We use _get() and _put() for device ref-counting in the kernel. However,
hci_conn_put() is _not_ used for ref-counting, hence, rename it to
hci_conn_drop() so we can later fix ref-counting and introduce
hci_conn_put().
hci_conn_hold() and hci_conn_put() are currently used to manage how long a
connection should be held alive. When the last user drops the connection,
we spawn a delayed work that performs the disconnect. Obviously, this has
nothing to do with ref-counting for the _object_ but rather for the
keep-alive of the connection.
But we really _need_ proper ref-counting for the _object_ to allow
connection-users like rfcomm-tty, HIDP or others.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
We need to verify that the given sockets actually are l2cap sockets. If
they aren't, we are not supposed to access bt_sk(sock) and we shouldn't
start the session if the offsets turn out to be valid local BT addresses.
That is, if someone passes a TCP socket to HIDCONNADD, then we access some
random offset in the TCP socket (which isn't even guaranteed to be valid).
Fix this by checking that the socket is an l2cap socket.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
The driver init queue is no longer needed. This can be all handled
inside the drivers now. So remove it.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Some drivers require a special stage for their early init. This is
always specific to the driver or transport. So call back into driver to
allow bringing up the device.
The advantage with this stage is that the Bluetooth core is actually
handling the HCI layer now. This means that command and event processing
is available.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
This patch adds a __hci_cmd_sync_ev function, analogous to
__hci_cmd_sync except that it also takes an event parameter to indicate
that the command completes with a special event instead of command
complete. Internally this new function takes advantage of the
hci_req_add_ev function introduced in the previous patch.
The primary expected user of this new function are the setup routines of
HCI drivers which may want to send custom commands and return only when
they have completed.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
This patch adds support for having commands within HCI requests that do
not result in a command complete but some other event. This is at least
needed for some vendor specific commands to be issued in the
hdev->setup() procecure, but might also be useful for other commands.
The way that the support is implemented is by extending the skb control
buffer to have a field to indicate that the command is expected to
terminate with a special event. After sending the command each received
event can then be compared against this field through hdev->sent_cmd.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
This patch adds a helper function for sending a single HCI command
waiting for its completion and then returning back the parameters in the
resulting command complete event (if there was one).
The implementation is very similar to that of hci_req_sync() except that
instead of invocing a callback for sending HCI commands the function
constructs and sends one itself and after being woken up picks the last
received event from hdev->recv_evt (if it matches the right criteria)
and returns it.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
This patch adds tracking of received HCI events to the hci_dev struct.
This is necessary so that a subsequent patch can implement a function
for sending a single command synchronously and returning the resulting
command complete parameters in the function return value.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
This patch removes the hci_req_cmd_status function since it is not
used anymore. The HCI request framework now considers the HCI command
has complete once the Command Status or Command Complete Event is
received.
Signed-off-by: Andre Guedes <andre.guedes@openbossa.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Sometimes queues are flushed in the middle of
operation, which can lead to driver issues.
Stop queues temporarily, while flushing, to
avoid transmitting new packets while they are
being flushed.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
There are a number of situations in which mac80211 only
really needs to flush queues for one virtual interface,
and in fact during this frames might be transmitted on
other virtual interfaces. Calculate and pass a queue
bitmap to the driver so it knows which queues to flush.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This basically reverts commit b207cdb07f.
Now is possible to use drv_{add,remove}_interface() and vif->debugfs_dir
to create/remove per interface debugfs files. Remove redundant
callbacks.
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
There is need create driver own per interface debugfs files. This is
currently done by drv_{add,remove}_interface_debugfs() callbacks. But it
is possible that after we remove interface from the driver (i.e.
on suspend) we call drv_remove_interface_debugfs() function. Fixing this
problem will require to add call drv_{add,remove}_interface_debugfs()
anytime we create and remove interface in mac80211. So it's better to
add debugfs dir dentry to vif structure to allow to create/remove
custom debugfs driver files on drv_{add,remove}_interface().
