headers
Michael Wu | 1 Mar 2007 03:14

[PATCH] d80211: Remove tx_timeout callback

On Wednesday 28 February 2007 15:39, Michael Wu wrote:
> This prevents drivers from crashing if there is a TX timeout and the reset
> callback isn't implemented.
>
Actually, I don't think we should bother setting the TX timeout callback at 
all.

--

d80211: Remove tx_timeout callback

From: Michael Wu <flamingice@...>

This never worked in the first place and we can't use the network watchdog
anyway since that checks if the queue is stopped, but wireless devices can
have many queues. TX timeouts on virtual interfaces don't really make sense
either since they can't stall.

Signed-off-by: Michael Wu <flamingice@...>
---

 net/mac80211/ieee80211.c |   15 ---------------
 1 files changed, 0 insertions(+), 15 deletions(-)

diff --git a/net/mac80211/ieee80211.c b/net/mac80211/ieee80211.c
index bbdf928..cdc8df4 100644
--- a/net/mac80211/ieee80211.c
+++ b/net/mac80211/ieee80211.c
 <at>  <at>  -2078,19 +2078,6  <at>  <at>  static int ieee80211_change_mtu_apdev(st
 }
(Continue reading)

Permalink | Reply | 1 comment |
headers
Michael Wu | 1 Mar 2007 06:43

Re: [PATCH 5/5] d80211: switch STA interfaces to PS mode during scan

On Wednesday 28 February 2007 16:44, Michael Wu wrote:
> On Wednesday 28 February 2007 16:42, Michael Wu wrote:
> > This makes scans switch STA interfaces into PS mode so the AP queues
> > frames destined for us while we are scanning. This is achieved by sending
> > a nullfunc data frame with the PS mode bit set before scanning commences,
> > and a PS poll frame after scanning is completed.
>
> Er. I should update the description too..
>
And drop the ps_poll part since we don't use the ps poll frame now..

--

d80211: switch STA interfaces to PS mode during scan

From: Michael Wu <flamingice@...>

This makes scans switch STA interfaces into PS mode so the AP queues frames
destined for us while we are scanning. This is achieved by sending a
nullfunc data frame with the PS mode bit set before scanning commences,
and a nullfunc data frame without the PS mode bit set after scanning is
completed.

Signed-off-by: Michael Wu <flamingice@...>
---

 net/mac80211/ieee80211_sta.c |   51 +++++++++++++++++++++++++++++++++++-------
 1 files changed, 43 insertions(+), 8 deletions(-)

diff --git a/net/mac80211/ieee80211_sta.c b/net/mac80211/ieee80211_sta.c
(Continue reading)

Permalink | Reply | 3 comments |
headers
Michael Wu | 1 Mar 2007 06:40

Re: [PATCH 4/5] d80211: Stop virtual interfaces during scan

On Wednesday 28 February 2007 15:39, Michael Wu wrote:
> From: Michael Wu <flamingice@...>
>
> This prevents data frames from being queued on the master device if it is
> in the midst of a scan. It also makes both master and virtual interfaces
> properly set trans_start when frames are sent so the network watchdog does
> not try to reset the interfaces. tx_queue_len is left as the default on
> virtual interfaces to allow frames to be queued while the device is
> scanning.
>
And now that all STA interfaces are woken up after a scan, we also can remove 
the scan_dev check.

--

d80211: Stop virtual interfaces during scan

From: Michael Wu <flamingice@...>

This prevents data frames from being queued on the master device if it is
in the midst of a scan. It also makes both master and virtual interfaces
properly set trans_start when frames are sent. tx_queue_len is left as the 
default on virtual interfaces to allow frames to be queued while the device 
is scanning.

Signed-off-by: Michael Wu <flamingice@...>
---

 net/mac80211/ieee80211.c     |    5 +++--
 net/mac80211/ieee80211_sta.c |   28 +++++++++++++++++++---------
(Continue reading)

Permalink | Reply | 1 comment |
headers
Johannes Berg | 1 Mar 2007 12:29
Favicon

Re: [PATCH] allow wiphy renaming

On Wed, 2007-02-28 at 15:15 +0100, Johannes Berg wrote:
> This patch enables userspace to rename 802.11 PHYs via nl80211. It adds
> the first multicast group and notifies it of these renames as well.

> Tested and works :)

Well, I forgot to test the corner cases. The patch is somewhat broken.

