Jump to content

Intel Wireless driver

jalavoui

By jalavoui,
in LAN and Wireless

 Share
3992 posts in this topic

Recommended Posts

moseschrist
moseschrist
Posted
Posted
After around 3 hours of uptime, I checked Activity Monitor and saw that nsGUI had taken 175 mb of system memory, and it was steadily rising. Does this means that there is a memory leak, or hopefully it is just not showing the correct memory usage? I'm using the latest ipi2200 as of June 6, from the google code page.

 

hmmm could be...

i will check that out...

jalavoui
jalavoui
Posted
  • Author
Posted

iwi3945 rev 463

http://code.google.com/p/iwidarwin/

fixes for last bugs

try option (1) of networkselector

 

 

moses,

check activity monitor after loading nsgui - it keeps eating memory

can you check the ioalloc/iofree code?

this reminds me the old mbuf cluster bug - i don't feel like going into this again...

the bug might be in the timer routine

moseschrist
moseschrist
Posted
Posted
iwi3945 rev 463

http://code.google.com/p/iwidarwin/

fixes for last bugs

try option (1) of networkselector

moses,

check activity monitor after loading nsgui - it keeps eating memory

can you check the ioalloc/iofree code?

this reminds me the old mbuf cluster bug - i don't feel like going into this again...

the bug might be in the timer routine

 

first of all, i want to congretulate "endlesssnowfall" for finding this bug, great job :huh:

the bug was found and fixed, here is an explantion of what went wrong:

 

priv=priv0;

sp=sizeof(priv);

int result = getsockopt( fd, SYSPROTO_CONTROL, 0, &priv, &sp);

priv.ieee = &ieee;

sp=sizeof(*priv.ieee);

result = getsockopt( fd, SYSPROTO_CONTROL, 1, priv.ieee, &sp);

priv.ieee->dev = &net_dev;

sp=sizeof(*priv.ieee->dev);

result = getsockopt( fd, SYSPROTO_CONTROL, 5, priv.ieee->dev, &sp);

priv.ieee->networks = (struct ieee80211_network*)malloc(MAX_NETWORK_COUNT * sizeof(struct ieee80211_network));

 

since we have two privs, priv & priv0 where priv0 is an empty template of priv, this way we don't need to create a new one everytime.

so before we do anything we refresh priv by adding the first line (priv=priv0), and in the last line we can see that we allocate a data structure inside priv. now all of that happnes inside a function which happnes every few seconds and when that happnes without freeing the old networks structure, well thats a big boo boo.

 

so here is the fix for the memory leak, we add a static variable to tell us when all the code wirtten above happned and if so we first release priv.ieee->networks and only then we overwrite priv with priv0.

 

i'm uploading it now to the svn.

 

edit:

i'm getting errors with the dmg, after using the installer i still have to fix permissions on my own, i don't know what it happnes.

i'm also getting errors when i try to upload to the svn, so i'm uploading it to the google code page.

jalavoui, please fix the error in the installer script [or tell me what i'm doing wrong, i'm compiling as root] and also commit the changes to the svn

 

http://iwidarwin.googlecode.com/files/iwi2200_fix_leak.zip

KemenAran
KemenAran
Posted
Posted

Here is an other log for iwi3945 rev 463, with additionnal descriptions of the events.

When the card is turned on for the first time, an uCode HW error is received - but after a turn off/on cycle, it's a SW interrupt error that comes.

 

I didn't get much time last week (real-time stuff and exams), but I can now work at the driver again. Does someone think the HW error is significant ? Do you know where we could start to look ?

dmesg.txt

jalavoui
jalavoui
Posted
  • Author
Posted
first of all, i want to congretulate "endlesssnowfall" for finding this bug, great job :)

the bug was found and fixed, here is an explantion of what went wrong:

 

jalavoui, please fix the error in the installer script [or tell me what i'm doing wrong, i'm compiling as root] and also commit the changes to the svn

 

one less bug :)

 

before build the .dmg you need to delete the old one. sometimes the script fails to build the .dmg and worst - sometimes it doesn't copy all files inside the .dmg (need to check manually)

 

sometimes i have problems uploading the .dmg to svn - google problem?. in this case it's better to try to upload later

 

moses, try to change the script - it's supposed to work for all user accounts

 

the script is in xcode - targets - iwi2200 - run script

did you commit all source code to svn? any files missing?

