Skip to contentSkip to conversation

fs/btrfs: Fix uninitialized variable

PATCH In review Khaled ROMDHANI 5 participants 8 messages May 1, 2021 – May 17, 2021

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

PATCH 1/1 fs/btrfs: Fix uninitialized variable

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---
 fs/btrfs/zoned.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/zoned.c +1 -1
index 8250ab3f0868..23da9d8dc184 100644
@@ -145,7 +145,7 @@ static inline u32 sb_zone_number(int shift, int mirror)
 	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
 	default:
 		ASSERT((u32)mirror < 3);
-		break;
+		return 0;
 	}
 
 	ASSERT(zone <= U32_MAX);

base-commit: b5c294aac8a6164ddf38bfbdd1776091b4a1eeba

Signature · 1 line
2.17.1
#1 lore raw

Le 02/05/2021 à 00:50, Khaled ROMDHANI a écrit :

Quoted 28 lines

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---
  fs/btrfs/zoned.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/zoned.c +1 -1
index 8250ab3f0868..23da9d8dc184 100644
@@ -145,7 +145,7 @@ static inline u32 sb_zone_number(int shift, int mirror)
  	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
  	default:
  		ASSERT((u32)mirror < 3);
-		break;
+		return 0;
  	}
 
  	ASSERT(zone <= U32_MAX);
Quoted 2 lines

base-commit: b5c294aac8a6164ddf38bfbdd1776091b4a1eeba

Hi,

just a few comments.

If I understand correctly, what you try to do is to silence a compiler warning if no case branch is taken.

First, all your proposals are based on the previous one.
I find it hard to follow because we don't easily see what are the differences since the beginning.

The "base-commit" at the bottom of your mail, is related to your own local tree, I guess. It can't be used by any-one.

My understanding it that a patch, should it be v2, v3..., must apply to the current tree. (In my case, it is the latest linux-next) This is not the case here and you have to apply each step to see the final result.

Should this version be fine, a maintainer wouldn't be able to apply it as-is.

You also try to take into account previous comments to check for incorrect negative values for minor and catch (the can't happen today) cases, should BTRFS_SUPER_MIRROR_MAX change and this function remain the same.

So, why hard-coding '3'?
The reason of magic numbers are hard to remember. You should avoid them or add a comment about it.

My own personal variation would be something like the code below (untested).

Hope this helps.

CJ

fs/btrfs/zoned.c +4 -1
index 70b23a0d03b1..75fe5f001d8b 100644
@@ -138,11 +138,14 @@ static inline u32 sb_zone_number(int shift, int mirror)
 {
 	u64 zone;
 
-	ASSERT(mirror < BTRFS_SUPER_MIRROR_MAX);
+	ASSERT(mirror >= 0 && mirror < BTRFS_SUPER_MIRROR_MAX);
 	switch (mirror) {
 	case 0: zone = 0; break;
 	case 1: zone = 1ULL << (BTRFS_SB_LOG_FIRST_SHIFT - shift); break;
 	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
+	default:
+		ASSERT(! "mirror < BTRFS_SUPER_MIRROR_MAX but not handled above.");
+		return 0;
 	}
 
 	ASSERT(zone <= U32_MAX);
#2 lore raw
Quoted 87 lines · Christophe JAILLET

Le 02/05/2021 à 00:50, Khaled ROMDHANI a écrit :

Quoted 27 lines

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---
  fs/btrfs/zoned.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/zoned.c +1 -1
index 8250ab3f0868..23da9d8dc184 100644
@@ -145,7 +145,7 @@ static inline u32 sb_zone_number(int shift, int mirror)
  	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
  	default:
  		ASSERT((u32)mirror < 3);
-		break;
+		return 0;
  	}
  	ASSERT(zone <= U32_MAX);
 
Quoted 2 lines

base-commit: b5c294aac8a6164ddf38bfbdd1776091b4a1eeba

Hi,

just a few comments.

If I understand correctly, what you try to do is to silence a compiler warning if no case branch is taken.

First, all your proposals are based on the previous one. I find it hard to follow because we don't easily see what are the differences since the beginning.

The "base-commit" at the bottom of your mail, is related to your own local tree, I guess. It can't be used by any-one.

My understanding it that a patch, should it be v2, v3..., must apply to the current tree. (In my case, it is the latest linux-next)
This is not the case here and you have to apply each step to see the final result.

Should this version be fine, a maintainer wouldn't be able to apply it as-is.

You also try to take into account previous comments to check for incorrect negative values for minor and catch (the can't happen today) cases, should BTRFS_SUPER_MIRROR_MAX change and this function remain the same.

So, why hard-coding '3'?
The reason of magic numbers are hard to remember. You should avoid them or add a comment about it.

My own personal variation would be something like the code below (untested).

Hope this helps.

