Ilpo Järvinen [Tue, 13 May 2008 09:54:19 +0000 (02:54 -0700)]
tcp FRTO: work-around inorder receivers
If receiver consumes segments successfully only in-order, FRTO
fallback to conventional recovery produces RTO loop because
FRTO's forward transmissions will always get dropped and need to
be resent, yet by default they're not marked as lost (which are
the only segments we will retransmit in CA_Loss).
Price to pay about this is occassionally unnecessarily
retransmitting the forward transmission(s). SACK blocks help
a bit to avoid this, so it's mainly a concern for NewReno case
though SACK is not fully immune either.
This change has a side-effect of fixing SACKFRTO problem where
it didn't have snd_nxt of the RTO time available anymore when
fallback become necessary (this problem would have only occured
when RTO would occur for two or more segments and ECE arrives
in step 3; no need to figure out how to fix that unless the
TODO item of selective behavior is considered in future).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Reported-by: Damon L. Chesser <damon@damtek.com> Tested-by: Damon L. Chesser <damon@damtek.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Tue, 13 May 2008 09:53:26 +0000 (02:53 -0700)]
tcp FRTO: Fix fallback to conventional recovery
It seems that commit 009a2e3e4ec ("[TCP] FRTO: Improve
interoperability with other undo_marker users") run into
another land-mine which caused fallback to conventional
recovery to break:
1. Cumulative ACK arrives after FRTO retransmission
2. tcp_try_to_open sees zero retrans_out, clears retrans_stamp
which should be kept like in CA_Loss state it would be
3. undo_marker change allowed tcp_packet_delayed to return
true because of the cleared retrans_stamp once FRTO is
terminated causing LossUndo to occur, which means all loss
markings FRTO made are reverted.
This means that the conventional recovery basically recovered
one loss per RTT, which is not that efficient. It was quite
unobvious that the undo_marker change broken something like
this, I had a quite long session to track it down because of
the non-intuitiviness of the bug (luckily I had a trivial
reproducer at hand and I was also able to learn to use kprobes
in the process as well :-)).
This together with the NewReno+FRTO fix and FRTO in-order
workaround this fixes Damon's problems, this and the first
mentioned are enough to fix Bugzilla #10063.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Reported-by: Damon L. Chesser <damon@damtek.com> Tested-by: Damon L. Chesser <damon@damtek.com> Tested-by: Sebastian Hyrwall <zibbe@cisko.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Enrico Scholz [Thu, 8 May 2008 10:35:13 +0000 (11:35 +0100)]
DM9000: Update and fix driver debugging messages
There was a missing newline in a dev_dbg() message. Values read
from/written into PHY registers might be for interest too, so I added
new dbg messages there.
Signed-off-by: Enrico Scholz <enrico.scholz@sigma-chemnitz.de> Signed-off-by: Ben Dooks <ben-linux@fluff.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Ben Hutchings [Wed, 7 May 2008 11:55:13 +0000 (12:55 +0100)]
[netdrvr] sfc: Add phy_flash_cfg module parameter and implementation
The 10Xpress PHY supports flash upgrades through MDIO, but needs to be
put in upgrade mode at power-up. This adds a module parameter and other
logic to support that.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Ben Hutchings [Wed, 7 May 2008 11:51:12 +0000 (12:51 +0100)]
[netdrvr] sfc: Add TSO support
The SFC4000 controller does not have hardware support for TSO, and the
core GSO code incurs a high cost in allocating and freeing skbs. This
TSO implementation uses lightweight packet header structures and is
substantially faster.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Hannes Hering [Wed, 7 May 2008 12:43:36 +0000 (14:43 +0200)]
ehea: Add DLPAR memory remove support
The eHEA driver uses the recently modified walk_memory_resource for powerpc
functionality to detect the memory layout. It further uses the memory hotplug
notifiers to catch memory hotplug events.
Signed-off-by: Hannes Hering <hering2@de.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Divy Le Ray [Wed, 7 May 2008 02:26:01 +0000 (19:26 -0700)]
cxgb3 - fix EEH
Reset the chip when the PCI link goes down.
Preserve the napi structure when a sge qset's resources are freed.
Replay only HW initialization when the chip comes out of reset.