 

 

Here is an other log for iwi3945 rev 463, with additionnal descriptions of the events.

 

someone think the HW error is significant ? Do you know where we could start to look ?

 

if all iwi3945 useres do this with their logs the driver will get ready much faster :)

 

you can try to set the p_disable=0. maybe it helps

i also suspect of the qeue_te/td functions

 

thanks

moseschrist
moseschrist
Posted
Posted
one less bug :whistle:

 

before build the .dmg you need to delete the old one. sometimes the script fails to build the .dmg and worst - sometimes it doesn't copy all files inside the .dmg (need to check manually)

 

sometimes i have problems uploading the .dmg to svn - google problem?. in this case it's better to try to upload later

 

moses, try to change the script - it's supposed to work for all user accounts

 

the dmg is building ok, i have tried adding this line in the script "sudo chmod -R 755 /build/Debug/iwi2200.kext" and still no go...

it always complains after i reboot that the kext has wrong permissions...

i have no idea why it happnes... maybe were missing something...

ivanpac
ivanpac
Posted
Posted

HI all i'm totally new to this. I was following the instructions and installed the iwi2200. nothing worked as described above. Then i wanted to get my ethernet to work at least so i followed the steps here: http://forum.insanelymac.com/lofiversion/i...php/t41245.html.

Then, after ethernet was not even detected i went back to the wireless utility and tried to connect to an unsecured network...to my surprise i watched an ethernet adapter pop up in sys pref>network and connect with a local LAN address...now the funny thing is the ethernet cable was nowhere near the computer. I cannot connect to any secure networks, but unsecured i'm browsing the internet!

jalavoui
jalavoui
Posted
  • Author
Posted

moses,

 

you're new version was made on old iwi code

download from svn tree and rewrite the "fix_leak" code

i've added adhoc and other fixes (iwi2200and mainmenu.nib) that doesn't exist in your version

 

after you rewrite the code you'll have no problems commiting to svn

 

i think the file permissions can bet set with other command rather than chown

anyone can help on this?

 

 

http://code.google.com/p/iwidarwin/

 

iwi3945 rev 464

-fixes for firmware errors

-disabled free(). don't use kextunload

 

please see KemenAran last log and try to post your logs as he as done

 

iwi2100 rev 465

-panic fixes

better to boot with -s and try to load the .kext after typing sh /etc/rc

Danl
Danl
Posted
Posted
iwi2100 rev 465

-panic fixes

better to boot with -s and try to load the .kext after typing sh /etc/rc

 

OSX 10.4.8 with 8.8.1 (Jas) - Gateway 200arc laptop, intel pro/wireless 2100 3b card.

 

I have tried to it this way and still get a kernel panic here is a horrible picture of the screen and logs:

 

th_0614071418a.jpg

 

(dmesg.txt was empty and therefore will not upload)

 

edit:

Decided I should be a little more clear. When I install the kext and restart, I get an error saying i need to power off or restart the system. If I boot with the -s option, I can actually see the kernel panic (in the picture). If i remove the kext, boot with -s, and then attempt to load the kext, I get the same kernel panic instantly.

system.txt

ioreg.txt

reeiit
reeiit
Posted
Posted

Wow, seems like everything OK.

Enable PCI Ethernet in Control Panel, then run Networkselector.

try to turn radio on (option 1), get

Jun 15 21:03:37 ree-rss-computer kernel[0]: iwi3945: radio on 0x40000 = 0x0

Jun 15 21:03:37 ree-rss-computer kernel[0]: iwi3945: Setting scan to on

 

try to scan networks:

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: request scan called when driver not ready.

