[patch 0/4] remove nopfn

[patch 0/4] remove nopfn

[Nick Piggin]

Hi,

Can we pretty please get this into the current merge window? All it takes
is a quick review and test ;) I think mspec was already tested, but it
wouldn't hurt to verify again...

It shaves about .5K off mm/memory.o so it is pretty significant.
Mispredicted branches are also actually a significant cost in the fault
path which I'm trying to reduce (merging fault with page_mkwrite should
help with this further).

--
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 1/4] mm: allow pfnmap ->fault()s

[Nick Piggin]

Take out an assertion to allow ->fault handlers to service PFNMAP regions.
This is required to reimplement .nopfn handlers with .fault handlers and
subsequently remove nopfn.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Jes Sorensen <jes@sgi.com>
---

 mm/memory.c |    2 --
 1 file changed, 2 deletions(-)

Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -2275,8 +2275,6 @@ static int __do_fault(struct mm_struct *
 	vmf.flags = flags;
 	vmf.page = NULL;
 
-	BUG_ON(vma->vm_flags & VM_PFNMAP);
-
 	ret = vma->vm_ops->fault(vma, &vmf);
 	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))
 		return ret;

--
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 2/4] mspec: convert nopfn to fault

[Nick Piggin]

---

 drivers/char/mspec.c |   22 ++++++++++++++--------
 1 file changed, 14 insertions(+), 8 deletions(-)

Index: linux-2.6/drivers/char/mspec.c
===================================================================
--- linux-2.6.orig/drivers/char/mspec.c
+++ linux-2.6/drivers/char/mspec.c
@@ -193,25 +193,24 @@ mspec_close(struct vm_area_struct *vma)
 }
 
 /*
- * mspec_nopfn
+ * mspec_fault
  *
  * Creates a mspec page and maps it to user space.
  */
-static unsigned long
-mspec_nopfn(struct vm_area_struct *vma, unsigned long address)
+static int
+mspec_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 	unsigned long paddr, maddr;
 	unsigned long pfn;
+	pgoff_t index = vmf->pgoff;
 	int index;
 	struct vma_data *vdata = vma->vm_private_data;
 
-	BUG_ON(address < vdata->vm_start || address >= vdata->vm_end);
-	index = (address - vdata->vm_start) >> PAGE_SHIFT;
 	maddr = (volatile unsigned long) vdata->maddr[index];
 	if (maddr == 0) {
 		maddr = uncached_alloc_page(numa_node_id(), 1);
 		if (maddr == 0)
-			return NOPFN_OOM;
+			return VM_FAULT_OOM;
 
 		spin_lock(&vdata->lock);
 		if (vdata->maddr[index] == 0) {
@@ -231,13 +230,20 @@ mspec_nopfn(struct vm_area_struct *vma, 
 
 	pfn = paddr >> PAGE_SHIFT;
 
-	return pfn;
+	/*
+	 * vm_insert_pfn can fail with -EBUSY, but in that case it will
+	 * be because another thread has installed the pte first, so it
+	 * is no problem.
+	 */
+	vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);
+
+	return VM_FAULT_NOPAGE;
 }
 
 static struct vm_operations_struct mspec_vm_ops = {
 	.open = mspec_open,
 	.close = mspec_close,
-	.nopfn = mspec_nopfn
+	.fault = mspec_fault,
 };
 
 /*

--
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 2/4] mspec: convert nopfn to fault

[Robin Holt]

Sorry, I missed this original post.  Saw the add this morning and went
back to the archives.

> -static unsigned long
> -mspec_nopfn(struct vm_area_struct *vma, unsigned long address)
> +static int
> +mspec_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
>  {
>  	unsigned long paddr, maddr;
>  	unsigned long pfn;
> +	pgoff_t index = vmf->pgoff;
>  	int index;

I think this will cause problems.  Two definitions of index.  I removed
the int index and tested.  This appears to work fine.  Sorry for the
delay.

