[PATCH 2/5] mm: Refactor remap_pfn_range()

[PATCH 2/5] mm: Refactor remap_pfn_range()

[Chris Wilson]

In preparation for exporting very similar functionality through another
interface, gut the current remap_pfn_range(). The motivating factor here
is to reuse the PGB/PUD/PMD/PTE walker, but allow back progation of
errors rather than BUG_ON.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: linux-mm@kvack.org
---
 mm/memory.c | 102 +++++++++++++++++++++++++++++++++---------------------------
 1 file changed, 57 insertions(+), 45 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index 97839f5c8c30..acb06f40d614 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1614,71 +1614,81 @@ int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
 }
 EXPORT_SYMBOL(vm_insert_mixed);
 
+struct remap_pfn {
+	struct mm_struct *mm;
+	unsigned long addr;
+	unsigned long pfn;
+	pgprot_t prot;
+};
+
 /*
  * maps a range of physical memory into the requested pages. the old
  * mappings are removed. any references to nonexistent pages results
  * in null mappings (currently treated as "copy-on-access")
  */
-static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
-			unsigned long addr, unsigned long end,
-			unsigned long pfn, pgprot_t prot)
+static inline int remap_pfn(struct remap_pfn *r, pte_t *pte)
+{
+	if (!pte_none(*pte))
+		return -EBUSY;
+
+	set_pte_at(r->mm, r->addr, pte,
+		   pte_mkspecial(pfn_pte(r->pfn, r->prot)));
+	r->pfn++;
+	r->addr += PAGE_SIZE;
+	return 0;
+}
+
+static int remap_pte_range(struct remap_pfn *r, pmd_t *pmd, unsigned long end)
 {
 	pte_t *pte;
 	spinlock_t *ptl;
+	int err;
 
-	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
+	pte = pte_alloc_map_lock(r->mm, pmd, r->addr, &ptl);
 	if (!pte)
 		return -ENOMEM;
+
 	arch_enter_lazy_mmu_mode();
 	do {
-		BUG_ON(!pte_none(*pte));
-		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
-		pfn++;
-	} while (pte++, addr += PAGE_SIZE, addr != end);
+		err = remap_pfn(r, pte++);
+	} while (err == 0 && r->addr < end);
 	arch_leave_lazy_mmu_mode();
+
 	pte_unmap_unlock(pte - 1, ptl);
-	return 0;
+	return err;
 }
 
-static inline int remap_pmd_range(struct mm_struct *mm, pud_t *pud,
-			unsigned long addr, unsigned long end,
-			unsigned long pfn, pgprot_t prot)
+static inline int remap_pmd_range(struct remap_pfn *r, pud_t *pud, unsigned long end)
 {
 	pmd_t *pmd;
-	unsigned long next;
+	int err;
 
-	pfn -= addr >> PAGE_SHIFT;
-	pmd = pmd_alloc(mm, pud, addr);
+	pmd = pmd_alloc(r->mm, pud, r->addr);
 	if (!pmd)
 		return -ENOMEM;
 	VM_BUG_ON(pmd_trans_huge(*pmd));
+
 	do {
-		next = pmd_addr_end(addr, end);
-		if (remap_pte_range(mm, pmd, addr, next,
-				pfn + (addr >> PAGE_SHIFT), prot))
-			return -ENOMEM;
-	} while (pmd++, addr = next, addr != end);
-	return 0;
+		err = remap_pte_range(r, pmd++, pmd_addr_end(r->addr, end));
+	} while (err == 0 && r->addr < end);
+
+	return err;
 }
 
-static inline int remap_pud_range(struct mm_struct *mm, pgd_t *pgd,
-			unsigned long addr, unsigned long end,
-			unsigned long pfn, pgprot_t prot)
+static inline int remap_pud_range(struct remap_pfn *r, pgd_t *pgd, unsigned long end)
 {
 	pud_t *pud;
-	unsigned long next;
+	int err;
 
-	pfn -= addr >> PAGE_SHIFT;
-	pud = pud_alloc(mm, pgd, addr);
+	pud = pud_alloc(r->mm, pgd, r->addr);
 	if (!pud)
 		return -ENOMEM;
+
 	do {
-		next = pud_addr_end(addr, end);
-		if (remap_pmd_range(mm, pud, addr, next,
-				pfn + (addr >> PAGE_SHIFT), prot))
-			return -ENOMEM;
-	} while (pud++, addr = next, addr != end);
-	return 0;
+		err = remap_pmd_range(r, pud++, pud_addr_end(r->addr, end));
+	} while (err == 0 && r->addr < end);
+
+	return err;
 }
 