Signed-off-by: Divy Le ray <divy@chelsio.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Brice Goglin [Fri, 9 May 2008 00:20:47 +0000 (02:20 +0200)]
myri10ge: fix the number of interrupt slots
Fix a long-standing bug/misunderstanding between the
driver and the firmware. The size of the interrupt
queue must be set to the number of rx slots (big + small),
and it should never have been a tunable.
Setting it too small results in chaos.
Signed-off-by: Brice Goglin <brice@myri.com> Signed-off-by: Andrew Gallatin <gallatin@myri.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Brice Goglin [Fri, 9 May 2008 00:18:45 +0000 (02:18 +0200)]
myri10ge: add barrier in myri10ge_send_cmd
Add a barrier() in the usleep() loop in myri10ge_send_cmd().
Without the barrier, some mips machine never notices that the
firmware has DMA'ed the response.
Signed-off-by: Brice Goglin <brice@myri.com> Signed-off-by: Andrew Gallatin <gallatin@myri.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Brice Goglin [Fri, 9 May 2008 00:18:24 +0000 (02:18 +0200)]
myri10ge: report FIBER in ethtool for XFP based NIC
Make ethtool report FIBER for XFP based NIC's port type.
Don't bother to poke around and try to find out what is in
the XFP cage, since Linux does not have separate media types
for -SR -LR, etc.
Signed-off-by: Brice Goglin <brice@myri.com> Signed-off-by: Andrew Gallatin <gallatin@myri.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Brice Goglin [Fri, 9 May 2008 00:17:16 +0000 (02:17 +0200)]
myri10ge: increase and fix handoff timeout
Increase the handoff timeout to 512ms so as to give the aeluros based
NICs sufficient time to handoff without relying on the msleep() being
sloppy, and accidentally sleeping way longer than the 20ms we specified
in 20 separate 1ms sleeps.
Fix typo in the handoff sleep delay, which made it additive, not
exponential.
Signed-off-by: Brice Goglin <brice@myri.com> Signed-off-by: Andrew Gallatin <gallatin@myri.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Johannes Berg [Tue, 13 May 2008 03:48:31 +0000 (20:48 -0700)]
net: Allow netdevices to specify needed head/tailroom
This patch adds needed_headroom/needed_tailroom members to struct
net_device and updates many places that allocate sbks to use them. Not
all of them can be converted though, and I'm sure I missed some (I
mostly grepped for LL_RESERVED_SPACE)
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Helmut Schaa [Thu, 8 May 2008 11:34:07 +0000 (13:34 +0200)]
mac80211: fix association with some APs
Some APs refuse association if the supported rates contained in the
association request do not match its own supported rates. This patch
introduces a new function which builds the intersection between the AP's
supported rates and the client's supported rates to work around such
problems. The same approach is already used in ipw2200 for example.
Signed-off-by: Helmut Schaa <hschaa@suse.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes a bug in RX path, the frequency was wrongly set in the
ieee80211_rx_status. This bug led to an empty scan list in A band.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com> Signed-off-by: Tomas Winkler <tomas.winkler@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes Berg [Wed, 7 May 2008 23:43:59 +0000 (01:43 +0200)]
zd1211rw: fix potential use-after-free bug
zd_mac_tx_to_dev() could potentially free the skb, or hand it off
to mac80211 which might free it. Hence, this code needs to get the
usb pointer out of skb->cb before handing it off to that function.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Pavel Emelyanov [Tue, 6 May 2008 14:46:36 +0000 (18:46 +0400)]
Fix GFP_KERNEL allocation under read lock.
The mesh_path_add() read-locks the pathtbl_resize_lock and calls
kmalloc with GFP_KERNEL mask.
Fix it and move the endadd2 label lower. It should be _before_ the
if() beyond, but it makes no sense for it being there, so I move it
right after this if().
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Without this patch, if xmit_skb is null but net_ratelimit() returns 0 we would
go to the else branch and access the null xmit_skb. Pointed out by Johannes
Berg.
Signed-off-by: Luis Carlos Cobo <luisca@cozybit.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Mon, 5 May 2008 18:40:35 +0000 (20:40 +0200)]
mac80211: Don't encrypt beacons
mac80211 should set the IEEE80211_TX_CTL_DO_NOT_ENCRYPT flag in tx_control
structure to inform drivers not to encrypt the beacon. Drivers that only check
for that flag before accessing the hw_key field, will otherwise cause a NULL
pointer dereference since that field is not configured for beacons.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Mon, 5 May 2008 15:24:03 +0000 (17:24 +0200)]
rt2x00: Clean up error handling of PCI queue DMA allocation.
