Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1401638 > unrolled thread

[PATCH v2 0/5] of: generic infrastructure fixes

Back to article view | Back to linux.kernel

Contents

  [PATCH v2 0/5] of: generic infrastructure fixes Pantelis Antoniou <pantelis.antoniou@konsulko.com> - 2016-05-16 19:00 +0200
    [PATCH v2 3/5] of: unittest: hashed phandles unitest Pantelis Antoniou <pantelis.antoniou@konsulko.com> - 2016-05-16 19:00 +0200
      Re: [PATCH v2 3/5] of: unittest: hashed phandles unitest Geert Uytterhoeven <geert@linux-m68k.org> - 2016-05-16 21:10 +0200
      Re: [PATCH v2 3/5] of: unittest: hashed phandles unitest Rob Herring <robherring2@gmail.com> - 2016-05-16 21:40 +0200
    [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays. Pantelis Antoniou <pantelis.antoniou@konsulko.com> - 2016-05-16 19:00 +0200
      Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays. Geert Uytterhoeven <geert@linux-m68k.org> - 2016-05-16 21:10 +0200
        Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays. Pantelis Antoniou <pantelis.antoniou@konsulko.com> - 2016-05-16 21:30 +0200
          Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays. Geert Uytterhoeven <geert@linux-m68k.org> - 2016-05-16 21:50 +0200
            Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays. Geert Uytterhoeven <geert@linux-m68k.org> - 2016-05-17 08:40 +0200

#1401638 — [PATCH v2 0/5] of: generic infrastructure fixes

The first patch renames the *_node_sysfs methods to _node_post since
this is more accurate when more work takes place besides sysfs
tweaking when the next patch is using a hashtable for phandles.
This is a win for very large blobs with a large number of phandles.
Current we use an exhaustive search which is not optimal at all.

The third patch adds a unittest/benchmark for hashed phandles
along with figures about performance.

A long standing request of applied overlays populating
the symbol list is implemented; this allows staggered overlay
application with a symbol from a previous overlay being used
by another after it.

Finally a pr_fmt for the overlay files is used for
uncluttering the printout a bit.

Changes since v1:
* Reworked phandle hash according to maintainer requests
* Added unittest/benchmark for phandle hash
* Reworked the overlay symbol patch
* Added pr_fmt patch


Pantelis Antoniou (5):
  of: rename *_node_sysfs to _node_post
  of: Support hashtable lookups for phandles
  of: unittest: hashed phandles unitest
  of: overlay: Pick up label symbols from overlays.
  of: overlay: Add pr_fmt for clarity

 drivers/of/base.c       |  45 +++++++++++--
 drivers/of/dynamic.c    |  18 +++--
 drivers/of/of_private.h |  37 ++++++++++-
 drivers/of/overlay.c    | 173 ++++++++++++++++++++++++++++++++++++------------
 drivers/of/unittest.c   |  72 +++++++++++++++++++-
 include/linux/of.h      |   2 +
 6 files changed, 289 insertions(+), 58 deletions(-)

-- 
1.7.12

[toc] | [next] | [standalone]

#1401639 — [PATCH v2 3/5] of: unittest: hashed phandles unitest

Add a benchmarking hashed phandles unittest which report what kind
of speed up we get switching to hashed phandle lookups.

 ### dt-test ### the hash method is 8.2 times faster than the original

On the beaglebone we perform about 1877 phandle lookups until that
point in the unittest. Each non-hashed lookup takes about 23us when
the cash is hot, while the hash lookup takes about 3us.

For those 1877 lookup we get a speedup in the boot sequence of
1877 * (23 - 3) = 37.5ms, which is not spectacular but there's no
point in wasting cycles and energy.

Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
---
 drivers/of/unittest.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 68 insertions(+)

diff --git a/drivers/of/unittest.c b/drivers/of/unittest.c
index 7ea3689..59cad84 100644
--- a/drivers/of/unittest.c
+++ b/drivers/of/unittest.c
@@ -25,6 +25,9 @@
 
 #include <linux/bitops.h>
 
+#include <linux/timekeeping.h>
+#include <linux/random.h>
+
 #include "of_private.h"
 
 static struct unittest_results {
@@ -2266,6 +2269,70 @@ out:
 static inline void __init of_unittest_overlay(void) { }
 #endif
 
+#define PHANDLE_LOOKUPS	1000
+
+static void __init of_unittest_phandle_hash(void)
+{
+	struct device_node *node;
+	phandle max_phandle;
+	u32 ph;
+	unsigned long flags;
+	int i, j, total;
+	ktime_t start, end;
+	s64 dur[2];
+	int dec, frac;
+
+	/* test only available when hashing is available */
+	if (!of_phandle_ht_available()) {
+		pr_warn("phandle hash test requires hash to be initialized\n");
+		return;
+	}
+
+	/* find the maximum phandle of the tree */
+	raw_spin_lock_irqsave(&devtree_lock, flags);
+	max_phandle = 0;
+	total = 0;
+	for_each_of_allnodes(node) {
+		if (node->phandle != (phandle)-1U &&
+				node->phandle > max_phandle)
+			max_phandle = node->phandle;
+		total++;
+	}
+	raw_spin_unlock_irqrestore(&devtree_lock, flags);
+	max_phandle++;
+
+	pr_debug("phandle: max-phandle #%u, #%d total nodes\n",
+			(u32)max_phandle, total);
+
+	/* perform random lookups using the hash */
+	for (j = 0; j < 2; j++) {
+
+		/* disabled for pass #0, enabled for pass #1 */
+		of_phandle_ht_is_disabled = j == 0;
+
+		start = ktime_get_raw();
+		for (i = 0; i < PHANDLE_LOOKUPS; i++) {
+			ph = prandom_u32() % max_phandle;
+			node = of_find_node_by_phandle(ph);
+			of_node_put(node);
+		}
+		end = ktime_get_raw();
+
+		dur[j] = ktime_to_us(end) - ktime_to_us(start);
+		pr_debug("#%d lookups in %lld us (%s)\n",
+				PHANDLE_LOOKUPS, dur[j],
+				j == 0 ? "original" : "hashed");
+	}
+
+	unittest(dur[0] > dur[1], "Non hashing phandles are faster!?");
+
+	dec = (int)div64_s64(dur[0] * 10 + 5, dur[1]);
+	frac = dec % 10;
+	dec /= 10;
+	pr_info("the hash method is %d.%d times faster than the original\n",
+			dec, frac);
+}
+
 static int __init of_unittest(void)
 {
 	struct device_node *np;
@@ -2300,6 +2367,7 @@ static int __init of_unittest(void)
 	of_unittest_match_node();
 	of_unittest_platform_populate();
 	of_unittest_overlay();
+	of_unittest_phandle_hash();
 
 	/* Double check linkage after removing testcase data */
 	of_unittest_check_tree_linkage();
-- 
1.7.12

[toc] | [prev] | [next] | [standalone]

#1401713 — Re: [PATCH v2 3/5] of: unittest: hashed phandles unitest

On Mon, May 16, 2016 at 6:52 PM, Pantelis Antoniou
<pantelis.antoniou@konsulko.com> wrote:
> Add a benchmarking hashed phandles unittest which report what kind
> of speed up we get switching to hashed phandle lookups.
>
>  ### dt-test ### the hash method is 8.2 times faster than the original
>
> On the beaglebone we perform about 1877 phandle lookups until that
> point in the unittest. Each non-hashed lookup takes about 23us when
> the cash is hot, while the hash lookup takes about 3us.

cache

> For those 1877 lookup we get a speedup in the boot sequence of
> 1877 * (23 - 3) = 37.5ms, which is not spectacular but there's no
> point in wasting cycles and energy.
>
> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
> ---
>  drivers/of/unittest.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 68 insertions(+)
>
> diff --git a/drivers/of/unittest.c b/drivers/of/unittest.c
> index 7ea3689..59cad84 100644
> --- a/drivers/of/unittest.c
> +++ b/drivers/of/unittest.c
> @@ -25,6 +25,9 @@
>
>  #include <linux/bitops.h>
>
> +#include <linux/timekeeping.h>
> +#include <linux/random.h>
> +
>  #include "of_private.h"
>
>  static struct unittest_results {
> @@ -2266,6 +2269,70 @@ out:
>  static inline void __init of_unittest_overlay(void) { }
>  #endif
>
> +#define PHANDLE_LOOKUPS        1000
> +
> +static void __init of_unittest_phandle_hash(void)
> +{
> +       struct device_node *node;
> +       phandle max_phandle;
> +       u32 ph;
> +       unsigned long flags;
> +       int i, j, total;

unsigned int

> +       ktime_t start, end;
> +       s64 dur[2];

No idea why ktime_to_us() returns s64 i.s.o. u64...

