Fairly complex code in iwlagn_tx_status_reply_tx handle the status reports for
aggregated packet batches sent by the NIC. This code aims to handle the case
where the NIC retransmits failed packets from a previous batch; the status
information for these packets can sometimes be inserted in the middle of a
batch and are actually not in order by sequence number! (I verified this can
happen with printk's in the function.)
The code in question adaptively identifies the "first" frame of the batch,
taking into account that it may not be the one corresponding to the first agg
status report, and also handles the case when the set of sent packets wraps the
256-character entry buffer. It generates the agg->bitmap field of sent packets
which is later compared to the BlockAck response from the receiver to see which
frames of those sent in this batch were ACKed. A small logic error (wrapping by
0xff==255 instead of 0x100==256) was causing the agg->bitmap to be set
incorrectly.
Fix this wrapping code, and add extensive comments to clarify what is going on.
Signed-off-by: Daniel Halperin <dhalperi@cs.washington.edu>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Compressed BlockAck frames store the ACKs/NACKs in a 64-bit bitmap that starts
at the sequence number of the first frame sent in the aggregated batch. Note
that this is a selective ACKnowledgement following selective retransmission;
e.g., if frames 1,4-5 in a batch are ACKed then the next transmission will
include frames 2-3,6-10 (7 frames). In this latter case, the Compressed
BlockAck will not have all meaningful information in the low order bits -- the
semantically meaningful bits of the BA will be 0x1f3 (where the low-order frame
is seq 2).
The driver code originally just looked at the lower (in this case, 7) bits of
the BlockAck. In this case, the lower 7 bits of 0x1f3 => only 5 packets,
maximum, could ever be ACKed. In reality it should be looking at all of the
bits, filtered by those corresponding to packets that were actually sent. This
flaw meant that the number of correctly ACked packets could be significantly
underreported and might result in asynchronous state between TX and RX sides as
well as driver and uCode.
Fix this and also add a shortcut that doesn't require the code to loop through
all 64 bits of the bitmap but rather stops when no higher packets are ACKed.
In my experiments this fix greatly reduces throughput swing, making throughput
stable and high. It is also likely related to some of the stalls observed in
aggregation mode and maybe some of the buffer underruns observed, e.g.,
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=1968http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2098http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2018
Signed-off-by: Daniel Halperin <dhalperi@cs.washington.edu>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
The internal scanning created a problem where
when userspace tries to scan, the scan gets
rejected. Instead of doing that, queue up the
user-initiated scan when doing an internal
scan.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
In "iwlwifi: make scan antenna forcing more generic"
I introduced generic scan RX antenna forcing, which
here I rename to make it more evident. Also add scan
TX antenna forcing, since I will need that as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Channel switch host command do not need to allocate huge command buffer
since its size is already included in the iwl_device_cmd structure.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Currently, the driver allocates up to 19 skb pointers
for each TFD, of which we have 256 per queue. This
means that for each TX queue, we allocate 19k/38k
(an order 4 or 5 allocation on 32/64 bit respectively)
just for each queue's "txb" array, which contains only
the SKB pointers.
However, due to the way we use these pointers only the
first one can ever be assigned. When the driver was
initially written, the idea was that it could be
passed multiple SKBs for each TFD and attach all
those to implement gather DMA. However, due to
constraints in the userspace API and lack of TCP/IP
level checksumming in the device, this is in fact not
possible. And even if it were, the SKBs would be
chained, and we wouldn't need to keep pointers to
each anyway.
Change this to only keep track of one SKB per TFD,
and thereby reduce memory consumption to just one
pointer per TFD, which is an order 0 allocation per
transmit queue.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
When we allocate queues, we currently don't
use kzalloc() right now. When we then free
those queues again without having used all
entries, we may end up trying to free random
pointers found in the txb array since it was
never initialised. This fixes it simply by
using kzalloc().
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
The iwl_hw_txq_free_tfd() function can be
called from contexts with IRQs disabled,
so it must not call dev_kfree_skb() but
rather dev_kfree_skb_any() instead.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
When we free a txq that had no txb array allocated,
we still try to access it. Fix that, and also free
all SKBs that may be in the txb array (although it
can just be a single one).
