Skip to contentSkip to conversation

RISC-V: Fix /proc/cpuinfo cpumask warning

PATCH Superseded Andrew Jones 4 participants 6 messages Oct 12, 2022 RISC-V

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 RISC-V: Fix /proc/cpuinfo cpumask warning

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +3 -0
index 4aa8cd749441..63138b880b92 100644
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+	if (*pos >= nr_cpu_ids)
+		return NULL;
+
 	*pos = cpumask_next(*pos - 1, cpu_online_mask);
 	if ((*pos) < nr_cpu_ids)
 		return (void *)(uintptr_t)(1 + *pos);
Signature · 1 line
2.37.3
Reviewed-by on v2 1/1
#1 lore raw
Quoted 16 lines · Andrew Jones

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }
Quoted 8 lines

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

Looks good to me.

Regards,
Anup

Quoted 29 lines

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +3 -1
index 4aa8cd749441..63138b880b92 100644
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+       if (*pos >= nr_cpu_ids)
+               return NULL;
+
        *pos = cpumask_next(*pos - 1, cpu_online_mask);
        if ((*pos) < nr_cpu_ids)
                return (void *)(uintptr_t)(1 + *pos);
--

2.37.3

Reviewed-by Tested-by on v2 1/1
#2 lore raw
In reply to Andrew Jones
Quoted 15 lines · Andrew Jones

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }
Quoted 8 lines

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

Thanks

Quoted 29 lines

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +3 -0
index 4aa8cd749441..63138b880b92 100644
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+	if (*pos >= nr_cpu_ids)
+		return NULL;
+
 	*pos = cpumask_next(*pos - 1, cpu_online_mask);
 	if ((*pos) < nr_cpu_ids)
 		return (void *)(uintptr_t)(1 + *pos);
Signature · 1 line
2.37.3
#3 lore raw
In reply to Andrew Jones
Quoted 5 lines · Andrew Jones

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

Quoted 9 lines
  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }

Can you instead of sudo-code print show the real control flow? What function hosts the infinite loop?

Quoted 25 lines

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +0 -0
index 4aa8cd749441..63138b880b92 100644

arch/riscv/kernel/cpu.c static void print_mmu(struct seq_file *f) · line 174

Quoted 9 lines
arch/riscv/kernel/cpu.c +3 -0
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+	if (*pos >= nr_cpu_ids)
+		return NULL;
+
 	*pos = cpumask_next(*pos - 1, cpu_online_mask);
 	if ((*pos) < nr_cpu_ids)
 		return (void *)(uintptr_t)(1 + *pos);

OK, as far as I understood your explanations, *pos == nr_cpu_ids is a valid index because it's used as stop-code for traversing.

However, you're completely silencing cpumask_check(), including those cases where *pos > nr_cpu_ids. I suspect there's no valid cases for it. If so, the patch should look like:

+ if (*pos == nr_cpu_ids)
+ return NULL;
+

The same for x86 patch.

If it comes to v3, can you send both as a series?

Thanks,
Yury

#4 lore raw
Quoted 16 lines · Yury Norov
Quoted 5 lines · Andrew Jones

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

Quoted 9 lines
  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }
Quoted 2 lines

Can you instead of sudo-code print show the real control flow? What function hosts the infinite loop?

The function is seq_read_iter(), which is pointed out above. I'd rather not reproduce / describe more than what I've done here, as the function is large. I'd be happy for reviewers to double check my pseudocode to make sure I got it and the analysis right, though.

Quoted 39 lines
Quoted 26 lines

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +0 -0
index 4aa8cd749441..63138b880b92 100644

arch/riscv/kernel/cpu.c static void print_mmu(struct seq_file *f) · line 174

Quoted 9 lines
arch/riscv/kernel/cpu.c +3 -0
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+	if (*pos >= nr_cpu_ids)
+		return NULL;
+
 	*pos = cpumask_next(*pos - 1, cpu_online_mask);
 	if ((*pos) < nr_cpu_ids)
 		return (void *)(uintptr_t)(1 + *pos);

OK, as far as I understood your explanations, *pos == nr_cpu_ids is a valid index because it's used as stop-code for traversing.

Quoted 7 lines

However, you're completely silencing cpumask_check(), including those cases where *pos > nr_cpu_ids. I suspect there's no valid cases for it. If so, the patch should look like:

+ if (*pos == nr_cpu_ids)
+ return NULL;
+

That makes sense and it's probably worth a v3. I'll wait and see if more comments roll in before sending though.

Quoted 4 lines

The same for x86 patch.

If it comes to v3, can you send both as a series?

OK. I'll write a cover letter trying to explain that I don't expect them to both go through the same tree.

Thanks,
drew

#5 lore raw
Quoted 77 lines · Andrew Jones
Quoted 16 lines · Yury Norov
Quoted 5 lines · Andrew Jones

Commit 78e5a3399421 ("cpumask: fix checking valid cpu range") has started issuing warnings[*] when cpu indices equal to nr_cpu_ids - 1 are passed to cpumask_next* functions. seq_read_iter() and cpuinfo's start and next seq operations implement a pattern like

Quoted 9 lines
  n = cpumask_next(n - 1, mask);
  show(n);
  while (1) {
      ++n;
      n = cpumask_next(n - 1, mask);
      if (n >= nr_cpu_ids)
          break;
      show(n);
  }
Quoted 2 lines

Can you instead of sudo-code print show the real control flow? What function hosts the infinite loop?

The function is seq_read_iter(), which is pointed out above. I'd rather not reproduce / describe more than what I've done here, as the function is large. I'd be happy for reviewers to double check my pseudocode to make sure I got it and the analysis right, though.

Quoted 39 lines
Quoted 26 lines

which will issue the warning when reading /proc/cpuinfo. Ensure no warning is generated by validating the cpu index before calling cpumask_next().

[*] Warnings will only appear with DEBUG_PER_CPU_MAPS enabled.

---
v2:
- Got comments on the x86 equivalent patch and made the same changes to this one
- Added all the information I should have in the first place to the commit message [Boris]
- Changed style of fix [Boris]

 arch/riscv/kernel/cpu.c | 3 +++
 1 file changed, 3 insertions(+)
arch/riscv/kernel/cpu.c +0 -0
index 4aa8cd749441..63138b880b92 100644

arch/riscv/kernel/cpu.c static void print_mmu(struct seq_file *f) · line 174

Quoted 9 lines
arch/riscv/kernel/cpu.c +3 -0
@@ -166,6 +166,9 @@ static void print_mmu(struct seq_file *f)
 
 static void *c_start(struct seq_file *m, loff_t *pos)
 {
+   if (*pos >= nr_cpu_ids)
+           return NULL;
+
    *pos = cpumask_next(*pos - 1, cpu_online_mask);
    if ((*pos) < nr_cpu_ids)
            return (void *)(uintptr_t)(1 + *pos);

OK, as far as I understood your explanations, *pos == nr_cpu_ids is a valid index because it's used as stop-code for traversing.

Quoted 7 lines

However, you're completely silencing cpumask_check(), including those cases where *pos > nr_cpu_ids. I suspect there's no valid cases for it. If so, the patch should look like:

+ if (*pos == nr_cpu_ids)
+ return NULL;
+

That makes sense and it's probably worth a v3. I'll wait and see if more comments roll in before sending though.

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

The same for x86 patch.

If it comes to v3, can you send both as a series?

OK. I'll write a cover letter trying to explain that I don't expect them to both go through the same tree.

I can take it in my tree, if it helps.