Thanks,
Robin

Index: remove_nopfn/drivers/char/mspec.c
===================================================================
--- remove_nopfn.orig/drivers/char/mspec.c	2008-05-06 07:07:30.000000000 -0500
+++ remove_nopfn/drivers/char/mspec.c	2008-05-06 07:17:01.784314587 -0500
@@ -203,7 +203,6 @@ mspec_fault(struct vm_area_struct *vma, 
 	unsigned long paddr, maddr;
 	unsigned long pfn;
 	pgoff_t index = vmf->pgoff;
-	int index;
 	struct vma_data *vdata = vma->vm_private_data;
 
 	maddr = (volatile unsigned long) vdata->maddr[index];

--
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 3/4] spufs: convert nopfn to fault

[Nick Piggin]

---

 arch/powerpc/platforms/cell/spufs/file.c     |   91 ++++++++++++---------------
 arch/powerpc/platforms/cell/spufs/sputrace.c |    8 +-
 2 files changed, 46 insertions(+), 53 deletions(-)

Index: linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/file.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
@@ -237,11 +237,10 @@ spufs_mem_write(struct file *file, const
 	return size;
 }
 
-static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int spufs_mem_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 	struct spu_context *ctx	= vma->vm_file->private_data;
-	unsigned long pfn, offset, addr0 = address;
+	unsigned long pfn, offset, address = (unsigned long)vmf->virtual_address;
 #ifdef CONFIG_SPU_FS_64K_LS
 	struct spu_state *csa = &ctx->csa;
 	int psize;
@@ -259,15 +258,14 @@ static unsigned long spufs_mem_mmap_nopf
 	}
 #endif /* CONFIG_SPU_FS_64K_LS */
 
-	offset = (address - vma->vm_start) + (vma->vm_pgoff << PAGE_SHIFT);
+	offset = vmf->pgoff << PAGE_SHIFT;
 	if (offset >= LS_SIZE)
-		return NOPFN_SIGBUS;
+		return VM_FAULT_SIGBUS;
 
-	pr_debug("spufs_mem_mmap_nopfn address=0x%lx -> 0x%lx, offset=0x%lx\n",
-		 addr0, address, offset);
+	pr_debug("spufs_mem_mmap_fault address=0x%lx, offset=0x%lx\n", address, offset);
 
 	if (spu_acquire(ctx))
-		return NOPFN_REFAULT;
+		return VM_FAULT_NOPAGE;
 
 	if (ctx->state == SPU_STATE_SAVED) {
 		vma->vm_page_prot = __pgprot(pgprot_val(vma->vm_page_prot)
@@ -278,16 +276,16 @@ static unsigned long spufs_mem_mmap_nopf
 					     | _PAGE_NO_CACHE);
 		pfn = (ctx->spu->local_store_phys + offset) >> PAGE_SHIFT;
 	}
-	vm_insert_pfn(vma, address, pfn);
+	vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);
 
 	spu_release(ctx);
 
-	return NOPFN_REFAULT;
+	return VM_FAULT_NOPAGE;
 }
 
 
 static struct vm_operations_struct spufs_mem_mmap_vmops = {
-	.nopfn = spufs_mem_mmap_nopfn,
+	.fault = spufs_mem_mmap_fault,
 };
 
 static int spufs_mem_mmap(struct file *file, struct vm_area_struct *vma)
@@ -350,20 +348,19 @@ static const struct file_operations spuf
 #endif
 };
 
-static unsigned long spufs_ps_nopfn(struct vm_area_struct *vma,
-				    unsigned long address,
+static int spufs_ps_fault(struct vm_area_struct *vma,
+				    struct vm_fault *vmf,
 				    unsigned long ps_offs,
 				    unsigned long ps_size)
 {
 	struct spu_context *ctx = vma->vm_file->private_data;
-	unsigned long area, offset = address - vma->vm_start;
+	unsigned long area, offset = vmf->pgoff << PAGE_SHIFT;
 	int ret = 0;
 
-	spu_context_nospu_trace(spufs_ps_nopfn__enter, ctx);
+	spu_context_nospu_trace(spufs_ps_fault__enter, ctx);
 
-	offset += vma->vm_pgoff << PAGE_SHIFT;
 	if (offset >= ps_size)
-		return NOPFN_SIGBUS;
+		return VM_FAULT_SIGBUS;
 