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
This variable is now no longer used, so it
can be removed completely.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
use TIME_UNIT define for beacon internal calculation
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Support channel switch in driver as a separated mac80211 callback
function instead of part of mac_config callback; by moving to this
approach, uCode can have more control of channel switch timing.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Move the ucode beacon formation related helper function from 3945 to
iwlcore, so both _3945 and _agn devices can utilize those functions.
When driver pass the beacon related timing information to uCode in both
spectrum measurement and channel switch commands, the beacon timing
parameter require in uCode beacon format; those helper functions will do
the conversation from uSec to the correct uCode format
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
restrict_refcnt is no longer used, remove it from iwl_priv
structure
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Chain noise calibration data are cleared after the calibration is done
in iwlagn_gain_computation() and iwl4965_gain_computation(). This cause
the debugfs entries for those data useless. To provide valid debugging
info, clear those data right before starting the calibration instead.
Signed-off-by: Shanyu Zhao <shanyu.zhao@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
This can now be much better achieved using
tracing and post-processing of the trace,
rather than doing the processing in place
in the driver, so remove a lot of code.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
For probe request frames sent during scan, we
should use the virtual interface's mac address
that the scan was initiated on to avoid issues
when the wrong address is used.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
When we do not have an interface, priv->mac_addr
is all zeroes, so the memcpy() is not useful as
the RXON buffer has been cleared previously.
Therefore, use the interface's address that we
are setting up the RXON for, if available.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
priv->mac_addr is the address of the operating
interface, not the permanent MAC address. They
are usually the same, but the user can override
the operating address, so we shouldn't set the
variable to the permanent one, it is assigned
when an interface is added.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
There's no microcode that actually uses this
variable, and it is reserved for functionality
that the driver doesn't support anyway, so we
shouldn't be setting it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
iwl_connection_init_rx_config() will already
have set up the entire RXON command, so these
assignments are duplicate.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Add "Wey-Yi Guy" to maintainers list for iwlwifi.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Port of internal scan to iwl3945 missed introduction
of iwl3945_get_single_channel_for_scan.
Fix the following bug by introducing the iwl3945_get_single_channel_for_scan
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2208
Signed-off-by: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Enable GRO by default for performance.
Signed-off-by: Shirley Ma <xma@us.ibm.com>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With mtu=9000, mld_newpack() use order-2 GFP_ATOMIC allocations, that
are very unreliable, on machines where PAGE_SIZE=4K
Limit allocated skbs to be at most one page. (order-0 allocations)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch compile tested only.
Convert legacy PM hooks to dev_pm_ops
Signed-off-by: Denis Kirjanov <dkirjanov@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
v2: changed bonding module version, modified to apply on top of changes
from previous patch in series, and updated documentation to elaborate on
multiqueue awareness that now exists in bonding driver.
This patch give the user the ability to control the output slave for
round-robin and active-backup bonding. Similar functionality was
discussed in the past, but Jay Vosburgh indicated he would rather see a
feature like this added to existing modes rather than creating a
completely new mode. Jay's thoughts as well as Neil's input surrounding
some of the issues with the first implementation pushed us toward a
design that relied on the queue_mapping rather than skb marks.
Round-robin and active-backup modes were chosen as the first users of
this slave selection as they seemed like the most logical choices when
considering a multi-switch environment.
Round-robin mode works without any modification, but active-backup does
require inclusion of the first patch in this series and setting
the 'all_slaves_active' flag. This will allow reception of unicast traffic on
any of the backup interfaces.
This was tested with IPv4-based filters as well as VLAN-based filters
with good results.
More information as well as a configuration example is available in the
patch to Documentation/networking/bonding.txt.
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
v2: changed parameter name from 'keep_all' to 'all_slaves_active' and
skipped setting slaves to inactive rather than creating a new flag at
Jay's suggestion.
In an effort to suppress duplicate frames on certain bonding modes
(specifically the modes that do not require additional configuration on
the switch or switches connected to the host), code was added in the
generic receive patch in 2.6.16. The current behavior works quite well
for most users, but there are some times it would be nice to restore old
functionality and allow all frames to make their way up the stack.
