex_unlock(&out->f_mapping->host->i_mutex);
+	return ret;
+
+}
+
+ssize_t generic_file_splice_write(struct inode *inode, struct file *out,
+				  size_t len, unsigned int flags)
+{
+	return move_from_pipe(inode, out, len, flags, pipe_to_file);
+}
+
+ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
+				size_t len, unsigned int flags)
+{
+	return move_from_pipe(inode, out, len, flags, pipe_to_sendpage);
+}
+
+static long do_splice_from(struct inode *pipe, struct file *out, size_t len,
+			   unsigned int flags)
+{
+	loff_t pos;
+	int ret;
+
+	if (!out->f_op || !out->f_op->splice_write)
+		return -EINVAL;
+
+	if (!(out->f_mode & FMODE_WRITE))
+		return -EBADF;
+
+	pos = out->f_pos;
+	ret = rw_verify_area(WRITE, out, &pos, len);
+	if (unlikely(ret < 0))
+		return ret;
+
+	return out->f_op->splice_write(pipe, out, len, flags);
+}
+
+static long do_splice_to(struct file *in, struct inode *pipe, size_t len,
+			 unsigned int flags)
+{
+	loff_t pos, isize, left;
+	int ret;
+
+	if (!in->f_op || !in->f_op->splice_read)
+		return -EINVAL;
+
+	if (!(in->f_mode & FMODE_READ))
+		return -EBADF;
+
+	pos = in->f_pos;
+	ret = rw_verify_area(READ, in, &pos, len);
+	if (unlikely(ret < 0))
+		return ret;
+
+	isize = i_size_read(in->f_mapping->host);
+	if (unlikely(in->f_pos >= isize))
+		return 0;
+	
+	left = isize - in->f_pos;
+	if (left < len)
+		len = left;
+
+	return in->f_op->splice_read(in, pipe, len, flags);
+}
+
+static long do_splice(struct file *in, struct file *out, size_t len,
+		      unsigned int flags)
+{
+	struct inode *pipe;
+
+	pipe = in->f_dentry->d_inode;
+	if (pipe->i_pipe)
+		return do_splice_from(pipe, out, len, flags);
+
+	pipe = out->f_dentry->d_inode;
+	if (pipe->i_pipe)
+		return do_splice_to(in, pipe, len, flags);
+
+	return -EINVAL;
+}
+
+asmlinkage long sys_splice(int fdin, int fdout, size_t len, unsigned int flags)
+{
+	long error;
+	struct file *in, *out;
+	int fput_in, fput_out;
+
+	if (unlikely(!len))
+		return 0;
+
+	error = -EBADF;
+	in = fget_light(fdin, &fput_in);
+	if (in) {
+		if (in->f_mode & FMODE_READ) {
+			out = fget_light(fdout, &fput_out);
+			if (out) {
+				if (out->f_mode & FMODE_WRITE)
+					error = do_splice(in, out, len, flags);
+				fput_light(out, fput_out);
+			}
+		}
+
+		fput_light(in, fput_in);
+	}
+
+	return error;
+}
diff --git a/include/asm-i386/unistd.h b/include/asm-i386/unistd.h
index 014e3562895b..789e9bdd0a40 100644
--- a/include/asm-i386/unistd.h
+++ b/include/asm-i386/unistd.h
@@ -318,8 +318,9 @@
 #define __NR_unshare		310
 #define __NR_set_robust_list	311
 #define __NR_get_robust_list	312
+#define __NR_sys_splice		313
 
-#define NR_syscalls 313
+#define NR_syscalls 314
 
 /*
  * user-visible error numbers are in the range -1 - -128: see
diff --git a/include/asm-ia64/unistd.h b/include/asm-ia64/unistd.h
index 019956c613e4..36070c1014d8 100644
--- a/include/asm-ia64/unistd.h
+++ b/include/asm-ia64/unistd.h
@@ -285,12 +285,13 @@
 #define __NR_faccessat			1293
 /* 1294, 1295 reserved for pselect/ppoll */
 #define __NR_unshare			1296
+#define __NR_splice			1297
 
 #ifdef __KERNEL__
 
 #include <linux/config.h>
 
-#define NR_syscalls			273 /* length of syscall table */
+#define NR_syscalls			274 /* length of syscall table */
 