 /**
@@ -1694,10 +1704,9 @@ static inline int remap_pud_range(struct mm_struct *mm, pgd_t *pgd,
 int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
 		    unsigned long pfn, unsigned long size, pgprot_t prot)
 {
-	pgd_t *pgd;
-	unsigned long next;
 	unsigned long end = addr + PAGE_ALIGN(size);
-	struct mm_struct *mm = vma->vm_mm;
+	struct remap_pfn r;
+	pgd_t *pgd;
 	int err;
 
 	/*
@@ -1731,19 +1740,22 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
 	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
 
 	BUG_ON(addr >= end);
-	pfn -= addr >> PAGE_SHIFT;
-	pgd = pgd_offset(mm, addr);
 	flush_cache_range(vma, addr, end);
+
+	r.mm = vma->vm_mm;
+	r.addr = addr;
+	r.pfn = pfn;
+	r.prot = prot;
+
+	pgd = pgd_offset(r.mm, addr);
 	do {
-		next = pgd_addr_end(addr, end);
-		err = remap_pud_range(mm, pgd, addr, next,
-				pfn + (addr >> PAGE_SHIFT), prot);
-		if (err)
-			break;
-	} while (pgd++, addr = next, addr != end);
+		err = remap_pud_range(&r, pgd++, pgd_addr_end(r.addr, end));
+	} while (err == 0 && r.addr < end);
 
-	if (err)
+	if (err) {
 		untrack_pfn(vma, pfn, PAGE_ALIGN(size));
+		BUG_ON(err == -EBUSY);
+	}
 
 	return err;
 }
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/5] mm: Refactor remap_pfn_range()

[Andrew Morton]

On Tue,  7 Apr 2015 17:31:36 +0100 Chris Wilson <chris@chris-wilson.co.uk> wrote:

> In preparation for exporting very similar functionality through another
> interface, gut the current remap_pfn_range(). The motivating factor here
> is to reuse the PGB/PUD/PMD/PTE walker, but allow back progation of
> errors rather than BUG_ON.

I'm not on intel-gfx and for some reason these patches didn't show up on
linux-mm.  I wanted to comment on "mutex: Export an interface to wrap a
mutex lock" but
http://lists.freedesktop.org/archives/intel-gfx/2015-April/064063.html
doesn't tell me which mailing lists were cc'ed and I can't find that
patch on linux-kernel.

Can you please do something to make this easier for us??

And please fully document all the mutex interfaces which you just
added.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/5] mm: Refactor remap_pfn_range()

[Peter Zijlstra]

On Tue, Apr 07, 2015 at 01:27:21PM -0700, Andrew Morton wrote:
> On Tue,  7 Apr 2015 17:31:36 +0100 Chris Wilson <chris@chris-wilson.co.uk> wrote:
> 
> > In preparation for exporting very similar functionality through another
> > interface, gut the current remap_pfn_range(). The motivating factor here
> > is to reuse the PGB/PUD/PMD/PTE walker, but allow back progation of
> > errors rather than BUG_ON.
> 
> I'm not on intel-gfx and for some reason these patches didn't show up on
> linux-mm.  I wanted to comment on "mutex: Export an interface to wrap a
> mutex lock" but
> http://lists.freedesktop.org/archives/intel-gfx/2015-April/064063.html
> doesn't tell me which mailing lists were cc'ed and I can't find that
> patch on linux-kernel.
> 
> Can you please do something to make this easier for us??
> 
> And please fully document all the mutex interfaces which you just
> added.

Also, please Cc locking people if you poke at mutexes..

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/5] mm: Refactor remap_pfn_range()

[Joonas Lahtinen]

On ti, 2015-04-07 at 17:31 +0100, Chris Wilson wrote:
> In preparation for exporting very similar functionality through another
> interface, gut the current remap_pfn_range(). The motivating factor here
> is to reuse the PGB/PUD/PMD/PTE walker, but allow back progation of
> errors rather than BUG_ON.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Rik van Riel <riel@redhat.com>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Cyrill Gorcunov <gorcunov@gmail.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: linux-mm@kvack.org
> ---
>  mm/memory.c | 102 +++++++++++++++++++++++++++++++++---------------------------
>  1 file changed, 57 insertions(+), 45 deletions(-)
> 
> diff --git a/mm/memory.c b/mm/memory.c
> index 97839f5c8c30..acb06f40d614 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -1614,71 +1614,81 @@ int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
>  }
>  EXPORT_SYMBOL(vm_insert_mixed);
>  
> +struct remap_pfn {
> +	struct mm_struct *mm;
> +	unsigned long addr;
> +	unsigned long pfn;
> +	pgprot_t prot;
> +};
> +
>  /*
>   * maps a range of physical memory into the requested pages. the old
>   * mappings are removed. any references to nonexistent pages results
>   * in null mappings (currently treated as "copy-on-access")
>   */
> -static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
> -			unsigned long addr, unsigned long end,
> -			unsigned long pfn, pgprot_t prot)
> +static inline int remap_pfn(struct remap_pfn *r, pte_t *pte)

