staging: rtl8723bs: refactor BSS Coexistence channel report logic
PATCH 1/1 staging: rtl8723bs: refactor BSS Coexistence channel report logic
Refactor the 'ICS' array in issue_action_BSSCoexistPacket() to improve readability and maintainability. This addresses technical debt related to magic numbers and ambiguous array usage.
The original implementation used a multi-purpose 2D array (ICS[8][15]) where magic numbers were prevalent and the first element of each row was overloaded as a status flag. This patch:
- Introduces descriptive macros: BSS_COEX_MAX_CLASSES, BSS_COEX_MAX_CHANNELS, and BSS_COEX_MAX_INFO_LEN.
- Splits the overloaded array into two distinct boolean arrays:
'class_active' (for group status) and 'ch_present' (for channel data).
- Converts the logic to use 'bool' types and 0-indexed loops, conforming to standard C programming practices.
- Adds defensive boundary checks (ch > 0 && ch < MAX) to ensure robustness against unexpected channel data.
This refactoring maintains the current 2.4GHz reporting behavior while providing a structured and extensible foundation for future Operating Class support according to IEEE 802.11 specifications.
--- drivers/staging/rtl8723bs/core/rtw_mlme_ext.c | 41 +++++++------------ .../staging/rtl8723bs/include/rtw_mlme_ext.h | 3 ++ 2 files changed, 18 insertions(+), 26 deletions(-)
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c +15 -26
@@ -3587,8 +3587,9 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter)struct mlme_ext_priv *pmlmeext = &(padapter->mlmeextpriv);struct mlme_ext_info *pmlmeinfo = &pmlmeext->mlmext_info;struct __queue *queue = &(pmlmepriv->scanned_queue); -u8 InfoContent[16 ] = {0 }; -u8 ICS [8 ][15 ]; +u8 InfoContent[BSS_COEX_MAX_INFO_LEN ] = {0 }; +bool class_active[BSS_COEX_MAX_CLASSES ] = {false}; +bool ch_present[BSS_COEX_MAX_CLASSES ][BSS_COEX_MAX_CHANNELS ] = {{false}};if ((pmlmepriv->num_FortyMHzIntolerant ==0 ) || (pmlmepriv->num_sta_no_ht ==0 ))return ;
@@ -3642,7 +3643,6 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter)/* */ -memset (ICS ,0 ,sizeof (ICS ));if (pmlmepriv->num_sta_no_ht >0 ) {int i;
@@ -3667,46 +3667,35 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter) p =rtw_get_ie (pbss_network->ies + _FIXED_IE_LENGTH_,WLAN_EID_HT_CAPABILITY , &len, pbss_network->ie_length - _FIXED_IE_LENGTH_);if (!p || len ==0 ) {/* non-HT */ +u8 ch = pbss_network->configuration.ds_config; -if (pbss_network->configuration.ds_config <=0 ) -continue ; - -ICS [0 ][pbss_network->configuration.ds_config] =1 ; - -if (ICS [0 ][0 ] ==0 ) -ICS [0 ][0 ] =1 ; +if (ch >0 && ch <BSS_COEX_MAX_CHANNELS ) { + ch_present[0 ][ch] = true; + class_active[0 ] = true; + } } }spin_unlock_bh (&(pmlmepriv->scanned_queue.lock)); - -for (i =0 ; i <8 ; i++) { +for (i =0 ; i <BSS_COEX_MAX_CLASSES ; i++) {int j, k =0 ; -if (ICS [i][0 ] !=1 ) +if (!class_active[i])continue ; - InfoContent[k] = i; + InfoContent[k++] = i;/* SET_BSS_INTOLERANT_ELE_REG_CLASS(InfoContent, i); */ - k++; -for (j =1 ; j <=14 ; j++) { -if (ICS [i][j] !=1 ) -continue ; - -if (k <16 ) { - InfoContent[k] = j;/* channel number */ -/* SET_BSS_INTOLERANT_ELE_CHANNEL(InfoContent+k, j); */ - k++; +for (j =0 ; j <BSS_COEX_MAX_CHANNELS ; j++) { +if (ch_present[i][j]) { +if (k <BSS_COEX_MAX_INFO_LEN ) + InfoContent[k++] = j;/* channel number */ } } - pframe =rtw_set_ie (pframe,WLAN_EID_BSS_INTOLERANT_CHL_REPORT , k, InfoContent, &(pattrib->pktlen)); - } - }