 #define __ARCH_WANT_SYS_RT_SIGACTION
 
diff --git a/include/asm-powerpc/unistd.h b/include/asm-powerpc/unistd.h
index 1e990747dce7..536ba0873052 100644
--- a/include/asm-powerpc/unistd.h
+++ b/include/asm-powerpc/unistd.h
@@ -301,8 +301,9 @@
 #define __NR_pselect6		280
 #define __NR_ppoll		281
 #define __NR_unshare		282
+#define __NR_splice		283
 
-#define __NR_syscalls		283
+#define __NR_syscalls		284
 
 #ifdef __KERNEL__
 #define __NR__exit __NR_exit
diff --git a/include/asm-x86_64/unistd.h b/include/asm-x86_64/unistd.h
index fcc516353087..f21ff2c1e960 100644
--- a/include/asm-x86_64/unistd.h
+++ b/include/asm-x86_64/unistd.h
@@ -609,8 +609,10 @@ __SYSCALL(__NR_unshare,	sys_unshare)
 __SYSCALL(__NR_set_robust_list, sys_set_robust_list)
 #define __NR_get_robust_list	274
 __SYSCALL(__NR_get_robust_list, sys_get_robust_list)
+#define __NR_splice		275
+__SYSCALL(__NR_splice, sys_splice)
 
-#define __NR_syscall_max __NR_get_robust_list
+#define __NR_syscall_max __NR_splice
 
 #ifndef __NO_STUBS
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 408fe89498f4..20fa5f6d7269 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1032,6 +1032,8 @@ struct file_operations {
 	int (*check_flags)(int);
 	int (*dir_notify)(struct file *filp, unsigned long arg);
 	int (*flock) (struct file *, int, struct file_lock *);
+	ssize_t (*splice_write)(struct inode *, struct file *, size_t, unsigned int);
+	ssize_t (*splice_read)(struct file *, struct inode *, size_t, unsigned int);
 };
 
 struct inode_operations {
@@ -1609,6 +1611,8 @@ extern ssize_t generic_file_sendfile(struct file *, loff_t *, size_t, read_actor
 extern void do_generic_mapping_read(struct address_space *mapping,
 				    struct file_ra_state *, struct file *,
 				    loff_t *, read_descriptor_t *, read_actor_t);
+extern ssize_t generic_file_splice_read(struct file *, struct inode *, size_t, unsigned int);
+extern ssize_t generic_file_splice_write(struct inode *, struct file *, size_t, unsigned int);
 extern void
 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
 extern ssize_t generic_file_readv(struct file *filp, const struct iovec *iov, 
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index e487e3b60f60..e78ffc7d5b56 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -569,5 +569,7 @@ asmlinkage long compat_sys_newfstatat(unsigned int dfd, char __user * filename,
 asmlinkage long compat_sys_openat(unsigned int dfd, const char __user *filename,
 				   int flags, int mode);
 asmlinkage long sys_unshare(unsigned long unshare_flags);
+asmlinkage long sys_splice(int fdin, int fdout, size_t len,
+				unsigned int flags);
 
 #endif
diff --git a/net/socket.c b/net/socket.c
index fcd77eac0ccf..b13042f68c02 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -119,6 +119,9 @@ static ssize_t sock_writev(struct file *file, const struct iovec *vector,
 static ssize_t sock_sendpage(struct file *file, struct page *page,
 			     int offset, size_t size, loff_t *ppos, int more);
 
+extern ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
+				size_t len, unsigned int flags);
+
 
 /*
  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
@@ -141,7 +144,8 @@ static struct file_operations socket_file_ops = {
 	.fasync =	sock_fasync,
 	.readv =	sock_readv,
 	.writev =	sock_writev,
-	.sendpage =	sock_sendpage
+	.sendpage =	sock_sendpage,
+	.splice_write = generic_splice_sendpage,
 };
 
 /*
-- 
cgit v1.2.3


From 5abc97aa25b2c41413b3a520faee83f2282d9f18 Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@suse.de>
Date: Thu, 30 Mar 2006 15:16:46 +0200
Subject: [PATCH] splice: add support for SPLICE_F_MOVE flag

This enables the caller to migrate pages from one address space page
cache to another.  In buzz word marketing, you can do zero-copy file
copies!

