Skip to contentSkip to conversation

sysctl: Reduce dput(child) calls in proc_sys_fill_cache()

PATCH Superseded Markus Elfring 2 participants 8 messages Sep 26, 2024 – Oct 31, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sysctl: Reduce dput(child) calls in proc_sys_fill_cache()

From: Markus Elfring <[email protected]>
Date: Thu, 26 Sep 2024 10:10:33 +0200

A dput(child) call was immediately used after an error pointer check for a d_splice_alias() call in this function implementation. Thus call such a function instead directly before the check.

This issue was transformed by using the Coccinelle software.

---
 fs/proc/proc_sysctl.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/proc/proc_sysctl.c +4 -4
index d11ebc055ce0..97547de58218 100644
@@ -698,11 +698,11 @@ static bool proc_sys_fill_cache(struct file *file,
 			res = d_splice_alias(inode, child);
 			d_lookup_done(child);
 			if (unlikely(res)) {
-				if (IS_ERR(res)) {
-					dput(child);
-					return false;
-				}
 				dput(child);
+
+				if (IS_ERR(res))
+					return false;
+
 				child = res;
 			}
 		}

--
2.46.1

maintainer
#1 lore raw
Quoted 6 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Thu, 26 Sep 2024 10:10:33 +0200

A dput(child) call was immediately used after an error pointer check for a d_splice_alias() call in this function implementation. Thus call such a function instead directly before the check.

This message reads funny, please re-write for your v2. Here is how I would write it.

"
Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This is ok because dput gets called regardless of the value returned by IS_ERR(res).
"

Quoted 2 lines

This issue was transformed by using the Coccinelle software.

How long is the coccinelle script? If it is a reasonable size, can you please append it to the commit message. If in doubt of what "reasonable" means, just share it to the list before doing your V2.

Quoted 29 lines
---
 fs/proc/proc_sysctl.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/proc/proc_sysctl.c +4 -5
index d11ebc055ce0..97547de58218 100644
@@ -698,11 +698,11 @@ static bool proc_sys_fill_cache(struct file *file,
 			res = d_splice_alias(inode, child);
 			d_lookup_done(child);
 			if (unlikely(res)) {
-				if (IS_ERR(res)) {
-					dput(child);
-					return false;
-				}
 				dput(child);
+
+				if (IS_ERR(res))
+					return false;
+
 				child = res;
 			}
 		}
--

2.46.1

Signature · 1 line
Joel Granados
#2 lore raw
Quoted 11 lines
Quoted 3 lines

A dput(child) call was immediately used after an error pointer check for a d_splice_alias() call in this function implementation. Thus call such a function instead directly before the check.

This message reads funny, please re-write for your v2. Here is how I would write it.

"
Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This is ok because dput gets called regardless of the value returned by IS_ERR(res).
"

Do you prefer the mentioned macro name over the wording “error pointer check”?

Quoted 2 lines
Quoted 1 line

This issue was transformed by using the Coccinelle software.

How long is the coccinelle script? …

A related script for the semantic patch language was presented already according to the clarification approach “Generalising a transformation with SmPL?”.
https://lore.kernel.org/kernel-janitors/[email protected]/
https://lkml.org/lkml/2024/9/14/464
https://sympa.inria.fr/sympa/arc/cocci/2024-09/msg00004.html

Will further development ideas evolve accordingly?

Regards,
Markus

maintainer
#3 lore raw
Quoted 6 lines · Markus Elfring
Quoted 6 lines
Quoted 3 lines

A dput(child) call was immediately used after an error pointer check for a d_splice_alias() call in this function implementation. Thus call such a function instead directly before the check.

This message reads funny, please re-write for your v2. Here is how I would write it.

Quoted 7 lines
Quoted 5 lines

"
Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This is ok because dput gets called regardless of the value returned by IS_ERR(res).
"

Do you prefer the mentioned macro name over the wording “error pointer check”?

yes.

Quoted 10 lines
Quoted 4 lines
Quoted 3 lines

This issue was transformed by using the Coccinelle software.

How long is the coccinelle script? …

A related script for the semantic patch language was presented already according to the clarification approach “Generalising a transformation with SmPL?”.
https://lore.kernel.org/kernel-janitors/[email protected]/
https://lkml.org/lkml/2024/9/14/464
https://sympa.inria.fr/sympa/arc/cocci/2024-09/msg00004.html

There where several scripts in these links but non of them where too long. Can you please append the one you used for this patch to the commit message.

Thx

Signature · 1 line
Joel Granados
#4 lore raw

PATCH v2 1/1 sysctl: Reduce dput(child) calls in proc_sys_fill_cache()

From: Markus Elfring <[email protected]>
Date: Wed, 23 Oct 2024 16:54:59 +0200

Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This transformation can be performed because dput() gets called regardless of the value returned by IS_ERR(res).

This issue was transformed by using a script for the semantic patch language like the following. <SmPL>
@extended_adjustment@
expression e, f != { mutex_unlock }, x, y; @@
+f(e);
if (...)
{
<+... when != \( e = x \| y(..., &e, ...) \)
- f(e);
...+>
}
-f(e);
</SmPL>

---

V2:
* This update suggestion was rebased on source files of the software “Linux next-20241023”.

