hugetlbfs: fix hugetlb_get_unmapped_area

PATCH In review Xiao Guangrong 2 participants 13 messages Jan 13, 2012 – Mar 8, 2012

Subsystems (5)

Memory management x86 x86 MM Memory mapping +1 more subsystems

lore

Share & tools

lore

Xiao Guangrong

author

2012-01-13 11:46 UTC #0 lore raw

PATCH 1/5 hugetlbfs: fix hugetlb_get_unmapped_area

Using/updating cached_hole_size and free_area_cache properly to speedup find free region

Signed-off-by Xiao Guangrong <[email protected]>

---
 fs/hugetlbfs/inode.c |   15 +++++++++++----
 1 files changed, 11 insertions(+), 4 deletions(-)

fs/hugetlbfs/inode.c +11 -4

index e425ad9..9e0794a 100644
@@ -154,10 +154,12 @@ hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
 			return addr;
 	}
 
-	start_addr = mm->free_area_cache;
-
-	if (len <= mm->cached_hole_size)
+	if (len > mm->cached_hole_size)
+		start_addr = mm->free_area_cache;
+	else {
 		start_addr = TASK_UNMAPPED_BASE;
+		mm->cached_hole_size = 0;
+	}
 
 full_search:
 	addr = ALIGN(start_addr, huge_page_size(h));

@@ -171,13 +173,18 @@ full_search:
 			 */
 			if (start_addr != TASK_UNMAPPED_BASE) {
 				start_addr = TASK_UNMAPPED_BASE;
+				mm->cached_hole_size = 0;
 				goto full_search;
 			}
 			return -ENOMEM;
 		}
 
-		if (!vma || addr + len <= vma->vm_start)
+		if (!vma || addr + len <= vma->vm_start) {
+			mm->free_area_cache = addr + len;
 			return addr;
+		}
+		if (addr + mm->cached_hole_size < vma->vm_start)
+			mm->cached_hole_size = vma->vm_start - addr;
 		addr = ALIGN(vma->vm_end, huge_page_size(h));
 	}
 }

Signature · 1 line

1.7.7.5

Xiao Guangrong

author

2012-01-13 11:47 UTC #1 lore raw

PATCH 2/5 hugetlb: drop prev_vma in hugetlb_get_unmapped_area_topdown

Afte call find_vma_prev(mm, addr, &prev_vma), following condition is always true:

	!prev_vma || (addr >= prev_vma->vm_end)

it can be happily drop prev_vma and use find_vma instead of find_vma_prev

Signed-off-by Xiao Guangrong <[email protected]>

---
 arch/x86/mm/hugetlbpage.c |   21 ++++++---------------
 1 files changed, 6 insertions(+), 15 deletions(-)

arch/x86/mm/hugetlbpage.c +6 ~~-15~~

index f581a18..e12debc 100644
@@ -308,7 +308,7 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 {
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
-	struct vm_area_struct *vma, *prev_vma;
+	struct vm_area_struct *vma;
 	unsigned long base = mm->mmap_base, addr = addr0;
 	unsigned long largest_hole = mm->cached_hole_size;
 	int first_time = 1;

@@ -333,24 +333,15 @@ try_again:
 		 * Lookup failure means no vma is above this address,
 		 * i.e. return with success:
 		 */
-		if (!(vma = find_vma_prev(mm, addr, &prev_vma)))
-			return addr;
-
-		/*
-		 * new region fits between prev_vma->vm_end and
-		 * vma->vm_start, use it:
-		 */
-		if (addr + len <= vma->vm_start &&
-		            (!prev_vma || (addr >= prev_vma->vm_end))) {
+		vma = find_vma(mm, addr);
+		if (!vma || addr + len <= vma->vm_start) {
 			/* remember the address as a hint for next time */
 		        mm->cached_hole_size = largest_hole;
 		        return (mm->free_area_cache = addr);
-		} else {
+		} else if (mm->free_area_cache == vma->vm_end) {
 			/* pull free_area_cache down to the first hole */
-		        if (mm->free_area_cache == vma->vm_end) {
-				mm->free_area_cache = vma->vm_start;
-				mm->cached_hole_size = largest_hole;
-			}
+			mm->free_area_cache = vma->vm_start;
+			mm->cached_hole_size = largest_hole;
 		}
 
 		/* remember the largest hole we saw so far */

Signature · 8 lines

1.7.7.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Andrew Morton

maintainer

2012-03-07 22:01 UTC #2 lore raw

Re: [PATCH 2/5] hugetlb: drop prev_vma in hugetlb_get_unmapped_area_topdown

Quoted 4 lines · On Fri, 13 Jan 2012 19:44:53 +0800 Xiao Guangrong…

Afte call find_vma_prev(mm, addr, &prev_vma), following condition is always true:
	!prev_vma || (addr >= prev_vma->vm_end)
it can be happily drop prev_vma and use find_vma instead of find_vma_prev

I had to rework this patch due to 097d59106a8e4b ("vm: avoid using find_vma_prev() unnecessarily") in mainline. Can you please check my handiwork?