 	/*
 	 * Because we release the mmap_sem, the context may be destroyed while
@@ -377,7 +374,7 @@ static unsigned long spufs_ps_nopfn(stru
 	 * pages to hand out to the user, but we don't want to wait
 	 * with the mmap_sem held.
 	 * It is possible to drop the mmap_sem here, but then we need
-	 * to return NOPFN_REFAULT because the mappings may have
+	 * to return VM_FAULT_NOPAGE because the mappings may have
 	 * hanged.
 	 */
 	if (spu_acquire(ctx))
@@ -385,14 +382,15 @@ static unsigned long spufs_ps_nopfn(stru
 
 	if (ctx->state == SPU_STATE_SAVED) {
 		up_read(&current->mm->mmap_sem);
-		spu_context_nospu_trace(spufs_ps_nopfn__sleep, ctx);
+		spu_context_nospu_trace(spufs_ps_fault__sleep, ctx);
 		ret = spufs_wait(ctx->run_wq, ctx->state == SPU_STATE_RUNNABLE);
-		spu_context_trace(spufs_ps_nopfn__wake, ctx, ctx->spu);
+		spu_context_trace(spufs_ps_fault__wake, ctx, ctx->spu);
 		down_read(&current->mm->mmap_sem);
 	} else {
 		area = ctx->spu->problem_phys + ps_offs;
-		vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);
-		spu_context_trace(spufs_ps_nopfn__insert, ctx, ctx->spu);
+		vm_insert_pfn(vma, (unsigned long)vmf->virtual_address,
+					(area + offset) >> PAGE_SHIFT);
+		spu_context_trace(spufs_ps_fault__insert, ctx, ctx->spu);
 	}
 
 	if (!ret)
@@ -400,18 +398,18 @@ static unsigned long spufs_ps_nopfn(stru
 
 refault:
 	put_spu_context(ctx);
-	return NOPFN_REFAULT;
+	return VM_FAULT_NOPAGE;
 }
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_cntl_mmap_nopfn(struct vm_area_struct *vma,
-					   unsigned long address)
+static int spufs_cntl_mmap_fault(struct vm_area_struct *vma,
+					   struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x4000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x4000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_cntl_mmap_vmops = {
-	.nopfn = spufs_cntl_mmap_nopfn,
+	.fault = spufs_cntl_mmap_fault,
 };
 
 /*
@@ -1096,23 +1094,22 @@ static ssize_t spufs_signal1_write(struc
 	return 4;
 }
 
-static unsigned long spufs_signal1_mmap_nopfn(struct vm_area_struct *vma,
-					      unsigned long address)
+static int spufs_signal1_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 #if PAGE_SIZE == 0x1000
-	return spufs_ps_nopfn(vma, address, 0x14000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x14000, 0x1000);
 #elif PAGE_SIZE == 0x10000
 	/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
 	 * signal 1 and 2 area
 	 */
-	return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+	return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
 #else
 #error unsupported page size
 #endif
 }
 
 static struct vm_operations_struct spufs_signal1_mmap_vmops = {
-	.nopfn = spufs_signal1_mmap_nopfn,
+	.fault = spufs_signal1_mmap_fault,
 };
 
 static int spufs_signal1_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1233,23 +1230,22 @@ static ssize_t spufs_signal2_write(struc
 }
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_signal2_mmap_nopfn(struct vm_area_struct *vma,
-					      unsigned long address)
+static int spufs_signal2_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 #if PAGE_SIZE == 0x1000
-	return spufs_ps_nopfn(vma, address, 0x1c000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x1c000, 0x1000);
 #elif PAGE_SIZE == 0x10000
 	/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
 	 * signal 1 and 2 area
 	 */
-	return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+	return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
 #else
 #error unsupported page size
 #endif
 }
 
 static struct vm_operations_struct spufs_signal2_mmap_vmops = {
-	.nopfn = spufs_signal2_mmap_nopfn,
+	.fault = spufs_signal2_mmap_fault,
 };
 
 static int spufs_signal2_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1361,14 +1357,13 @@ DEFINE_SPUFS_ATTRIBUTE(spufs_signal2_typ
 		       spufs_signal2_type_set, "%llu\n", SPU_ATTR_ACQUIRE);
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_mss_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int spufs_mss_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x0000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x0000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_mss_mmap_vmops = {
-	.nopfn = spufs_mss_mmap_nopfn,
+	.fault = spufs_mss_mmap_fault,
 };
 
 /*
@@ -1423,14 +1418,13 @@ static const struct file_operations spuf
 	.mmap	 = spufs_mss_mmap,
 };
 