When, for some reason, the rt2x00pci module fails to allocate DMA memory for
the queues, it tries to undo the complete initialization of the PCI device,
including freeing of the irq. This results in the following error in dmesg, as
the irq hadn't been requested yet:
[ 78.123456] Trying to free already-free IRQ 17
Fix this by implementing proper error handling code, instead of just using the
full uninitialization function.
Signed-off-by: Gertjan van Wingerde <gwingerde@kpnplanet.nl> Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Mon, 5 May 2008 15:23:47 +0000 (17:23 +0200)]
rt2x00: Fix broken recover-on-error path
During initialization the initialize() callback function
in rt2x00pci and rt2x00usb will cleanup the mess they made.
rt2x00lib shouldn't call uninitialize because the callback function already
cleaned up _and_ the DEVICE_INITIALIZED isn't set which causes the
rt2x00lib_uninitialize() to halt directly anyway. All that is required
to be cleaned up by rt2x00lib is the queue, and that can be done by
calling rt2x00queue_uninitialize() directly.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn [Mon, 5 May 2008 15:23:31 +0000 (17:23 +0200)]
rt2x00: Don't use pskb_expand_head()
rt2x00pci allocates DMA for descriptor and data,
rt61pci doesn't use this for the beacon, but it can
use the descriptor part as temporary buffer instead
of using pskb_expand_head().
Using this temporary buffer is obviously much better
then reallocating the skb buffer...
At the same time we can set the data length for the
beacon queue at 0, to make sure no DMA is allocated for
data (but just for the descriptor).
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes Berg [Fri, 2 May 2008 23:04:47 +0000 (01:04 +0200)]
mac80211: fix debugfs default key oops
Under certain circumstances (in AP mode) the debugfs function
that is supposed to add the default key symlink can encounter
a NULL default_key pointer. This patch makes it handle that
situtation gracefully.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Steven Rostedt [Thu, 1 May 2008 22:58:17 +0000 (15:58 -0700)]
fix irq flags in mac80211 code
A file in the net/mac80211 directory uses "int" for flags. This can cause
hard to find bugs on some architectures. This patch converts the flags to use
"long" instead.
This bug was discovered by doing an allyesconfig make on the -rt kernel where
checks are done to ensure all flags are of size sizeof(long).
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
prism54: fix regression with missing carrier in AP-mode
This fixes a regression introduced by commit 7b463ced6 (prism54: set
carrier flags correctly) which causes the device to come up without
a carrier in AP-mode.
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Julia Lawall [Mon, 12 May 2008 22:43:46 +0000 (15:43 -0700)]
pppol2tp: Remove null pointer dereference.
If session is NULL, it is not possible to access its name field. So I
have split apart the printing of the error message to drop the
printing of the name field in this case.
The macro PRINTK actually only evaluates its arguments starting with
the third one if the bitwise conjunction of the first two is non-zero.
Normally, this conjunction would only be non-zero if debugging mode
were turned on, but when session is NULL, the first argument in both
the old and new code is -1, and thus the bitwise conjunction is true.
Perhaps a different strategy is desired, such as using tunnel->debug,
which session->debug is initialized to, but tunnel can also be NULL,
so this does not completely solve the problem.
This problem was found using the following semantic match
(http://www.emn.fr/x-info/coccinelle/)
David S. Miller [Mon, 12 May 2008 10:29:11 +0000 (03:29 -0700)]
syncppp: Fix crashes.
The syncppp layer wants a mid-level netdev private pointer.
It was using netdev->priv but that only worked by accident,
and thus this scheme was broken when the device private
allocation strategy changed.
Add a proper mid-layer private pointer for uses like this,
update syncppp and all users, and remove the HDLC_PPP broken
tag from drivers/net/wan/Kconfig
Signed-off-by: David S. Miller <davem@davemloft.net>
Wei Yongjun [Fri, 9 May 2008 22:11:53 +0000 (15:11 -0700)]
sctp: Add address type check while process paramaters of ASCONF chunk
If socket is create by AF_INET type, add IPv6 address to asoc will cause
kernel panic while packet is transmitted on that transport.
