| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch splits the red_parms structure into two components.
One holding the RED 'constant' parameters, and one containing the
variables.
This permits a size reduction of GRED qdisc, and is a preliminary step
to add an optional RED unit to SFQ.
SFQRED will have a single red_parms structure shared by all flows, and a
private red_vars per flow.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Dave Taht <dave.taht@gmail.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|\
| |
| |
| |
| |
| |
| | |
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next into for-davem
Conflicts:
drivers/net/wireless/b43legacy/dma.c
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The nfcid1 is the NFC-A identifier.
It is exported as an attribute of the target info
(returned as a response to NFC_CMD_GET_TARGET).
Signed-off-by: Ilan Elias <ilane@ti.com>
Acked-by: Samuel Ortiz <sameo@linux.intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add support for NCI Interface Error Notification.
When this notification is received and we're during a
data exchange transaction, indicate an error to the NFC
core layer via the data exchange callback.
Signed-off-by: Ilan Elias <ilane@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Addition, deletion, and modification of NCI constants.
Changes in NCI commands, responses, and notifications structures.
Signed-off-by: Ilan Elias <ilane@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Recently Dave noticed that a test we did in ipv6_add_addr to see if we next hop
route for the interface we're adding an addres to was wrong (see commit
7ffbcecbeed91e5874e9a1cfc4c0cbb07dac3069). for one, it never triggers, and two,
it was completely wrong to begin with. This test was meant to cover this
section of RFC 4429:
3.3 Modifications to RFC 2462 Stateless Address Autoconfiguration
* (modifies section 5.5) A host MAY choose to configure a new address
as an Optimistic Address. A host that does not know the SLLAO
of its router SHOULD NOT configure a new address as Optimistic.
A router SHOULD NOT configure an Optimistic Address.
This patch should bring us into proper compliance with the above clause. Since
we only add a SLAAC address after we've received a RA which may or may not
contain a source link layer address option, we can pass a pointer to that option
to addrconf_prefix_rcv (which may be null if the option is not present), and
only set the optimistic flag if the option was found in the RA.
Change notes:
(v2) modified the new parameter to addrconf_prefix_rcv to be a bool rather than
a pointer to make its use more clear as per request from davem.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: "David S. Miller" <davem@davemloft.net>
CC: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|\ \
| |/
| |
| |
| |
| |
| |
| | |
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next into for-davem
Conflicts:
drivers/net/wireless/b43/dma.c
drivers/net/wireless/brcm80211/brcmfmac/dhd_linux.c
|
| |
| |
| |
| |
| |
| |
| |
| | |
They don't need to disable interrupts anymore, we only run in process
context now.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| | |
It was never used, so removing it.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
To achive Man-In-The-Middle (MITM) level security with Low Energy,
we have to enable User Passkey Comparison. This commit modifies the
hard-coded JUST-WORKS pairing mechanism to support query via the MGMT
interface of Passkey comparison and User Confirmation.
Signed-off-by: Brian Gix <bgix@codeaurora.org>
Acked-by: Marcel Holtmann<marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When cancelling a delayed work (timer) in L2CAP we can not sleep holding
the sock mutex otherwise we might deadlock with an L2CAP timer handler.
This is possible because RX/TX and L2CAP timers run in different workqueues.
The scenario below illustrates the problem. Thus we are now avoiding to
sleep on the timers locks.