-static unsigned long spufs_psmap_mmap_nopfn(struct vm_area_struct *vma,
-					    unsigned long address)
+static int spufs_psmap_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x0000, 0x20000);
+	return spufs_ps_fault(vma, vmf, 0x0000, 0x20000);
 }
 
 static struct vm_operations_struct spufs_psmap_mmap_vmops = {
-	.nopfn = spufs_psmap_mmap_nopfn,
+	.fault = spufs_psmap_mmap_fault,
 };
 
 /*
@@ -1483,14 +1477,13 @@ static const struct file_operations spuf
 
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_mfc_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int spufs_mfc_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x3000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x3000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_mfc_mmap_vmops = {
-	.nopfn = spufs_mfc_mmap_nopfn,
+	.fault = spufs_mfc_mmap_fault,
 };
 
 /*
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/sputrace.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
@@ -182,10 +182,10 @@ struct spu_probe spu_probes[] = {
 	{ "spu_yield__enter", "ctx %p", spu_context_nospu_event },
 	{ "spu_deactivate__enter", "ctx %p", spu_context_nospu_event },
 	{ "__spu_deactivate__unload", "ctx %p spu %p", spu_context_event },
-	{ "spufs_ps_nopfn__enter", "ctx %p", spu_context_nospu_event },
-	{ "spufs_ps_nopfn__sleep", "ctx %p", spu_context_nospu_event },
-	{ "spufs_ps_nopfn__wake", "ctx %p spu %p", spu_context_event },
-	{ "spufs_ps_nopfn__insert", "ctx %p spu %p", spu_context_event },
+	{ "spufs_ps_fault__enter", "ctx %p", spu_context_nospu_event },
+	{ "spufs_ps_fault__sleep", "ctx %p", spu_context_nospu_event },
+	{ "spufs_ps_fault__wake", "ctx %p spu %p", spu_context_event },
+	{ "spufs_ps_fault__insert", "ctx %p spu %p", spu_context_event },
 	{ "spu_acquire_saved__enter", "ctx %p", spu_context_nospu_event },
 	{ "destroy_spu_context__enter", "ctx %p", spu_context_nospu_event },
 	{ "spufs_stop_callback__enter", "ctx %p spu %p", spu_context_event },

--
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 3/4] spufs: convert nopfn to fault

[Jeremy Kerr]

Hi Nick,

> -static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct
> *vma, -					  unsigned long address)

Aside from the > 80 character lines, all is OK here.

Acked-by: Jeremy Kerr <jk@ozlabs.org>

Cheers,


Jeremy

--
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 3/4] spufs: convert nopfn to fault

[Nick Piggin]

On Fri, May 02, 2008 at 02:06:38PM +1000, Jeremy Kerr wrote:
> Hi Nick,
> 
> > -static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct
> > *vma, -					  unsigned long address)
> 
> Aside from the > 80 character lines, all is OK here.
> 
> Acked-by: Jeremy Kerr <jk@ozlabs.org>

Great, thanks very much!

--
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 3/4] spufs: convert nopfn to fault

[Jeremy Kerr]

Hi Nick,

> > Acked-by: Jeremy Kerr <jk@ozlabs.org>
>
> Great, thanks very much!

After more testing, it looks like these patches cause a huge increase in 
load (ie, system is unresponsive for large amounts of time) for various 
tests which depend on the fault path.

I need to get some quantitative numbers, but it looks like oprofile is 
broken at the moment. More debugging coming..

Cheers,


Jeremy

--
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 3/4] spufs: convert nopfn to fault

[Nick Piggin]

On Fri, May 02, 2008 at 07:43:53PM +1000, Jeremy Kerr wrote:
> Hi Nick,
> 
> > > Acked-by: Jeremy Kerr <jk@ozlabs.org>
> >
> > Great, thanks very much!
> 
> After more testing, it looks like these patches cause a huge increase in 
> load (ie, system is unresponsive for large amounts of time) for various 
> tests which depend on the fault path.
> 
> I need to get some quantitative numbers, but it looks like oprofile is 
> broken at the moment. More debugging coming..

OK, thanks for testing that... It _should_ be 100% equivalent really,
so it must be some problem in the conversion. Don't worry too much
about getting exact numbers because any noticable difference would be
a bug.

Hmm, in spufs_mem_mmap_fault, vm_insert_pfn should just take
address (corrected for 64K), rather than the uncorrected address I
gave it...

Can't see any other problems though. Is it getting stuck looping in
faults somehow?

