Skip to contentSkip to conversation

UDF: possible usage of uninitialized variables

PATCH In review Cyrill Gorcunov 3 participants 5 messages May 22, 2007 – May 23, 2007

lore
Share & tools
lore

Cyrill Gorcunov

author
#0 lore raw

PATCH 1/1 UDF: possible usage of uninitialized variables

A few variables could be used without being explicitly initialized. Fixed.

Signed-off-by Cyrill Gorcunov <[email protected]>

---

 balloc.c |    6 +++++-
 super.c  |    5 ++++-
 2 files changed, 9 insertions(+), 2 deletions(-)
balloc.c +5 -1
index 4cec910..be37393 100644
@@ -744,7 +744,11 @@ static int udf_table_new_block(struct super_block * sb,
 	uint32_t spread = 0xFFFFFFFF, nspread = 0xFFFFFFFF;
 	uint32_t newblock = 0, adsize;
 	uint32_t elen, goal_elen = 0;
-	kernel_lb_addr eloc, goal_eloc;
+	kernel_lb_addr eloc;
+	kernel_lb_addr goal_eloc = {
+		.logicalBlockNum = 0,
+		.partitionReferenceNum = 0,
+	};
 	struct extent_position epos, goal_epos;
 	int8_t etype;
super.c +4 -1
index 9b8644a..068a99c 100644
@@ -1358,7 +1358,10 @@ udf_load_partition(struct super_block *sb, kernel_lb_addr *fileset)
 			case UDF_VIRTUAL_MAP15:
 			case UDF_VIRTUAL_MAP20:
 			{
-				kernel_lb_addr ino;
+				kernel_lb_addr ino = {
+					.logicalBlockNum = 0,
+					.partitionReferenceNum = 0,
+				};
 
 				if (!UDF_SB_LASTBLOCK(sb))
 				{
#1 lore raw
Quoted 2 lines · On Tue, 22 May 2007 21:52:23 +0400 Cyrill Gorcunov <[email protected]> wrote:

A few variables could be used without being explicitly initialized. Fixed.

Are you sure? I'm expecting that this is a gcc mistake.

balloc.c

Quoted 13 lines
Signed-off-by Cyrill Gorcunov <[email protected]>

---

 balloc.c |    6 +++++-
 super.c  |    5 ++++-
 2 files changed, 9 insertions(+), 2 deletions(-)
balloc.c +0 -0
index 4cec910..be37393 100644

Please always use `patch -p1' form from the top-level directory. This should have been

--- a/fs/udf/balloc.c
+++ b/fs/udf/balloc.c

balloc.c static int udf_table_new_block(struct super_block * sb, · line 751

Quoted 10 lines
balloc.c +5 -1
@@ -744,7 +744,11 @@ static int udf_table_new_block(struct super_block * sb,
 	uint32_t spread = 0xFFFFFFFF, nspread = 0xFFFFFFFF;
 	uint32_t newblock = 0, adsize;
 	uint32_t elen, goal_elen = 0;
-	kernel_lb_addr eloc, goal_eloc;
+	kernel_lb_addr eloc;
+	kernel_lb_addr goal_eloc = {
+		.logicalBlockNum = 0,
+		.partitionReferenceNum = 0,
+	};

	kernel_lb_addr goal_eloc = { };

is equivalent and sufficient.

#2 lore raw
In reply to Cyrill Gorcunov
Quoted 18 lines

A few variables could be used without being explicitly initialized. Fixed.

Signed-off-by Cyrill Gorcunov <[email protected]>

---

 balloc.c |    6 +++++-
 super.c  |    5 ++++-
 2 files changed, 9 insertions(+), 2 deletions(-)
balloc.c +0 -0
index 4cec910..be37393 100644
@@ -744,7 +744,11 @@ static int udf_table_new_block(struct super_block * sb,
 	uint32_t spread = 0xFFFFFFFF, nspread = 0xFFFFFFFF;

balloc.c static int udf_table_new_block(struct super_block * sb, · line 753

Quoted 10 lines
balloc.c +5 -1
 	uint32_t newblock = 0, adsize;
 	uint32_t elen, goal_elen = 0;
-	kernel_lb_addr eloc, goal_eloc;
+	kernel_lb_addr eloc;
+	kernel_lb_addr goal_eloc = {
+		.logicalBlockNum = 0,
+		.partitionReferenceNum = 0,
+	};
 	struct extent_position epos, goal_epos;
 	int8_t etype;

As I read the code, I don't think it can happen goal_eloc is used
unitialized. Also please diff the tree completely, not just the udf directory..

Quoted 6 lines
super.c +0 -0
index 9b8644a..068a99c 100644
@@ -1358,7 +1358,10 @@ udf_load_partition(struct super_block *sb, kernel_lb_addr *fileset)
 			case UDF_VIRTUAL_MAP15:

super.c udf_load_partition(struct super_block *sb, kernel_lb_addr *fileset) · line 1367

Quoted 10 lines
super.c +4 -1
 			case UDF_VIRTUAL_MAP20:
 			{
-				kernel_lb_addr ino;
+				kernel_lb_addr ino = {
+					.logicalBlockNum = 0,
+					.partitionReferenceNum = 0,
+				};
 
 				if (!UDF_SB_LASTBLOCK(sb))
 				{

This one seems correct. Thanks for fixing this.