From: Xiao Guangrong <[email protected]>
Subject: hugetlb: drop prev_vma in hugetlb_get_unmapped_area_topdown()

After the call find_vma_prev(mm, addr, &prev_vma), the following condition is always true:

	!prev_vma || (addr >= prev_vma->vm_end)

it can happily drop prev_vma and use find_vma() instead of find_vma_prev().

Signed-off-by Xiao Guangrong <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Ingo Molnar <[email protected]>

Cc "H. Peter Anvin" <[email protected]>

Cc Michal Hocko <[email protected]>

Cc Hillf Danton <[email protected]>

Cc Andrea Arcangeli <[email protected]>

Cc KAMEZAWA Hiroyuki <[email protected]>

Signed-off-by Andrew Morton <[email protected]>

---

 arch/x86/mm/hugetlbpage.c |   23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)

arch/x86/mm/hugetlbpage.c renamed from arch/x86/mm/hugetlbpage.c~hugetlb-drop-prev_vma-in-hugetlb_get_unmapped_area_topdown +7 ~~-16~~

diff -puN arch/x86/mm/hugetlbpage.c~hugetlb-drop-prev_vma-in-hugetlb_get_unmapped_area_topdown arch/x86/mm/hugetlbpage.c
@@ -308,7 +308,7 @@ static unsigned long hugetlb_get_unmappe
 {
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
-	struct vm_area_struct *vma, *prev_vma;
+	struct vm_area_struct *vma;
 	unsigned long base = mm->mmap_base, addr = addr0;
 	unsigned long largest_hole = mm->cached_hole_size;
 	int first_time = 1;

@@ -334,25 +334,16 @@ try_again:
 		 * i.e. return with success:
 		 */
 		vma = find_vma(mm, addr);
-		if (!vma)
-			return addr;
-
-		/*
-		 * new region fits between prev_vma->vm_end and
-		 * vma->vm_start, use it:
-		 */
-		prev_vma = vma->vm_prev;
-		if (addr + len <= vma->vm_start &&
-		            (!prev_vma || (addr >= prev_vma->vm_end))) {
+		if (vma)
+			prev_vma = vma->vm_prev;
+		if (!vma || addr + len <= vma->vm_start) {
 			/* remember the address as a hint for next time */
 		        mm->cached_hole_size = largest_hole;
 		        return (mm->free_area_cache = addr);
-		} else {
+		} else if (mm->free_area_cache == vma->vm_end) {
 			/* pull free_area_cache down to the first hole */
-		        if (mm->free_area_cache == vma->vm_end) {
-				mm->free_area_cache = vma->vm_start;
-				mm->cached_hole_size = largest_hole;
-			}
+			mm->free_area_cache = vma->vm_start;
+			mm->cached_hole_size = largest_hole;
 		}
 
 		/* remember the largest hole we saw so far */

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Xiao Guangrong

author

2012-03-08 02:29 UTC #3 lore raw

Re: [PATCH 2/5] hugetlb: drop prev_vma in hugetlb_get_unmapped_area_topdown

Quoted 12 lines · Andrew Morton

Quoted 4 lines · On Fri, 13 Jan 2012 19:44:53 +0800 Xiao Guangrong…
Afte call find_vma_prev(mm, addr, &prev_vma), following condition is always true:
	!prev_vma || (addr >= prev_vma->vm_end)
it can be happily drop prev_vma and use find_vma instead of find_vma_prev
I had to rework this patch due to 097d59106a8e4b ("vm: avoid using find_vma_prev() unnecessarily") in mainline. Can you please check my handiwork?

It looks good to me, thanks Andrew!