Thanks,
Nick

--
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 3/4] spufs: convert nopfn to fault

[Jeremy Kerr]

Hi Nick,

> Hmm, in spufs_mem_mmap_fault, vm_insert_pfn should just take
> address (corrected for 64K), rather than the uncorrected address I
> gave it...

Yep, using the 'address' var for vm_insert_pfn fixes the problem for me.

Cheers,


Jeremy

--
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 3/4] spufs: convert nopfn to fault

[Nick Piggin]

On Tue, May 06, 2008 at 01:01:26PM +1000, Jeremy Kerr wrote:
> Hi Nick,
> 
> > Hmm, in spufs_mem_mmap_fault, vm_insert_pfn should just take
> > address (corrected for 64K), rather than the uncorrected address I
> > gave it...
> 
> Yep, using the 'address' var for vm_insert_pfn fixes the problem for me.

Ah, thanks for testing. Will send an updated patch also with the warning
you noticed fixed.

Thanks,
Nick

--
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 3/4] spufs: convert nopfn to fault

[Nick Piggin]

On Fri, May 02, 2008 at 02:06:38PM +1000, Jeremy Kerr wrote:
> Hi Nick,
> 
> > -static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct
> > *vma, -					  unsigned long address)
> 
> Aside from the > 80 character lines, all is OK here.

And here is an update with <= 80 column lines...


spufs: convert nopfn to fault

From: Nick Piggin <npiggin@suse.de>

Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Jeremy Kerr <jk@ozlabs.org>
---

 arch/powerpc/platforms/cell/spufs/file.c     |   91 ++++++++++++---------------
 arch/powerpc/platforms/cell/spufs/sputrace.c |    8 +-
 2 files changed, 46 insertions(+), 53 deletions(-)

Index: linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/file.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
@@ -237,11 +237,14 @@ spufs_mem_write(struct file *file, const
 	return size;
 }
 
-static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int
+spufs_mem_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 	struct spu_context *ctx	= vma->vm_file->private_data;
-	unsigned long pfn, offset, addr0 = address;
+	unsigned long pfn, offset, address;
+
+	address = (unsigned long)vmf->virtual_address;
+
 #ifdef CONFIG_SPU_FS_64K_LS
 	struct spu_state *csa = &ctx->csa;
 	int psize;
@@ -259,15 +262,15 @@ static unsigned long spufs_mem_mmap_nopf
 	}
 #endif /* CONFIG_SPU_FS_64K_LS */
 
-	offset = (address - vma->vm_start) + (vma->vm_pgoff << PAGE_SHIFT);
+	offset = vmf->pgoff << PAGE_SHIFT;
 	if (offset >= LS_SIZE)
-		return NOPFN_SIGBUS;
+		return VM_FAULT_SIGBUS;
 
-	pr_debug("spufs_mem_mmap_nopfn address=0x%lx -> 0x%lx, offset=0x%lx\n",
-		 addr0, address, offset);
+	pr_debug("spufs_mem_mmap_fault address=0x%lx, offset=0x%lx\n",
+			address, offset);
 
 	if (spu_acquire(ctx))
-		return NOPFN_REFAULT;
+		return VM_FAULT_NOPAGE;
 
 	if (ctx->state == SPU_STATE_SAVED) {
 		vma->vm_page_prot = __pgprot(pgprot_val(vma->vm_page_prot)
@@ -278,16 +281,16 @@ static unsigned long spufs_mem_mmap_nopf
 					     | _PAGE_NO_CACHE);
 		pfn = (ctx->spu->local_store_phys + offset) >> PAGE_SHIFT;
 	}
-	vm_insert_pfn(vma, address, pfn);
+	vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);
 
 	spu_release(ctx);
 
-	return NOPFN_REFAULT;
+	return VM_FAULT_NOPAGE;
 }
 
 
 static struct vm_operations_struct spufs_mem_mmap_vmops = {
-	.nopfn = spufs_mem_mmap_nopfn,
+	.fault = spufs_mem_mmap_fault,
 };
 
 static int spufs_mem_mmap(struct file *file, struct vm_area_struct *vma)
@@ -350,20 +353,19 @@ static const struct file_operations spuf
 #endif
 };
 
