Skip to contentSkip to conversation

fs: ocfs2: Fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

PATCH In review Jia-Ju Bai 3 participants 5 messages Jul 26, 2019 – Aug 6, 2019 OCFS2

lore
Share & tools
lore

Jia-Ju Bai

author
#0 lore raw

PATCH 3/3 fs: ocfs2: Fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283 to check whether inode_alloc is NULL:
if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:

    ocfs2_inode_lock(inode_alloc, &bh, 0);
        ocfs2_inode_lock_full_nested(inode, ...)
            struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

This bug is found by a static analysis tool STCheck written by us.

---
 fs/ocfs2/ioctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ocfs2/ioctl.c +1 -1
index d6f7b299eb23..efeea208fdeb 100644
@@ -283,7 +283,7 @@ static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 	if (inode_alloc)
 		inode_lock(inode_alloc);
 
-	if (o2info_coherent(&fi->ifi_req)) {
+	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
 		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
 		if (status < 0) {
 			mlog_errno(status);
Signature · 1 line
2.17.0
maintainer Reviewed-by on v1 3/3
#1 lore raw
Quoted 9 lines · Jia-Ju Bai

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283 to check whether inode_alloc is NULL:
if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:

    ocfs2_inode_lock(inode_alloc, &bh, 0);
        ocfs2_inode_lock_full_nested(inode, ...)
            struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
Quoted 7 lines

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

This bug is found by a static analysis tool STCheck written by us.

Looks good.

Quoted 18 lines
---
 fs/ocfs2/ioctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ocfs2/ioctl.c +1 -1
index d6f7b299eb23..efeea208fdeb 100644
@@ -283,7 +283,7 @@ static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 	if (inode_alloc)
 		inode_lock(inode_alloc);
 
-	if (o2info_coherent(&fi->ifi_req)) {
+	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
 		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
 		if (status < 0) {
 			mlog_errno(status);
#2 lore raw

PATCH 3/3 fs: ocfs2: Fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

Hi Jia-Ju and Joseph,

Busy with a training session last week, so late for reviewing this patch.

I think this patch might not be necessary and a little violates original logic.

Please check out my comments inline.

Quoted 32 lines · Joseph Qi
Quoted 9 lines · Jia-Ju Bai

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283 to check whether inode_alloc is NULL:
if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:

     ocfs2_inode_lock(inode_alloc, &bh, 0);
         ocfs2_inode_lock_full_nested(inode, ...)
             struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
Quoted 7 lines

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

This bug is found by a static analysis tool STCheck written by us.

Looks good.

Quoted 12 lines
---
  fs/ocfs2/ioctl.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
fs/ocfs2/ioctl.c +0 -0
index d6f7b299eb23..efeea208fdeb 100644
@@ -283,7 +283,7 @@ static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
  	if (inode_alloc)
  		inode_lock(inode_alloc);
 
Quoted 2 lines

fs/ocfs2/ioctl.c static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb, · line 286

Quoted 2 lines
fs/ocfs2/ioctl.c +1 -1
-	if (o2info_coherent(&fi->ifi_req)) {
+	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {

If caller is asking for strict coherency but *inode_alloc* is NULL, we fallback to non-coherency branch?

I think this is destroying the file semantics. Returning error to caller is better?

On the other hand, from the code path:

ocfs2_info_handle_freeinode()

??? if o2info_coherent() then inode_alloc =
ocfs2_get_system_file_inode() ---> here we can ensure that being passed *inode_alloc* can't be NULL

??? ocfs2_info_scan_inode_alloc()

Thanks,

Changwei

Quoted 8 lines

fs/ocfs2/ioctl.c static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb, · line 289

Quoted 4 lines
fs/ocfs2/ioctl.c +0 -0
  		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
  		if (status < 0) {
  			mlog_errno(status);

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel

maintainer
#3 lore raw

PATCH 3/3 fs: ocfs2: Fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

Quoted 51 lines · Changwei Ge

Hi Jia-Ju and Joseph,

Busy with a training session last week, so late for reviewing this patch.

I think this patch might not be necessary and a little violates original logic.

Please check out my comments inline.

