Mohamed Abbas [Thu, 29 Nov 2007 03:10:15 +0000 (11:10 +0800)]
iwlwifi: avoid firmware command sending if rfkill is enabled
This patch fixed a ucode timeout issue and worked code with suspend
to disk.
Signed-off-by: Mohamed Abbas <mabbas@linux.intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Mohamed Abbas [Thu, 29 Nov 2007 03:10:14 +0000 (11:10 +0800)]
iwlwifi: fix ucode assertion for RX queue overrun
This patch allows the driver to restock the RX queue early if the RX
queue is almost empty. This will help on avoiding any ucode assert
for the RX overrun problem.
Signed-off-by: Mohamed Abbas <mabbas@linux.intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Mohamed Abbas [Thu, 29 Nov 2007 03:10:13 +0000 (11:10 +0800)]
iwlwifi: enhance WPA authenication stability
This patch enhanced WPA authenication stability by avoiding scan
immediately followed by association. We don't do any scanning right
after association in next several seconds. This will allow WPA
authentication to take place without any interruption.
Signed-off-by: Mohamed Abbas <mabbas@linux.intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cahill, Ben M [Thu, 29 Nov 2007 03:09:55 +0000 (11:09 +0800)]
iwlwifi: add comments to iwl*-base.c
Add comments to iwlXXXX-base.c
Signed-off-by: Cahill, Ben M <ben.m.cahill@intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cahill, Ben M [Thu, 29 Nov 2007 03:09:54 +0000 (11:09 +0800)]
iwlwifi: add comments, mostly on Tx queues
Add comments, mostly on Tx queues
Signed-off-by: Cahill, Ben M <ben.m.cahill@intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
- Remove HT code from iwl-3945.h - it is not needed here as 3945
does not support HT. The code ended up here during the header file
split.
- Modify a few places where the CONFIG variables were named
incorrectly: all changes are to comments only.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Tomas Winkler [Thu, 29 Nov 2007 03:09:41 +0000 (11:09 +0800)]
iwlwifi: Support for uCode without init and bsm section
This patch enables loading fw w/o init and bsm section. It also provides
general cleanup of the rutine.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com> Signed-off-by: Gregory Greenman <gregory.greenman@intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Tue, 27 Nov 2007 20:51:57 +0000 (21:51 +0100)]
rt2x00: Cleanup rfkill
The label exit_free_polldev is no longer used and can be removed.
input_free_polled_device() also calls input_free_device(), so
don't call it seperatly.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Larry Finger [Tue, 27 Nov 2007 20:51:39 +0000 (21:51 +0100)]
rt2x00: Remove redundant code in rfkill setup
In rt2x00rfkill.c, routine input_allocate_device() is called even though
input_allocate_polled_device(), which was just called, includes a call to
that routine. This patch, which has not been tested, removes the redundant code.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Mattias Nissler [Tue, 27 Nov 2007 20:50:26 +0000 (21:50 +0100)]
rt2x00: Only update rssi average approximation on receiving beacon frames.
Restrict rssi average updating to beacon frames of the bssid the
interface is
associated with. Without this restriction, strong signals belonging to other
BSS, e.g. beacon frames coming from a nearby AP, would cause incorrectly high
rssi approximation values. This would then cause the link tuner to reduce
sensitivity, resulting in transmissions from the BSS associated to to be
missed.
Signed-off-by: Mattias Nissler <mattias.nissler@gmx.de> Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Tue, 27 Nov 2007 20:49:29 +0000 (21:49 +0100)]
rt2x00: Add TX/RX frame dumping facility
This adds TX/RX frame dumping capabilities through debugfs.
The intention is that with this approach debugging of rt2x00 is
simplified since _all_ frames going in and out of the device
are send to debugfs as well along with additional information
like the hardware descriptor.
Based on the patch by Mattias Nissler.
Mattias also has some tools that will make the dumped frames
available to wireshark: http://www-user.rhrk.uni-kl.de/~nissler/rt2x00/
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Thu, 24 Jan 2008 09:56:25 +0000 (01:56 -0800)]
rt2x00: Add skb descriptor
Use the skb->cb field to add a frame description that can be used
to transfer information passed each rt2x00 layer. This reduces the
required arguments for rt2x00lib_write_tx_desc().
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ivo van Doorn [Tue, 27 Nov 2007 20:47:56 +0000 (21:47 +0100)]
rt2x00: Extend PLCP descriptor definition for rt2400pci
Extend word field definitions for the PLCP words
in the TX descriptor to contain the BBP fields as well.
This will remove rt2400pci_write_tx_desc() from the
checkstack script.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes Berg [Sat, 24 Nov 2007 20:11:09 +0000 (21:11 +0100)]
b43legacy: include full 64-bit timestamp in monitor mode
When monitor mode is enabled, this will make b43legacy read out
the full 64-bit MAC time from the chip for each received packet.