This patch add address type check before process paramaters of ASCONF
chunk. If peer is not support this address type, return with error
invald parameter.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com> Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Wei Yongjun [Fri, 9 May 2008 22:11:17 +0000 (15:11 -0700)]
sctp: Do not enable peer IPv6 address support on PF_INET socket
If socket is create by PF_INET type, it can not used IPv6 address to
send/recv DATA, So we can not used IPv6 address even if peer tell us it
support IPv6 address.
This patch fix to only enabled peer IPv6 address support on PF_INET6 socket.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com> Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Matheos Worku [Mon, 12 May 2008 10:10:59 +0000 (03:10 -0700)]
niu: Determine the # of ports from the card's VPD data
Determine the number of physical ports from the card's VPD data.
Previous fix failed on Maramba platform which doesn't have the
"board-model" property. This fix uses the "model" property which
exists on all cards and Neptune based motherboards.
cstyle cleanup included.
Signed-off-by: Matheos Worku <matheos.worku@sun.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Oliver Hartkopp [Thu, 8 May 2008 09:49:55 +0000 (02:49 -0700)]
can: Fix can_send() handling on dev_queue_xmit() failures
The tx packet counting and the local loopback of CAN frames should
only happen in the case that the CAN frame has been enqueued to the
netdevice tx queue successfully.
Thanks to Andre Naujoks <nautsch@gmail.com> for reporting this issue.
Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net> Signed-off-by: Urs Thuermann <urs@isnogud.escape.de> Signed-off-by: David S. Miller <davem@davemloft.net>
Pavel Emelyanov [Thu, 8 May 2008 08:24:25 +0000 (01:24 -0700)]
netns: Fix arbitrary net_device-s corruptions on net_ns stop.
When a net namespace is destroyed, some devices (those, not killed
on ns stop explicitly) are moved back to init_net.
The problem, is that this net_ns change has one point of failure -
the __dev_alloc_name() may be called if a name collision occurs (and
this is easy to trigger). This allocator performs a likely-to-fail
GFP_ATOMIC allocation to find a suitable number. Other possible
conditions that may cause error (for device being ns local or not
registered) are always false in this case.
So, when this call fails, the device is unregistered. But this is
*not* the right thing to do, since after this the device may be
released (and kfree-ed) improperly. E. g. bridges require more
actions (sysfs update, timer disarming, etc.), some other devices
want to remove their private areas from lists, etc.
I. e. arbitrary use-after-free cases may occur.
The proposed fix is the following: since the only reason for the
dev_change_net_namespace to fail is the name generation, we may
give it a unique fall-back name w/o %d-s in it - the dev<ifindex>
one, since ifindexes are still unique.
So make this change, raise the failure-case printk loglevel to
EMERG and replace the unregister_netdevice call with BUG().
[ Use snprintf() -DaveM ]
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Thu, 8 May 2008 08:15:21 +0000 (01:15 -0700)]
netfilter: nf_conntrack_sip: restrict RTP expect flushing on error to last request
Some Inovaphone PBXs exhibit very stange behaviour: when dialing for
example "123", the device sends INVITE requests for "1", "12" and
"123" back to back. The first requests will elicit error responses
from the receiver, causing the SIP helper to flush the RTP
expectations even though we might still see a positive response.
Note the sequence number of the last INVITE request that contained a
media description and only flush the expectations when receiving a
negative response for that sequence number.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Thu, 8 May 2008 08:13:31 +0000 (01:13 -0700)]
macvlan: Fix memleak on device removal/crash on module removal
As noticed by Ben Greear, macvlan crashes the kernel when unloading the
module. The reason is that it tries to clean up the macvlan_port pointer
on the macvlan device itself instead of the underlying device. A non-NULL
pointer is taken as indication that the macvlan_handle_frame_hook is
valid, when receiving the next packet on the underlying device it tries
to call the NULL hook and crashes.
Clean up the macvlan_port on the correct device to fix this.