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SuSE CR Labs
author
#3 lore raw

[Jan Kara - Tue, May 22, 2007 at 10:29:38PM +0200]

| > A few variables could be used without being explicitly initialized.
| > Fixed.
| > 
| > Signed-off-by: Cyrill Gorcunov <[email protected]>
| > ---
| > 
| > 
| >  balloc.c |    6 +++++-
| >  super.c  |    5 ++++-
| >  2 files changed, 9 insertions(+), 2 deletions(-)
| > 
| > 
| > diff --git a/balloc.c b/balloc.c
| > index 4cec910..be37393 100644
| > --- a/balloc.c
| > +++ b/balloc.c
| > @@ -744,7 +744,11 @@ static int udf_table_new_block(struct super_block * sb,
| >  	uint32_t spread = 0xFFFFFFFF, nspread = 0xFFFFFFFF;
| >  	uint32_t newblock = 0, adsize;
| >  	uint32_t elen, goal_elen = 0;
| > -	kernel_lb_addr eloc, goal_eloc;
| > +	kernel_lb_addr eloc;
| > +	kernel_lb_addr goal_eloc = {
| > +		.logicalBlockNum = 0,
| > +		.partitionReferenceNum = 0,
| > +	};
| >  	struct extent_position epos, goal_epos;
| >  	int8_t etype;
|   As I read the code, I don't think it can happen goal_eloc is used
| unitialized. Also please diff the tree completely, not just the udf
| directory..
| 
| > diff --git a/super.c b/super.c
| > index 9b8644a..068a99c 100644
| > --- a/super.c
| > +++ b/super.c
| > @@ -1358,7 +1358,10 @@ udf_load_partition(struct super_block *sb, kernel_lb_addr *fileset)
| >  			case UDF_VIRTUAL_MAP15:
| >  			case UDF_VIRTUAL_MAP20:
| >  			{
| > -				kernel_lb_addr ino;
| > +				kernel_lb_addr ino = {
| > +					.logicalBlockNum = 0,
| > +					.partitionReferenceNum = 0,
| > +				};
| >  
| >  				if (!UDF_SB_LASTBLOCK(sb))
| >  				{
|   This one seems correct. Thanks for fixing this.
| 
| 								Honza
| -- 
| Jan Kara <[email protected]>
| SuSE CR Labs
|

Sorry, that's my fault :(

Cyrill

author
#4 lore raw
In reply to Jan Kara

[Jan Kara - Tue, May 22, 2007 at 10:29:38PM +0200]

| > A few variables could be used without being explicitly initialized.
| > Fixed.
| > 
| > Signed-off-by: Cyrill Gorcunov <[email protected]>
| > ---
| > 
| > 
| >  balloc.c |    6 +++++-
| >  super.c  |    5 ++++-
| >  2 files changed, 9 insertions(+), 2 deletions(-)
| > 
| > 
| > diff --git a/balloc.c b/balloc.c
| > index 4cec910..be37393 100644
| > --- a/balloc.c
| > +++ b/balloc.c
| > @@ -744,7 +744,11 @@ static int udf_table_new_block(struct super_block * sb,
| >  	uint32_t spread = 0xFFFFFFFF, nspread = 0xFFFFFFFF;
| >  	uint32_t newblock = 0, adsize;
| >  	uint32_t elen, goal_elen = 0;
| > -	kernel_lb_addr eloc, goal_eloc;
| > +	kernel_lb_addr eloc;
| > +	kernel_lb_addr goal_eloc = {
| > +		.logicalBlockNum = 0,
| > +		.partitionReferenceNum = 0,
| > +	};
| >  	struct extent_position epos, goal_epos;
| >  	int8_t etype;
|   As I read the code, I don't think it can happen goal_eloc is used
| unitialized. Also please diff the tree completely, not just the udf
| directory..
| 
| > diff --git a/super.c b/super.c
| > index 9b8644a..068a99c 100644
| > --- a/super.c
| > +++ b/super.c
| > @@ -1358,7 +1358,10 @@ udf_load_partition(struct super_block *sb, kernel_lb_addr *fileset)
| >  			case UDF_VIRTUAL_MAP15:
| >  			case UDF_VIRTUAL_MAP20:
| >  			{
| > -				kernel_lb_addr ino;
| > +				kernel_lb_addr ino = {
| > +					.logicalBlockNum = 0,
| > +					.partitionReferenceNum = 0,
| > +				};
| >  
| >  				if (!UDF_SB_LASTBLOCK(sb))
| >  				{
|   This one seems correct. Thanks for fixing this.
| 
| 								Honza
| -- 
| Jan Kara <[email protected]>
| SuSE CR Labs
|

Jan, you know I've been messed up. The variables I've touched in my patch became initialized in any case.

Andrew, just drop my patch (you're right and I was wrong ;).

Thanks you both for comments.

Cyrill