Xiao Guangrong

author

2012-01-13 11:47 UTC #4 lore raw

PATCH 3/5 hugetlb: try to search again if it is really needed

In reply to Xiao Guangrong

Search again only if some holes may be skipped in the first time

Signed-off-by Xiao Guangrong <[email protected]>

---
 arch/x86/mm/hugetlbpage.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

arch/x86/mm/hugetlbpage.c +4 -4

index e12debc..6bf5735 100644
@@ -309,9 +309,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
-	unsigned long base = mm->mmap_base, addr = addr0;
+	unsigned long base = mm->mmap_base, addr = addr0, start_addr;
 	unsigned long largest_hole = mm->cached_hole_size;
-	int first_time = 1;
 
 	/* don't allow allocations above current base */
 	if (mm->free_area_cache > base)

@@ -322,6 +321,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 		mm->free_area_cache  = base;
 	}
 try_again:
+	start_addr = mm->free_area_cache;
+
 	/* make sure it can fit in the remaining address space */
 	if (mm->free_area_cache < len)
 		goto fail;

@@ -357,10 +358,9 @@ fail:
 	 * if hint left us with no space for the requested
 	 * mapping then try again:
 	 */
-	if (first_time) {
+	if (start_addr != base) {
 		mm->free_area_cache = base;
 		largest_hole = 0;
-		first_time = 0;
 		goto try_again;
 	}
 	/*

Signature · 8 lines

1.7.7.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Andrew Morton

maintainer

2012-02-01 22:43 UTC #5 lore raw

Re: [PATCH 3/5] hugetlb: try to search again if it is really needed

Quoted 12 lines · On Fri, 13 Jan 2012 19:45:45 +0800 Xiao Guangrong…

Search again only if some holes may be skipped in the first time
Signed-off-by Xiao Guangrong <[email protected]>
---
 arch/x86/mm/hugetlbpage.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
arch/x86/mm/hugetlbpage.c +0 -0
index e12debc..6bf5735 100644

arch/x86/mm/hugetlbpage.c static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, · line 312

Quoted 5 lines

arch/x86/mm/hugetlbpage.c +1 -1

@@ -309,9 +309,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
-	unsigned long base = mm->mmap_base, addr = addr0;
+	unsigned long base = mm->mmap_base, addr = addr0, start_addr;

grr. The multiple-definitions-per-line thing is ugly, makes for more patch conflicts and reduces opportunities to add useful comments.

arch/x86/mm/hugetlbpage.c renamed from arch/x86/mm/hugetlbpage.c~hugetlb-try-to-search-again-if-it-is-really-needed-fix +3 -1

@@ -309,7 +309,9 @@ static unsigned long hugetlb_get_unmappe
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
-	unsigned long base = mm->mmap_base, addr = addr0, start_addr;
+	unsigned long base = mm->mmap_base;
+	unsigned long addr = addr0;
+	unsigned long start_addr;
 	unsigned long largest_hole = mm->cached_hole_size;
 
 	/* don't allow allocations above current base */

Quoted 15 lines

arch/x86/mm/hugetlbpage.c +2 -1

 	unsigned long largest_hole = mm->cached_hole_size;
-	int first_time = 1;
 
 	/* don't allow allocations above current base */
 	if (mm->free_area_cache > base)

@@ -322,6 +321,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 		mm->free_area_cache  = base;
 	}
 try_again:
+	start_addr = mm->free_area_cache;
+
 	/* make sure it can fit in the remaining address space */
 	if (mm->free_area_cache < len)
 		goto fail;

arch/x86/mm/hugetlbpage.c fail: · line 364

Quoted 9 lines

arch/x86/mm/hugetlbpage.c +1 -2

@@ -357,10 +358,9 @@ fail:
 	 * if hint left us with no space for the requested
 	 * mapping then try again:
 	 */
-	if (first_time) {
+	if (start_addr != base) {
 		mm->free_area_cache = base;
 		largest_hole = 0;
-		first_time = 0;
 		goto try_again;

The code used to retry a single time. With this change the retrying is potentially infinite. What is the reason for this change? What is the potential for causing a lockup?

Xiao Guangrong

author

2012-02-02 05:22 UTC #6 lore raw

Re: [PATCH 3/5] hugetlb: try to search again if it is really needed

Hi Andrew,

Thanks for your review!

Quoted 10 lines · Andrew Morton

Quoted 7 lines · On Fri, 13 Jan 2012 19:45:45 +0800 Xiao Guangrong…
Search again only if some holes may be skipped in the first time
Signed-off-by Xiao Guangrong <[email protected]>
---
 arch/x86/mm/hugetlbpage.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

Quoted 13 lines

arch/x86/mm/hugetlbpage.c static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, · line 312
Quoted 10 lines
arch/x86/mm/hugetlbpage.c +1 -1
index e12debc..6bf5735 100644
@@ -309,9 +309,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 	struct hstate *h = hstate_file(file);
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
-	unsigned long base = mm->mmap_base, addr = addr0;
+	unsigned long base = mm->mmap_base, addr = addr0, start_addr;
grr. The multiple-definitions-per-line thing is ugly, makes for more patch conflicts and reduces opportunities to add useful comments.

