staging: r8188eu: clean up rtw_rf.c
cover staging: r8188eu: clean up rtw_rf.c
This set removes 5 GHz frequencies from the channel-frequency map in rtw_rf.c and simplifies the rtw_ch2freq() function.
Tested on x86_64 with Inter-Tech DMG-02.
Michael Straube (3):
staging: r8188eu: remove 5 GHz channels from ch_freq_map
staging: r8188eu: refactor rtw_ch2freq()
staging: r8188eu: clean up rtw_rf.c
drivers/staging/r8188eu/core/rtw_rf.c | 61 +++++++-------------------- 1 file changed, 16 insertions(+), 45 deletions(-)
Signature · 1 line
2.35.1
PATCH 1/3 staging: r8188eu: remove 5 GHz channels from ch_freq_map
This driver is for chips that operate only in the 2.4 GHz band. Frequencies for 5 GHz channels can be removed from the ch_freq_map array.
--- drivers/staging/r8188eu/core/rtw_rf.c | 16 ---------------- 1 file changed, 16 deletions(-)
drivers/staging/r8188eu/core/rtw_rf.c +0 -16
@@ -17,22 +17,6 @@ static struct ch_freq ch_freq_map[] = { {1, 2412}, {2, 2417}, {3, 2422}, {4, 2427}, {5, 2432}, {6, 2437}, {7, 2442}, {8, 2447}, {9, 2452}, {10, 2457}, {11, 2462}, {12, 2467}, {13, 2472}, {14, 2484}, -/* UNII */ - {36, 5180}, {40, 5200}, {44, 5220}, {48, 5240}, {52, 5260}, - {56, 5280}, {60, 5300}, {64, 5320}, {149, 5745}, {153, 5765}, - {157, 5785}, {161, 5805}, {165, 5825}, {167, 5835}, {169, 5845}, - {171, 5855}, {173, 5865}, -/* HiperLAN2 */ - {100, 5500}, {104, 5520}, {108, 5540}, {112, 5560}, {116, 5580}, - {120, 5600}, {124, 5620}, {128, 5640}, {132, 5660}, {136, 5680}, - {140, 5700}, -/* Japan MMAC */ - {34, 5170}, {38, 5190}, {42, 5210}, {46, 5230}, -/* Japan */ - {184, 4920}, {188, 4940}, {192, 4960}, {196, 4980}, - {208, 5040},/* Japan, means J08 */ - {212, 5060},/* Japan, means J12 */ - {216, 5080},/* Japan, means J16 */ };static int ch_freq_map_num =ARRAY_SIZE (ch_freq_map);
Signature · 1 line
2.35.1
PATCH 2/3 staging: r8188eu: refactor rtw_ch2freq()
In reply to Michael StraubeSince we have only channel values from 1 to 14 we can convert the array ch_freq_map to a simple integer array and use the indices as channel numbers. This simplifies the code and avoids looping through the array to get the frequency.
--- drivers/staging/r8188eu/core/rtw_rf.c | 40 +++++++++++---------------- 1 file changed, 16 insertions(+), 24 deletions(-)
drivers/staging/r8188eu/core/rtw_rf.c +16 -24
@@ -8,32 +8,24 @@#include "../include/recv_osdep.h" #include "../include/xmit_osdep.h" -struct ch_freq { -u32 channel; -u32 frequency; +static const u32 ch_freq_map[] = { +2412 , +2417 , +2422 , +2427 , +2432 , +2437 , +2442 , +2447 , +2452 , +2457 , +2462 , +2467 , +2472 , +2484 }; -static struct ch_freq ch_freq_map[] = { - {1 ,2412 }, {2 ,2417 }, {3 ,2422 }, {4 ,2427 }, {5 ,2432 }, - {6 ,2437 }, {7 ,2442 }, {8 ,2447 }, {9 ,2452 }, {10 ,2457 }, - {11 ,2462 }, {12 ,2467 }, {13 ,2472 }, {14 ,2484 }, -}; - -static int ch_freq_map_num =ARRAY_SIZE (ch_freq_map); -u32 rtw_ch2freq (u32 channel) { -u8 i; -u32 freq =0 ; - -for (i =0 ; i < ch_freq_map_num; i++) { -if (channel == ch_freq_map[i].channel) { - freq = ch_freq_map[i].frequency; -break ; - } - } -if (i == ch_freq_map_num) - freq =2412 ; - -return freq; +return ch_freq_map[channel -1 ]; }
Signature · 1 line
2.35.1
PATCH 3/3 staging: r8188eu: clean up rtw_rf.c
In reply to Michael StraubeRemove unnecessary includes and the unused define _RTW_RF_C_ from rtw_rf.c.