Quoted 24 lines · Joseph Qi
Quoted 9 lines · Jia-Ju Bai

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283 to check whether inode_alloc is NULL:
???? if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:
???? ocfs2_inode_lock(inode_alloc, &bh, 0); ???????? ocfs2_inode_lock_full_nested(inode, ...) ???????????? struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

Quoted 7 lines

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

This bug is found by a static analysis tool STCheck written by us.

Looks good.

Quoted 4 lines

---
? fs/ocfs2/ioctl.c | 2 +-
? 1 file changed, 1 insertion(+), 1 deletion(-)

Quoted 9 lines

fs/ocfs2/ioctl.c static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb, · line 286

Quoted 9 lines
fs/ocfs2/ioctl.c +1 -0
index d6f7b299eb23..efeea208fdeb 100644
@@ -283,7 +283,7 @@ static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 ????? if (inode_alloc)
 ????????? inode_lock(inode_alloc);
 ? -??? if (o2info_coherent(&fi->ifi_req)) {
+??? if (inode_alloc && o2info_coherent(&fi->ifi_req)) {

If caller is asking for strict coherency but *inode_alloc* is NULL, we fallback to non-coherency branch?

I think this is destroying the file semantics. Returning error to caller is better?

Quoted 8 lines

On the other hand, from the code path:

ocfs2_info_handle_freeinode()

??? if o2info_coherent() then inode_alloc = ocfs2_get_system_file_inode() ---> here we can ensure that being passed *inode_alloc* can't be NULL

??? ocfs2_info_scan_inode_alloc()

You are right, the caller make sure inode_alloc won't be NULL if o2info_coherent(). So adding this check just make it more explicitly, right?
I don't get how it breaks the original logic.

Thanks,
Joseph

#4 lore raw

PATCH 3/3 fs: ocfs2: Fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

Morning,

Quoted 55 lines · Joseph Qi
Quoted 53 lines · Changwei Ge

Hi Jia-Ju and Joseph,

Busy with a training session last week, so late for reviewing this patch.

I think this patch might not be necessary and a little violates original logic.

Please check out my comments inline.

Quoted 23 lines · Joseph Qi
Quoted 9 lines · Jia-Ju Bai

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283 to check whether inode_alloc is NULL:
???? if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:
???? ocfs2_inode_lock(inode_alloc, &bh, 0); ???????? ocfs2_inode_lock_full_nested(inode, ...) ???????????? struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

Quoted 7 lines

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

This bug is found by a static analysis tool STCheck written by us.

Looks good.

Quoted 4 lines
---
 ? fs/ocfs2/ioctl.c | 2 +-

? 1 file changed, 1 insertion(+), 1 deletion(-)

Quoted 9 lines

fs/ocfs2/ioctl.c static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb, · line 286

Quoted 9 lines
fs/ocfs2/ioctl.c +1 -0
index d6f7b299eb23..efeea208fdeb 100644
@@ -283,7 +283,7 @@ static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 ????? if (inode_alloc)
 ????????? inode_lock(inode_alloc);
 ? -??? if (o2info_coherent(&fi->ifi_req)) {
+??? if (inode_alloc && o2info_coherent(&fi->ifi_req)) {

If caller is asking for strict coherency but *inode_alloc* is NULL, we fallback to non-coherency branch?

I think this is destroying the file semantics. Returning error to caller is better?

On the other hand, from the code path:

ocfs2_info_handle_freeinode()

Quoted 7 lines
Quoted 4 lines

??? if o2info_coherent() then inode_alloc = ocfs2_get_system_file_inode() ---> here we can ensure that being passed *inode_alloc* can't be NULL

??? ocfs2_info_scan_inode_alloc()

You are right, the caller make sure inode_alloc won't be NULL if o2info_coherent(). So adding this check just make it more explicitly, right?
I don't get how it breaks the original logic.

As user? is asking for a strict coherent filesystem info with OCFS2_INFO_FL_NON_COHERENT cleared, we have to lock the inode cluster-wide via DLM.

If somehow, *inode_alloc* is NULL(actually it can't be), I think, we should still comply above strict coherency sentiments and can't just jump to the non-coherency branch.

Perhaps, return error to caller is better under above situation(can't exist)?

Thanks,

Changwei

Quoted 3 lines

Thanks,
Joseph