CJ

fs/btrfs/zoned.c +4 -1
index 70b23a0d03b1..75fe5f001d8b 100644
@@ -138,11 +138,14 @@ static inline u32 sb_zone_number(int shift, int
 mirror)
 {
 	u64 zone;
 
-	ASSERT(mirror < BTRFS_SUPER_MIRROR_MAX);
+	ASSERT(mirror >= 0 && mirror < BTRFS_SUPER_MIRROR_MAX);
 	switch (mirror) {
 	case 0: zone = 0; break;
 	case 1: zone = 1ULL << (BTRFS_SB_LOG_FIRST_SHIFT - shift); break;
 	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
+	default:
+		ASSERT(! "mirror < BTRFS_SUPER_MIRROR_MAX but not handled above.");
+		return 0;
 	}
 
Quoted 2 lines

ASSERT(zone <= U32_MAX);

Thank you for all of your comments. Yes, of course, they will help me. I will try to handle that more properly.
Thanks again.

#3 lore raw
In reply to Khaled ROMDHANI
Quoted 11 lines · Khaled ROMDHANI

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---

This is not how you send a v4 patch... v2 patches have to apply to the original code and not on top of the patched code.

I sort of think you should find a different thing to work on. This code works fine as-is. Just leave it and try to find a real bug and fix that instead.

regards,
dan carpenter

#4 lore raw
Quoted 16 lines · Dan Carpenter
Quoted 11 lines · Khaled ROMDHANI

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---

This is not how you send a v4 patch... v2 patches have to apply to the original code and not on top of the patched code.

Quoted 7 lines

I sort of think you should find a different thing to work on. This code works fine as-is. Just leave it and try to find a real bug and fix that instead.

regards,
dan carpenter

Sorry, I was wrong and I shall send a proper V4.

Yes, this code works fine just a warning caught by Coverity scan analysis. So the idea behind the patch is to fix the warning. To do that and as suggested by David Sterba, it would be rather to add a default case. So we fix the warning through the enhancement of the switch statement (some sort of 2*1).

Yes, I will always try to find other bugs. It is a pleasure for me to do that.

Thanks.

#5 lore raw
Quoted 21 lines · Khaled Romdhani
Quoted 20 lines · Dan Carpenter
Quoted 11 lines · Khaled ROMDHANI

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---

This is not how you send a v4 patch... v2 patches have to apply to the original code and not on top of the patched code.

I sort of think you should find a different thing to work on. This code works fine as-is. Just leave it and try to find a real bug and fix that instead.

Quoted 7 lines
Quoted 3 lines

regards,
dan carpenter

Sorry, I was wrong and I shall send a proper V4.

Yes, this code works fine just a warning caught by Coverity scan analysis.

We're going to a lot of work to silence a static checker false positive. As a rule, I tell people to ignore the static checker when it is wrong.

Btw, Smatch parses this code correctly and understands that the callers only pass valid values for "mirror".

regards,
dan carpenter

#6 lore raw
Quoted 28 lines · Dan Carpenter
Quoted 27 lines · Khaled Romdhani
Quoted 16 lines · Dan Carpenter
Quoted 11 lines · Khaled ROMDHANI

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

---

This is not how you send a v4 patch... v2 patches have to apply to the original code and not on top of the patched code.

Quoted 7 lines

I sort of think you should find a different thing to work on. This code works fine as-is. Just leave it and try to find a real bug and fix that instead.

regards,
dan carpenter

Sorry, I was wrong and I shall send a proper V4.

Quoted 7 lines
Quoted 1 line

Yes, this code works fine just a warning caught by Coverity scan analysis.

We're going to a lot of work to silence a static checker false positive. As a rule, I tell people to ignore the static checker when it is wrong.

Btw, Smatch parses this code correctly and understands that the callers only pass valid values for "mirror".

..and Coverity does report a lot of false positives, so one needs to be really sure the issue is a genuine issue rather than a warning that can be ignore.

Colin

Quoted 4 lines

regards,
dan carpenter

#7 lore raw
In reply to Khaled ROMDHANI
Quoted 10 lines · Khaled ROMDHANI

Fix the warning: variable 'zone' is used uninitialized whenever '?:' condition is true.

Fix that by preventing the code to reach the last assertion. If the variable 'mirror' is invalid, the assertion fails and we return immediately.

This took several rounds and none of them was close to what I'd consider a proper fix, for something that's not really important. As Dan said, smatch does understand the values passed from the callers and the function is a static inline so the complete information is available. No tricky analysis is required, so why does not coverity see that too?

We use assertions to namely catch programmer errors and API misuse, anything that can happen at runtime or depends on input needs proper checks and error handling. But for the super block copies, the constant won't change so all we want is to catch the stupid errors.

Quoted 9 lines
---
 fs/btrfs/zoned.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/zoned.c +0 -0
index 8250ab3f0868..23da9d8dc184 100644

fs/btrfs/zoned.c static inline u32 sb_zone_number(int shift, int mirror) · line 148

Quoted 5 lines
fs/btrfs/zoned.c +1 -1
@@ -145,7 +145,7 @@ static inline u32 sb_zone_number(int shift, int mirror)
 	case 2: zone = 1ULL << (BTRFS_SB_LOG_SECOND_SHIFT - shift); break;
 	default:
 		ASSERT((u32)mirror < 3);
-		break;
+		return 0;

It's been pointed out that this does not apply on the current code but on top of previous versions, so it's not making it easy for me to apply the patch and do maybe some tweaks only.

I don't mind merging trivial patches, people can learn the process and few iterations are not a big deal. What I also hope for is to get some understanding of the code being changed and not just silencing some tools' warnings.