--- drivers/staging/r8188eu/core/rtw_rf.c | 5 ----- 1 file changed, 5 deletions(-)
drivers/staging/r8188eu/core/rtw_rf.c +0 -5
@@ -1,12 +1,7 @@// SPDX-License-Identifier: GPL-2.0 /* Copyright(c) 2007 - 2011 Realtek Corporation. */ -#define _RTW_RF_C_ - -#include "../include/osdep_service.h" #include "../include/drv_types.h" -#include "../include/recv_osdep.h" -#include "../include/xmit_osdep.h" static const u32 ch_freq_map[] = {2412 ,
Signature · 1 line
2.35.1
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
In reply to Michael StraubeHi Michael,
Quoted 9 lines · Michael Straube
drivers/staging/r8188eu/core/rtw_rf.c +0
-7-static int ch_freq_map_num =ARRAY_SIZE (ch_freq_map); -u32 rtw_ch2freq (u32 channel) { -u8 i; -u32 freq =0 ; - -for (i =0 ; i < ch_freq_map_num; i++) { -if (channel == ch_freq_map[i].channel) {
drivers/staging/r8188eu/core/rtw_rf.c line 31
Quoted 10 lines
drivers/staging/r8188eu/core/rtw_rf.c +1
-8- freq = ch_freq_map[i].frequency; -break ; - } - } -if (i == ch_freq_map_num) - freq =2412 ; - -return freq; +return ch_freq_map[channel -1 ]; }
What if channel has wrong value? The old code returned some default value, but with new one we will hit OOB.
With regards,
Pavel Skripkin
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
Quoted 16 lines · Pavel Skripkin
Hi Michael,
Quoted 13 lines · Michael Straube
(quoted diff) +0
-2-static int ch_freq_map_num = ARRAY_SIZE(ch_freq_map); -u32
rtw_ch2freq(u32 channel)
{
- u8 i;
- u32 freq = 0;
-
- for (i = 0; i <ch_freq_map_num; i++) {
- if (channel == ch_freq_map[i].channel) {
- freq = ch_freq_map[i].frequency;
- break;
- }
- }
Quoted 9 lines
Quoted 6 lines
- if (i ==
ch_freq_map_num)
- freq = 2412;
-
- return freq;
+ return ch_freq_map[channel - 1];
}What if channel has wrong value? The old code returned some default value, but with new one we will hit OOB.
Hi Pavel,
thanks for reviewing. Yeah, I thought about adding a check for channel value between 1 and 14. But I did not add it because I think if this function will ever be called with channel < 1 or channel > 14, then the calling code must be wrong.
Would be nice to see what others think about this.
Thanks,
Michael
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
Quoted 28 lines · Michael Straube
Quoted 17 lines · Pavel Skripkin
Hi Michael,
Quoted 14 lines · Michael Straube
(quoted diff) +0
-2-static int ch_freq_map_num = ARRAY_SIZE(ch_freq_map); -u32
rtw_ch2freq(u32 channel)
{
- u8 i;
- u32 freq = 0;
-
- for (i = 0; i <ch_freq_map_num; i++) {
- if (channel == ch_freq_map[i].channel) {
- freq = ch_freq_map[i].frequency;
- break;
- }
- }
- if (i ==ch_freq_map_num)Quoted 9 lines
Quoted 5 lines
- freq = 2412;
-
- return freq;
+ return ch_freq_map[channel - 1];
}What if channel has wrong value? The old code returned some default value, but with new one we will hit OOB.