> +       int dec, frac;

unsigned int?

> +       /* test only available when hashing is available */
> +       if (!of_phandle_ht_available()) {
> +               pr_warn("phandle hash test requires hash to be initialized\n");
> +               return;
> +       }
> +
> +       /* find the maximum phandle of the tree */
> +       raw_spin_lock_irqsave(&devtree_lock, flags);
> +       max_phandle = 0;
> +       total = 0;
> +       for_each_of_allnodes(node) {
> +               if (node->phandle != (phandle)-1U &&

Drop the "U" suffix?

> +                               node->phandle > max_phandle)
> +                       max_phandle = node->phandle;
> +               total++;
> +       }
> +       raw_spin_unlock_irqrestore(&devtree_lock, flags);
> +       max_phandle++;
> +
> +       pr_debug("phandle: max-phandle #%u, #%d total nodes\n",
> +                       (u32)max_phandle, total);

phandle is already u32, so no need for the cast.

> +
> +       /* perform random lookups using the hash */
> +       for (j = 0; j < 2; j++) {
> +
> +               /* disabled for pass #0, enabled for pass #1 */
> +               of_phandle_ht_is_disabled = j == 0;
> +
> +               start = ktime_get_raw();
> +               for (i = 0; i < PHANDLE_LOOKUPS; i++) {
> +                       ph = prandom_u32() % max_phandle;
> +                       node = of_find_node_by_phandle(ph);
> +                       of_node_put(node);
> +               }
> +               end = ktime_get_raw();
> +
> +               dur[j] = ktime_to_us(end) - ktime_to_us(start);
> +               pr_debug("#%d lookups in %lld us (%s)\n",

$u

> +                               PHANDLE_LOOKUPS, dur[j],
> +                               j == 0 ? "original" : "hashed");
> +       }
> +
> +       unittest(dur[0] > dur[1], "Non hashing phandles are faster!?");
> +
> +       dec = (int)div64_s64(dur[0] * 10 + 5, dur[1]);

I'd expect div64_u64(), if not for ktime_to_us() returning s64...

> +       frac = dec % 10;
> +       dec /= 10;
> +       pr_info("the hash method is %d.%d times faster than the original\n",

%u.%u once dec and frac are unsigned.

> +                       dec, frac);
> +}

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[toc] | [prev] | [next] | [standalone]

#1401724 — Re: [PATCH v2 3/5] of: unittest: hashed phandles unitest

On Mon, May 16, 2016 at 11:52 AM, Pantelis Antoniou
<pantelis.antoniou@konsulko.com> wrote:
> Add a benchmarking hashed phandles unittest which report what kind
> of speed up we get switching to hashed phandle lookups.
>
>  ### dt-test ### the hash method is 8.2 times faster than the original
>
> On the beaglebone we perform about 1877 phandle lookups until that
> point in the unittest. Each non-hashed lookup takes about 23us when
> the cash is hot, while the hash lookup takes about 3us.
>
> For those 1877 lookup we get a speedup in the boot sequence of
> 1877 * (23 - 3) = 37.5ms, which is not spectacular but there's no
> point in wasting cycles and energy.
>
> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
> ---
>  drivers/of/unittest.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 68 insertions(+)
>
> diff --git a/drivers/of/unittest.c b/drivers/of/unittest.c
> index 7ea3689..59cad84 100644
> --- a/drivers/of/unittest.c
> +++ b/drivers/of/unittest.c
> @@ -25,6 +25,9 @@
>
>  #include <linux/bitops.h>
>
> +#include <linux/timekeeping.h>
> +#include <linux/random.h>
> +
>  #include "of_private.h"
>
>  static struct unittest_results {
> @@ -2266,6 +2269,70 @@ out:
>  static inline void __init of_unittest_overlay(void) { }
>  #endif
>
> +#define PHANDLE_LOOKUPS        1000
> +
> +static void __init of_unittest_phandle_hash(void)
> +{
> +       struct device_node *node;
> +       phandle max_phandle;
> +       u32 ph;
> +       unsigned long flags;
> +       int i, j, total;
> +       ktime_t start, end;
> +       s64 dur[2];
> +       int dec, frac;
> +
> +       /* test only available when hashing is available */
> +       if (!of_phandle_ht_available()) {
> +               pr_warn("phandle hash test requires hash to be initialized\n");
> +               return;

As the point of the unittest is to test the core DT code, this should
be a test fail.