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Multiple concurrent scan requests in parallel. Ignoring second request.

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Initiating direct scan for <hidden>.

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 1 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 2 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 3 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 4 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 5 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 6 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 7 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 8 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 9 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 10 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 11 [ACTIVE 20]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 12 [PASSIVE 120]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: Scanning 13 [PASSIVE 120]

Jun 15 21:03:40 ree-rss-computer kernel[0]: iwi3945: total channels to scan 13

 

try to radio off:

Jun 15 21:04:54 ree-rss-computer kernel[0]: iwi3945: radio off 0x40000 = 0x0

 

Other options - nothing happens for me.

 

How to search networks (router Zyxel HT 660)?

 

thanks

dmesg.txt

ioreg.txt

system.txt

evis
evis
Posted
Posted

iwi3945

 

logs

 

the card recognized at the network configurations.

 

in the network selector it says turn card on/off, but I don't see the led turning on...

and the other options don't work.

 

dmseg.txt is empty somehow?

 

do I need the Network Frameworks of 10.4.7?

 

thanks for the great job.

system.txt

ioreg.txt

Danl
Danl
Posted
Posted
iwi2100 rev 469

changed ipw_up()

 

thanks for the logs

 

I will need to further test when I get home, and upload logs, but from my initial test of attempting to straight boot, then a -s boot with a kextload of the kext, both end in the exact same kernel panic. Will upload logs shortly.

itsmehere
itsmehere
Posted
Posted

iwi 3945 latest version... no panic

3945abg VEN 8086 DEVID 4222 REV_02

ifconfig en1 up/down for this interface don't working

networkselector enable/disable interface

but ifconfig show status en1:inactive

 

router Zyxel 660HW EE

thanks for the hard job.

dmesg.txt

ioreg.txt

system.txt

KemenAran
KemenAran
Posted
Posted

Logs for iwi3945 rev 464

 

As reeiit said, is some conditions, the driver starts to build a scan command - but it is only because the driver state check (just before running the scan) has been disabled. Thus the driver is actually not ready to scan, and when the scan command is sent, nothing happens.

 

It seems the driver is not ready because it is not GEO_CONFIGURED - and it would be only if a READY_ALIVE interrupt was received before starting the scan. I can't figure out why this interrupt isn't fired.

system.log.txt

jalavoui
jalavoui
Posted
  • Author
Posted

moses,

 

can you check the code MainController.m? this is probablly wrong:

 

if ((priv.ieee->networks[ii].capability & WLAN_CAPABILITY_IBSS) ? 1 : 0)

{

NSString *temp = [[NSBundle mainBundle] pathForResource:@"enc" ofType:@"tif"];

prot = [[NSImage alloc] initWithContentsOfFile:temp];

}

else prot = [[NSImage alloc] init];

 

if ((priv.ieee->networks[ii].capability & WLAN_CAPABILITY_PRIVACY) ? 1 : 0)

{

adhoc=@"x";

}

else adhoc=@"";

 

 

WLAN_CAPABILITY_IBSS is shifted with WLAN_CAPABILITY_PRIVACY

or you can try to include the ibss detection on the "info" column (remove the adhoc column)- but to do this you'll have to make new pictures for:

- bss secure net (this is ok)

- ibss net

- ibss secure net

 

 

 

 

KemenAran,

 

you're right - the scan() hack can't work like this

i just notice the the ipw_commit_rxon() is wrong (can lead to fw errors) i'll revert it and upload to svn

 

iwi3945 rev 471

http://code.google.com/p/iwidarwin/

William the Conquerer
William the Conquerer
Posted
Posted

I'm having problems with the iwi2200 drivers... everything works ok until it has to get an IP address. It gets stuck there and just tells me "getting ip address..." and never ends up doing anything... I'm using uphuck's 10.4.9 v1.3 release.

Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...