drivers/staging/rtl8723bs/include/rtw_mlme_ext.h +3 -0
@@ -41,6 +41,9 @@#define _HW_STATE_STATION_ 0x02#define _HW_STATE_AP_ 0x03 +#define BSS_COEX_MAX_CLASSES 8 +#define BSS_COEX_MAX_CHANNELS 15 +#define BSS_COEX_MAX_INFO_LEN 16#define _1M_RATE_ 0#define _2M_RATE_ 1
Signature · 1 line
2.43.0
Quoted 32 lines · Michael Huang
Refactor the 'ICS' array in
issue_action_BSSCoexistPacket()to improve readability and maintainability. This addresses technical debt related to magic numbers and ambiguous array usage.The original implementation used a multi-purpose 2D array (ICS[8][15]) where magic numbers were prevalent and the first element of each row was overloaded as a status flag. This patch:
- Introduces descriptive macros:
BSS_COEX_MAX_CLASSES,BSS_COEX_MAX_CHANNELS, andBSS_COEX_MAX_INFO_LEN.
- Splits the overloaded array into two distinct boolean arrays:
'class_active' (for group status) and 'ch_present' (for channel data).
- Converts the logic to use 'bool' types and 0-indexed loops, conforming to standard C programming practices.
- Adds defensive boundary checks (ch > 0 && ch < MAX) to ensure robustness against unexpected channel data.This refactoring maintains the current 2.4GHz reporting behavior while providing a structured and extensible foundation for future Operating Class support according to IEEE 802.11 specifications.
Signed-off-by Michael Huang <[email protected]>--- drivers/staging/rtl8723bs/core/rtw_mlme_ext.c | 41 +++++++------------ .../staging/rtl8723bs/include/rtw_mlme_ext.h | 3 ++ 2 files changed, 18 insertions(+), 26 deletions(-)drivers/staging/rtl8723bs/core/rtw_mlme_ext.c +0
-0
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c static void issue_action_BSSCoexistPacket(struct adapter *padapter) · line 3591
Quoted 7 lines
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c +2
-2@@ -3587,8 +3587,9 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter)struct mlme_ext_priv *pmlmeext = &(padapter->mlmeextpriv);struct mlme_ext_info *pmlmeinfo = &pmlmeext->mlmext_info;struct __queue *queue = &(pmlmepriv->scanned_queue); -u8 InfoContent[16 ] = {0 }; -u8 ICS [8 ][15 ]; +u8 InfoContent[BSS_COEX_MAX_INFO_LEN ] = {0 }; +bool class_active[BSS_COEX_MAX_CLASSES ] = {false};
We only ever set the first element in the class_active[] array. #puzzled
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c static void issue_action_BSSCoexistPacket(struct adapter *padapter) · line 3592
Quoted 1 line
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c +1
-0+bool ch_present[BSS_COEX_MAX_CLASSES ][BSS_COEX_MAX_CHANNELS ] = {{false}};
Btw, use "= {};" to initialize arrays to zero. It's not standard C, but it's kernel C.
regards,
dan carpenter
Quoted 4 lines
We only ever set the first element in the class_active[] array. #puzzled
Quoted 8 lines
Quoted 1 line
+ bool ch_present[BSS_COEX_MAX_CLASSES][BSS_COEX_MAX_CHANNELS] = {{false}};
Btw, use "= {};" to initialize arrays to zero. It's not standard C, but it's kernel C.
regards,
dan carpenter
Hi Dan,
Thanks for your time!! I understand the confusion. I kept the array structure because it reflects the IEEE 802.11 Intolerant Channel Report format, which is organized by Regulatory Classes.