* The change description was adjusted according to the wording preferences by Joel Granados.

* An SmPL script example was appended.

 fs/proc/proc_sysctl.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/proc/proc_sysctl.c +4 -4
index 9b9dfc450cb3..b277a1ca392e 100644
@@ -698,11 +698,11 @@ static bool proc_sys_fill_cache(struct file *file,
 			res = d_splice_alias(inode, child);
 			d_lookup_done(child);
 			if (unlikely(res)) {
-				if (IS_ERR(res)) {
-					dput(child);
-					return false;
-				}
 				dput(child);
+
+				if (IS_ERR(res))
+					return false;
+
 				child = res;
 			}
 		}

--
2.47.0

maintainer
#5 lore raw
Quoted 10 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Wed, 23 Oct 2024 16:54:59 +0200

Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This transformation can be performed because dput() gets called regardless of the value returned by IS_ERR(res).

This issue was transformed by using a script for the semantic patch language like the following. <SmPL>

Aren't you missing a "virtual patch" here?

Is there another way to run it besides this command:
make coccicheck MODE=patch SPFLAGS="--in-place --include-headers --smpl-spacing --jobs=16" COCCI=SCRIPT

Best

Quoted 52 lines

@extended_adjustment@
expression e, f != { mutex_unlock }, x, y;
@@

(quoted diff) +1 -2
+f(e);
 if (...)
 {
 <+... when != \( e = x \| y(..., &e, ...) \)
-   f(e);
 ...+>
 }
-f(e);

</SmPL>

---

V2:
* This update suggestion was rebased on source files of the software “Linux next-20241023”.

* The change description was adjusted according to the wording preferences by Joel Granados.

* An SmPL script example was appended.

 fs/proc/proc_sysctl.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/proc/proc_sysctl.c +4 -5
index 9b9dfc450cb3..b277a1ca392e 100644
@@ -698,11 +698,11 @@ static bool proc_sys_fill_cache(struct file *file,
 			res = d_splice_alias(inode, child);
 			d_lookup_done(child);
 			if (unlikely(res)) {
-				if (IS_ERR(res)) {
-					dput(child);
-					return false;
-				}
 				dput(child);
+
+				if (IS_ERR(res))
+					return false;
+
 				child = res;
 			}
 		}
--

2.47.0

Signature · 1 line
Joel Granados
#6 lore raw

…

Quoted 4 lines
Quoted 3 lines

This issue was transformed by using a script for the semantic patch language like the following. <SmPL>

Aren't you missing a "virtual patch" here?

Yes.

I presented an SmPL script example which would not be directly used for the tool “coccicheck”.

Quoted 1 line

Is there another way to run it besides this command:

Another command example:
time spatch -timeout 23 -j4 --chunksize 1 --no-loops -dir . …/Projekte/Coccinelle/janitor/avoid_duplicate_function_call2.cocci > …/Projekte/Bau/Linux/scripts/Coccinelle/tuning1/next/20240913/avoid_duplicate_function_call2-no_loops.diff 2> …/Projekte/Bau/Linux/scripts/Coccinelle/tuning1/next/20240913/avoid_duplicate_function_call2-no_loops-errors.txt

Regards,
Markus

maintainer Reviewed-by on v2 1/1
#7 lore raw
In reply to Markus Elfring
Quoted 2 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Wed, 23 Oct 2024 16:54:59 +0200

Small comment: Remember to send out a new patch-set email in order to make it easier for tools like b4.

Quoted 58 lines

Replace two dput(child) calls with one that occurs immediately before the IS_ERR evaluation. This transformation can be performed because dput() gets called regardless of the value returned by IS_ERR(res).

This issue was transformed by using a script for the semantic patch language like the following. <SmPL>
@extended_adjustment@
expression e, f != { mutex_unlock }, x, y; @@

(quoted diff) +1 -2
+f(e);
 if (...)
 {
 <+... when != \( e = x \| y(..., &e, ...) \)
-   f(e);
 ...+>
 }
-f(e);

</SmPL>

---

V2:
* This update suggestion was rebased on source files of the software “Linux next-20241023”.

* The change description was adjusted according to the wording preferences by Joel Granados.

* An SmPL script example was appended.

 fs/proc/proc_sysctl.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/proc/proc_sysctl.c +4 -5
index 9b9dfc450cb3..b277a1ca392e 100644
@@ -698,11 +698,11 @@ static bool proc_sys_fill_cache(struct file *file,
 			res = d_splice_alias(inode, child);
 			d_lookup_done(child);
 			if (unlikely(res)) {
-				if (IS_ERR(res)) {
-					dput(child);
-					return false;
-				}
 				dput(child);
+
+				if (IS_ERR(res))
+					return false;
+
 				child = res;
 			}
 		}
--
Quoted 2 lines

2.47.0

This looks good to me. I'll push it to sysctl-next. Thx for the patch

Best

Signature · 1 line
Joel Granados