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
These parameters are related to the "fast connectable" mode that can be
changed through the mgmt interface. Not all controllers properly reset
these values with HCI_Reset so they need to be read in order to be able
to verify whether the values are correct or not before enabling page
scan.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
In order to be able to represent fast connectable mode in the mgmt
settings we need to have a HCI dev flag for it. This patch adds the flag
and makes sure its value is changed whenever a mgmt_set_fast_connectable
command completes.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
For proper control of the AD update and the related HCI commands it's
best to run the AD update through an async request instead of a
standalone HCI command. This patch changes the hci_update_ad() function
to take a request pointer and updates its users appropriately. E.g. the
function is no longer called after the init sequence but during stage 3
of the init sequence.
The TX power is read during the init sequence, so we don't need an
explicit update whenever it is read and the AD update based on the local
name should be done through the local name mgmt handler. The only other
user is the update based on enabling advertising. This part is still
kept as there is no mgmt API to enable it.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
We'll need to use this mask also when powering off the HCI device
so it's better to have this in a single and visible place.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Now that class related operations are tracked through asynchronous HCI
requests this flag is no longer needed.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Some devices have multicast filter capability for each individual
virtual interface rather than just a global one. Add an interface
specific driver callback allowing such drivers to configure this.
Signed-off-by: Alexander Bondar <alexander.bondar@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Since no one checks the returning value of hci_req_add and HCI
request errors are now handled in hci_req_run, we can make hci_
req_add returning void.
Signed-off-by: Andre Guedes <andre.guedes@openbossa.org>
Acked-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
When we are building a HCI request with more than one HCI command
and one of the hci_req_add calls fail, we should have some cleanup
routine so the HCI commands already queued on HCI request can be
deleted. Otherwise, we will face some memory leaks issues.
This patch implements the HCI request error handling which is the
following: If a hci_req_add fails, we save the error code in hci_
request. Once hci_req_run is called, we verify the error field. If
it is different from zero, we delete all HCI commands already queued
and return the error code.
Signed-off-by: Andre Guedes <andre.guedes@openbossa.org>
Acked-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
This variable is no longer needed (due to async HCI request support and
the conversion of hci_req_sync to use it), so it can be safely removed.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
This patch converts the hci_req_sync() procedure to internaly use the
asynchronous HCI requests.
The hci_req_sync mechanism relies on hci_req_complete() calls from
hci_event.c into hci_core.c whenever a HCI command completes. This is
very similar to what asynchronous requests do and makes the conversion
fairly straight forward by converting hci_req_complete into a request
complete callback. By this change hci_req_complete (renamed to
hci_req_sync_complete) becomes private to hci_core.c and all calls to it
can be removed from hci_event.c.
The commands in each hci_req_sync procedure are collected into their own
request by passing the hci_request pointer to the request callback
(instead of the hci_dev pointer). The one slight exception is the HCI
init request which has the special handling of HCI driver specific
initialization commands. These commands are run in their own request
prior to the "main" init request.
One other extra change that this patch must contain is the handling of
spontaneous HCI reset complete events that some controllers exhibit.
These were previously handled in the hci_req_complete function but the
right place for them now becomes the hci_req_cmd_complete function.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
This patch introduces functions to process the HCI request state when
receiving HCI Command Status or Command Complete events. Some HCI
commands, like Inquiry do not result in a Command complete event so
special handling is needed for them. Inquiry is a particularly important
one since it is the only forseeable "non-cmd_complete" command that will
make good use of the request functionality, and its completion is either
indicated by an Inquiry Complete event of a successful Command Complete
for HCI_Inquiry_Cancel.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
This function is analogous to hci_send_cmd() but instead of directly
queuing the command to hdev->cmd_q it adds it to the local queue of the
asynchronous HCI request being build (inside struct hci_request).
