Skip to contentSkip to conversation

staging: wlan-ng: Use flexible-array for one / zero-length arrays

PATCH In review Deepak R Varma 2 participants 8 messages Oct 28, 2022 – Nov 1, 2022 Staging

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 staging: wlan-ng: Use flexible-array for one / zero-length arrays

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/staging/wlan-ng/p80211mgmt.h +4 -4
index 1ef30d3f3159..d6fe52de2c8f 100644
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */
 } __packed;
 
 /*-- Supported Rates  -----------------------------*/
 struct wlan_ie_supp_rates {
 	u8 eid;
 	u8 len;
-	u8 rates[1];		/* had better be at LEAST one! */
+	u8 rates[];		/* had better be at LEAST one! */
 } __packed;
 
 /*-- FH Parameter Set  ----------------------------*/
@@ -274,7 +274,7 @@ struct wlan_ie_tim {
 	u8 dtim_cnt;
 	u8 dtim_period;
 	u8 bitmap_ctl;
-	u8 virt_bm[1];
+	u8 virt_bm[];
 } __packed;
 
 /*-- IBSS Parameter Set ---------------------------*/
@@ -288,7 +288,7 @@ struct wlan_ie_ibss_parms {
 struct wlan_ie_challenge {
 	u8 eid;
 	u8 len;
-	u8 challenge[1];
+	u8 challenge[];
 } __packed;
 
 /*-------------------------------------------------*/
drivers/staging/wlan-ng/p80211types.h +1 -1
index 6486612a8f31..b2ffd09881b0 100644
@@ -234,7 +234,7 @@ struct p80211pstr32 {
 /* MAC address array */
 struct p80211macarray {
 	u32 cnt;
-	u8 data[1][MAXLEN_PSTR6];
+	u8 data[][MAXLEN_PSTR6];
 } __packed;
 
 /* prototype template */

--
2.34.1

#1 lore raw
Quoted 21 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/staging/wlan-ng/p80211mgmt.h +0 -0
index 1ef30d3f3159..d6fe52de2c8f 100644

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 232

Quoted 5 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */

How have you ensured that changing this from a 1 byte array to a zero byte array does not break anything? It's not uncommon for a people to do math like "size - 1 + length". The "- 1" would be to take the 1 element into consideration.

I was trying to read through this code to check your work, but then you sent a second patch which also does not explain how you are auditing your changes. Can you go a bit slower?

regards,
dan carpenter

#2 lore raw
Quoted 17 lines · Dan Carpenter
Quoted 16 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
Quoted 15 lines

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 232

Quoted 10 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
index 1ef30d3f3159..d6fe52de2c8f 100644
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */

How have you ensured that changing this from a 1 byte array to a zero byte array does not break anything? It's not uncommon for a people to do math like "size - 1 + length". The "- 1" would be to take the 1 element into consideration.

Hi Dan,
I did a code review to understand how this structure member is used and did not find any computations you mentioned. I would certainly like to receive your feedback as well.

Quoted 4 lines

I was trying to read through this code to check your work, but then you sent a second patch which also does not explain how you are auditing your changes. Can you go a bit slower?

My apologies for rushing patches in. I will hold on for feedback on these patches before turning in any new patch involving similar change. I hope it is okay to send a different type of patch though. Please correct if I am wrong.

Thank you for looking into the patches. Appreciate your time and feedback.

./drv

Quoted 4 lines

regards,
dan carpenter

#3 lore raw
Quoted 34 lines · Deepak R Varma
Quoted 17 lines · Dan Carpenter
Quoted 16 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
Quoted 15 lines

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 232

Quoted 10 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
index 1ef30d3f3159..d6fe52de2c8f 100644
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */

How have you ensured that changing this from a 1 byte array to a zero byte array does not break anything? It's not uncommon for a people to do math like "size - 1 + length". The "- 1" would be to take the 1 element into consideration.

Quoted 5 lines

Hi Dan,
I did a code review to understand how this structure member is used and did not find any computations you mentioned. I would certainly like to receive your feedback as well.

That would be useful information for the commit message. Even if it goes under the --- cut off.

---
Compile tested only. I audited all fourty two uses of the wlan_ie_ssid struct and nothing checks sizeof(struct wlan_ie_ssid) and changing this from 1 to 0 will not break anything.

Quoted 8 lines
Quoted 4 lines

I was trying to read through this code to check your work, but then you sent a second patch which also does not explain how you are auditing your changes. Can you go a bit slower?