======================================================
[ INFO: possible circular locking dependency detected ]
3.1.0-05270-ga978dc7-dirty #239
-------------------------------------------------------
kworker/1:1/873 is trying to acquire lock:
(sk_lock-AF_BLUETOOTH-BTPROTO_L2CAP){+.+...}, at: [<ffffffffa002ceac>] l2cap_chan_timeout+0x3c/0xe0 [bluetooth]
but task is already holding lock:
((&(&chan->chan_timer)->work)){+.+...}, at: [<ffffffff81051a86>] process_one_work+0x126/0x450
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 ((&(&chan->chan_timer)->work)){+.+...}:
[<ffffffff8106b276>] check_prevs_add+0xf6/0x170
[<ffffffff8106b903>] validate_chain+0x613/0x790
[<ffffffff8106dfee>] __lock_acquire+0x4be/0xac0
[<ffffffff8106ec2d>] lock_acquire+0x8d/0xb0
[<ffffffff81052a6f>] wait_on_work+0x4f/0x160
[<ffffffff81052ca3>] __cancel_work_timer+0x73/0x80
[<ffffffff81052cbd>] cancel_delayed_work_sync+0xd/0x10
[<ffffffffa002f2ed>] l2cap_chan_connect+0x22d/0x470 [bluetooth]
[<ffffffffa002fb51>] l2cap_sock_connect+0xb1/0x140 [bluetooth]
[<ffffffff8130811b>] kernel_connect+0xb/0x10
[<ffffffffa00cf98a>] rfcomm_session_create+0x12a/0x1c0 [rfcomm]
[<ffffffffa00cfbe7>] __rfcomm_dlc_open+0x1c7/0x240 [rfcomm]
[<ffffffffa00d07c2>] rfcomm_dlc_open+0x42/0x70 [rfcomm]
[<ffffffffa00d3b03>] rfcomm_sock_connect+0x103/0x150 [rfcomm]
[<ffffffff8130bd7e>] sys_connect+0xae/0xc0
[<ffffffff813368d2>] compat_sys_socketcall+0xb2/0x220
[<ffffffff813b2089>] sysenter_dispatch+0x7/0x30
-> #0 (sk_lock-AF_BLUETOOTH-BTPROTO_L2CAP){+.+...}:
[<ffffffff8106b16d>] check_prev_add+0x6cd/0x6e0
[<ffffffff8106b276>] check_prevs_add+0xf6/0x170
[<ffffffff8106b903>] validate_chain+0x613/0x790
[<ffffffff8106dfee>] __lock_acquire+0x4be/0xac0
[<ffffffff8106ec2d>] lock_acquire+0x8d/0xb0
[<ffffffff8130d91a>] lock_sock_nested+0x8a/0xa0
[<ffffffffa002ceac>] l2cap_chan_timeout+0x3c/0xe0 [bluetooth]
[<ffffffff81051ae4>] process_one_work+0x184/0x450
[<ffffffff8105276e>] worker_thread+0x15e/0x340
[<ffffffff81057bb6>] kthread+0x96/0xa0
[<ffffffff813b1ef4>] kernel_thread_helper+0x4/0x10
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock((&(&chan->chan_timer)->work));
lock(sk_lock-AF_BLUETOOTH-BTPROTO_L2CAP);
lock((&(&chan->chan_timer)->work));
lock(sk_lock-AF_BLUETOOTH-BTPROTO_L2CAP);
*** DEADLOCK ***
2 locks held by kworker/1:1/873:
#0: (events){.+.+.+}, at: [<ffffffff81051a86>] process_one_work+0x126/0x450
#1: ((&(&chan->chan_timer)->work)){+.+...}, at: [<ffffffff81051a86>] process_one_work+0x126/0x450
stack backtrace:
Pid: 873, comm: kworker/1:1 Not tainted 3.1.0-05270-ga978dc7-dirty #239
Call Trace:
[<ffffffff813a0f6e>] print_circular_bug+0xd2/0xe3
[<ffffffff8106b16d>] check_prev_add+0x6cd/0x6e0
[<ffffffff8106b276>] check_prevs_add+0xf6/0x170
[<ffffffff8106b903>] validate_chain+0x613/0x790
[<ffffffff8106dfee>] __lock_acquire+0x4be/0xac0
[<ffffffff8130d8f6>] ? lock_sock_nested+0x66/0xa0
[<ffffffff8106ea30>] ? lock_release_nested+0x100/0x110
[<ffffffff8130d8f6>] ? lock_sock_nested+0x66/0xa0
[<ffffffff8106ec2d>] lock_acquire+0x8d/0xb0
[<ffffffffa002ceac>] ? l2cap_chan_timeout+0x3c/0xe0 [bluetooth]
[<ffffffff8130d91a>] lock_sock_nested+0x8a/0xa0
[<ffffffffa002ceac>] ? l2cap_chan_timeout+0x3c/0xe0 [bluetooth]
[<ffffffff81051a86>] ? process_one_work+0x126/0x450
[<ffffffffa002ceac>] l2cap_chan_timeout+0x3c/0xe0 [bluetooth]
[<ffffffff81051ae4>] process_one_work+0x184/0x450
[<ffffffff81051a86>] ? process_one_work+0x126/0x450
[<ffffffffa002ce70>] ? l2cap_security_cfm+0x4e0/0x4e0 [bluetooth]
[<ffffffff8105276e>] worker_thread+0x15e/0x340
[<ffffffff81052610>] ? manage_workers+0x110/0x110
[<ffffffff81057bb6>] kthread+0x96/0xa0
[<ffffffff813b1ef4>] kernel_thread_helper+0x4/0x10
[<ffffffff813af69d>] ? retint_restore_args+0xe/0xe
[<ffffffff81057b20>] ? __init_kthread_worker+0x70/0x70
[<ffffffff813b1ef0>] ? gs_change+0xb/0xb
Signed-off-by: Ulisses Furquim <ulisses@profusion.mobi>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The struct hci_proto and all related register/unregister and dispatching
code was removed. HCI core code now call directly the SCO and L2CAP
event functions.