Quoted 8 lines
Hi Pavel,
thanks for reviewing. Yeah, I thought about adding a check for channel value between 1 and 14. But I did not add it because I think if this function will ever be called with channel < 1 or channel > 14, then the calling code must be wrong.
Would be nice to see what others think about this.
I'm glad that Pavel noticed this change. This is a risky thing and should have been noted in the commit message.
Just from a review stand point it would be best to leave the original behavior.
I have audited this change and I do not think it is safe. It seems to me that one way this can be controlled is via module_param(rtw_channel, int, 0644); in drivers/staging/r8188eu/os_dep/os_intfs.c. I don't see any checking on that.
regards,
dan carpenter
Greg KH
maintainerRe: [PATCH 1/3] staging: r8188eu: remove 5 GHz channels from ch_freq_map
In reply to Michael StraubeQuoted 1 line · Michael Straube
This driver is for chips that operate only in the 2.4 GHz band.
How do we know that?
thanks,
greg k-h
Re: [PATCH 1/3] staging: r8188eu: remove 5 GHz channels from ch_freq_map
Quoted 5 lines · Greg KH
Quoted 1 line · Michael Straube
This driver is for chips that operate only in the 2.4 GHz band.
How do we know that?
The documentation that I've found for RTL8188EU and RTL8188ETV says that both are 2.4 Ghz only chips.
Also the old rtl8188eu driver had a TODO item to remove code that is only valid for 5 GHz.
So I thought this can be removed. Did I miss something?
regards,
Michael
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
In reply to Dan CarpenterQuoted 36 lines · Dan Carpenter
Quoted 35 lines · Michael Straube
Quoted 17 lines · Pavel Skripkin
Hi Michael,
Quoted 14 lines · Michael Straube
drivers/staging/r8188eu/core/rtw_rf.c +0
-12-static int ch_freq_map_num =ARRAY_SIZE (ch_freq_map); - u32 rtw_ch2freq(u32 channel) { - u8 i; - u32 freq = 0; - - for (i = 0; i < ch_freq_map_num; i++) { - if (channel == ch_freq_map[i].channel) { - freq = ch_freq_map[i].frequency; - break; - } - } - if (i == ch_freq_map_num)Quoted 9 lines
Quoted 5 lines
- freq = 2412;
-
- return freq;
+ return ch_freq_map[channel - 1];
}What if channel has wrong value? The old code returned some default value, but with new one we will hit OOB.
Hi Pavel,
thanks for reviewing. Yeah, I thought about adding a check for channel value between 1 and 14. But I did not add it because I think if this function will ever be called with channel < 1 or channel > 14, then the calling code must be wrong.
Quoted 8 lines
Quoted 1 line
Would be nice to see what others think about this.
I'm glad that Pavel noticed this change. This is a risky thing and should have been noted in the commit message.
Just from a review stand point it would be best to leave the original behavior.
Do you mean to leave the whole original code including the 5 GHz frequencies? Or returning a default value if we have a channel value < 1 or > 14?
I'm a bit confused now, because Greg asked how we know that the driver is only for 2.4 GHz chips.
Quoted 6 lines
I have audited this change and I do not think it is safe. It seems to me that one way this can be controlled is via
module_param(rtw_channel, int, 0644); indrivers/staging/r8188eu/os_dep/os_intfs.c. I don't see any checking on that.
Thank you Dan!
I missed that and blindly assumed the function will never be called with channel values OOB. That was not good, sorry.
regards,
Michael
Re: [PATCH 1/3] staging: r8188eu: remove 5 GHz channels from ch_freq_map
In reply to Greg KHQuoted 6 lines · Greg KH
Quoted 1 line · Michael Straube
This driver is for chips that operate only in the 2.4 GHz band.
How do we know that?
thanks,
Realtek (https://www.realtek.com/en/products/communications-network-ics/item/rtl8188eus) states the following:
"The Realtek RTL8188EUS is an 802.11bgn 2.4G single-chip that integrates Wireless LAN (WLAN) and a network USB interface (USB 1.0/1.1/2.0 compatible) controller. It combines a WLAN MAC, a 1T1R capable WLAN baseband, and WLAN RF in a single chip. The RTL8188EUS provides a complete solution for high-throughput performance and low power consumption integrated wireless LAN devices."
