| Summary: | net.* fails to start rt61 wireless if because of missing /proc/net/wireless entry | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Martin Kramer <kramer.martin> |
| Component: | New packages | Assignee: | Mobile Herd (OBSOLETE) <mobile+disabled> |
| Status: | RESOLVED WONTFIX | ||
| Severity: | major | CC: | base-system |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | 183085 | ||
| Bug Blocks: | |||
|
Description
Martin Kramer
2007-10-08 15:34:41 UTC
sounds like the kernel wireless drivers arent doing enough ? (In reply to comment #1) > sounds like the kernel wireless drivers arent doing enough ? > That#s whatI fear too, but honestly, I cannot wait for the driver devs to fix that (if they would ever fix it, because it isn't anything that reduces the functionality of the driver). I think it's something that should be fixed on the script side. As I said iwconfig without params could be used to determine if a card has wireless extensions. # iwfonfig eth0 no wireless extensions. ra0 no wireless extensions. # modprobe rt61 # iwconfig eth0 no wireless extensions. ra0 RT61 Wireless ESSID:"" Nickname:"" Mode:Auto Frequency:2.412 GHz Bit Rate=54 Mb/s RTS thr:off Fragment thr:off Encryption key:off So as soon as you modprobe the driver the "no wireless extensions." goes away for ra0 in iwconfig. I checked the baselayout-2 files, but the method of checking for wireless extensions is the same as in the current baselayout. iwconfig output is not reliable enough. Sorry, but I got sick of specific driver hacks for iwconfig, so we enforce either /proc/net/wireless or something in /sys. If there isn't anything in either to show it's wireless then the driver will have to be patched. (In reply to comment #3) > iwconfig output is not reliable enough. > > Sorry, but I got sick of specific driver hacks for iwconfig, so we enforce > either /proc/net/wireless or something in /sys. If there isn't anything in > either to show it's wireless then the driver will have to be patched. > That would be the best solution I guess. But where can I file a BUG report with RaLink? As I said. It's the ralink driver not the rt61 or rt2x00 (which seams to lack of entries in &proc/net/wireless as well...) one. Our mobile team could try and patch the driver. If they can't then they will close this as UPSTREAM. You could try on the rt2x00 forums. I've done some more research, and as far as I can tell ralink-rt61 desperately tries to be a "normal" network card rather then being a wlan card. I've compared the /sys/class/net/${dev}/ directory of both the ra0 and eth0 interface. They look practically the same
# ls -li /sys/class/net/{eth0,ra0}/
/sys/class/net/eth0/:
insgesamt 0
5603 -r--r--r-- 1 root root 4096 14. Okt 12:19 address
5597 -r--r--r-- 1 root root 4096 14. Okt 12:19 addr_len
5604 -r--r--r-- 1 root root 4096 14. Okt 12:19 broadcast
5605 -r--r--r-- 1 root root 4096 14. Okt 12:19 carrier
5595 lrwxrwxrwx 1 root root 0 14. Okt 12:19 device -> ../../../devices/pci0000:00/0000:00:04.0
5606 -r--r--r-- 1 root root 4096 14. Okt 12:19 dormant
5600 -r--r--r-- 1 root root 4096 14. Okt 12:19 features
5609 -rw-r--r-- 1 root root 4096 14. Okt 12:19 flags
5599 -r--r--r-- 1 root root 4096 14. Okt 12:19 ifindex
5598 -r--r--r-- 1 root root 4096 14. Okt 12:19 iflink
5602 -r--r--r-- 1 root root 4096 14. Okt 12:19 link_mode
5608 -rw-r--r-- 1 root root 4096 14. Okt 12:19 mtu
5607 -r--r--r-- 1 root root 4096 14. Okt 12:19 operstate
5636 drwxr-xr-x 2 root root 0 14. Okt 12:19 power
5612 drwxr-xr-x 2 root root 0 14. Okt 12:19 statistics
5594 lrwxrwxrwx 1 root root 0 14. Okt 12:19 subsystem -> ../../../class/net
5610 -rw-r--r-- 1 root root 4096 14. Okt 12:19 tx_queue_len
5601 -r--r--r-- 1 root root 4096 14. Okt 12:19 type
5593 -rw-r--r-- 1 root root 4096 14. Okt 12:19 uevent
5611 -rw-r--r-- 1 root root 4096 14. Okt 12:19 weight
/sys/class/net/ra0/:
insgesamt 0