I think add a brief own comment for this function and keep it below old
comment not to cause unnecessary noise.

Otherwise looks good.

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

> +{
> +	if (!pte_none(*pte))
> +		return -EBUSY;
> +
> +	set_pte_at(r->mm, r->addr, pte,
> +		   pte_mkspecial(pfn_pte(r->pfn, r->prot)));
> +	r->pfn++;
> +	r->addr += PAGE_SIZE;
> +	return 0;
> +}
> +
> +static int remap_pte_range(struct remap_pfn *r, pmd_t *pmd, unsigned long end)
>  {
>  	pte_t *pte;
>  	spinlock_t *ptl;
> +	int err;
>  
> -	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
> +	pte = pte_alloc_map_lock(r->mm, pmd, r->addr, &ptl);
>  	if (!pte)
>  		return -ENOMEM;
> +
>  	arch_enter_lazy_mmu_mode();
>  	do {
> -		BUG_ON(!pte_none(*pte));
> -		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
> -		pfn++;
> -	} while (pte++, addr += PAGE_SIZE, addr != end);
> +		err = remap_pfn(r, pte++);
> +	} while (err == 0 && r->addr < end);
>  	arch_leave_lazy_mmu_mode();
> +
>  	pte_unmap_unlock(pte - 1, ptl);
> -	return 0;
> +	return err;
>  }
>  
> -static inline int remap_pmd_range(struct mm_struct *mm, pud_t *pud,
> -			unsigned long addr, unsigned long end,
> -			unsigned long pfn, pgprot_t prot)
> +static inline int remap_pmd_range(struct remap_pfn *r, pud_t *pud, unsigned long end)
>  {
>  	pmd_t *pmd;
> -	unsigned long next;
> +	int err;
>  
> -	pfn -= addr >> PAGE_SHIFT;
> -	pmd = pmd_alloc(mm, pud, addr);
> +	pmd = pmd_alloc(r->mm, pud, r->addr);
>  	if (!pmd)
>  		return -ENOMEM;
>  	VM_BUG_ON(pmd_trans_huge(*pmd));
> +
>  	do {
> -		next = pmd_addr_end(addr, end);
> -		if (remap_pte_range(mm, pmd, addr, next,
> -				pfn + (addr >> PAGE_SHIFT), prot))
> -			return -ENOMEM;
> -	} while (pmd++, addr = next, addr != end);
> -	return 0;
> +		err = remap_pte_range(r, pmd++, pmd_addr_end(r->addr, end));
> +	} while (err == 0 && r->addr < end);
> +
> +	return err;
>  }
>  
> -static inline int remap_pud_range(struct mm_struct *mm, pgd_t *pgd,
> -			unsigned long addr, unsigned long end,
> -			unsigned long pfn, pgprot_t prot)
> +static inline int remap_pud_range(struct remap_pfn *r, pgd_t *pgd, unsigned long end)
>  {
>  	pud_t *pud;
> -	unsigned long next;
> +	int err;
>  
> -	pfn -= addr >> PAGE_SHIFT;
> -	pud = pud_alloc(mm, pgd, addr);
> +	pud = pud_alloc(r->mm, pgd, r->addr);
>  	if (!pud)
>  		return -ENOMEM;
> +
>  	do {
> -		next = pud_addr_end(addr, end);
> -		if (remap_pmd_range(mm, pud, addr, next,
> -				pfn + (addr >> PAGE_SHIFT), prot))
> -			return -ENOMEM;
> -	} while (pud++, addr = next, addr != end);
> -	return 0;
> +		err = remap_pmd_range(r, pud++, pud_addr_end(r->addr, end));
> +	} while (err == 0 && r->addr < end);
> +
> +	return err;
>  }
>  
>  /**
> @@ -1694,10 +1704,9 @@ static inline int remap_pud_range(struct mm_struct *mm, pgd_t *pgd,
>  int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
>  		    unsigned long pfn, unsigned long size, pgprot_t prot)
>  {
> -	pgd_t *pgd;
> -	unsigned long next;
>  	unsigned long end = addr + PAGE_ALIGN(size);
> -	struct mm_struct *mm = vma->vm_mm;
> +	struct remap_pfn r;
> +	pgd_t *pgd;
>  	int err;
>  
>  	/*
> @@ -1731,19 +1740,22 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
>  	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
>  
>  	BUG_ON(addr >= end);
> -	pfn -= addr >> PAGE_SHIFT;
> -	pgd = pgd_offset(mm, addr);
>  	flush_cache_range(vma, addr, end);
> +
> +	r.mm = vma->vm_mm;
> +	r.addr = addr;
> +	r.pfn = pfn;
> +	r.prot = prot;
> +
> +	pgd = pgd_offset(r.mm, addr);
>  	do {
> -		next = pgd_addr_end(addr, end);
> -		err = remap_pud_range(mm, pgd, addr, next,
> -				pfn + (addr >> PAGE_SHIFT), prot);
> -		if (err)
> -			break;
> -	} while (pgd++, addr = next, addr != end);
> +		err = remap_pud_range(&r, pgd++, pgd_addr_end(r.addr, end));
> +	} while (err == 0 && r.addr < end);
>  
> -	if (err)
> +	if (err) {
>  		untrack_pfn(vma, pfn, PAGE_ALIGN(size));
> +		BUG_ON(err == -EBUSY);
> +	}
>  
>  	return err;
>  }