This is the main function used for building asynchronous requests and
there should be one or more calls to it between calls to hci_req_init
and hci_req_run.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
This patch adds the initial definitions and functions for asynchronous
HCI requests. Asynchronous requests are essentially a group of HCI
commands together with an optional completion callback. The request is
tracked through the already existing command queue by having the
necessary context information as part of the control buffer of each skb.
The only information needed in the skb control buffer is a flag for
indicating that the skb is the start of a request as well as the
optional complete callback that should be used when the request is
complete (this will be found in the last skb of the request).
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Previous commits have improved the handling of the RFCOMM session
timer and the RFCOMM session pointers such that freed RFCOMM
session structures should no longer be erroneously accessed. The
RFCOMM session refcnt now has no purpose and will be deleted by
this commit.
Note that the RFCOMM session is now deleted as soon as the
RFCOMM control channel link is no longer required. This makes the
lifetime of the RFCOMM session deterministic and absolute.
Previously with the refcnt, there was uncertainty about when
the session structure would be deleted because the relative
refcnt prevented the session structure from being deleted at will.
It was noted that the refcnt could malfunction under very heavy
real-time processor loading in embedded SMP environments. This
could cause premature RFCOMM session deletion or double session
deletion that could result in kernel crashes. Removal of the
refcnt prevents this issue.
There are 4 connection / disconnection RFCOMM session scenarios:
host initiated control link ---> host disconnected control link
host initiated ctrl link ---> remote device disconnected ctrl link
remote device initiated ctrl link ---> host disconnected ctrl link
remote device initiated ctrl link ---> remote device disc'ed ctrl link
The control channel connection procedures are independent of the
disconnection procedures. Strangely, the RFCOMM session refcnt was
applying special treatment so erroneously combining connection and
disconnection events. This commit fixes this issue by removing
some session code that used the "initiator" member of the session
structure that was intended for use with the data channels.
Signed-off-by: Dean Jenkins <Dean_Jenkins@mentor.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Unfortunately, the design retains local copies of the s RFCOMM
session pointer in various code blocks and this invites the erroneous
access to a freed RFCOMM session structure.
Therefore, return the RFCOMM session pointer back up the call stack
to avoid accessing a freed RFCOMM session structure. When the RFCOMM
session is deleted, NULL is passed up the call stack.
If active DLCs exist when the rfcomm session is terminating,
avoid a memory leak of rfcomm_dlc structures by ensuring that
rfcomm_session_close() is used instead of rfcomm_session_del().
Signed-off-by: Dean Jenkins <Dean_Jenkins@mentor.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
There is no reason a caller ever wants to check the return type of this
call. _Iff_ a user successfully called bt_sock_register(), they're allowed
to call bt_sock_unregister().
All other calls in the kernel (device_del, device_unregister, kfree(), ..)
that are logically equivalent return void. Lets not make callers think
they have to check the return type of this call and instead simply return
void.
We guarantee that after bt_sock_unregister() is called, the socket type
_is_ unregistered. If that is not what the caller wants, they're using the
wrong function, anyway.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
As hci_acl_disconn function basically sends the HCI Disconnect Command
and it is used to disconnect ACL, SCO and LE links, renaming it to
hci_disconnect is more suitable.
Signed-off-by: Andre Guedes <andre.guedes@openbossa.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
The irqsafe version ieee80211_sta_eosp_irqsafe() exists, but
drivers must not mix calls to any irqsafe/non-irqsafe function.
Both ath9k and iwlwifi, the likely first users of this interface,
use non-irqsafe RX/TX/TX status so must also use a non-irqsafe
version of this function. Since no driver uses the _irqsafe()
version, remove that.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The mesh station types used to refer to whether the
station was secure or nonsecure. Really the salient
information is whether it is managed by the kernel or
userspace
Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Secure mesh had the implicit requirement that the Mesh
Peering Management entity be in userspace. However
userspace might want to implement an open MPM as well, so
specify a mesh setup parameter to indicate this.
Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Implement restricting peer VHT capabilities to the device's own
capabilities. This is useful when a single driver supports more
than one device and the devices have different capabilities
(often they will differ in the number of spatial streams), but
in particular is also necessary for VHT capability overrides to
work correctly -- otherwise it'd be possible to e.g. advertise,
due to overrides, that TX-STBC is not supported, but then still
use it to TX to the AP because it supports RX-STBC.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
There's no driver using this flag, so it seems
that all drivers support HW crypto with WMM or
don't support it at all. Remove the flag and
code setting it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Add NL80211_CMD_UPDATE_FT_IES to support update of FT IEs to the WLAN
driver and NL80211_CMD_FT_EVENT to send FT events from the WLAN driver.
This will carry the target AP's MAC address along with the relevant
Information Elements. This event is used to report received FT IEs
(MDIE, FTIE, RSN IE, TIE, RICIE). These changes allow FT to be supported
with drivers that use an internal SME instead of user space option (like
FT implementation in wpa_supplicant with mac80211-based drivers).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Some devices can handle remain on channel requests differently
based on the request type/priority. Add support to
differentiate between different ROC types, i.e., indicate that
the ROC is required for sending managment frames.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
For testing it's sometimes useful to be able to
override certain VHT capability advertisement,
add the ability to do that in cfg80211.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The station change API isn't being checked properly before
drivers are called, and as a result it is difficult to see
what should be allowed and what not.
In order to comprehensively check the API parameters parse
everything first, and then have the driver call a function
(cfg80211_check_station_change()) with the additionally
information about the kind of station that is being changed;
this allows the function to make better decisions than the
old code could.
While at it, also add a few checks, particularly in mesh
and clarify the TDLS station lifetime in documentation.
To be able to reduce a few checks, ignore any flag set bits
when the mask isn't set, they shouldn't be applied then.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
All the pointers point right into the skb data and
not to anything that would be useful to change, so
make them const.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Make the ability to leave the plink_state unchanged not use a
magic -1 variable that isn't in the enum, but an explicit change
flag; reject invalid plink states or actions and move the needed
constants for plink actions to the right header file. Also
reject plink_state changes for non-mesh interfaces.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Pull networking fixes from David Miller:
"A moderately sized pile of fixes, some specifically for merge window
introduced regressions although others are for longer standing items
and have been queued up for -stable.
I'm kind of tired of all the RDS protocol bugs over the years, to be
honest, it's way out of proportion to the number of people who
actually use it.
1) Fix missing range initialization in netfilter IPSET, from Jozsef
Kadlecsik.
2) ieee80211_local->tim_lock needs to use BH disabling, from Johannes
Berg.
3) Fix DMA syncing in SFC driver, from Ben Hutchings.
4) Fix regression in BOND device MAC address setting, from Jiri
Pirko.
5) Missing usb_free_urb in ISDN Hisax driver, from Marina Makienko.
6) Fix UDP checksumming in bnx2x driver for 57710 and 57711 chips,
fix from Dmitry Kravkov.
7) Missing cfgspace_lock initialization in BCMA driver.
8) Validate parameter size for SCTP assoc stats getsockopt(), from
Guenter Roeck.
9) Fix SCTP association hangs, from Lee A Roberts.
10) Fix jumbo frame handling in r8169, from Francois Romieu.
11) Fix phy_device memory leak, from Petr Malat.
12) Omit trailing FCS from frames received in BGMAC driver, from Hauke
Mehrtens.
13) Missing socket refcount release in L2TP, from Guillaume Nault.
14) sctp_endpoint_init should respect passed in gfp_t, rather than use
GFP_KERNEL unconditionally. From Dan Carpenter.
15) Add AISX AX88179 USB driver, from Freddy Xin.
16) Remove MAINTAINERS entries for drivers deleted during the merge
window, from Cesar Eduardo Barros.
17) RDS protocol can try to allocate huge amounts of memory, check
that the user's request length makes sense, from Cong Wang.
18) SCTP should use the provided KMALLOC_MAX_SIZE instead of it's own,
bogus, definition. From Cong Wang.