This patch has been ported from b43.
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it> Cc: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Use initialized_var() to properly fix a bogus gcc warning.
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it> Cc: Frank Lichtenheld <frank@lichtenheld.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Larry Finger [Mon, 26 Nov 2007 16:29:47 +0000 (10:29 -0600)]
b43: Changes to enable BCM4311 rev 02 with wireless core revision 13
The BCM94311MCG rev 02 chip has an 802.11 core with revision 13 and
has not been supported until now. The changes include the following:
(1) Add the 802.11 rev 13 device to the ssb_device_id table to load b43.
(2) Add PHY revision 9 to the supported list.
(3) Change the 2-bit routing code for address extensions to 0b10 rather
than the 0b01 used for the 32-bit case.
(4) Remove some magic numbers in the DMA setup.
The DMA implementation for this chip supports full 64-bit addressing with
one exception. Whenever the Descriptor Ring Buffer is in high memory, a
fatal DMA error occurs. This problem was not present in 2.6.24-rc2 due
to code to "Bias the placement of kernel pages at lower PFNs". When
commit 44048d70 reverted that code, the DMA error appeared. As a "fix",
use the GFP_DMA flag when allocating the buffer for 64-bit DMA. At present,
this problem is thought to arise from a hardware error.
This patch has been tested on my system and by Cédric Caumont
<icare40@hotmail.com>.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Acked-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Stefano Brivio [Sun, 25 Nov 2007 10:10:33 +0000 (11:10 +0100)]
b43: reinit on too many PHY TX errors
Restart the hardware on too many PHY TX errors. A thousand PHY TX errors
per 15 seconds means we won't be able to recover for sure.
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it> Signed-off-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes Berg [Sat, 24 Nov 2007 14:06:33 +0000 (15:06 +0100)]
b43: include FCS in frames handed to mac80211
Sometimes it can be useful to see the FCS, especially when
bad-FCS frames are shown. Pass the FCS to mac80211 and let
it worry about snipping it off when required.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes Berg [Fri, 23 Nov 2007 13:50:51 +0000 (14:50 +0100)]
b43: include full 64-bit timestamp in monitor mode
When monitor mode is enabled, this will make b43 read out the
full 64-bit MAC time from the chip for each received packet.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Wed, 28 Nov 2007 08:15:11 +0000 (09:15 +0100)]
libertas: less eventcause shifts
* only shift eventcause once
* convert mac events to decimal, as this is what the firmware
manual uses in section 6.1, too
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Wed, 28 Nov 2007 16:30:21 +0000 (17:30 +0100)]
libertas: remove user-specified channel list
Remove the ability to specify channels to scan via debugfs
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Wed, 28 Nov 2007 16:29:36 +0000 (17:29 +0100)]
libertas: remove numprobes
Remove the ability to specify number of probes via debugfs
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Wed, 28 Nov 2007 13:08:12 +0000 (14:08 +0100)]
libertas: tweak tx path debugging
Make two functions in the TX packet path emit
their debug messages with LBS_DEB_TX, not LBS_DEB_MAIN.
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Wed, 28 Nov 2007 13:05:02 +0000 (14:05 +0100)]
libertas: tweak association debug output
Change debug output codes from LBS_DEB_JOIN to LBS_DEB_ASSOC
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Mon, 26 Nov 2007 09:07:14 +0000 (10:07 +0100)]
libertas: rework event subscription
This patch moves re-works the implementation of event subscription
via debugfs. For this:
* it tells cmd.c and cmdresp.c about CMD_802_11_SUBSCRIBE_EVENT
* removes lots of low-level cmd stuff from debugfs.c
* create unified functions to read/write snr, rssi, bcnmiss and
failcount
* introduces #define's for subscription event bitmask values
* add a function to search for a specific element in an IE
(a.k.a. TLV)
* add a function to find out the size of the TLV. This is needed
because lbs_prepare_and_send_command() has an argument for a
data buffer, but not for it's lengths and TLVs can be, by
definition, vary in size.
* fix a bug where it was not possible to disable an event
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Fri, 23 Nov 2007 14:43:44 +0000 (15:43 +0100)]
libertas: remove arbitrary typedefs
New typedefs are usually frowned upon. This patch changes
libertas_adapter -> struct libertas_adapter
libertas_priv -> struct libertas_priv
While passing, make everything checkpatch.pl-clean that gets touches.