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 3/5] io-mapping: Always create a struct to hold metadata about the io-mapping

[Chris Wilson]

Currently, we only allocate a structure to hold metadata if we need to
allocate an ioremap for every access, such as on x86-32. However, it
would be useful to store basic information about the io-mapping, such as
its page protection, on all platforms.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: linux-mm@kvack.org
---
 include/linux/io-mapping.h | 52 ++++++++++++++++++++++++++++------------------
 1 file changed, 32 insertions(+), 20 deletions(-)

diff --git a/include/linux/io-mapping.h b/include/linux/io-mapping.h
index 657fab4efab3..e053011f50bb 100644
--- a/include/linux/io-mapping.h
+++ b/include/linux/io-mapping.h
@@ -31,16 +31,17 @@
  * See Documentation/io-mapping.txt
  */
 
-#ifdef CONFIG_HAVE_ATOMIC_IOMAP
-
-#include <asm/iomap.h>
-
 struct io_mapping {
 	resource_size_t base;
 	unsigned long size;
 	pgprot_t prot;
+	void __iomem *iomem;
 };
 
+
+#ifdef CONFIG_HAVE_ATOMIC_IOMAP
+
+#include <asm/iomap.h>
 /*
  * For small address space machines, mapping large objects
  * into the kernel virtual space isn't practical. Where
@@ -119,48 +120,59 @@ io_mapping_unmap(void __iomem *vaddr)
 #else
 
 #include <linux/uaccess.h>
-
-/* this struct isn't actually defined anywhere */
-struct io_mapping;
+#include <asm/pgtable_types.h>
 
 /* Create the io_mapping object*/
 static inline struct io_mapping *
 io_mapping_create_wc(resource_size_t base, unsigned long size)
 {
-	return (struct io_mapping __force *) ioremap_wc(base, size);
+	struct io_mapping *iomap;
+
+	iomap = kmalloc(sizeof(*iomap), GFP_KERNEL);
+	if (!iomap)
+		return NULL;
+
+	iomap->base = base;
+	iomap->size = size;
+	iomap->iomem = ioremap_wc(base, size);
+	iomap->prot = pgprot_writecombine(PAGE_KERNEL_IO);
+
+	return iomap;
 }
 
 static inline void
 io_mapping_free(struct io_mapping *mapping)
 {
-	iounmap((void __force __iomem *) mapping);
+	iounmap(mapping->iomem);
+	kfree(mapping);
 }
 
-/* Atomic map/unmap */
+/* Non-atomic map/unmap */
 static inline void __iomem *
-io_mapping_map_atomic_wc(struct io_mapping *mapping,
-			 unsigned long offset)
+io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
 {
-	pagefault_disable();
-	return ((char __force __iomem *) mapping) + offset;
+	return mapping->iomem + offset;
 }
 
 static inline void
-io_mapping_unmap_atomic(void __iomem *vaddr)
+io_mapping_unmap(void __iomem *vaddr)
 {
-	pagefault_enable();
 }
 