19) Fix deadlocks in FEC driver by moving TX reclaim into NAPI poll,
from Frank Li. Also, fix a build error introduced in the merge
window.
20) Fix bogus purging of default routes in ipv6, from Lorenzo Colitti.
21) Don't double count RTT measurements when we leave the TCP receive
fast path, from Neal Cardwell."
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (61 commits)
tcp: fix double-counted receiver RTT when leaving receiver fast path
CAIF: fix sparse warning for caif_usb
rds: simplify a warning message
net: fec: fix build error in no MXC platform
net: ipv6: Don't purge default router if accept_ra=2
net: fec: put tx to napi poll function to fix dead lock
sctp: use KMALLOC_MAX_SIZE instead of its own MAX_KMALLOC_SIZE
rds: limit the size allocated by rds_message_alloc()
MAINTAINERS: remove eexpress
MAINTAINERS: remove drivers/net/wan/cycx*
MAINTAINERS: remove 3c505
caif_dev: fix sparse warnings for caif_flow_cb
ax88179_178a: ASIX AX88179_178A USB 3.0/2.0 to gigabit ethernet adapter driver
sctp: use the passed in gfp flags instead GFP_KERNEL
ipv[4|6]: correct dropwatch false positive in local_deliver_finish
l2tp: Restore socket refcount when sendmsg succeeds
net/phy: micrel: Disable asymmetric pause for KSZ9021
bgmac: omit the fcs
phy: Fix phy_device_free memory leak
bnx2x: Fix KR2 work-around condition
...
Pull more VFS bits from Al Viro:
"Unfortunately, it looks like xattr series will have to wait until the
next cycle ;-/
This pile contains 9p cleanups and fixes (races in v9fs_fid_add()
etc), fixup for nommu breakage in shmem.c, several cleanups and a bit
more file_inode() work"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
constify path_get/path_put and fs_struct.c stuff
fix nommu breakage in shmem.c
cache the value of file_inode() in struct file
9p: if v9fs_fid_lookup() gets to asking server, it'd better have hashed dentry
9p: make sure ->lookup() adds fid to the right dentry
9p: untangle ->lookup() a bit
9p: double iput() in ->lookup() if d_materialise_unique() fails
9p: v9fs_fid_add() can't fail now
v9fs: get rid of v9fs_dentry
9p: turn fid->dlist into hlist
9p: don't bother with private lock in ->d_fsdata; dentry->d_lock will do just fine
more file_inode() open-coded instances
selinux: opened file can't have NULL or negative ->f_path.dentry
(In the meantime, the hlist traversal macros have changed, so this
required a semantic conflict fixup for the newly hlistified fid->dlist)
TCP prequeue mechanism purpose is to let incoming packets
being processed by the thread currently blocked in tcp_recvmsg(),
instead of behalf of the softirq handler, to better adapt flow
control on receiver host capacity to schedule the consumer.
But in typical request/answer workloads, we send request, then
block to receive the answer. And before the actual answer, TCP
stack receives the ACK packets acknowledging the request.
Processing pure ACK on behalf of the thread blocked in tcp_recvmsg()
is a waste of resources, as thread has to immediately sleep again
because it got no payload.
This patch avoids the extra context switches and scheduler overhead.