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Brajesh Dave [Tue, 20 Nov 2007 22:44:28 +0000 (17:44 -0500)]
libertas: configurable beacon interval
Requires firmware version 5.110.19.p0 or newer, available here:
http://dev.laptop.org/pub/firmware/libertas/
Signed-off-by: Ashish Shukla <ashishs@marvell.com> Signed-off-by: Javier Cardona <javier@cozybit.com> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
This feature requires firmware version 5.110.19.p0 or newer, available
here: http://dev.laptop.org/pub/firmware/libertas/
Signed-off-by: Ashish Shukla <ashishs@marvell.com> Signed-off-by: Javier Cardona <javier@cozybit.com> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Andres Salomon [Tue, 20 Nov 2007 22:44:04 +0000 (17:44 -0500)]
libertas: nuke useless variable usbdriver_name and useless comments
I think it was pretty obvious what fields in if_usb_driver are...
Signed-off-by: Andres Salomon <dilinger@debian.org> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Andres Salomon [Tue, 20 Nov 2007 22:43:55 +0000 (17:43 -0500)]
libertas: reset devices upon disconnect rather than module unloading
1) Do not reset libertas devices upon module unload. We're unloading
the module, we're not killing off devices.
2) Instead, reset libertas devices inside if_usb_disconnect, as we're
killing off interfaces and so on.
3) Resetting via disconnect callback means we no longer need to keep
track of probed libertas devices; kill off that list (and its wonderful
lack of locking..), as well.
Drop a useless comment as well.
Signed-off-by: Andres Salomon <dilinger@debian.org> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Andres Salomon [Tue, 20 Nov 2007 22:43:45 +0000 (17:43 -0500)]
libertas: mark module_init/exit functions as __init/__exit
Signed-off-by: Andres Salomon <dilinger@debian.org> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Andres Salomon [Tue, 20 Nov 2007 22:43:32 +0000 (17:43 -0500)]
libertas: drop useless default_fw_name variable
Signed-off-by: Andres Salomon <dilinger@debian.org> Signed-off-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Daniel Drake [Mon, 19 Nov 2007 15:00:29 +0000 (15:00 +0000)]
zd1211rw: port to mac80211
This seems to be working smoothly now. Let's not hold back the mac80211
transition any further. This patch ports the existing driver from softmac
to mac80211.
Many thanks to everyone who helped out with the porting efforts.
Signed-off-by: Daniel Drake <dsd@gentoo.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Holger Schurig [Thu, 15 Nov 2007 23:05:47 +0000 (18:05 -0500)]
libertas: move to uniform lbs_/LBS_ namespace
This patch unifies the namespace of variables, functions defines
and structures. It does:
- rename libertas_XXX to lbs_XXX
- rename LIBERTAS_XXX to lbs_XXX
- rename wlan_XXX to lbs_XXX
- rename WLAN_XXX to LBS_XXX (but only those that were
defined in libertas-local *.h files, e.g. not defines
from net/ieee80211.h)
While passing, I fixed some checkpatch.pl errors too.
Signed-off-by: Holger Schurig <hs4233@mail.mn-solutions.de> Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Larry Finger [Fri, 9 Nov 2007 22:56:25 +0000 (16:56 -0600)]
ssb: Convert to use of the new SPROM structure
In disagreement with the SPROM specs, revision 3 devices appear to have
moved the MAC address.
Change ssb to handle the revision 4 SPROM, which is a different size.
This change in size is handled by adding a new variable to the ssb_sprom
struct and using it whenever possible. For those routines that do not
have access to this structure, a 'u16 size' argument is added.
The new PCI_ID for the BCM4328 is also added.
Testing of the Revision 4 SPROM, which is used on the BCM4328, was done
by Michael Gerdau <mgerdau@tiscali.de>.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Larry Finger [Fri, 9 Nov 2007 22:54:45 +0000 (16:54 -0600)]
ssb: Add new SPROM structure while keeping the old
The SPROM's for various devices utilizing the Sonics Silicon Backplane come
with various revisions. The Revision 2 SPROM inherited the data layout of 1, and
Revision 3 inherited the layout of 2. The first instance of Revision 4 has
now been found in a BCM4328 wireless LAN card. This device does not inherit any
layout from previous versions. Although it was possible to create a data
structure that kept all the old layouts, we decided to start fresh, keep only
those SPROM variables that are used by the drivers that utilize ssb, and to
do the conversion in such a manner that neither compilation or execution will
be affected if a bisection lands in the middle of these changes, while keeping
the patches as small as possible.
In this patch, the sprom structures are changed while maintaining the old ones.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Mon, 12 Nov 2007 14:02:40 +0000 (15:02 +0100)]
rt2x00: Move duplicate code into rt2x00pci_txdone()
rt2400pci, rt2500pci and rt61 require different
txdone handling, but the code that pushes the frame
upstream and cleans up the entry is identical to
all of them.