The reason there are 5G snippets in the driver is because Realtek uses the base code to generate drivers for other chip families such as RTL8192DU that have radios for both 2.4 and 5G bands. The RTL81288EU and variants have only a single radio.
Larry
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
In reply to Michael StraubeHi Michael,
Quoted 11 lines · Michael Straube
Quoted 6 lines
I'm glad that Pavel noticed this change. This is a risky thing and should have been noted in the commit message.
Just from a review stand point it would be best to leave the original behavior.
Do you mean to leave the whole original code including the 5 GHz frequencies? Or returning a default value if we have a channel value < 1 or > 14?
IMO, your version is much cleaner than previous one. This table walk seems really unreasonable, since 5 GHz support is really redundant (I saw it in other thread)
I'd put just sanity check and return the default value from previous version. Maybe even wrapped with unlikely() if we sure, that in normal state we won't hit it ;)
With regards,
Pavel Skripkin
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
In reply to Michael StraubeQuoted 11 lines · Michael Straube
Quoted 6 lines · Dan Carpenter
Quoted 1 line · Michael Straube
On 2/20/22 17:20, Pavel Skripkin wrote:
Just from a review stand point it would be best to leave the original behavior.
Do you mean to leave the whole original code including the 5 GHz frequencies? Or returning a default value if we have a channel value < 1 or > 14?
I meant what Pavel said:
if (channel == 0 || channel > ARRAY_SIZE(ch_freq_map)) return 2412; return ch_freq_map[channel - 1];
Quoted 4 lines
I'm a bit confused now, because Greg asked how we know that the driver is only for 2.4 GHz chips.
Haven't you been deleting all the 5GHz code for some time now? Greg isn't saying you have to keep that if there is no reason. Just put the reason in the commit message and resend.
regards,
dan carpenter
Re: [PATCH 2/3] staging: r8188eu: refactor rtw_ch2freq()
In reply to Pavel SkripkinQuoted 15 lines · Pavel Skripkin
Hi Michael,
Quoted 11 lines · Michael Straube
Quoted 6 lines
I'm glad that Pavel noticed this change. This is a risky thing and should have been noted in the commit message.
Just from a review stand point it would be best to leave the original behavior.
Do you mean to leave the whole original code including the 5 GHz frequencies? Or returning a default value if we have a channel value < 1 or > 14?
Quoted 7 lines
IMO, your version is much cleaner than previous one. This table walk seems really unreasonable, since 5 GHz support is really redundant (I saw it in other thread)
I'd put just sanity check and return the default value from previous version. Maybe even wrapped with
unlikely()if we sure, that in normal state we won't hit it ;)
Adding likely()/unlikely() annotations hurt readability. Do not do that unless you have benchmark data to prove that it is worth it. (Hint: it is not worth it here).
regards,
dan carpenter
Greg KH
maintainerRe: [PATCH 1/3] staging: r8188eu: remove 5 GHz channels from ch_freq_map
In reply to Larry FingerQuoted 17 lines · Larry Finger
Quoted 6 lines · Greg KH
Quoted 1 line · Michael Straube
This driver is for chips that operate only in the 2.4 GHz band.
How do we know that?
thanks,
Realtek (https://www.realtek.com/en/products/communications-network-ics/item/rtl8188eus)
states the following:"The Realtek RTL8188EUS is an 802.11bgn 2.4G single-chip that integrates Wireless LAN (WLAN) and a network USB interface (USB 1.0/1.1/2.0 compatible) controller. It combines a WLAN MAC, a 1T1R capable WLAN baseband, and WLAN RF in a single chip. The RTL8188EUS provides a complete solution for high-throughput performance and low power consumption integrated wireless LAN devices."
Quoted 4 lines
The reason there are 5G snippets in the driver is because Realtek uses the base code to generate drivers for other chip families such as RTL8192DU that have radios for both 2.4 and 5G bands. The RTL81288EU and variants have only a single radio.
Ah, ok, that's good to know. And should belong in the changelog text so that we are not confused :)
thanks,
greg k-h