Before patch :
a:~# echo 0 >/proc/sys/net/ipv4/tcp_low_latency
a:~# perf stat ./super_netperf 300 -t TCP_RR -l 10 -H 7.7.7.84 -- -r 8k,8k
231676
Performance counter stats for './super_netperf 300 -t TCP_RR -l 10 -H 7.7.7.84 -- -r 8k,8k':
116251.501765 task-clock # 11.369 CPUs utilized
5,025,463 context-switches # 0.043 M/sec
1,074,511 CPU-migrations # 0.009 M/sec
216,923 page-faults # 0.002 M/sec
311,636,972,396 cycles # 2.681 GHz
260,507,138,069 stalled-cycles-frontend # 83.59% frontend cycles idle
155,590,092,840 stalled-cycles-backend # 49.93% backend cycles idle
100,101,255,411 instructions # 0.32 insns per cycle
# 2.60 stalled cycles per insn
16,535,930,999 branches # 142.243 M/sec
646,483,591 branch-misses # 3.91% of all branches
10.225482774 seconds time elapsed
After patch :
a:~# echo 0 >/proc/sys/net/ipv4/tcp_low_latency
a:~# perf stat ./super_netperf 300 -t TCP_RR -l 10 -H 7.7.7.84 -- -r 8k,8k
233297
Performance counter stats for './super_netperf 300 -t TCP_RR -l 10 -H 7.7.7.84 -- -r 8k,8k':
91084.870855 task-clock # 8.887 CPUs utilized
2,485,916 context-switches # 0.027 M/sec
815,520 CPU-migrations # 0.009 M/sec
216,932 page-faults # 0.002 M/sec
245,195,022,629 cycles # 2.692 GHz
202,635,777,041 stalled-cycles-frontend # 82.64% frontend cycles idle
124,280,372,407 stalled-cycles-backend # 50.69% backend cycles idle
83,457,289,618 instructions # 0.34 insns per cycle
# 2.43 stalled cycles per insn
13,431,472,361 branches # 147.461 M/sec
504,470,665 branch-misses # 3.76% of all branches
10.249594448 seconds time elapsed
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull networking fixes from David Miller:
1) ping_err() ICMP error handler looks at wrong ICMP header, from Li
Wei.
2) TCP socket hash function on ipv6 is too weak, from Eric Dumazet.
3) netif_set_xps_queue() forgets to drop mutex on errors, fix from
Alexander Duyck.
4) sum_frag_mem_limit() can deadlock due to lack of BH disabling, fix
from Eric Dumazet.
5) TCP SYN data is miscalculated in tcp_send_syn_data(), because the
amount of TCP option space was not taken into account properly in
this code path. Fix from yuchung Cheng.
6) MLX4 driver allocates device queues with the wrong size, from Kleber
Sacilotto.
7) sock_diag can access past the end of the sock_diag_handlers[] array,
from Mathias Krause.
8) vlan_set_encap_proto() makes incorrect assumptions about where
skb->data points, rework the logic so that it works regardless of
where skb->data happens to be. From Jesse Gross.
9) Fix gianfar build failure with NET_POLL enabled, from Paul
Gortmaker.
10) Fix Ipv4 ID setting and checksum calculations in GRE driver, from
Pravin B Shelar.
11) bgmac driver does:
int i;
for (i = 0; ...; ...) {
...
for (i = 0; ...; ...) {
effectively corrupting the outer loop index, use a seperate
variable for the inner loops. From Rafał Miłecki.
12) Fix suspend bugs in smsc95xx driver, from Ming Lei.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (35 commits)
usbnet: smsc95xx: rename FEATURE_AUTOSUSPEND
usbnet: smsc95xx: fix broken runtime suspend
usbnet: smsc95xx: fix suspend failure
bgmac: fix indexing of 2nd level loops
b43: Fix lockdep splat on module unload
Revert "ip_gre: propogate target device GSO capability to the tunnel device"
IP_GRE: Fix GRE_CSUM case.
VXLAN: Use tunnel_ip_select_ident() for tunnel IP-Identification.
IP_GRE: Fix IP-Identification.
net/pasemi: Fix missing coding style
vmxnet3: fix ethtool ring buffer size setting
vmxnet3: make local function static
bnx2x: remove dead code and make local funcs static
gianfar: fix compile fail for NET_POLL=y due to struct packing
vlan: adjust vlan_set_encap_proto() for its callers
sock_diag: Simplify sock_diag_handlers[] handling in __sock_diag_rcv_msg
sock_diag: Fix out-of-bounds access to sock_diag_handlers[]
vxlan: remove depends on CONFIG_EXPERIMENTAL
mlx4_en: fix allocation of CPU affinity reverse-map
mlx4_en: fix allocation of device tx_cq
...