-/* Non-atomic map/unmap */
+/* Atomic map/unmap */
 static inline void __iomem *
-io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
+io_mapping_map_atomic_wc(struct io_mapping *mapping,
+			 unsigned long offset)
 {
-	return ((char __force __iomem *) mapping) + offset;
+	pagefault_disable();
+	return io_mapping_map_wc(mapping, offset);
 }
 
 static inline void
-io_mapping_unmap(void __iomem *vaddr)
+io_mapping_unmap_atomic(void __iomem *vaddr)
 {
+	io_mapping_unmap(vaddr);
+	pagefault_enable();
 }
 
 #endif /* HAVE_ATOMIC_IOMAP */
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/5] io-mapping: Always create a struct to hold metadata about the io-mapping

[Joonas Lahtinen]

On ti, 2015-04-07 at 17:31 +0100, Chris Wilson wrote:
> Currently, we only allocate a structure to hold metadata if we need to
> allocate an ioremap for every access, such as on x86-32. However, it
> would be useful to store basic information about the io-mapping, such as
> its page protection, on all platforms.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

> Cc: linux-mm@kvack.org
> ---
>  include/linux/io-mapping.h | 52 ++++++++++++++++++++++++++++------------------
>  1 file changed, 32 insertions(+), 20 deletions(-)
> 
> diff --git a/include/linux/io-mapping.h b/include/linux/io-mapping.h
> index 657fab4efab3..e053011f50bb 100644
> --- a/include/linux/io-mapping.h
> +++ b/include/linux/io-mapping.h
> @@ -31,16 +31,17 @@
>   * See Documentation/io-mapping.txt
>   */
>  
> -#ifdef CONFIG_HAVE_ATOMIC_IOMAP
> -
> -#include <asm/iomap.h>
> -
>  struct io_mapping {
>  	resource_size_t base;
>  	unsigned long size;
>  	pgprot_t prot;
> +	void __iomem *iomem;
>  };
>  
> +
> +#ifdef CONFIG_HAVE_ATOMIC_IOMAP
> +
> +#include <asm/iomap.h>
>  /*
>   * For small address space machines, mapping large objects
>   * into the kernel virtual space isn't practical. Where
> @@ -119,48 +120,59 @@ io_mapping_unmap(void __iomem *vaddr)
>  #else
>  
>  #include <linux/uaccess.h>
> -
> -/* this struct isn't actually defined anywhere */
> -struct io_mapping;
> +#include <asm/pgtable_types.h>
>  
>  /* Create the io_mapping object*/
>  static inline struct io_mapping *
>  io_mapping_create_wc(resource_size_t base, unsigned long size)
>  {
> -	return (struct io_mapping __force *) ioremap_wc(base, size);
> +	struct io_mapping *iomap;
> +
> +	iomap = kmalloc(sizeof(*iomap), GFP_KERNEL);
> +	if (!iomap)
> +		return NULL;
> +
> +	iomap->base = base;
> +	iomap->size = size;
> +	iomap->iomem = ioremap_wc(base, size);
> +	iomap->prot = pgprot_writecombine(PAGE_KERNEL_IO);
> +
> +	return iomap;
>  }
>  
>  static inline void
>  io_mapping_free(struct io_mapping *mapping)
>  {
> -	iounmap((void __force __iomem *) mapping);
> +	iounmap(mapping->iomem);
> +	kfree(mapping);
>  }
>  
> -/* Atomic map/unmap */
> +/* Non-atomic map/unmap */
>  static inline void __iomem *
> -io_mapping_map_atomic_wc(struct io_mapping *mapping,
> -			 unsigned long offset)
> +io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
>  {
> -	pagefault_disable();
> -	return ((char __force __iomem *) mapping) + offset;
> +	return mapping->iomem + offset;
>  }
>  
>  static inline void
> -io_mapping_unmap_atomic(void __iomem *vaddr)
> +io_mapping_unmap(void __iomem *vaddr)
>  {
> -	pagefault_enable();
>  }
>  
> -/* Non-atomic map/unmap */
> +/* Atomic map/unmap */
>  static inline void __iomem *
> -io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
> +io_mapping_map_atomic_wc(struct io_mapping *mapping,
> +			 unsigned long offset)
>  {
> -	return ((char __force __iomem *) mapping) + offset;
> +	pagefault_disable();
> +	return io_mapping_map_wc(mapping, offset);
>  }
>  
>  static inline void
> -io_mapping_unmap(void __iomem *vaddr)
> +io_mapping_unmap_atomic(void __iomem *vaddr)
>  {
> +	io_mapping_unmap(vaddr);
> +	pagefault_enable();
>  }
>  
>  #endif /* HAVE_ATOMIC_IOMAP */