Although the driver currently only reports for the 2.4GHz class (index 0), using an array makes the code semantically aligned with the specification and easier to extend when other operating classes need to be supported in the future.
I have submitted a V2 which uses the '= {}' initialization you suggested.
PATCH v2 1/1 staging: rtl8723bs: refactor BSS Coexistence channel report logic
In reply to Dan CarpenterRefactor the 'ICS' array in issue_action_BSSCoexistPacket() to improve readability and maintainability. This addresses technical debt related to magic numbers and ambiguous array usage.
The original implementation used a multi-purpose 2D array (ICS[8][15]) where magic numbers were prevalent and the first element of each row was overloaded as a status flag. This patch:
- Introduces descriptive macros: BSS_COEX_MAX_CLASSES, BSS_COEX_MAX_CHANNELS, and BSS_COEX_MAX_INFO_LEN.
- Splits the overloaded array into two distinct boolean arrays:
'class_active' (for group status) and 'ch_present' (for channel data).
- Converts the logic to use 'bool' types and 0-indexed loops, conforming to standard C programming practices.
- Adds defensive boundary checks (ch > 0 && ch < MAX) to ensure robustness against unexpected channel data.
This refactoring maintains the current 2.4GHz reporting behavior while providing a structured and extensible foundation for future Operating Class support according to IEEE 802.11 specifications.
---
v2:
- Use kernel-style = {} for array initialization per Dan Carpenter.
- Separate flags from data using bool arrays per Joe Perches.
- Fix magic numbers with proper defines.
- Retained the 2D structure for 'ch_present' despite only using class 0 for now. This aligns with the IEEE 802.11 Regulatory Class/Channel List hierarchy, making the code easier to map to the specification and extensible for future operating class support.
v1:
- Initial refactoring of the ICS array logic.
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c | 41 +++++++------------ .../staging/rtl8723bs/include/rtw_mlme_ext.h | 3 ++ 2 files changed, 18 insertions(+), 26 deletions(-)
drivers/staging/rtl8723bs/core/rtw_mlme_ext.c +15 -26
@@ -3587,8 +3587,9 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter)struct mlme_ext_priv *pmlmeext = &(padapter->mlmeextpriv);struct mlme_ext_info *pmlmeinfo = &pmlmeext->mlmext_info;struct __queue *queue = &(pmlmepriv->scanned_queue); -u8 InfoContent[16 ] = {0 }; -u8 ICS [8 ][15 ]; +u8 InfoContent[BSS_COEX_MAX_INFO_LEN ] = {}; +bool class_active[BSS_COEX_MAX_CLASSES ] = {}; +bool ch_present[BSS_COEX_MAX_CLASSES ][BSS_COEX_MAX_CHANNELS ] = {};if ((pmlmepriv->num_FortyMHzIntolerant ==0 ) || (pmlmepriv->num_sta_no_ht ==0 ))return ;
@@ -3642,7 +3643,6 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter)/* */ -memset (ICS ,0 ,sizeof (ICS ));if (pmlmepriv->num_sta_no_ht >0 ) {int i;