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
---
 fs/pipe.c                 |   8 +++
 fs/splice.c               | 121 ++++++++++++++++++++++++++++++++--------------
 include/linux/pipe_fs_i.h |   8 +++
 3 files changed, 100 insertions(+), 37 deletions(-)

(limited to 'include')

diff --git a/fs/pipe.c b/fs/pipe.c
index 2414bf270db6..109a102c150d 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -121,11 +121,19 @@ static void anon_pipe_buf_unmap(struct pipe_inode_info *info, struct pipe_buffer
 	kunmap(buf->page);
 }
 
+static int anon_pipe_buf_steal(struct pipe_inode_info *info,
+			       struct pipe_buffer *buf)
+{
+	buf->stolen = 1;
+	return 0;
+}
+
 static struct pipe_buf_operations anon_pipe_buf_ops = {
 	.can_merge = 1,
 	.map = anon_pipe_buf_map,
 	.unmap = anon_pipe_buf_unmap,
 	.release = anon_pipe_buf_release,
+	.steal = anon_pipe_buf_steal,
 };
 
 static ssize_t
diff --git a/fs/splice.c b/fs/splice.c
index efa47c1c4e13..4a026f95884f 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -21,6 +21,7 @@
 #include <linux/pagemap.h>
 #include <linux/pipe_fs_i.h>
 #include <linux/mm_inline.h>
+#include <linux/swap.h>
 
 /*
  * Passed to the actors
@@ -32,11 +33,37 @@ struct splice_desc {
 	loff_t pos;			/* file position */
 };
 
+static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
+				     struct pipe_buffer *buf)
+{
+	struct page *page = buf->page;
+
+	WARN_ON(!PageLocked(page));
+	WARN_ON(!PageUptodate(page));
+
+	if (!remove_mapping(page_mapping(page), page))
+		return 1;
+
+	if (PageLRU(page)) {
+		struct zone *zone = page_zone(page);
+
+		spin_lock_irq(&zone->lru_lock);
+		BUG_ON(!PageLRU(page));
+		__ClearPageLRU(page);
+		del_page_from_lru(zone, page);
+		spin_unlock_irq(&zone->lru_lock);
+	}
+
+	buf->stolen = 1;
+	return 0;
+}
+
 static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
 					struct pipe_buffer *buf)
 {
 	page_cache_release(buf->page);
 	buf->page = NULL;
+	buf->stolen = 0;
 }
 
 static void *page_cache_pipe_buf_map(struct file *file,
@@ -63,7 +90,8 @@ static void *page_cache_pipe_buf_map(struct file *file,
 static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
 				      struct pipe_buffer *buf)
 {
-	unlock_page(buf->page);
+	if (!buf->stolen)
+		unlock_page(buf->page);
 	kunmap(buf->page);
 }
 
@@ -72,6 +100,7 @@ static struct pipe_buf_operations page_cache_pipe_buf_ops = {
 	.map = page_cache_pipe_buf_map,
 	.unmap = page_cache_pipe_buf_unmap,
 	.release = page_cache_pipe_buf_release,
+	.steal = page_cache_pipe_buf_steal,
 };
 
 static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
@@ -336,8 +365,8 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	struct address_space *mapping = file->f_mapping;
 	unsigned int offset;
 	struct page *page;
-	char *src, *dst;
 	pgoff_t index;
+	char *src;
 	int ret;
 
 	/*
@@ -350,40 +379,54 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	index = sd->pos >> PAGE_CACHE_SHIFT;
 	offset = sd->pos & ~PAGE_CACHE_MASK;
 