> +       }
> +
> +       /* find the maximum phandle of the tree */
> +       raw_spin_lock_irqsave(&devtree_lock, flags);
> +       max_phandle = 0;
> +       total = 0;
> +       for_each_of_allnodes(node) {
> +               if (node->phandle != (phandle)-1U &&
> +                               node->phandle > max_phandle)
> +                       max_phandle = node->phandle;
> +               total++;
> +       }
> +       raw_spin_unlock_irqrestore(&devtree_lock, flags);
> +       max_phandle++;
> +
> +       pr_debug("phandle: max-phandle #%u, #%d total nodes\n",
> +                       (u32)max_phandle, total);
> +
> +       /* perform random lookups using the hash */
> +       for (j = 0; j < 2; j++) {
> +
> +               /* disabled for pass #0, enabled for pass #1 */
> +               of_phandle_ht_is_disabled = j == 0;

I'm not wild about having this variable leaked from the core code just
for the unit test. Yet another step away from the unittest being a
module.

I think you should just measure current performance and users can
revert the hash table if they want to measure the slow path.

Rob

[toc] | [prev] | [next] | [standalone]

#1401640 — [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays.

Insert overlay symbols to the base tree when applied.
This makes it possible to apply an overlay that references a label
that a previously inserted overlay had.

Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
---
 drivers/of/overlay.c | 93 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)

diff --git a/drivers/of/overlay.c b/drivers/of/overlay.c
index a274d65..89ebb70 100644
--- a/drivers/of/overlay.c
+++ b/drivers/of/overlay.c
@@ -8,6 +8,7 @@
  * modify it under the terms of the GNU General Public License
  * version 2 as published by the Free Software Foundation.
  */
+
 #undef DEBUG
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -514,6 +515,91 @@ static int of_free_overlay_info(struct of_overlay *ov)
 	return 0;
 }
 
+static int of_overlay_add_symbols(
+		struct device_node *tree,
+		struct of_overlay *ov)
+{
+	struct of_overlay_info *ovinfo;
+	struct device_node *root_sym = NULL;
+	struct device_node *child = NULL;
+	struct property *prop;
+	const char *path, *s;
+	char *new_path;
+	int i, len, err;
+
+	/* both may fail (if no fixups are required) */
+	root_sym = of_find_node_by_path("/__symbols__");
+	child = of_get_child_by_name(tree, "__symbols__");
+
+	err = 0;
+	/* do nothing if either is NULL */
+	if (!root_sym || !child)
+		goto out;
+
+	for_each_property_of_node(child, prop) {
+
+		/* skip properties added automatically */
+		if (of_prop_cmp(prop->name, "name") == 0)
+			continue;
+
+		err = of_property_read_string(child,
+				prop->name, &path);
+		if (err != 0) {
+			pr_err("Could not find symbol '%s'\n", prop->name);
+			continue;
+		}
+
+		/* now find fragment index */
+		s = path;
+
+		/* compare paths to find fragment index */
+		for (i = 0, ovinfo = NULL, len = -1; i < ov->count; i++) {
+			ovinfo = &ov->ovinfo_tab[i];
+
+			pr_debug("#%d: overlay->name=%s target->name=%s\n",
+					i, ovinfo->overlay->full_name,
+					ovinfo->target->full_name);
+
+			len = strlen(ovinfo->overlay->full_name);
+			if (strncasecmp(path, ovinfo->overlay->full_name,
+						len) == 0 && path[len] == '/')
+				break;
+		}
+
+		if (i >= ov->count)
+			continue;
+
+		pr_debug("found target at #%d\n", i);
+		new_path = kasprintf(GFP_KERNEL, "%s%s",
+				ovinfo->target->full_name,
+				path + len);
+		if (!new_path) {
+			pr_err("Failed to allocate propname for \"%s\"\n",
+					prop->name);
+			err = -ENOMEM;
+			break;
+		}
+
+		err = of_changeset_add_property_string(&ov->cset, root_sym,
+				prop->name, new_path);
+
+		/* free always */
+		kfree(new_path);
+
+		if (err) {
+			pr_err("Failed to add property for \"%s\"\n",
+					prop->name);
+			break;
+		}
+	}
+
+out:
+	of_node_put(child);
+	of_node_put(root_sym);
+
+	return err;
+}
+
 static LIST_HEAD(ov_list);
 static DEFINE_IDR(ov_idr);
 