This patch adds support for a new module option and sysfs file called
'all_slaves_active' that will restore pre-2.6.16 functionality if the
user desires. The default value is '0' and retains existing behavior,
but the user can set it to '1' and allow all frames up if desired.
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can avoid an unecessary cache miss by checking if the skb is non-linear
before accessing gso_size/gso_type in skb_warn_if_lro, the same can also be
done to avoid a cache miss on nr_frags if data_len is 0.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hi,
__pppoe_xmit() in drivers/net/pppoe always returns 1.
When the methods fails (via goto abort), it should return 0 and not 1.
Regards,
Rami Rosen
Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- ipv6 msstab: account for ipv6 header size
- ipv4 msstab: add mss for Jumbograms.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
caller: if (!th->rst && !th->syn && th->ack)
callee: if (!th->ack)
make the caller only check for !syn (common for 3whs), and move
the !rst / ack test to the callee.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
both syn_flood_warning functions print a message, but
ipv4 version only prints a warning if CONFIG_SYN_COOKIES=y.
Make the v4 one behave like the v6 one.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove duplicated #include('s) in drivers/net/wan/x25_asy.c
Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Its better to make a route lookup in appropriate namespace.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I believe a moderate SYN flood attack can corrupt RFS flow table
(rps_sock_flow_table), making RPS/RFS much less effective.
Even in a normal situation, server handling short lived sessions suffer
from bad steering for the first data packet of a session, if another SYN
packet is received for another session.
We do following action in tcp_v4_rcv() :
sock_rps_save_rxhash(sk, skb->rxhash);
We should _not_ do this if sk is a LISTEN socket, as about each
packet received on a LISTEN socket has a different rxhash than
previous one.
-> RPS_NO_CPU markers are spread all over rps_sock_flow_table.
Also, it makes sense to protect sk->rxhash field changes with socket
lock (We currently can change it even if user thread owns the lock
and might use rxhash)
This patch moves sock_rps_save_rxhash() to a sock locked section,
and only for non LISTEN sockets.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix bug in multilink fragment size calculation introduced by
commit 9c705260fe
"ppp: ppp_mp_explode() redesign"
Signed-off-by: Ben McKeegan <ben@netservers.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
syncookies default to on since
e994b7c901
(tcp: Don't make syn cookies initial setting depend on CONFIG_SYSCTL).
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only check pfc bits in hang logic if PFC is enabled. Previously,
if DCB was enabled but PFC was disabled the incorrect pause
bits would be checked.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
xfrm triggers a warning if dst_pop() drops a refcount
on a noref dst. This patch changes dst_pop() to
skb_dst_pop(). skb_dst_pop() drops the refcnt only
on a refcounted dst. Also we don't clone the child
dst_entry, so it is not refcounted and we can use
skb_dst_set_noref() in xfrm_output_one().
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now they are unnecessary. We can use the generic DMA API with any bus.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Michael Buesch <mb@bu3sch.de>
Cc: Gary Zambrano <zambrano@broadcom.com>
Cc: Stefano Brivio <stefano.brivio@polimi.it>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: John W. Linville <linville@tuxdriver.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Note that dma_sync_single_for_device and dma_sync_single_for_cpu support a
partial sync.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Gary Zambrano <zambrano@broadcom.com>
Acked-by: Michael Buesch <mb@bu3sch.de>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Stefano Brivio <stefano.brivio@polimi.it>
Cc: John W. Linville <linville@tuxdriver.com>
Acked-by: Michael Buesch <mb@bu3sch.de>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stefano Brivio <stefano.brivio@polimi.it>
Cc: John W. Linville <linville@tuxdriver.com>
Acked-by: Michael Buesch <mb@bu3sch.de>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add dma_dev, a pointer to struct device, to struct ssb_device. We pass it
to the generic DMA API with SSB_BUSTYPE_PCI and SSB_BUSTYPE_SSB.
ssb_devices_register() sets up it properly.
This is preparation for replacing the ssb bus specific DMA API (ssb_dma_*)
with the generic DMA API.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Michael Buesch <mb@bu3sch.de>
Cc: Gary Zambrano <zambrano@broadcom.com>
Cc: Stefano Brivio <stefano.brivio@polimi.it>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: John W. Linville <linville@tuxdriver.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