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 4/5] mm: Export remap_io_mapping()

[Chris Wilson]

This is similar to remap_pfn_range(), and uses the recently refactor
code to do the page table walking. The key difference is that is back
propagates its error as this is required for use from within a pagefault
handler. The other difference, is that it combine the page protection
from io-mapping, which is known from when the io-mapping is created,
with the per-vma page protection flags. This avoids having to walk the
entire system description to rediscover the special page protection
established for the io-mapping.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: linux-mm@kvack.org
---
 include/linux/mm.h |  4 ++++
 mm/memory.c        | 46 ++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 50 insertions(+)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 47a93928b90f..3dfecd58adb0 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2083,6 +2083,10 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
 			unsigned long pfn, unsigned long size, pgprot_t);
+struct io_mapping;
+int remap_io_mapping(struct vm_area_struct *,
+		     unsigned long addr, unsigned long pfn, unsigned long size,
+		     struct io_mapping *iomap);
 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
 			unsigned long pfn);
diff --git a/mm/memory.c b/mm/memory.c
index acb06f40d614..83bc5df3fafc 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -61,6 +61,7 @@
 #include <linux/string.h>
 #include <linux/dma-debug.h>
 #include <linux/debugfs.h>
+#include <linux/io-mapping.h>
 
 #include <asm/io.h>
 #include <asm/pgalloc.h>
@@ -1762,6 +1763,51 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
 EXPORT_SYMBOL(remap_pfn_range);
 
 /**
+ * remap_io_mapping - remap an IO mapping to userspace
+ * @vma: user vma to map to
+ * @addr: target user address to start at
+ * @pfn: physical address of kernel memory
+ * @size: size of map area
+ * @iomap: the source io_mapping
+ *
+ *  Note: this is only safe if the mm semaphore is held when called.
+ */
+int remap_io_mapping(struct vm_area_struct *vma,
+		     unsigned long addr, unsigned long pfn, unsigned long size,
+		     struct io_mapping *iomap)
+{
+	unsigned long end = addr + PAGE_ALIGN(size);
+	struct remap_pfn r;
+	pgd_t *pgd;
+	int err;
+
+	if (WARN_ON(addr >= end))
+		return -EINVAL;
+
+#define MUST_SET (VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP)
+	BUG_ON(is_cow_mapping(vma->vm_flags));
+	BUG_ON((vma->vm_flags & MUST_SET) != MUST_SET);
+#undef MUST_SET
+
+	r.mm = vma->vm_mm;
+	r.addr = addr;
+	r.pfn = pfn;
+	r.prot = __pgprot((pgprot_val(iomap->prot) & _PAGE_CACHE_MASK) |
+			  (pgprot_val(vma->vm_page_prot) & ~_PAGE_CACHE_MASK));
+
+	pgd = pgd_offset(r.mm, addr);
+	do {
+		err = remap_pud_range(&r, pgd++, pgd_addr_end(r.addr, end));
+	} while (err == 0 && r.addr < end);
+
+	if (err)
+		zap_page_range_single(vma, addr, r.addr - addr, NULL);
+
+	return err;
+}
+EXPORT_SYMBOL(remap_io_mapping);
+
+/**
  * vm_iomap_memory - remap memory to userspace
  * @vma: user vma to map to
  * @start: start of area
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 4/5] mm: Export remap_io_mapping()

[Joonas Lahtinen]

On ti, 2015-04-07 at 17:31 +0100, Chris Wilson wrote:
> This is similar to remap_pfn_range(), and uses the recently refactor
> code to do the page table walking. The key difference is that is back
> propagates its error as this is required for use from within a pagefault
> handler. The other difference, is that it combine the page protection
> from io-mapping, which is known from when the io-mapping is created,
> with the per-vma page protection flags. This avoids having to walk the
> entire system description to rediscover the special page protection
> established for the io-mapping.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Rik van Riel <riel@redhat.com>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Cyrill Gorcunov <gorcunov@gmail.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: linux-mm@kvack.org
> ---
>  include/linux/mm.h |  4 ++++
>  mm/memory.c        | 46 ++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 50 insertions(+)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 47a93928b90f..3dfecd58adb0 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2083,6 +2083,10 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
>  struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
>  int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
>  			unsigned long pfn, unsigned long size, pgprot_t);
> +struct io_mapping;

This is unconditional code, so just move the struct forward declaration
to the top of the file after "struct writeback_control" and others.