-static unsigned long spufs_ps_nopfn(struct vm_area_struct *vma,
-				    unsigned long address,
+static int spufs_ps_fault(struct vm_area_struct *vma,
+				    struct vm_fault *vmf,
 				    unsigned long ps_offs,
 				    unsigned long ps_size)
 {
 	struct spu_context *ctx = vma->vm_file->private_data;
-	unsigned long area, offset = address - vma->vm_start;
+	unsigned long area, offset = vmf->pgoff << PAGE_SHIFT;
 	int ret = 0;
 
-	spu_context_nospu_trace(spufs_ps_nopfn__enter, ctx);
+	spu_context_nospu_trace(spufs_ps_fault__enter, ctx);
 
-	offset += vma->vm_pgoff << PAGE_SHIFT;
 	if (offset >= ps_size)
-		return NOPFN_SIGBUS;
+		return VM_FAULT_SIGBUS;
 
 	/*
 	 * Because we release the mmap_sem, the context may be destroyed while
@@ -377,7 +379,7 @@ static unsigned long spufs_ps_nopfn(stru
 	 * pages to hand out to the user, but we don't want to wait
 	 * with the mmap_sem held.
 	 * It is possible to drop the mmap_sem here, but then we need
-	 * to return NOPFN_REFAULT because the mappings may have
+	 * to return VM_FAULT_NOPAGE because the mappings may have
 	 * hanged.
 	 */
 	if (spu_acquire(ctx))
@@ -385,14 +387,15 @@ static unsigned long spufs_ps_nopfn(stru
 
 	if (ctx->state == SPU_STATE_SAVED) {
 		up_read(&current->mm->mmap_sem);
-		spu_context_nospu_trace(spufs_ps_nopfn__sleep, ctx);
+		spu_context_nospu_trace(spufs_ps_fault__sleep, ctx);
 		ret = spufs_wait(ctx->run_wq, ctx->state == SPU_STATE_RUNNABLE);
-		spu_context_trace(spufs_ps_nopfn__wake, ctx, ctx->spu);
+		spu_context_trace(spufs_ps_fault__wake, ctx, ctx->spu);
 		down_read(&current->mm->mmap_sem);
 	} else {
 		area = ctx->spu->problem_phys + ps_offs;
-		vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);
-		spu_context_trace(spufs_ps_nopfn__insert, ctx, ctx->spu);
+		vm_insert_pfn(vma, (unsigned long)vmf->virtual_address,
+					(area + offset) >> PAGE_SHIFT);
+		spu_context_trace(spufs_ps_fault__insert, ctx, ctx->spu);
 	}
 
 	if (!ret)
@@ -400,18 +403,18 @@ static unsigned long spufs_ps_nopfn(stru
 
 refault:
 	put_spu_context(ctx);
-	return NOPFN_REFAULT;
+	return VM_FAULT_NOPAGE;
 }
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_cntl_mmap_nopfn(struct vm_area_struct *vma,
-					   unsigned long address)
+static int spufs_cntl_mmap_fault(struct vm_area_struct *vma,
+					   struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x4000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x4000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_cntl_mmap_vmops = {
-	.nopfn = spufs_cntl_mmap_nopfn,
+	.fault = spufs_cntl_mmap_fault,
 };
 
 /*
@@ -1096,23 +1099,23 @@ static ssize_t spufs_signal1_write(struc
 	return 4;
 }
 
-static unsigned long spufs_signal1_mmap_nopfn(struct vm_area_struct *vma,
-					      unsigned long address)
+static int
+spufs_signal1_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 #if PAGE_SIZE == 0x1000
-	return spufs_ps_nopfn(vma, address, 0x14000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x14000, 0x1000);
 #elif PAGE_SIZE == 0x10000
 	/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
 	 * signal 1 and 2 area
 	 */
-	return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+	return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
 #else
 #error unsupported page size
 #endif
 }
 
 static struct vm_operations_struct spufs_signal1_mmap_vmops = {
-	.nopfn = spufs_signal1_mmap_nopfn,
+	.fault = spufs_signal1_mmap_fault,
 };
 
 static int spufs_signal1_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1233,23 +1236,23 @@ static ssize_t spufs_signal2_write(struc
 }
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_signal2_mmap_nopfn(struct vm_area_struct *vma,
-					      unsigned long address)
+static int
+spufs_signal2_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 #if PAGE_SIZE == 0x1000
-	return spufs_ps_nopfn(vma, address, 0x1c000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x1c000, 0x1000);
 #elif PAGE_SIZE == 0x10000
 	/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
 	 * signal 1 and 2 area
 	 */
-	return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+	return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
 #else
 #error unsupported page size
 #endif
 }
 