My apologies for rushing patches in. I will hold on for feedback on these patches before turning in any new patch involving similar change. I hope it is okay to send a different type of patch though. Please correct if I am wrong.

I'm not saying you have to wait for days for reviewers to get through your patches... I guess what I'm really trying to say is please group your patches together. Pavel Skripkin's review comments seem reasonable so if the patches are grouped then he only has to reply once but now someone has to reply to every email.

Also if you write 10 patches maybe you will notice or think of something part way through and which forces you to re-write the earlier patches.

When you send patches every ten minutes then everyone can see you have not tested it at all. I don't really expect people to test their staging patches, but let's not make it obvious. Wait overnight so people think "Ah, maybe they were tested". It gives people (false) confidence that this is quality work.

regards,
dan carpenter

#4 lore raw
In reply to Deepak R Varma
Quoted 21 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/staging/wlan-ng/p80211mgmt.h +0 -0
index 1ef30d3f3159..d6fe52de2c8f 100644

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 233

Quoted 6 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */
 } __packed;

When you're writing this change look at where struct wlan_ie_ssid is used. It's included in five different structs:

	struct wlan_ie_ssid *ssid;

But none of those *ssid pointers are used. So instead of doing this just delete the wlan_ie_ssid struct and delete every reference to it. Send that as a patch by itself and then repeat that process for the next struct.

regards,
dan carpenter

#5 lore raw
Quoted 32 lines · Dan Carpenter
Quoted 21 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/staging/wlan-ng/p80211mgmt.h +0 -0
index 1ef30d3f3159..d6fe52de2c8f 100644

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 233

Quoted 6 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */
 } __packed;

When you're writing this change look at where struct wlan_ie_ssid is used. It's included in five different structs:

Quoted 6 lines
	struct wlan_ie_ssid *ssid;

But none of those *ssid pointers are used. So instead of doing this just delete the wlan_ie_ssid struct and delete every reference to it. Send that as a patch by itself and then repeat that process for the next struct.

Thank you Dan. I had the same observation, but was not sure if that clean up should be included as part of current patch scope. Now I will, as you advised.

Also, there are a few function declarations that I do not think are used in the code anywhere.
File: drivers/staging/wlan-ng/p80211mgmt.h
Line number : 500 through 519

Can you please comment if it is safe to remove these functions and include this clean-up as well in the proposed patch?

Thank you,
./drv

Quoted 4 lines

regards,
dan carpenter

#6 lore raw
Quoted 7 lines · Deepak R Varma

Also, there are a few function declarations that I do not think are used in the code anywhere.
File: drivers/staging/wlan-ng/p80211mgmt.h
Line number : 500 through 519

Can you please comment if it is safe to remove these functions and include this clean-up as well in the proposed patch?

Always feel free to delete code that is not used.

(But send the patch tomorrow and as part of a patchset).

regards,
dan carpenter

#7 lore raw
In reply to Dan Carpenter
Quoted 32 lines · Dan Carpenter
Quoted 21 lines · Deepak R Varma

Flexible-array member should be used instead of one or zero member to meet the need for having a dynamically sized trailing elements in a structure. Refer to links [1] and [2] for detailed guidance on this suggestion.

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays

Issue identified using coccicheck.

---
 drivers/staging/wlan-ng/p80211mgmt.h  | 8 ++++----
 drivers/staging/wlan-ng/p80211types.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/staging/wlan-ng/p80211mgmt.h +0 -0
index 1ef30d3f3159..d6fe52de2c8f 100644

drivers/staging/wlan-ng/p80211mgmt.h struct wlan_ie { · line 233

Quoted 6 lines
drivers/staging/wlan-ng/p80211mgmt.h +1 -1
@@ -229,14 +229,14 @@ struct wlan_ie {
 struct wlan_ie_ssid {
 	u8 eid;
 	u8 len;
-	u8 ssid[1];		/* may be zero, ptrs may overlap */
+	u8 ssid[];		/* may be zero, ptrs may overlap */
 } __packed;

When you're writing this change look at where struct wlan_ie_ssid is used. It's included in five different structs:

Quoted 6 lines
	struct wlan_ie_ssid *ssid;

But none of those *ssid pointers are used. So instead of doing this just delete the wlan_ie_ssid struct and delete every reference to it. Send that as a patch by itself and then repeat that process for the next struct.

Hello Dan and Pavel,
I have built a patch set based on your feedback on this (and the other similar patch for r8188eu driver). I have only covered one driver for now and would wait for feedback before I do the same thing for the other driver.

Thank you,
./drv

Quoted 4 lines

regards,
dan carpenter