Yes, thanks.

Quoted 12 lines

arch/x86/mm/hugetlbpage.c +3 -1

 try_again:
+	start_addr = mm->free_area_cache;
+
 	/* make sure it can fit in the remaining address space */
 	if (mm->free_area_cache < len)
 		goto fail;

@@ -357,10 +358,9 @@ fail:
 	 * if hint left us with no space for the requested
 	 * mapping then try again:
 	 */
-	if (first_time) {
+	if (start_addr != base) {

Quoted 9 lines

arch/x86/mm/hugetlbpage.c fail: · line 364
Quoted 4 lines
arch/x86/mm/hugetlbpage.c +0 -1
 		mm->free_area_cache = base;
 		largest_hole = 0;
-		first_time = 0;
 		goto try_again;
The code used to retry a single time. With this change the retrying is potentially infinite. What is the reason for this change? What is the potential for causing a lockup?

Actually, it is not infinite, after retry, we set "mm->free_area_cache = base", "start_addr" will set to "base" in the second search, so the condition of "goto try_again" will be unsatisfied.

Xiao Guangrong

author

2012-01-13 11:49 UTC #7 lore raw

PATCH 4/5 mm: do not reset cached_hole_size when vma is unmapped

In reply to Xiao Guangrong

In current code, cached_hole_size is set to the maximal value if the unmapped vma is under free_area_cache, next search will search from the base addr

Actually, we can keep cached_hole_size so that if next required size is more that cached_hole_size, it can search from free_area_cache

Signed-off-by Xiao Guangrong <[email protected]>

---
 mm/mmap.c |    4 +---
 1 files changed, 1 insertions(+), 3 deletions(-)

mm/mmap.c +1 -3

index 3f758c7..970f572 100644
@@ -1423,10 +1423,8 @@ void arch_unmap_area(struct mm_struct *mm, unsigned long addr)
 	/*
 	 * Is this a new hole at the lowest possible address?
 	 */
-	if (addr >= TASK_UNMAPPED_BASE && addr < mm->free_area_cache) {
+	if (addr >= TASK_UNMAPPED_BASE && addr < mm->free_area_cache)
 		mm->free_area_cache = addr;
-		mm->cached_hole_size = ~0UL;
-	}
 }
 
 /*

Signature · 8 lines

1.7.7.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Xiao Guangrong

author

2012-01-13 11:50 UTC #8 lore raw

PATCH 5/5 mm: search from free_area_cache for the bigger size

In reply to Xiao Guangrong

If the required size is bigger than cached_hole_size, we would better search from free_area_cache, it is more easier to get free region, specifically for the 64 bit process whose address space is large enough

Do it just as hugetlb_get_unmapped_area_topdown() in arch/x86/mm/hugetlbpage.c

Signed-off-by Xiao Guangrong <[email protected]>

---
 arch/x86/kernel/sys_x86_64.c |   34 +++++++++++++++++-----------------
 mm/mmap.c                    |   36 +++++++++++++++++++++---------------
 2 files changed, 38 insertions(+), 32 deletions(-)

arch/x86/kernel/sys_x86_64.c +17 ~~-17~~

index 0514890..ef59642 100644
@@ -195,7 +195,7 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 {
 	struct vm_area_struct *vma;
 	struct mm_struct *mm = current->mm;
-	unsigned long addr = addr0;
+	unsigned long addr = addr0, start_addr;
 
 	/* requested length too big for entire address space */
 	if (len > TASK_SIZE)

@@ -223,25 +223,14 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 		mm->free_area_cache = mm->mmap_base;
 	}
 
+try_again:
 	/* either no address requested or can't fit in requested address hole */
-	addr = mm->free_area_cache;
-
-	/* make sure it can fit in the remaining address space */
-	if (addr > len) {
-		unsigned long tmp_addr = align_addr(addr - len, filp,
-						    ALIGN_TOPDOWN);
-
-		vma = find_vma(mm, tmp_addr);
-		if (!vma || tmp_addr + len <= vma->vm_start)
-			/* remember the address as a hint for next time */
-			return mm->free_area_cache = tmp_addr;
-	}
-
-	if (mm->mmap_base < len)
-		goto bottomup;
+	start_addr = addr = mm->free_area_cache;
 
-	addr = mm->mmap_base-len;
+	if (addr < len)
+		goto fail;
 
+	addr -= len;
 	do {
 		addr = align_addr(addr, filp, ALIGN_TOPDOWN);

@@ -263,6 +252,17 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 		addr = vma->vm_start-len;
 	} while (len < vma->vm_start);
 