> +int remap_io_mapping(struct vm_area_struct *,
> +		     unsigned long addr, unsigned long pfn, unsigned long size,
> +		     struct io_mapping *iomap);
>  int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
>  int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
>  			unsigned long pfn);
> diff --git a/mm/memory.c b/mm/memory.c
> index acb06f40d614..83bc5df3fafc 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -61,6 +61,7 @@
>  #include <linux/string.h>
>  #include <linux/dma-debug.h>
>  #include <linux/debugfs.h>
> +#include <linux/io-mapping.h>
>  
>  #include <asm/io.h>
>  #include <asm/pgalloc.h>
> @@ -1762,6 +1763,51 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
>  EXPORT_SYMBOL(remap_pfn_range);
>  
>  /**
> + * remap_io_mapping - remap an IO mapping to userspace
> + * @vma: user vma to map to
> + * @addr: target user address to start at
> + * @pfn: physical address of kernel memory
> + * @size: size of map area
> + * @iomap: the source io_mapping
> + *
> + *  Note: this is only safe if the mm semaphore is held when called.
> + */
> +int remap_io_mapping(struct vm_area_struct *vma,
> +		     unsigned long addr, unsigned long pfn, unsigned long size,
> +		     struct io_mapping *iomap)
> +{
> +	unsigned long end = addr + PAGE_ALIGN(size);
> +	struct remap_pfn r;
> +	pgd_t *pgd;
> +	int err;
> +
> +	if (WARN_ON(addr >= end))
> +		return -EINVAL;
> +
> +#define MUST_SET (VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP)
> +	BUG_ON(is_cow_mapping(vma->vm_flags));
> +	BUG_ON((vma->vm_flags & MUST_SET) != MUST_SET);
> +#undef MUST_SET
> +

I think that is bit general for define name, maybe something along
REMAP_IO_NEEDED_FLAGS outside of the function... and then it doesn't
have to be #undeffed. And if it is kept inside function then at least _
prefix it. But I don't see why not make it available outside too.

Otherwise looking good.

Regards, Joonas

> +	r.mm = vma->vm_mm;
> +	r.addr = addr;
> +	r.pfn = pfn;
> +	r.prot = __pgprot((pgprot_val(iomap->prot) & _PAGE_CACHE_MASK) |
> +			  (pgprot_val(vma->vm_page_prot) & ~_PAGE_CACHE_MASK));
> +
> +	pgd = pgd_offset(r.mm, addr);
> +	do {
> +		err = remap_pud_range(&r, pgd++, pgd_addr_end(r.addr, end));
> +	} while (err == 0 && r.addr < end);
> +
> +	if (err)
> +		zap_page_range_single(vma, addr, r.addr - addr, NULL);
> +
> +	return err;
> +}
> +EXPORT_SYMBOL(remap_io_mapping);
> +
> +/**
>   * vm_iomap_memory - remap memory to userspace
>   * @vma: user vma to map to
>   * @start: start of area


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 5/5] drm/i915: Use remap_io_mapping() to prefault all PTE in a single pass

[Chris Wilson]

On an Ivybridge i7-3720qm with 1600MHz DDR3, with 32 fences,
Upload rate for 2 linear surfaces:  8134MiB/s -> 8154MiB/s
Upload rate for 2 tiled surfaces:   8625MiB/s -> 8632MiB/s
Upload rate for 4 linear surfaces:  8127MiB/s -> 8134MiB/s
Upload rate for 4 tiled surfaces:   8602MiB/s -> 8629MiB/s
Upload rate for 8 linear surfaces:  8124MiB/s -> 8137MiB/s
Upload rate for 8 tiled surfaces:   8603MiB/s -> 8624MiB/s
Upload rate for 16 linear surfaces: 8123MiB/s -> 8128MiB/s
Upload rate for 16 tiled surfaces:  8606MiB/s -> 8618MiB/s
Upload rate for 32 linear surfaces: 8121MiB/s -> 8128MiB/s
Upload rate for 32 tiled surfaces:  8605MiB/s -> 8614MiB/s
Upload rate for 64 linear surfaces: 8121MiB/s -> 8127MiB/s
Upload rate for 64 tiled surfaces:  3017MiB/s -> 5202MiB/s