-find_page:
-	ret = -ENOMEM;
-	page = find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
-	if (!page)
-		goto out;
-
 	/*
-	 * If the page is uptodate, it is also locked. If it isn't
-	 * uptodate, we can mark it uptodate if we are filling the
-	 * full page. Otherwise we need to read it in first...
+	 * reuse buf page, if SPLICE_F_MOVE is set
 	 */
-	if (!PageUptodate(page)) {
-		if (sd->len < PAGE_CACHE_SIZE) {
-			ret = mapping->a_ops->readpage(file, page);
-			if (unlikely(ret))
-				goto out;
-
-			lock_page(page);
-
-			if (!PageUptodate(page)) {
-				/*
-				 * page got invalidated, repeat
-				 */
-				if (!page->mapping) {
-					unlock_page(page);
-					page_cache_release(page);
-					goto find_page;
+	if (sd->flags & SPLICE_F_MOVE) {
+		if (buf->ops->steal(info, buf))
+			goto find_page;
+
+		page = buf->page;
+		if (add_to_page_cache_lru(page, mapping, index,
+						mapping_gfp_mask(mapping)))
+			goto find_page;
+	} else {
+find_page:
+		ret = -ENOMEM;
+		page = find_or_create_page(mapping, index,
+						mapping_gfp_mask(mapping));
+		if (!page)
+			goto out;
+
+		/*
+		 * If the page is uptodate, it is also locked. If it isn't
+		 * uptodate, we can mark it uptodate if we are filling the
+		 * full page. Otherwise we need to read it in first...
+		 */
+		if (!PageUptodate(page)) {
+			if (sd->len < PAGE_CACHE_SIZE) {
+				ret = mapping->a_ops->readpage(file, page);
+				if (unlikely(ret))
+					goto out;
+
+				lock_page(page);
+
+				if (!PageUptodate(page)) {
+					/*
+					 * page got invalidated, repeat
+					 */
+					if (!page->mapping) {
+						unlock_page(page);
+						page_cache_release(page);
+						goto find_page;
+					}
+					ret = -EIO;
+					goto out;
 				}
-				ret = -EIO;
-				goto out;
+			} else {
+				WARN_ON(!PageLocked(page));
+				SetPageUptodate(page);
 			}
-		} else {
-			WARN_ON(!PageLocked(page));
-			SetPageUptodate(page);
 		}
 	}
 
@@ -391,10 +434,13 @@ find_page:
 	if (ret)
 		goto out;
 
-	dst = kmap_atomic(page, KM_USER0);
-	memcpy(dst + offset, src + buf->offset, sd->len);
-	flush_dcache_page(page);
-	kunmap_atomic(dst, KM_USER0);
+	if (!buf->stolen) {
+		char *dst = kmap_atomic(page, KM_USER0);
+
+		memcpy(dst + offset, src + buf->offset, sd->len);
+		flush_dcache_page(page);
+		kunmap_atomic(dst, KM_USER0);
+	}
 
 	ret = mapping->a_ops->commit_write(file, page, 0, sd->len);
 	if (ret < 0)
@@ -405,7 +451,8 @@ find_page:
 out:
 	if (ret < 0)
 		unlock_page(page);
-	page_cache_release(page);
+	if (!buf->stolen)
+		page_cache_release(page);
 	buf->ops->unmap(info, buf);
 	return ret;
 }
diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h
index b12e59c75752..75c7f55023ab 100644
--- a/include/linux/pipe_fs_i.h
+++ b/include/linux/pipe_fs_i.h
@@ -9,6 +9,7 @@ struct pipe_buffer {
 	struct page *page;
 	unsigned int offset, len;
 	struct pipe_buf_operations *ops;
+	unsigned int stolen;
 };
 
 struct pipe_buf_operations {
@@ -16,6 +17,7 @@ struct pipe_buf_operations {
 	void * (*map)(struct file *, struct pipe_inode_info *, struct pipe_buffer *);
 	void (*unmap)(struct pipe_inode_info *, struct pipe_buffer *);
 	void (*release)(struct pipe_inode_info *, struct pipe_buffer *);
+	int (*steal)(struct pipe_inode_info *, struct pipe_buffer *);
 };
 
 struct pipe_inode_info {
@@ -53,4 +55,10 @@ void pipe_wait(struct inode * inode);
 struct inode* pipe_new(struct inode* inode);
 void free_pipe_info(struct inode* inode);
 
+/*
+ * splice is tied to pipes as a transport (at least for now), so we'll just
+ * add the splice flags here.
+ */
+#define SPLICE_F_MOVE	(0x01)	/* move pages instead of copying */
+
 #endif
-- 
cgit v1.2.3