@@ -3667,46 +3667,35 @@ static void issue_action_BSSCoexistPacket(struct adapter *padapter) p =rtw_get_ie (pbss_network->ies + _FIXED_IE_LENGTH_,WLAN_EID_HT_CAPABILITY , &len, pbss_network->ie_length - _FIXED_IE_LENGTH_);if (!p || len ==0 ) {/* non-HT */ +u8 ch = pbss_network->configuration.ds_config; -if (pbss_network->configuration.ds_config <=0 ) -continue ; - -ICS [0 ][pbss_network->configuration.ds_config] =1 ; - -if (ICS [0 ][0 ] ==0 ) -ICS [0 ][0 ] =1 ; +if (ch >0 && ch <BSS_COEX_MAX_CHANNELS ) { + ch_present[0 ][ch] = true; + class_active[0 ] = true; + } } }spin_unlock_bh (&(pmlmepriv->scanned_queue.lock)); - -for (i =0 ; i <8 ; i++) { +for (i =0 ; i <BSS_COEX_MAX_CLASSES ; i++) {int j, k =0 ; -if (ICS [i][0 ] !=1 ) +if (!class_active[i])continue ; - InfoContent[k] = i; + InfoContent[k++] = i;/* SET_BSS_INTOLERANT_ELE_REG_CLASS(InfoContent, i); */ - k++; -for (j =1 ; j <=14 ; j++) { -if (ICS [i][j] !=1 ) -continue ; - -if (k <16 ) { - InfoContent[k] = j;/* channel number */ -/* SET_BSS_INTOLERANT_ELE_CHANNEL(InfoContent+k, j); */ - k++; +for (j =0 ; j <BSS_COEX_MAX_CHANNELS ; j++) { +if (ch_present[i][j]) { +if (k <BSS_COEX_MAX_INFO_LEN ) + InfoContent[k++] = j;/* channel number */ } } - pframe =rtw_set_ie (pframe,WLAN_EID_BSS_INTOLERANT_CHL_REPORT , k, InfoContent, &(pattrib->pktlen)); - } - }
drivers/staging/rtl8723bs/include/rtw_mlme_ext.h +3 -0
@@ -41,6 +41,9 @@#define _HW_STATE_STATION_ 0x02#define _HW_STATE_AP_ 0x03 +#define BSS_COEX_MAX_CLASSES 8 +#define BSS_COEX_MAX_CHANNELS 15 +#define BSS_COEX_MAX_INFO_LEN 16#define _1M_RATE_ 0#define _2M_RATE_ 1
Signature · 1 line
2.43.0
Greg Kroah-Hartman
maintainerQuoted 8 lines · Michael Huang
Refactor the 'ICS' array in
issue_action_BSSCoexistPacket()to improve readability and maintainability. This addresses technical debt related to magic numbers and ambiguous array usage.The original implementation used a multi-purpose 2D array (ICS[8][15]) where magic numbers were prevalent and the first element of each row was overloaded as a status flag. This patch:
Quoted 8 lines
- Introduces descriptive macros:
BSS_COEX_MAX_CLASSES,BSS_COEX_MAX_CHANNELS, andBSS_COEX_MAX_INFO_LEN.
- Splits the overloaded array into two distinct boolean arrays:
'class_active' (for group status) and 'ch_present' (for channel data).
- Converts the logic to use 'bool' types and 0-indexed loops, conforming to standard C programming practices.
- Adds defensive boundary checks (ch > 0 && ch < MAX) to ensure robustness against unexpected channel data.
That's a lot of different things all in one single patch. SHouldn't this be split up into different ones?
And did you use AI to generate this patch?
Also, how was this tested? What prompted you to want to make this change in the first place?
thanks,
greg k-h
Quoted 2 lines · Greg Kroah-Hartman
That's a lot of different things all in one single patch. SHouldn't this be split up into different ones?
I kept these changes in one patch because they all work together to replace the old ICS logic and remove the magic numbers. Splitting them might break the build or make the code inconsistent halfway through. I was suggested renaming and using a bool array structure in the earlier patch here:
https://lore.kernel.org/all/[email protected]/
I kept the 2D structure to maintain parity with the IEEE 802.11 spec as original implementation. It makes the code semantically correct and easier to extend for other Operating Classes in the future.
Quoted 1 line
And did you use AI to generate this patch?
I used AI tools to polish the language of my commit message and to perform a parity check, ensuring the refactored boolean logic is functionally identical to the original code.
Quoted 2 lines
Also, how was this tested? What prompted you to want to make this change in the first place?