 static struct vm_operations_struct spufs_signal2_mmap_vmops = {
-	.nopfn = spufs_signal2_mmap_nopfn,
+	.fault = spufs_signal2_mmap_fault,
 };
 
 static int spufs_signal2_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1361,14 +1364,14 @@ DEFINE_SPUFS_ATTRIBUTE(spufs_signal2_typ
 		       spufs_signal2_type_set, "%llu\n", SPU_ATTR_ACQUIRE);
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_mss_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int
+spufs_mss_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x0000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x0000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_mss_mmap_vmops = {
-	.nopfn = spufs_mss_mmap_nopfn,
+	.fault = spufs_mss_mmap_fault,
 };
 
 /*
@@ -1423,14 +1426,14 @@ static const struct file_operations spuf
 	.mmap	 = spufs_mss_mmap,
 };
 
-static unsigned long spufs_psmap_mmap_nopfn(struct vm_area_struct *vma,
-					    unsigned long address)
+static int
+spufs_psmap_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x0000, 0x20000);
+	return spufs_ps_fault(vma, vmf, 0x0000, 0x20000);
 }
 
 static struct vm_operations_struct spufs_psmap_mmap_vmops = {
-	.nopfn = spufs_psmap_mmap_nopfn,
+	.fault = spufs_psmap_mmap_fault,
 };
 
 /*
@@ -1483,14 +1486,14 @@ static const struct file_operations spuf
 
 
 #if SPUFS_MMAP_4K
-static unsigned long spufs_mfc_mmap_nopfn(struct vm_area_struct *vma,
-					  unsigned long address)
+static int
+spufs_mfc_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
-	return spufs_ps_nopfn(vma, address, 0x3000, 0x1000);
+	return spufs_ps_fault(vma, vmf, 0x3000, 0x1000);
 }
 
 static struct vm_operations_struct spufs_mfc_mmap_vmops = {
-	.nopfn = spufs_mfc_mmap_nopfn,
+	.fault = spufs_mfc_mmap_fault,
 };
 
 /*
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/sputrace.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
@@ -182,10 +182,10 @@ struct spu_probe spu_probes[] = {
 	{ "spu_yield__enter", "ctx %p", spu_context_nospu_event },
 	{ "spu_deactivate__enter", "ctx %p", spu_context_nospu_event },
 	{ "__spu_deactivate__unload", "ctx %p spu %p", spu_context_event },
-	{ "spufs_ps_nopfn__enter", "ctx %p", spu_context_nospu_event },
-	{ "spufs_ps_nopfn__sleep", "ctx %p", spu_context_nospu_event },
-	{ "spufs_ps_nopfn__wake", "ctx %p spu %p", spu_context_event },
-	{ "spufs_ps_nopfn__insert", "ctx %p spu %p", spu_context_event },
+	{ "spufs_ps_fault__enter", "ctx %p", spu_context_nospu_event },
+	{ "spufs_ps_fault__sleep", "ctx %p", spu_context_nospu_event },
+	{ "spufs_ps_fault__wake", "ctx %p spu %p", spu_context_event },
+	{ "spufs_ps_fault__insert", "ctx %p spu %p", spu_context_event },
 	{ "spu_acquire_saved__enter", "ctx %p", spu_context_nospu_event },
 	{ "destroy_spu_context__enter", "ctx %p", spu_context_nospu_event },
 	{ "spufs_stop_callback__enter", "ctx %p spu %p", spu_context_event },

--
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 3/4] spufs: convert nopfn to fault

[Jeremy Kerr]

Hi Nick,

> -static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct
> *vma, -					  unsigned long address)
> +static int
> +spufs_mem_mmap_fault(struct vm_area_struct *vma, struct vm_fault
> *vmf) {
>  	struct spu_context *ctx	= vma->vm_file->private_data;
> -	unsigned long pfn, offset, addr0 = address;
> +	unsigned long pfn, offset, address;
> +
> +	address = (unsigned long)vmf->virtual_address;
> +
>  #ifdef CONFIG_SPU_FS_64K_LS
>  	struct spu_state *csa = &ctx->csa;
>  	int psize;

This will add a warning (you're "mixing declarations and code") if 
CONFIG_SPU_FS_64K_LS.

