Skip to contentSkip to conversation

padata: Honor the caller's alignment in case of chunk_size 0

PATCH In review Kamlesh Gurudasani 5 participants 9 messages Aug 21, 2024 – Aug 30, 2024

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 padata: Honor the caller's alignment in case of chunk_size 0

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
---
 kernel/padata.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +4 -8
index 0fa6c2895460..d8a51eff1581 100644
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
 	/*
 	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by
 	 * increasing the number of chunks, guarantee at least the minimum
 	 * chunk size from the caller, and honor the caller's alignment.
+	 * Ensure chunk_size is at least 1 to prevent divide-by-0
+	 * panic in padata_mt_helper().
 	 */
 	ps.chunk_size = job->size / (ps.nworks * load_balance_factor);
 	ps.chunk_size = max(ps.chunk_size, job->min_chunk);
+	ps.chunk_size = max(ps.chunk_size, 1ul);
 	ps.chunk_size = roundup(ps.chunk_size, job->align);
 
-	/*
-	 * chunk_size can be 0 if the caller sets min_chunk to 0. So force it
-	 * to at least 1 to prevent divide-by-0 panic in padata_mt_helper().`
-	 */
-	if (!ps.chunk_size)
-		ps.chunk_size = 1U;
-
 	list_for_each_entry(pw, &works, pw_list)
 		if (job->numa_aware) {
 			int old_node = atomic_read(&last_used_nid);

---
base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
change-id: 20240822-max-93c17adc6457

Best regards,

Signature · 1 line
Kamlesh Gurudasani <[email protected]>
#1 lore raw
Quoted 8 lines · On Thu, 22 Aug 2024 02:32:52 +0530 Kamlesh Gurudasani <[email protected]> wrote:

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Please (as always) describe the userspace-visible runtime effects of this change. This helps others to determine which kernel(s) should be patched. And it helps others decide whether this fix might address an issue which they are encountering.

#2 lore raw
Quoted 12 lines · Andrew Morton

This message was sent from outside of Texas Instruments.
Do not click links or open attachments unless you recognize the source of this email and know the content is safe. Report Suspicious

Quoted 6 lines · On Thu, 22 Aug 2024 02:32:52 +0530 Kamlesh Gurudasani <[email protected]> wrote:

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

Quoted 7 lines
Quoted 2 lines

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Please (as always) describe the userspace-visible runtime effects of this change. This helps others to determine which kernel(s) should be patched. And it helps others decide whether this fix might address an issue which they are encountering.

Thanks for the review, Andrew.
Honestly, I'm not sure the effects would be visble to userspace or not. or even how to reproduce it.

I have fixed according to discussion here,
https://patchwork.kernel.org/project/linux-crypto/patch/[email protected]/#25984314

Kamlesh

Acked-by on v1 1/1
#3 lore raw
In reply to Kamlesh Gurudasani
Quoted 47 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
---
  kernel/padata.c | 12 ++++--------
  1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +4 -9
index 0fa6c2895460..d8a51eff1581 100644
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
  	/*
  	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by
  	 * increasing the number of chunks, guarantee at least the minimum
  	 * chunk size from the caller, and honor the caller's alignment.
+	 * Ensure chunk_size is at least 1 to prevent divide-by-0
+	 * panic in padata_mt_helper().
  	 */
  	ps.chunk_size = job->size / (ps.nworks * load_balance_factor);
  	ps.chunk_size = max(ps.chunk_size, job->min_chunk);
+	ps.chunk_size = max(ps.chunk_size, 1ul);
  	ps.chunk_size = roundup(ps.chunk_size, job->align);
 
-	/*
-	 * chunk_size can be 0 if the caller sets min_chunk to 0. So force it
-	 * to at least 1 to prevent divide-by-0 panic in padata_mt_helper().`
-	 */
-	if (!ps.chunk_size)
-		ps.chunk_size = 1U;
-
  	list_for_each_entry(pw, &works, pw_list)
  		if (job->numa_aware) {
  			int old_node = atomic_read(&last_used_nid);
 
---
Quoted 2 lines

base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
change-id: 20240822-max-93c17adc6457

LGTM, my only nit is the use of "1ul" which is less common and harder to read than "1UL" as the former one may be misread as a "lul" variable.