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Testcase: igt/gem_fence_upload/performance
Testcase: igt/gem_mmap_gtt
Reviewed-by: Brad Volkin <bradley.d.volkin@intel.com>
Cc: linux-mm@kvack.org
---
 drivers/gpu/drm/i915/i915_gem.c | 23 ++++++-----------------
 1 file changed, 6 insertions(+), 17 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 7ab8e0039790..90d772f72276 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1667,25 +1667,14 @@ int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	pfn = dev_priv->gtt.mappable_base + i915_gem_obj_ggtt_offset(obj);
 	pfn >>= PAGE_SHIFT;
 
-	if (!obj->fault_mappable) {
-		unsigned long size = min_t(unsigned long,
-					   vma->vm_end - vma->vm_start,
-					   obj->base.size);
-		int i;
+	ret = remap_io_mapping(vma,
+			       vma->vm_start, pfn, vma->vm_end - vma->vm_start,
+			       dev_priv->gtt.mappable);
+	if (ret)
+		goto unpin;
 
-		for (i = 0; i < size >> PAGE_SHIFT; i++) {
-			ret = vm_insert_pfn(vma,
-					    (unsigned long)vma->vm_start + i * PAGE_SIZE,
-					    pfn + i);
-			if (ret)
-				break;
-		}
+	obj->fault_mappable = true;
 
-		obj->fault_mappable = true;
-	} else
-		ret = vm_insert_pfn(vma,
-				    (unsigned long)vmf->virtual_address,
-				    pfn + page_offset);
 unpin:
 	i915_gem_object_ggtt_unpin(obj);
 unlock:
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 5/5] drm/i915: Use remap_io_mapping() to prefault all PTE in a single pass

[Joonas Lahtinen]

On ti, 2015-04-07 at 17:31 +0100, Chris Wilson wrote:
> On an Ivybridge i7-3720qm with 1600MHz DDR3, with 32 fences,
> Upload rate for 2 linear surfaces:  8134MiB/s -> 8154MiB/s
> Upload rate for 2 tiled surfaces:   8625MiB/s -> 8632MiB/s
> Upload rate for 4 linear surfaces:  8127MiB/s -> 8134MiB/s
> Upload rate for 4 tiled surfaces:   8602MiB/s -> 8629MiB/s
> Upload rate for 8 linear surfaces:  8124MiB/s -> 8137MiB/s
> Upload rate for 8 tiled surfaces:   8603MiB/s -> 8624MiB/s
> Upload rate for 16 linear surfaces: 8123MiB/s -> 8128MiB/s
> Upload rate for 16 tiled surfaces:  8606MiB/s -> 8618MiB/s
> Upload rate for 32 linear surfaces: 8121MiB/s -> 8128MiB/s
> Upload rate for 32 tiled surfaces:  8605MiB/s -> 8614MiB/s
> Upload rate for 64 linear surfaces: 8121MiB/s -> 8127MiB/s
> Upload rate for 64 tiled surfaces:  3017MiB/s -> 5202MiB/s
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Testcase: igt/gem_fence_upload/performance
> Testcase: igt/gem_mmap_gtt
> Reviewed-by: Brad Volkin <bradley.d.volkin@intel.com>

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

> Cc: linux-mm@kvack.org
> ---
>  drivers/gpu/drm/i915/i915_gem.c | 23 ++++++-----------------
>  1 file changed, 6 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 7ab8e0039790..90d772f72276 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1667,25 +1667,14 @@ int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
>  	pfn = dev_priv->gtt.mappable_base + i915_gem_obj_ggtt_offset(obj);
>  	pfn >>= PAGE_SHIFT;
>  
> -	if (!obj->fault_mappable) {
> -		unsigned long size = min_t(unsigned long,
> -					   vma->vm_end - vma->vm_start,
> -					   obj->base.size);
> -		int i;
> +	ret = remap_io_mapping(vma,
> +			       vma->vm_start, pfn, vma->vm_end - vma->vm_start,
> +			       dev_priv->gtt.mappable);
> +	if (ret)
> +		goto unpin;
>  
> -		for (i = 0; i < size >> PAGE_SHIFT; i++) {
> -			ret = vm_insert_pfn(vma,
> -					    (unsigned long)vma->vm_start + i * PAGE_SIZE,
> -					    pfn + i);
> -			if (ret)
> -				break;
> -		}
> +	obj->fault_mappable = true;
>  
> -		obj->fault_mappable = true;
> -	} else
> -		ret = vm_insert_pfn(vma,
> -				    (unsigned long)vmf->virtual_address,
> -				    pfn + page_offset);
>  unpin:
>  	i915_gem_object_ggtt_unpin(obj);
>  unlock:


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>