Cheers,


Jeremy

--
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 4/4] mm: remove nopfn

[Nick Piggin]

There are no users of nopfn in the tree. Remove it.

Signed-off-by: Nick Piggin <npiggin@suse.de>
---

 include/linux/mm.h |    9 -------
 mm/memory.c        |   61 ++++-------------------------------------------------
 2 files changed, 5 insertions(+), 65 deletions(-)

Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -165,8 +165,6 @@ struct vm_operations_struct {
 	void (*open)(struct vm_area_struct * area);
 	void (*close)(struct vm_area_struct * area);
 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
-	unsigned long (*nopfn)(struct vm_area_struct *area,
-			unsigned long address);
 
 	/* notification that a previously read-only page is about to become
 	 * writable, if an error is returned it will cause a SIGBUS */
@@ -674,13 +672,6 @@ static inline int page_mapped(struct pag
 }
 
 /*
- * Error return values for the *_nopfn functions
- */
-#define NOPFN_SIGBUS	((unsigned long) -1)
-#define NOPFN_OOM	((unsigned long) -2)
-#define NOPFN_REFAULT	((unsigned long) -3)
-
-/*
  * Different kinds of faults, as returned by handle_mm_fault().
  * Used to decide whether a process gets delivered SIGBUS or
  * just gets major/minor fault counters bumped up.
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -1290,6 +1290,11 @@ out:
  *
  * This function should only be called from a vm_ops->fault handler, and
  * in that case the handler should return NULL.
+ *
+ * vma cannot be a COW mapping.
+ *
+ * As this is called only for pages that do not currently exist, we
+ * do not need to flush old virtual caches or the TLB.
  */
 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
 			unsigned long pfn)
@@ -2416,59 +2421,6 @@ static int do_linear_fault(struct mm_str
 	return __do_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
 }
 
-
-/*
- * do_no_pfn() tries to create a new page mapping for a page without
- * a struct_page backing it
- *
- * As this is called only for pages that do not currently exist, we
- * do not need to flush old virtual caches or the TLB.
- *
- * We enter with non-exclusive mmap_sem (to exclude vma changes,
- * but allow concurrent faults), and pte mapped but not yet locked.
- * We return with mmap_sem still held, but pte unmapped and unlocked.
- *
- * It is expected that the ->nopfn handler always returns the same pfn
- * for a given virtual mapping.
- *
- * Mark this `noinline' to prevent it from bloating the main pagefault code.
- */
-static noinline int do_no_pfn(struct mm_struct *mm, struct vm_area_struct *vma,
-		     unsigned long address, pte_t *page_table, pmd_t *pmd,
-		     int write_access)
-{
-	spinlock_t *ptl;
-	pte_t entry;
-	unsigned long pfn;
-
-	pte_unmap(page_table);
-	BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
-	BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
-
-	pfn = vma->vm_ops->nopfn(vma, address & PAGE_MASK);
-
-	BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn));
-
-	if (unlikely(pfn == NOPFN_OOM))
-		return VM_FAULT_OOM;
-	else if (unlikely(pfn == NOPFN_SIGBUS))
-		return VM_FAULT_SIGBUS;
-	else if (unlikely(pfn == NOPFN_REFAULT))
-		return 0;
-
-	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
-
-	/* Only go through if we didn't race with anybody else... */
-	if (pte_none(*page_table)) {
-		entry = pfn_pte(pfn, vma->vm_page_prot);
-		if (write_access)
-			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
-		set_pte_at(mm, address, page_table, entry);
-	}
-	pte_unmap_unlock(page_table, ptl);
-	return 0;
-}
-
 /*
  * Fault of a previously existing named mapping. Repopulate the pte
  * from the encoded file_pte if possible. This enables swappable
@@ -2529,9 +2481,6 @@ static inline int handle_pte_fault(struc
 				if (likely(vma->vm_ops->fault))
 					return do_linear_fault(mm, vma, address,
 						pte, pmd, write_access, entry);
-				if (unlikely(vma->vm_ops->nopfn))
-					return do_no_pfn(mm, vma, address, pte,
-							 pmd, write_access);
 			}
 			return do_anonymous_page(mm, vma, address,
 						 pte, pmd, write_access);

--
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>