+fail:
+	/*
+	 * if hint left us with no space for the requested
+	 * mapping then try again:
+	 */
+	if (start_addr != mm->mmap_base) {
+		mm->free_area_cache = mm->mmap_base;
+		mm->cached_hole_size = 0;
+		goto try_again;
+	}
+
 bottomup:
 	/*
 	 * A failed mmap() very likely causes application failure,

mm/mmap.c +21 ~~-15~~

index 970f572..e3c4b97 100644
@@ -1439,7 +1439,7 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 {
 	struct vm_area_struct *vma;
 	struct mm_struct *mm = current->mm;
-	unsigned long addr = addr0;
+	unsigned long addr = addr0, start_addr;
 
 	/* requested length too big for entire address space */
 	if (len > TASK_SIZE)

@@ -1463,22 +1463,14 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
  		mm->free_area_cache = mm->mmap_base;
  	}
 
+try_again:
 	/* either no address requested or can't fit in requested address hole */
-	addr = mm->free_area_cache;
+	start_addr = addr = mm->free_area_cache;
 
-	/* make sure it can fit in the remaining address space */
-	if (addr > len) {
-		vma = find_vma(mm, addr-len);
-		if (!vma || addr <= vma->vm_start)
-			/* remember the address as a hint for next time */
-			return (mm->free_area_cache = addr-len);
-	}
-
-	if (mm->mmap_base < len)
-		goto bottomup;
-
-	addr = mm->mmap_base-len;
+	if (addr < len)
+		goto fail;
 
+	addr -= len;
 	do {
 		/*
 		 * Lookup failure means no vma is above this address,

@@ -1498,7 +1490,21 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 		addr = vma->vm_start-len;
 	} while (len < vma->vm_start);
 
-bottomup:
+fail:
+	/*
+	 * if hint left us with no space for the requested
+	 * mapping then try again:
+	 *
+	 * Note: this is different with the case of bottomup
+	 * which does the fully line-search, but we use find_vma
+	 * here that causes some holes skipped.
+	 */
+	if (start_addr != mm->mmap_base) {
+		mm->free_area_cache = mm->mmap_base;
+		mm->cached_hole_size = 0;
+		goto try_again;
+	}
+
 	/*
 	 * A failed mmap() very likely causes application failure,
 	 * so fall back to the bottom-up function here. This scenario

Signature · 8 lines

1.7.7.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Andrew Morton

maintainer

2012-02-01 22:44 UTC #9 lore raw

Re: [PATCH 5/5] mm: search from free_area_cache for the bigger size

Quoted 5 lines · On Fri, 13 Jan 2012 19:47:31 +0800 Xiao Guangrong…

If the required size is bigger than cached_hole_size, we would better search from free_area_cache, it is more easier to get free region, specifically for the 64 bit process whose address space is large enough
Do it just as hugetlb_get_unmapped_area_topdown() in arch/x86/mm/hugetlbpage.c

Can this cause additional fragmentation of the virtual address region? If so, what might be the implications of this?

Xiao Guangrong

author

2012-02-02 07:10 UTC #10 lore raw

Re: [PATCH 5/5] mm: search from free_area_cache for the bigger size

Quoted 11 lines · Andrew Morton

Quoted 5 lines · On Fri, 13 Jan 2012 19:47:31 +0800 Xiao Guangrong…
If the required size is bigger than cached_hole_size, we would better search from free_area_cache, it is more easier to get free region, specifically for the 64 bit process whose address space is large enough
Do it just as hugetlb_get_unmapped_area_topdown() in arch/x86/mm/hugetlbpage.c
Can this cause additional fragmentation of the virtual address region? If so, what might be the implications of this?

Hmm, i think it is not bad since we have cached_hole_size, and, this way is also used in other functions and architectures(arch_get_unmapped_area, hugetlb_get_unmapped_area_bottomup, hugetlb_get_unmapped_area_topdown......).

Xiao Guangrong

author

2012-01-31 06:08 UTC #11 lore raw

In reply to Xiao Guangrong

Quoted 3 lines · Xiao Guangrong

Using/updating cached_hole_size and free_area_cache properly to speedup find free region

Ping...???

Andrew Morton

maintainer

2012-02-01 22:43 UTC #12 lore raw

Quoted 8 lines · On Tue, 31 Jan 2012 14:05:50 +0800 Xiao Guangrong…

Quoted 3 lines · Xiao Guangrong
Using/updating cached_hole_size and free_area_cache properly to speedup find free region
Ping...???

I'd saved the patches away, hoping that someone who works on hugetlb would comment. They didn't. We haven't heard from wli in a long time.