Signed-off-by; Patrick McHardy <kaber@trash.net> Tested-by: Ben Greear <greearb@candelatech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: J.H.M. Dassen (Ray) <jdassen@debian.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Thu, 8 May 2008 08:09:11 +0000 (01:09 -0700)]
tcp FRTO: SACK variant is errorneously used with NewReno
Note: there's actually another bug in FRTO's SACK variant, which
is the causing failure in NewReno case because of the error
that's fixed here. I'll fix the SACK case separately (it's
a separate bug really, though related, but in order to fix that
I need to audit tp->snd_nxt usage a bit).
There were two places where SACK variant of FRTO is getting
incorrectly used even if SACK wasn't negotiated by the TCP flow.
This leads to incorrect setting of frto_highmark with NewReno
if a previous recovery was interrupted by another RTO.
An eventual fallback to conventional recovery then incorrectly
considers one or couple of segments as forward transmissions
though they weren't, which then are not LOST marked during
fallback making them "non-retransmittable" until the next RTO.
In a bad case, those segments are really lost and are the only
one left in the window. Thus TCP needs another RTO to continue.
The next FRTO, however, could again repeat the same events
making the progress of the TCP flow extremely slow.
In order for these events to occur at all, FRTO must occur
again in FRTOs step 3 while the key segments must be lost as
well, which is not too likely in practice. It seems to most
frequently with some small devices such as network printers
that *seem* to accept TCP segments only in-order. In cases
were key segments weren't lost, things get automatically
resolved because those wrongly marked segments don't need to be
retransmitted in order to continue.
I found a reproducer after digging up relevant reports (few
reports in total, none at netdev or lkml I know of), some
cases seemed to indicate middlebox issues which seems now
to be a false assumption some people had made. Bugzilla
#10063 _might_ be related. Damon L. Chesser <damon@damtek.com>
had a reproducable case and was kind enough to tcpdump it
for me. With the tcpdump log it was quite trivial to figure
out.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
On a read error, e1000e might have returned uninitialized block of
eeprom data back to userspace. The convention is that 0xff is "empty",
so mark the entire eeprom as empty in case of an error.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Commit 9fb1e350e16164d56990dde036ae9c0a2fd3f634,
ucc_geth: use rx-clock-name and tx-clock-name device tree properties
Introduced a typo that made the driver use the RX clock
as TX clock, causing massive TX errors.
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Taken from http://bugzilla.kernel.org/show_bug.cgi?id=10577
I was unable to access a computer containing an Intel EtherExpress 16 network
card using IPv6.
I traced this to failure of neighbour discovery. When I used an "ip -6 neigh
add" command, on the computer attempting access, to insert a binding between
the IPv6 address of the computer with the Intel EtherExpress 16 network card
and the card's ethernet address, I was able to access that computer using
IPv6.
Neighbour discovery requires working multicast. The driver sources file
eexpress.c contains an approximately 30 line function eexp_setup_filter used
when loading multicast addresses.
I found 3 problems in this function
1) It wrote the number of multicast addresses to the card instead of the
number of bytes in the multicast addresses.
2) When loading multiple multicast addresses it loaded the first one
provided multiple times instead of loading each one once.
3) The setting of pointer 'data' from 'dmi->dmi_addr' occured before the
test for the error situation of 'dmi' being NULL.
Correcting these problems allows the computer with the Intel EtherExpress 16
network card to found by IPv6 neighbour discovery.
p.s. There is some information on the Intel EtherExpress 16 at
http://www.intel.com/support/etherexpress/vintage/sb/cs-013500.htm
Datasheet for the Intel 82586 ethernet controller used by the card
http://www.datasheetcatalog.com/datasheets_pdf/8/2/5/8/82586.shtml
Signed-off-by: Bruce Robson <bns_robson@hotmail.com> Cc: Jeff Garzik <jeff@garzik.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Bruce Allan [Tue, 29 Apr 2008 16:16:05 +0000 (09:16 -0700)]
e1000e: Add support for BM PHYs on ICH9
This patch adds support for the BM PHY, a new PHY model being used
on ICH9-based implementations.
This new PHY exposes issues in the ICH9 silicon when receiving
jumbo frames large enough to use more than a certain part of the
Rx FIFO, and this unfortunately breaks packet split jumbo receives.
For this reason we re-introduce (for affected adapters only) the
jumbo single-skb receive routine back so that people who do
wish to use jumbo frames on these ich9 platforms can do so.
