Skip to contentSkip to conversation

smb: client: Move an error code assignment in smb3_init_transform_rq()

PATCH In review Markus Elfring 2 participants 4 messages Oct 8, 2025 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 smb: client: Move an error code assignment in smb3_init_transform_rq()

From: Markus Elfring <[email protected]>
Date: Wed, 8 Oct 2025 18:48:28 +0200

Convert an initialisation for the variable “rc” into an error code assignment at the end of this function implementation.

---
 fs/smb/client/smb2ops.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
fs/smb/client/smb2ops.c +4 -2
index 7c3e96260fd4..2513270ac596 100644
@@ -4596,7 +4596,7 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 {
 	struct smb2_transform_hdr *tr_hdr = new_rq[0].rq_iov[0].iov_base;
 	unsigned int orig_len = 0;
-	int rc = -ENOMEM;
+	int rc;
 
 	for (int i = 1; i < num_rqst; i++) {
 		struct smb_rqst *old = &old_rq[i - 1];
@@ -4611,7 +4611,7 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 		if (size > 0) {
 			buffer = cifs_alloc_folioq_buffer(size);
 			if (!buffer)
-				goto err_free;
+				goto e_nomem;
 
 			new->rq_buffer = buffer;
 			iov_iter_folio_queue(&new->rq_iter, ITER_SOURCE,
@@ -4634,6 +4634,8 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 
 	return rc;
 
+e_nomem:
+	rc = -ENOMEM;
 err_free:
 	smb3_free_compound_rqst(num_rqst - 1, &new_rq[1]);
 	return rc;
Signature · 1 line
2.51.0
#1 lore raw

Hi Markus,

Quoted 34 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Wed, 8 Oct 2025 18:48:28 +0200

Convert an initialisation for the variable “rc” into an error code assignment at the end of this function implementation.

---
 fs/smb/client/smb2ops.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
fs/smb/client/smb2ops.c +2 -2
index 7c3e96260fd4..2513270ac596 100644
@@ -4596,7 +4596,7 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 {
 	struct smb2_transform_hdr *tr_hdr = new_rq[0].rq_iov[0].iov_base;
 	unsigned int orig_len = 0;
-	int rc = -ENOMEM;
+	int rc;
 
 	for (int i = 1; i < num_rqst; i++) {
 		struct smb_rqst *old = &old_rq[i - 1];
@@ -4611,7 +4611,7 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 		if (size > 0) {
 			buffer = cifs_alloc_folioq_buffer(size);
 			if (!buffer)
-				goto err_free;
+				goto e_nomem;
 
 			new->rq_buffer = buffer;
 			iov_iter_folio_queue(&new->rq_iter, ITER_SOURCE,

fs/smb/client/smb2ops.c smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst, · line 4641

Quoted 8 lines
fs/smb/client/smb2ops.c +2 -0
@@ -4634,6 +4634,8 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 
 	return rc;
 
+e_nomem:
+	rc = -ENOMEM;
 err_free:
 	smb3_free_compound_rqst(num_rqst - 1, &new_rq[1]);
 	return rc;

I don't think this change improves readability.

I understand that making the assignment explicit is good, but why not simply set rc to -ENOMEM if !buffer and then goto err_free?

Also, I think its a bit confusing having inconsistent naming styles e_ err_...

Signature · 2 lines
Henrique
SUSE Labs
#2 lore raw
Quoted 2 lines
Quoted 2 lines

Convert an initialisation for the variable “rc” into an error code assignment at the end of this function implementation.

…>> +++ b/fs/smb/client/smb2ops.c
…>> @@ -4611,7 +4611,7 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,

Quoted 20 lines
Quoted 9 lines
fs/smb/client/smb2ops.c +1 -1
 		if (size > 0) {
 			buffer = cifs_alloc_folioq_buffer(size);
 			if (!buffer)
-				goto err_free;
+				goto e_nomem;
 
 			new->rq_buffer = buffer;
 			iov_iter_folio_queue(&new->rq_iter, ITER_SOURCE,

fs/smb/client/smb2ops.c smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst, · line 4641

Quoted 8 lines
fs/smb/client/smb2ops.c +2 -0
@@ -4634,6 +4634,8 @@ smb3_init_transform_rq(struct TCP_Server_Info *server, int num_rqst,
 
 	return rc;
 
+e_nomem:
+	rc = -ENOMEM;
 err_free:
 	smb3_free_compound_rqst(num_rqst - 1, &new_rq[1]);
 	return rc;

I don't think this change improves readability.

Quoted 1 line

I understand that making the assignment explicit is good,

Thanks for this constructive feedback.

Quoted 1 line

but why not simply set rc to -ENOMEM if !buffer and then goto err_free?

I proposed to adjust the affected if branch in this way because there is no need to add curly brackets then.

Quoted 2 lines

Also, I think its a bit confusing having inconsistent naming styles e_ err_...

Which naming approach would you find more helpful for the marking of a variable assignment instead of a subsequent function call?

Regards,
Markus

A message that is not in the archive (sent off-list or not yet received): CAH2r5msZjL_krwN-nd1Ly3skxAHK9srJehPJ_vYXPLRFXpJ_yw@mail.gmail.com

#4 lore raw
Quoted 1 line

Also when a patch doesn't shrink code, but even grows it by one line, and doesn't make it much clearer, it is probably not worth changing (it complicates future backports of real fixes in the future eg). Let's limit these to those that shrink code and make code clearer (or ideally fix potential bugs)

Do you insist to keep an extra initialisation for the local variable “rc” despite of the implementation detail that the value “-ENOMEM” is needed only for a single if branch here?

Regards,
Markus