> +	/* prohibit calling the thing phy%d when %d is not its number */
> +	sscanf(newname, PHY_NAME "%n", &idx, &taken);
> +	if (taken == strlen(newname) && idx != rdev->idx)
> +		return -EINVAL;

sscanf has a bug that makes this use case not work. If it worked
correctly, this would also disallow things like 'phy007' when the index
is not 7.

The whole point in disallowing something other than the own number was
to simplify code in other places so I don't have to check for used
numbers when assigning new ones. Any comments on the whole thing? I
kinda fear that if I go through with this someone will show up who
really desperately needs to call something phy7 when it has index 22 and
I will be held responsible ;)

OTOH, this may be an incentive to not use phy* for stable names to begin
with, avoiding the problem we see with netdevs and people continually
complaining about unstable names...

johannes
(Continue reading)

Permalink | Reply | 2 comments |
headers
Johannes Berg | 1 Mar 2007 12:31
Favicon

Re: Patch: prism54usb, new device

I guess this should be on the radar of the prism54 people since it
applies against that driver and not p54?

Please check the MAINTAINERS file (also on
http://linuxwireless.org/en/developers/maintainers) next time.

On Wed, 2007-02-28 at 16:50 +0100, Ivo Couckuyt wrote:
> Using islsm/prism54usb  succesfully with the Philips CPWUA054 device for 
> a year now, made a little
> patch for prism54usb
> patched against
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git
> 
> 
> more info:
> http://www.p4c.philips.com/cgi-bin/dcbint/cpindex.pl?ctn=CPWUA054/00&scy=BE&slg=NL&grp=PC_PRODUCTS_GR_BE_CONSUMER&cat=PC_NETWORKING_CA_BE_CONSUMER&sct=WIRELESS_ADAPTERS_SU_BE_CONSUMER
> 
> regards,
> ilm
Permalink | Reply | 2 comments |
headers
Ivo Couckuyt | 1 Mar 2007 12:45
Picon
Favicon

Re: Patch: prism54usb, new device

i got the info from here:
http://bugs.gentoo.org/show_bug.cgi?id=142651
comments 17-19
19 suggested to sent it here

don't know about the actually driver name, it's known to me as islsm
which is not the same as the recently added p54? driver based on dscape 
(from wireless-dev)

regards,
ilm

Johannes Berg wrote:
> I guess this should be on the radar of the prism54 people since it
> applies against that driver and not p54?
>
> Please check the MAINTAINERS file (also on
> http://linuxwireless.org/en/developers/maintainers) next time.
>
> On Wed, 2007-02-28 at 16:50 +0100, Ivo Couckuyt wrote:
>   
>> Using islsm/prism54usb  succesfully with the Philips CPWUA054 device for 
>> a year now, made a little
>> patch for prism54usb
>> patched against
>>
>> git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git
>>
>>
>> more info:
(Continue reading)

Permalink | Reply | 1 comment |
headers
Johannes Berg | 1 Mar 2007 14:08
Favicon

[TRIAL BALLOON] remove rtnetlink wext

This patch removes rtnetlink wireless extensions. The rtnetlink
transport for wireless extensions is
 * a fairly large chunk of code
 * not enabled on most kernels
 * not required by any userspace tools
 * only about a year old
 * pointless

Thus, removing this code seems feasible.

Signed-off-by: Johannes Berg <johannes@...>

---
 drivers/net/wireless/Kconfig |   10 
 include/net/iw_handler.h     |   10 
 net/core/rtnetlink.c         |   30 --
 net/wireless/wext-old.c      |  608 -------------------------------------------
 4 files changed, 658 deletions(-)

--- wireless-dev.orig/drivers/net/wireless/Kconfig	2007-03-01 13:20:00.164246483 +0100
+++ wireless-dev/drivers/net/wireless/Kconfig	2007-03-01 13:34:02.404246483 +0100
 <at>  <at>  -25,16 +25,6  <at>  <at>  config NET_RADIO
 	  the tools from
 	  <http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/Tools.html>.