Part of this problem has to do with CPU sleep states and to make
sure that all the wake up timings are correctly we force them
with the recently merged pm_qos infrastructure written by Mark
Gross. (See http://lkml.org/lkml/2007/10/4/400).
To make code read a bit easier we introduce a _IS_ICH flag so
that we don't need to do mac type checks over the code.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Anton Vorontsov [Tue, 29 Apr 2008 15:53:18 +0000 (19:53 +0400)]
uli526x: fix endianness issues in the setup frame
This patch fixes uli526x driver's issues on a PowerPC boards: uli chip
is unable to receive the packets.
It appears that send_frame_filter prepares the setup frame in the
endianness unsafe manner. On a big endian machines we should shift
the address nibble by two bytes.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Anton Vorontsov [Tue, 29 Apr 2008 15:53:13 +0000 (19:53 +0400)]
uli526x: initialize the hardware prior to requesting interrupts
The firmware on MPC8610HPCD boards enables ULI ethernet and leaves it
in some funky state before booting Linux. For drivers, it's always good
idea to (re)initialize the hardware prior to requesting interrupts.
Andy Fleming [Fri, 2 May 2008 18:01:23 +0000 (13:01 -0500)]
ucc_geth: Fix a bunch of sparse warnings
ucc_geth didn't have anything marked as __iomem. It was also inconsistent
with its use of in/out accessors (using them sometimes, not using them other
times). Cleaning this up cuts the warnings down from hundreds to just over a
dozen.
Signed-off-by: Andy Fleming <afleming@freescale.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Andy Fleming [Fri, 2 May 2008 18:00:30 +0000 (13:00 -0500)]
gianfar: Fix a locking bug in gianfar's sysfs code
During sparse cleanup, found a locking bug. Some of the sysfs functions were
acquiring a lock, and then returning in the event of an error. We rearrange
the code so that the lock is released in error conditions, too.
Signed-off-by: Andy Fleming <afleming@freescale.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
bonding / ipv6: no addrconf for slaves separately from master
two steps were rearranged in the enslavement process: netdev_set_master
is now before the call to dev_open to open the slave.
This patch updates the error cases and unwind process at the
end of bond_enslave to match the new order. Without this patch, it is
possible for the enslavement to fail, but leave the slave with IFF_SLAVE
set in its flags.
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Pavel Emelyanov [Sat, 3 May 2008 00:49:39 +0000 (17:49 -0700)]
bonding: Deadlock between bonding_store_bonds and bond_destroy_sysfs.
The sysfs layer has an internal protection, that ensures, that
all the process sitting inside ->sore/->show callback exits
before the appropriate entry is unregistered (the calltraces
are rather big, but I can provide them if required).
On the other hand, bonding takes rtnl_lock in
a) the bonding_store_bonds, i.e. in ->store callback,
b) module exit before calling the sysfs unregister routines.
Thus, the classical AB-BA deadlock may occur. To reproduce run
# while :; do modprobe bonding; rmmod bonding; done
and
# while :; do echo '+bond%d' > /sys/class/net/bonding_masters ; done
in parallel.
The fix is to move the bond_destroy_sysfs out of the rtnl_lock,
but _before_ bond_free_all to make sure no bonding devices exist
after module unload.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Acked-by: Jay Vosburgh <fubar@us.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Jay Vosburgh [Sat, 3 May 2008 00:49:38 +0000 (17:49 -0700)]
bonding: fix error unwind in bonding_store_bonds
Fixed an error unwind in bonding_store_bonds that didn't release
the locks it held, and consolidated unwinds into a common block at the
end of the function. Bug reported by Pavel Emelyanov <xemul@openvz.org>,
who provided a different fix.
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Pavel Emelyanov [Sat, 3 May 2008 00:49:37 +0000 (17:49 -0700)]
bonding: Do not call free_netdev for already registered device.
If the call to bond_create_sysfs_entry in bond_create fails, the
proper rollback is to call unregister_netdevice, not free_netdev.
Otherwise - kernel BUG at net/core/dev.c:4057!
Checked with artificial failures injected into bond_create_sysfs_entry.
Pavel's original patch modified by Jay Vosburgh to move code around
for clarity (remove goto-hopping within the unwind block).
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: Jay Vosburgh <fubar@us.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
projects / linux-2.6-omap-h63xx.git / log