6986 -r--r--r-- 1 root root 4096 14. Okt 11:55 address
6980 -r--r--r-- 1 root root 4096 14. Okt 11:55 addr_len
6987 -r--r--r-- 1 root root 4096 14. Okt 11:55 broadcast
6988 -r--r--r-- 1 root root 4096 14. Okt 11:55 carrier
6989 -r--r--r-- 1 root root 4096 14. Okt 11:55 dormant
6983 -r--r--r-- 1 root root 4096 14. Okt 11:55 features
6992 -rw-r--r-- 1 root root 4096 14. Okt 11:55 flags
6982 -r--r--r-- 1 root root 4096 14. Okt 11:55 ifindex
6981 -r--r--r-- 1 root root 4096 14. Okt 11:55 iflink
6985 -r--r--r-- 1 root root 4096 14. Okt 11:55 link_mode
6991 -rw-r--r-- 1 root root 4096 14. Okt 11:55 mtu
6990 -r--r--r-- 1 root root 4096 14. Okt 11:55 operstate
7019 drwxr-xr-x 2 root root 0 14. Okt 11:55 power
6995 drwxr-xr-x 2 root root 0 14. Okt 11:55 statistics
6979 lrwxrwxrwx 1 root root 0 14. Okt 11:55 subsystem -> ../../../class/net
6993 -rw-r--r-- 1 root root 4096 14. Okt 11:55 tx_queue_len
6984 -r--r--r-- 1 root root 4096 14. Okt 11:55 type
6978 -rw-r--r-- 1 root root 4096 14. Okt 11:55 uevent
6994 -rw-r--r-- 1 root root 4096 14. Okt 11:55 weight
I'm no expert for WLAN devices, but shouldn't the ralink directory look totally or at least a bit different? As you see there is no wireless file at all. And it get's even better. Upon checking /proc/net/dev I found out that ra0 appears there. Totally wrong place if you ask me.
(In reply to comment #6) > I've done some more research, and as far as I can tell ralink-rt61 desperately > tries to be a "normal" network card rather then being a wlan card. I've > compared the /sys/class/net/${dev}/ directory of both the ra0 and eth0 > interface. They look practically the same That sucks ass > > I'm no expert for WLAN devices, but shouldn't the ralink directory look totally > or at least a bit different? As you see there is no wireless file at all. And > it get's even better. Upon checking /proc/net/dev I found out that ra0 appears > there. Totally wrong place if you ask me. Every network interface has an entry there, wireless or otherwise. >
> Every network interface has an entry there, wireless or otherwise.
>
OK. So the only real problem(s) is/are that ralink-rt61 refuses to create an entry in /proc/net/wireless and /sys/class/net/ra0/wireless.
Reading through tons of forums posts I came across problems that older versions of the rt61 driver used to produced error msgs like this: "Driver using old /proc/net/wireless support, please fix driver !" Is /proc/net/wirelss deprecated? And if so could it be RaLink kicked out support for it but "forgot" to implement the /sys/class/net stuff?
(In reply to comment #8) > Reading through tons of forums posts I came across problems that older versions > of the rt61 driver used to produced error msgs like this: "Driver using old > /proc/net/wireless support, please fix driver !" Is /proc/net/wirelss > deprecated? And if so could it be RaLink kicked out support for it but "forgot" > to implement the /sys/class/net stuff? > /proc/net/wireless is deprecated, but only if the same stuff can be found in /sys somewhere. That's why we check for both, and currently onlt some ra drivers provide neither. (In reply to comment #9) > > /proc/net/wireless is deprecated, but only if the same stuff can be found in > /sys somewhere. That's why we check for both, and currently onlt some ra > drivers provide neither. > But wouldn't it be best to implement some "pretend-wireless" option tot the init scripts that just deactivates these tests for such cards? (In reply to comment #10) > But wouldn't it be best to implement some "pretend-wireless" option tot the > init scripts that just deactivates these tests for such cards? No, it would be best to fix the driver. (In reply to comment #11) > > No, it would be best to fix the driver. > Sorry. didn't want to sound rude. I just tried to come up with some ideas for workarounds until the driver gets fixed (if it will ever get fixed). removed from tree, wontfix |