#4 lore raw
Quoted 51 lines · Waiman Long

This message was sent from outside of Texas Instruments.
Do not click links or open attachments unless you recognize the source of this email and know the content is safe. Report Suspicious

Quoted 46 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
---
  kernel/padata.c | 12 ++++--------
  1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +4 -8
index 0fa6c2895460..d8a51eff1581 100644
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
  	/*
  	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by
  	 * increasing the number of chunks, guarantee at least the minimum
  	 * chunk size from the caller, and honor the caller's alignment.
+	 * Ensure chunk_size is at least 1 to prevent divide-by-0
+	 * panic in padata_mt_helper().
  	 */
  	ps.chunk_size = job->size / (ps.nworks * load_balance_factor);
  	ps.chunk_size = max(ps.chunk_size, job->min_chunk);
+	ps.chunk_size = max(ps.chunk_size, 1ul);
  	ps.chunk_size = roundup(ps.chunk_size, job->align);
 
-	/*
-	 * chunk_size can be 0 if the caller sets min_chunk to 0. So force it
-	 * to at least 1 to prevent divide-by-0 panic in padata_mt_helper().`
-	 */
-	if (!ps.chunk_size)
-		ps.chunk_size = 1U;
-
  	list_for_each_entry(pw, &works, pw_list)
  		if (job->numa_aware) {
  			int old_node = atomic_read(&last_used_nid);
Quoted 8 lines
Quoted 3 lines

---
base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
change-id: 20240822-max-93c17adc6457

LGTM, my only nit is the use of "1ul" which is less common and harder to read than "1UL" as the former one may be misread as a "lul" variable.

Thanks for the Acked-by, Waiman. I understand your point, though Daniel seems to be okay with this, so will keep it as is this time.

Cheers,
Kamlesh

#5 lore raw
Quoted 59 lines · Kamlesh Gurudasani
Quoted 57 lines · Waiman Long

This message was sent from outside of Texas Instruments.
Do not click links or open attachments unless you recognize the source of this email and know the content is safe. Report Suspicious

Quoted 47 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
---
   kernel/padata.c | 12 ++++--------
   1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +4 -9
index 0fa6c2895460..d8a51eff1581 100644
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
   	/*
   	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by
   	 * increasing the number of chunks, guarantee at least the minimum
   	 * chunk size from the caller, and honor the caller's alignment.
+	 * Ensure chunk_size is at least 1 to prevent divide-by-0
+	 * panic in padata_mt_helper().
   	 */
   	ps.chunk_size = job->size / (ps.nworks * load_balance_factor);
   	ps.chunk_size = max(ps.chunk_size, job->min_chunk);
+	ps.chunk_size = max(ps.chunk_size, 1ul);
   	ps.chunk_size = roundup(ps.chunk_size, job->align);
 
-	/*
-	 * chunk_size can be 0 if the caller sets min_chunk to 0. So force it
-	 * to at least 1 to prevent divide-by-0 panic in padata_mt_helper().`
-	 */
-	if (!ps.chunk_size)
-		ps.chunk_size = 1U;
-
   	list_for_each_entry(pw, &works, pw_list)
   		if (job->numa_aware) {
   			int old_node = atomic_read(&last_used_nid);
 
---
Quoted 2 lines

base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
change-id: 20240822-max-93c17adc6457

LGTM, my only nit is the use of "1ul" which is less common and harder to read than "1UL" as the former one may be misread as a "lul" variable.

Quoted 3 lines
Quoted 1 line

Thanks for the Acked-by, Waiman. I understand your point, though Daniel seems to be okay with this, so will keep it as is this time.

This is just a suggestion in case you need to update your patch. I am fine with keeping it as is if no further update is needed.

Cheers,
Longman

maintainer Acked-by on v1 1/1
#6 lore raw
In reply to Kamlesh Gurudasani
Quoted 10 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")

Looks fine.

Quoted 10 lines
---
 kernel/padata.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +0 -0
index 0fa6c2895460..d8a51eff1581 100644

kernel/padata.c void __init padata_do_multithreaded(struct padata_mt_job *job) · line 512

Quoted 5 lines
kernel/padata.c +1 -1
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
 	/*
 	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by

Though whitespace changes like these just add noise...

#7 lore raw
Quoted 27 lines · Daniel Jordan
Quoted 10 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")

Looks fine.

Quoted 11 lines
---
 kernel/padata.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)
kernel/padata.c +0 -0
index 0fa6c2895460..d8a51eff1581 100644
@@ -509,21 +509,17 @@ void __init padata_do_multithreaded(struct padata_mt_job *job)
 
Quoted 6 lines

kernel/padata.c void __init padata_do_multithreaded(struct padata_mt_job *job) · line 512

Quoted 4 lines
kernel/padata.c +1 -1
 	/*
 	 * Chunk size is the amount of work a helper does per call to the
-	 * thread function.  Load balance large jobs between threads by
+	 * thread function. Load balance large jobs between threads by

Though whitespace changes like these just add noise...

Thanks for the Ack, would keep these in mind from next time onwards.

Kamlesh

maintainer
#8 lore raw
In reply to Kamlesh Gurudasani
Quoted 6 lines · Kamlesh Gurudasani

In the case where we are forcing the ps.chunk_size to be at least 1, we are ignoring the caller's alignment.

Move the forcing of ps.chunk_size to be at least 1 before rounding it up to caller's alignment, so that caller's alignment is honored.

Quoted 8 lines

While at it, use max() to force the ps.chunk_size to be at least 1 to improve readability.

Fixes 6d45e1c948a8 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
---
 kernel/padata.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

Patch applied. Thanks.

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt