Al Viro [Wed, 19 Dec 2007 23:56:37 +0000 (18:56 -0500)]
airo: sanitize BSSListRid handling
Stop byteswap-in-place in readBSSListRid(), annotate the sucker.
BTW, that had immediately found a bug - another codepath fetching
the same struct from card did _not_ byteswap, but used ->dBm the
same as everything else - host-endian. Fix in the next patch...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Al Viro [Sat, 29 Dec 2007 09:55:50 +0000 (04:55 -0500)]
ray_cs fixes
bugs galore:
* 0xf380 instead of htons(ETH_P_AARP), etc. Works only on l-e.
* back in 2.3.20 driver got readb() and friends instead of
direct dereferencing of iomem. Somebody got too enthusiatic and replaced
ntohs(p->mrx_overflow)
with
ntohs(read(&p->mrx_overflow)
without noticing that (a) the sucker is 16bit and (b) that expression can't possibly
be portable anyway (hell, on l-e it's always less than 256, on b-e it's always a
multiple of 256). Proper fix is
swab16(readw(&p->mrx_overflow)
taking into account the conversion done by readw() itself. That crap happened
in several places; the same fix applies.
* untranslate() assumes little-endian almost everywhere, except for
the code checking for IPX/AARP packets; there we forgot ntohs(), so that part
only works on big-endian.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Al Viro [Thu, 20 Dec 2007 03:38:33 +0000 (22:38 -0500)]
airo: bug in airo_interrupt() handling on incoming 802.11
On big-endian we end up with swapped first two bytes in packet,
due to earlier conversion to host-endian and forgotten conversion
back.
The code we calculated that host-endian for had been duplicated
several time - it finds the 802.11 MAC header length by the first
two bytes of packet; taken into a new helper (header_len(__le16 ctl)).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Al Viro [Wed, 19 Dec 2007 21:45:29 +0000 (16:45 -0500)]
airo: fix transmit_802_11_packet()
a) gaplen would better be stored little-endian
b) for control packets (shorter than 24-byte header) we ended up with
bap_write(ai, hdrlen == 30 ?
(const u16*)&gap.gaplen : (const u16*)&gap, 38 - hdrlen, BAP1);
passing to card the data past the end of gap (i.e. random stuff from stack)
and did _not_ feed the gaplen at the right offset.
c) sending the contents of uninitialized fields of struct is Not Nice(tm) either
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Al Viro [Sat, 29 Dec 2007 09:58:39 +0000 (04:58 -0500)]
ieee80211softmac_auth_resp() fix
The struct ieee8021_auth * passed to it comes straight from skb->data
without any conversions; members of the struct are little-endian, so
we'd better take that into account when doing switch by auth->algorithm,
etc.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Jan Engelhardt [Mon, 21 Jan 2008 11:18:08 +0000 (03:18 -0800)]
[IPV4]: Enable use of 240/4 address space.
This short patch modifies the IPv4 networking to enable use of the
240.0.0.0/4 (aka "class-E") address space as propsed in the internet
draft draft-fuller-240space-00.txt.
Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de> Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:47:43 +0000 (00:47 -0800)]
[NET_SCHED]: kill obsolete NET_CLS_POLICE option
The code is already gone for about half a year, the config option
has been kept around to select the replacement options for easier
upgrades. This seems long enough, people upgrading from older
kernels will have to reconfigure a lot anyway.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:47:08 +0000 (00:47 -0800)]
[MACVLAN]: Fix thinko in macvlan_transfer_operstate()
When the lower device's carrier is off, the macvlan devices's
carrier state should be checked to decide whether it needs to
be turned off. Currently the lower device's state is checked
a second time.
This still works, but unnecessarily tries to turn off the
carrier when its already off.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Pavel Emelyanov [Mon, 21 Jan 2008 08:28:03 +0000 (00:28 -0800)]
[VLAN]: Move protocol determination to seperate function
I think, that we can make this code flow easier to understand
by introducing the vlan_set_encap_proto() function (I hope the
name is good) to setup the skb proto and merge the paths calling
netif_rx() together.