This will create the function rt2x00pci_txdone()
to remove the duplicate code.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
use uninitialized_var() to avoid the following bogus warning:
CC [M] drivers/net/wireless/b43/debugfs.o
drivers/net/wireless/b43/debugfs.c: In function ‘b43_debugfs_read’:
drivers/net/wireless/b43/debugfs.c:355: warning: ‘ret’ may be used uninitialized in this function
Signed-off-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: Frank Lichtenheld <frank@lichtenheld.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Zhu Yi [Wed, 7 Nov 2007 06:06:27 +0000 (22:06 -0800)]
remove unused iwl4965_init_hw_rates function
The patch removes the unused function to fix a recursive bug caused
by namespace change.
Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Prefix all symbols with iwl3945_ or iwl4965_ and thus allow building
the driver into the kernel. Also remove all the useless default
statements in Kconfig while we're at it.
Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Mohamed Abbas [Wed, 7 Nov 2007 06:06:25 +0000 (22:06 -0800)]
iwl4965: fix cannot find a suitable rate issue
This patch fixes the iwl4965 problem for "Can not find a suitable rate
issues." by making rs_switch_to_mimo and rs_switch_to_siso functions
return -1 when CONFIG_IWL4965_HT is not selected. They used to return 0,
which means we can switch to HT rate causing the rate scale problem and
the error message.
The patch also fix another bug reported by Ben Cahill that it uses wrong
value for max_success_limit.
Signed-off-by: Mohamed Abbas <mabbas@linux.intel.com> Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Zhu Yi [Wed, 7 Nov 2007 06:06:24 +0000 (22:06 -0800)]
iwlwifi: replace 0x8086 with PCI_VENDOR_ID_INTEL
Replace 0x8086 with PCI_VENDOR_ID_INTEL for PCI_DEVICE declaration.
Signed-off-by: Zhu Yi <yi.zhu@intel.com> Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Stefano Brivio [Tue, 6 Nov 2007 21:49:05 +0000 (22:49 +0100)]
b43: rewrite A PHY initialization
Rewrite and sync A PHY initialization with specs, thus allowing for further
work to be done on 802.11a support. Note that A PHY initialization involves
G PHYs as well.
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it> Acked-by: Michael Buesch <mb@bu3sch.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Stefano Brivio [Tue, 6 Nov 2007 21:48:45 +0000 (22:48 +0100)]
b43legacy: rewrite and fix rfkill initialization
The rfkill subsystem doesn't like code like that
rfkill_allocate();
rfkill_register();
rfkill_unregister();
rfkill_register(); /* <- This will crash */
This sequence happens with
modprobe b43
ifconfig wlanX up
ifconfig wlanX down
ifconfig wlanX up
Fix this by always re-allocating the rfkill stuff before register.
The patch to b43 by Michael Buesch <mb@bu3sch.de> has been ported to
b43legacy.
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it> Signed-off-by: John W. Linville <linville@tuxdriver.com>
SGISEEQ: use cached memory access to make driver work on IP28
SGI IP28 machines would need special treatment (enable adding addtional
wait states) when accessing memory uncached. To avoid this pain I changed
the driver to use only cached access to memory.
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Olof Johansson [Thu, 29 Nov 2007 02:58:06 +0000 (20:58 -0600)]
pasemi_mac: Print warning when not attaching to a PHY
pasemi_mac: Print warning when not attaching to a PHY
Print a warning on the console when not connecting to a phy for an interface.
It turns out to be a pretty common problem when someone gets the MDIO info
wrong in their device tree, resulting in the macs running at a fixed 1Gbit FD.
Signed-off-by: Olof Johansson <olof@lixom.net> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Olof Johansson [Thu, 29 Nov 2007 02:57:45 +0000 (20:57 -0600)]
pasemi_mac: SKB unmap optimization
pasemi_mac: SKB unmap optimization
Avoid touching skb_shinfo() in the unmap path, since it turns out to
normally cause cache misses and delays. instead, save number of fragments
in the TX_RING_INFO structures since that's all that's needed anyway.
Signed-off-by: Olof Johansson <olof@lixom.net> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Olof Johansson [Thu, 29 Nov 2007 02:57:09 +0000 (20:57 -0600)]
pasemi_mac: Improve RX interrupt mitigation
pasemi_mac: Improve RX interrupt mitigation
Currently the receive side interrupts will go off on the reception of
a packet, NAPI will poll the ring and keep polling as long as there's
a decent amount of packets to receive.
This is less than optimal, especially for LRO where it's better if we
have a more substantial amount of packets to process at once, to get
the real LRO benefits.
So, set the count threshold to a higher value and use the timeout feature
that will give us an interrupt even if not enough packets have come in
to set off the count threshold.
FIXME: It'd be real nice to have ethtool support for users to tune this
at runtime.
Signed-off-by: Olof Johansson <olof@lixom.net> Signed-off-by: Jeff Garzik <jeff@garzik.org>
projects / linux-2.6 / log