-config NET_WIRELESS_RTNETLINK
-	bool "Wireless Extension API over RtNetlink"
-	depends on NET_RADIO
-	---help---
-	  Support the Wireless Extension API over the RtNetlink socket
(Continue reading)

Permalink | Reply | 0 comments |
headers
Johannes Berg | 1 Mar 2007 14:25
Favicon

[TRIAL BALLOON] rework wireless Kconfig

This patch
 * kills NET_RADIO
 * adds two new wireless driver menus for pre-802.11 and 802.11
 * makes WIRELESS_EXT visible (to avoid the arguments we had in
   commit c1783454a31e05b94774951b0b5d1eb9075ebfb4)
 * changes everything that depended on NET_RADIO to select WIRELESS_EXT

By removing NET_RADIO, these changes pave the way to making wireless
extensions optional when cfg80211 can fully take over for some drivers
and you don't have any older drivers that still require wext.

These conflict with other changesets (notable Ivo's) but they should be
easy to respin once those get in. This also depends on removing wext
netlink, but it's trivial to do them w/o removing netlink wext of
course. Though I think that the rtnetlink wext option shouldn't be under
drivers so I'd move it in the case we don't want to remove it.

johannes

---
 drivers/net/Makefile                           |    2 
 drivers/net/wireless/Kconfig                   |  100 +++++++++----------------
 drivers/net/wireless/bcm43xx/Kconfig           |    3 
 drivers/net/wireless/hostap/Kconfig            |    2 
 drivers/net/wireless/mac80211/adm8211/Kconfig  |    2 
 drivers/net/wireless/mac80211/bcm43xx/Kconfig  |    2 
 drivers/net/wireless/mac80211/p54/Kconfig      |    2 
 drivers/net/wireless/mac80211/rt2x00/Kconfig   |    2 
 drivers/net/wireless/mac80211/rtl818x/Kconfig  |    2 
 drivers/net/wireless/mac80211/zd1211rw/Kconfig |    2
(Continue reading)

Permalink | Reply | 2 comments |
headers
Johannes Berg | 1 Mar 2007 15:35
Favicon

[PATCH] schedule wext/rtnl for removal

Since wext is being replaced as fast as we can (it'll probably stick
around for legacy drivers though) and the wext/netlink stuff was never
really used, this schedules it for removal.

The removal schedule is tight but there are no users of the code, the
main user of the wext user interface are the wireless-tools, they only
have an alpha version using the netlink interface and even that is
incomplete.

Signed-off-by: Johannes Berg <johannes@...>

---
 Documentation/feature-removal-schedule.txt |   12 ++++++++++++
 1 file changed, 12 insertions(+)

--- linux-2.6.orig/Documentation/feature-removal-schedule.txt	2007-03-01 15:25:31.714246483 +0100
+++ linux-2.6/Documentation/feature-removal-schedule.txt	2007-03-01 15:32:23.904246483 +0100
 <at>  <at>  -304,3 +304,15  <at>  <at>  Why:	The code says it was obsolete when 
 Who:	Richard Purdie <rpurdie@...>

 ---------------------------
+
+What:	Wireless extensions over netlink (CONFIG_NET_WIRELESS_RTNETLINK)
+When:	with the merge of wireless-dev, 2.6.22 or later
+Why:	The option/code is
+	 * not enabled on most kernels
+	 * not required by any userspace tools (except an experimental one,
+	   and even there only for some parts, others use ioctl)
+	 * pointless since wext is no longer evolving and the ioctl
+	   interface needs to be kept
(Continue reading)

Permalink | Reply | 0 comments |
headers
Johannes Berg | 1 Mar 2007 15:16
Favicon

Re: [TRIAL BALLOON] rework wireless Kconfig

On Thu, 2007-03-01 at 09:11 -0500, Luis R. Rodriguez wrote:

> > --- wireless-dev.orig/net/wireless/Kconfig      2007-03-01 14:18:51.854246483 +0100
> > +++ wireless-dev/net/wireless/Kconfig   2007-03-01 14:19:19.614246483 +0100
> >  <at>  <at>  -29,3 +29,11  <at>  <at>  config NL80211
> >
> >           If unsure, say Y.
> >
> > +config WIRELESS_EXT
> > +       bool "Wireless extensions"
> > +       default n
> > +       ---help---
> > +         This compiles code for the wireless extensions.
> > +
> > +         Say N if you can unless you have external modules
> > +         that rely on it.
> 
> I'd say we put more descriptive text here and explain a little of how
> cfg80211 is supposed to replace it with nl80211.

Noted. I'll need to respin the patches anyway and just wanted some early
feedback.

johannes
Permalink | Reply | 0 comments |

Gmane