@@ -642,6 +728,13 @@ static int __of_overlay_create(struct device_node *tree,
 		goto err_abort_trans;
 	}
 
+	err = of_overlay_add_symbols(tree, ov);
+	if (err) {
+		pr_err("%s: of_overlay_add_symbols() failed for tree@%s\n",
+				__func__, tree->full_name);
+		goto err_abort_trans;
+	}
+
 	/* apply the changeset */
 	err = __of_changeset_apply(&ov->cset);
 	if (err) {
-- 
1.7.12

[toc] | [prev] | [next] | [standalone]

#1401709 — Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays.

On Mon, May 16, 2016 at 6:52 PM, Pantelis Antoniou
<pantelis.antoniou@konsulko.com> wrote:
> Insert overlay symbols to the base tree when applied.
> This makes it possible to apply an overlay that references a label
> that a previously inserted overlay had.
>
> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>

This patch hasn't changed, so I think you can keep my
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[toc] | [prev] | [next] | [standalone]

#1401721 — Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays.

Hi Geert,

> On May 16, 2016, at 22:06 , Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> 
> On Mon, May 16, 2016 at 6:52 PM, Pantelis Antoniou
> <pantelis.antoniou@konsulko.com> wrote:
>> Insert overlay symbols to the base tree when applied.
>> This makes it possible to apply an overlay that references a label
>> that a previously inserted overlay had.
>> 
>> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
> 
> This patch hasn't changed, so I think you can keep my

It’s been tweaked slightly that’s why I dropped your tested-by
signoff. Could you test this version again please to verify it
works for you?

> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
> 
> Gr{oetje,eeting}s,
> 
>                        Geert
> 

Regards

— Pantelis

> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                -- Linus Torvalds

[toc] | [prev] | [next] | [standalone]

#1401739 — Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays.

Hi Pantelis,

On Mon, May 16, 2016 at 9:27 PM, Pantelis Antoniou
<pantelis.antoniou@konsulko.com> wrote:
>> On May 16, 2016, at 22:06 , Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>> On Mon, May 16, 2016 at 6:52 PM, Pantelis Antoniou
>> <pantelis.antoniou@konsulko.com> wrote:
>>> Insert overlay symbols to the base tree when applied.
>>> This makes it possible to apply an overlay that references a label
>>> that a previously inserted overlay had.
>>>
>>> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
>>
>> This patch hasn't changed, so I think you can keep my
>
> It’s been tweaked slightly that’s why I dropped your tested-by
> signoff. Could you test this version again please to verify it
> works for you?

Oh, I missed that you did make some changes...

Will test again and report back to you...

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[toc] | [prev] | [next] | [standalone]

#1402131 — Re: [PATCH v2 4/5] of: overlay: Pick up label symbols from overlays.

On Mon, May 16, 2016 at 9:40 PM, Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
> On Mon, May 16, 2016 at 9:27 PM, Pantelis Antoniou
> <pantelis.antoniou@konsulko.com> wrote:
>>> On May 16, 2016, at 22:06 , Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>>> On Mon, May 16, 2016 at 6:52 PM, Pantelis Antoniou
>>> <pantelis.antoniou@konsulko.com> wrote:
>>>> Insert overlay symbols to the base tree when applied.
>>>> This makes it possible to apply an overlay that references a label
>>>> that a previously inserted overlay had.
>>>>
>>>> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
>>>
>>> This patch hasn't changed, so I think you can keep my
>>
>> It’s been tweaked slightly that’s why I dropped your tested-by
>> signoff. Could you test this version again please to verify it
>> works for you?
>
> Oh, I missed that you did make some changes...
>
> Will test again and report back to you...

Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web