[Patrick: Modified to apply on top of my previous patches]
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:25:50 +0000 (00:25 -0800)]
[VLAN]: Simplify vlan unregistration
Keep track of the number of VLAN devices in a vlan group. This allows
to have the caller sense when the group is going to be destroyed and
stop using it, which in turn allows to remove the wrapper around
unregister_vlan_dev for the NETDEV_UNREGISTER notifier and avoid
iterating over all possible VLAN ids whenever a device in unregistered.
Also fix what looks like a use-after-free (but is actually safe since
we're holding the RTNL), the real_dev reference should not be dropped
while we still use it.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:25:15 +0000 (00:25 -0800)]
[VLAN]: Clean up initialization code
- move module init/exit functions to end of file, remove some now unnecessary
forward declarations
- remove some obvious comments
- clean up proc init function and move a proc-related printk there
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:24:59 +0000 (00:24 -0800)]
[VLAN]: Remove non-implemented ioctls
The GET_VLAN_INGRESS_PRIORITY_CMD/GET_VLAN_EGRESS_PRIORITY_CMD ioctls are
not implemented and won't be, new functionality will be added to the netlink
interface. Remove the code and make the ioctl handler return -EOPNOTSUPP
for unknown commands instead of -EINVAL.
Also remove a comment about passing unknown commands to the underlying
device, that doesn't make any sense since its a VLAN specific ioctl and
if its not implemented here, its implemented nowhere.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:24:30 +0000 (00:24 -0800)]
[VLAN]: Clean up debugging and printks
- use pr_* functions and common prefix for non-device related messages
- remove VLAN_ printk levels
- kill lots of useless debugging statements
- remove a few unnecessary printks like for double VID registration (already
returns -EEXIST) and kill of a number of unnecessary checks in
vlan_proc_{add,rem}_dev() that are already performed by the caller
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:24:13 +0000 (00:24 -0800)]
[ETHER]: Bring back MAC_FMT
The print_mac function is not very suitable for debugging printks
in performance critical paths since without ifdefs it will always
get called. MAC_FMT can be used with pr_debug without any overhead
when debugging is disabled.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:18:26 +0000 (00:18 -0800)]
[VLAN]: Remove unnecessary structure declarations
- struct packet_type is not used
- struct vlan_group is declared later in the file before the first use
- struct net_device is not needed since netdevice.h is included
- struct vlan_collection does not exist
- struct vlan_dev_info is declared later in the file before the first use
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy [Mon, 21 Jan 2008 08:14:05 +0000 (00:14 -0800)]
[NET_SCHED]: sch_ingress: remove useless printk
The printk about ingress qdisc registration error can't be triggered
under normal circumstances. Since register_qdisc only fails for two
identical registrations, the only way to trigger it is by loading the
sch_ingress modules multiple times under different names, in which
case we already return -EEXIST to userspace.
Signed-off-by: Patrick McHardy <kaber@trash.net> Acked-by: Jamal Hadi Salim <hadi@cyberus.ca> Signed-off-by: David S. Miller <davem@davemloft.net>
SFQ doesn't need true random numbers, it is only using them to salt a
hash. Therefore it is better to use net_random() and avoid any
possible problems with depleting the entropy pool.
Signed-off-by: Stephen Hemminger <stephen.hemminger@vyatta.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
The perturbation timer used for re-keying can be deferred, it doesn't
need to be deterministic.
Signed-off-by: Stephen Hemminger <stephen.hemminger@vyatta.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vlad Yasevich [Sun, 20 Jan 2008 14:10:46 +0000 (06:10 -0800)]
[SCTP]: Correctly initialize error when parameter validation failed.
When parameter validation fails, there should be error causes that
specify what type of failure we've encountered. If the causes are not
there, we lacked memory to allocated them. Thus make that the default
value for the error.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Robie Basak [Sat, 19 Jan 2008 07:58:44 +0000 (23:58 -0800)]
[IrDA]: Frame length validation.
When using a stir4200-based USB adaptor to talk to a device that uses
an mcp2150, the stir4200 sometimes drops an incoming frame causing the
mcp2150 to try and retransmit the lost frame. In this combination, the
next frame received from the mcp2150 is often invalid - either an
empty i:rsp or an IrCOMM i:rsp with an invalid clen. These corner
cases are now checked.
Signed-off-by: Robie Basak <rb-oss-1@justgohome.co.uk> Signed-off-by: Samuel Ortiz <samuel@sortiz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Robie Basak [Sat, 19 Jan 2008 07:56:54 +0000 (23:56 -0800)]
[IrDA]: Resend frames on timeout.
When final timer expires, it might also mean that the i:cmd wasn't
received properly. If we have rejected frames, we can try to resend them.
Signed-off-by: Robie Basak <rb-oss-1@justgohome.co.uk> Signed-off-by: Samuel Ortiz <samuel@sortiz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Denis V. Lunev [Sat, 19 Jan 2008 07:55:19 +0000 (23:55 -0800)]
[NETNS]: Namespace stop vs 'ip r l' race.
During network namespace stop process kernel side netlink sockets
belonging to a namespace should be closed. They should not prevent
namespace to stop, so they do not increment namespace usage
counter. Though this counter will be put during last sock_put.
The raplacement of the correct netns for init_ns solves the problem
only partial as socket to be stoped until proper stop is a valid
netlink kernel socket and can be looked up by the user processes. This
is not a problem until it resides in initial namespace (no processes
inside this net), but this is not true for init_net.
So, hold the referrence for a socket, remove it from lookup tables and
only after that change namespace and perform a last put.
Signed-off-by: Denis V. Lunev <den@openvz.org> Tested-by: Alexey Dobriyan <adobriyan@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Create a specific helper for netlink kernel socket disposal. This just
let the code look better and provides a ground for proper disposal
inside a namespace.
Signed-off-by: Denis V. Lunev <den@openvz.org> Tested-by: Alexey Dobriyan <adobriyan@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Denis V. Lunev [Sat, 19 Jan 2008 07:54:15 +0000 (23:54 -0800)]
[NETNS]: Memory leak on network namespace stop.
Network namespace allocates 2 kernel netlink sockets, fibnl &
rtnl. These sockets should be disposed properly, i.e. by
sock_release. Plain sock_put is not enough.
Signed-off-by: Denis V. Lunev <den@openvz.org> Tested-by: Alexey Dobriyan <adobriyan@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Denis V. Lunev [Sat, 19 Jan 2008 07:53:31 +0000 (23:53 -0800)]
[NETNS]: Double free in netlink_release.
Netlink protocol table is global for all namespaces. Some netlink
protocols have been virtualized, i.e. they have per/namespace netlink
socket. This difference can easily lead to double free if more than 1
namespace is started. Count the number of kernel netlink sockets to
track that this table is not used any more.
Signed-off-by: Denis V. Lunev <den@openvz.org> Tested-by: Alexey Dobriyan <adobriyan@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Lezcano [Sat, 19 Jan 2008 07:52:35 +0000 (23:52 -0800)]
[IPV6]: Fix ip6_frag ctl
Alexey Dobriyan reported an oops when unsharing the network
indefinitely inside a loop. This is because the ip6_frag is not per
namespace while the ctls are.
That happens at the fragment timer expiration:
inet_frag_secret_rebuild function is called and this one restarts the
timer using the value stored inside the sysctl field.
"mod_timer(&f->secret_timer, now + f->ctl->secret_interval);"
When the network is unshared, ip6_frag.ctl is initialized with the new
sysctl instances, but ip6_frag has only one instance. A race in this
case will appear because f->ctl can be modified during the read access
in the timer callback.
Until the ip6_frag is not per namespace, I discard the assignation to
the ctl field of ip6_frags in ip6_frag_sysctl_init when the network
namespace is not the init net.
Signed-off-by: Daniel Lezcano <dlezcano@fr.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Stefan Roese [Wed, 16 Jan 2008 07:11:15 +0000 (08:11 +0100)]
net: NEWEMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's
Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
xmit. This routine doesn't check if the frame length exceeds the max.
MAL buffer size.
This patch now changes the driver to call emac_start_xmit_sg() on all
GigE platforms and not only the TAH equipped ones (440GX). This enables
an MTU of 9000 instead 4080.
Signed-off-by: Stefan Roese <sr@denx.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
projects / linux-2.6-omap-h63xx.git / log