The patch was tested by:
1. Cross-compile the module for the rtl8723bs driver (make M=drivers/staging/rtl8723bs) to ensure no build regressions.
2. Ran it through scripts/checkpatch.pl to confirm it meets kernel coding standards.
3. Since I do not have the physical hardware, I performed a manual check (by AI verification) to ensure the logic remains identical. Specifically:
- The original code used ICS[0][ch] to store channel data.
- The new code uses ch_present[0][ch] for the same purpose.
- The original ICS[i][0] flag is now explicitly class_active[i].
- Verified that the loop boundaries and index offsets (0-15)
remain exactly the same as the original implementation.
Motivation: This follows up on the cleanup tasks suggested during the previous discussion in
https://lore.kernel.org/all/[email protected].
I want to improve the readability of the driver by removing magic numbers.
Quoted 4 lines
thanks,
greg k-h
Thanks,
Michael
Greg Kroah-Hartman
maintainerQuoted 15 lines · Te-Hsiu Huang
Quoted 2 lines · Greg Kroah-Hartman
That's a lot of different things all in one single patch. SHouldn't this be split up into different ones?
I kept these changes in one patch because they all work together to replace the old ICS logic and remove the magic numbers. Splitting them might break the build or make the code inconsistent halfway through. I was suggested renaming and using a bool array structure in the earlier patch here:
https://lore.kernel.org/all/[email protected]/
I kept the 2D structure to maintain parity with the IEEE 802.11 spec as original implementation. It makes the code semantically correct and easier to extend for other Operating Classes in the future.
Quoted 4 lines
Quoted 1 line
And did you use AI to generate this patch?
I used AI tools to polish the language of my commit message and to perform a parity check, ensuring the refactored boolean logic is functionally identical to the original code.
Then as per our development rules, you must document this properly. Please read and follow them for when you resubmit this.
Quoted 10 lines
Quoted 3 lines
Also, how was this tested? What prompted you to want to make this change in the first place?
The patch was tested by:
1. Cross-compile the module for the rtl8723bs driver (make M=drivers/staging/rtl8723bs) to ensure no build regressions.
2. Ran it throughscripts/checkpatch.plto confirm it meets kernel coding standards.
3. Since I do not have the physical hardware, I performed a manual check (by AI verification) to ensure the logic remains identical.
AI can not "verify" anything, you must always check that, sorry.
greg k-h
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
And did you use AI to generate this patch?
I used AI tools to polish the language of my commit message and to perform a parity check, ensuring the refactored boolean logic is functionally identical to the original code.
Then as per our development rules, you must document this properly. Please read and follow them for when you resubmit this.
Quoted 6 lines
Quoted 2 lines
3. Since I do not have the physical hardware, I performed a manual check (by AI verification) to ensure the logic remains identical.
AI can not "verify" anything, you must always check that, sorry.
greg k-h
Hi Greg,
I apologize for the poor phrasing. I did not mean that I relied solely on AI to verify the code. To ensure the refactor was correct, I performed a manual line-by-line audit to verify that the logic remains identical to the original implementation.
I understand that I am responsible for the code's correctness. I will properly document the use of AI tools according to the rules and resubmit this as part of V3, including splitting the changes and providing more test details.
Thank you for the guidance!
Quoted 4 lines · Michael Huang
- Retained the 2D structure for '
ch_present' despite only using class 0 for now. This aligns with the IEEE 802.11 Regulatory Class/Channel List hierarchy, making the code easier to map to the specification and extensible for future operating class support.
No one can predit the future and adding unused cruft based on our horoscope is discouraged. Future developers can add it themselves if they want it.
regards,
dan carpenter
Quoted 7 lines
No one can predit the future and adding unused cruft based on our horoscope is discouraged. Future developers can add it themselves if they want it.
regards,
dan carpenter
Thank Greg and Dan for the feedback. I now have a much better understanding of the development philosophy behind the Linux kernel (My first time). This kind of 'culture shock' is exactly what I was looking for — it's a valuable learning experience that helps me transition toward writing more professional and maintainable kernel code :) .
Thanks,
Michael