Signed-off-by: Ulisses Furquim <ulisses@profusion.mobi>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Make code readable by removing magic numbers.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The handling of SCO audio links and the L2CAP protocol are essential to
any system with Bluetooth thus are always compiled in from now on.
Signed-off-by: Ulisses Furquim <ulisses@profusion.mobi>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| | |
It makes more sense this way, since info_timer is a timer using delayed
work API.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| | |
This one also needs to run in process context
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| | |
It is the only place where it is used.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| | |
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| | |
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When a peer changes SMPS state we should update
rate control so it doesn't have to detect it by
itself. It can't detect "dynamic" mode anyway
since that just requires rts-cts handshaking.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
If station info contains a beacon loss count, return
it to userspace.
Signed-off-by: Paul Stewart <pstew@chromium.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| |\
| | |
| | |
| | | |
git://git.kernel.org/pub/scm/linux/kernel/git/padovan/bluetooth-next
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch adds the necessary structs for the Confirm Name command. This
ensures that the protocol definitions are up to date with the latest
mgmt specification. The actual implementation of the command will follow
in a later patch-set.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch fixes the opcodes of the Block/Unblock device commands to
match with what user-space expects and to confirm with the latest mgmt
specification. The reason the values were wrong was a missing Confirm
Name command definition (which will be added by a subsequent patch).
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch adds a missing confirm_name field to mgmt_ev_device_found.
Support for setting the correct value for this field is not implemented
yet, but having it part of the struct definition ensures that user-space
gets correct sized device_found events and is thereby able to do at
least rudimentary parsing of them.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Implement block size read function. Use different variables for
packet-based and block-based flow control.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch updates the ordering and opcodes of mgmt messages to match
the latest API specification.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
We do not want the service cache to be enabled indefinitely after
mgmt_read_info is called. To solve this a timer is added which will
automatically disable the cache if mgmt_set_dev_class isn't called
within 5 seconds of calling mgmt_read_info.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Instead of having an explicit service cache command we can make the mgmt
API simpler by implicitly enabling the cache when mgmt_read_info is
called for the first time and disabling it when mgmt_set_dev_class is
called.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fast connectable is logically after the connectable property so that's
where it should show up in the code as well (it's also after connectable
in the settings bitfield).
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch updates the mgmt_read_info and related messages to the latest
management API which uses a bitfield of settings instead of individual
boolean values.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
As we run in process context now we don't need worqueue to add e del from
sysfs.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This move some checking code that was in l2cap_sock_connect() to
l2cap_chan_connect(). Thus we can invert the lock calls, i.e., call
lock_sock() before hci_dev_lock() to avoid a deadlock scenario.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Another step of remove interrupt context from Bluetooth Core.
Use the system workqueue.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This should simplify Bluetooth core processing a lot.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
As part of the moving on all the Bluetooth processing to Process context.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Handling hci_conn_hash with RCU make us avoid some locking and disable
tasklets.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
spin lock doesn't fit ok anymore on the new code based on workqueues.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
L2CAP timers also need to run in process context. As the works in l2cap
are small we are using the system worqueue.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
As HCI rx path is now done in process context it makes sense to do all the
timer in process context as well.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Bluetooth rx task runs now in a workqueue, so it a good approach run any
timer that share locking with process context code also in a workqueue.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Now we run everything in HCI in process context, so it's a better idea use
mutex instead spin_lock. The macro remains hci_dev_lock() (and I got rid
of hci_dev_lock_bh()), of course.
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Run recv process in workqueue helps a lot with our processing as the recv
path will also be in the process context, i.e., now all our tx and rx are
in process context.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | | |
It isn't used outside hci_conn.c
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | | |
le_num needs to be set to zero.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | | |
It isn't used anywhere.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Only the list member of the struct was used, so we now fold it into
hci_conn.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
We should not forget to try for real server with port 0
in the backup server when processing the sync message. We should
do it in all cases because the backup server can use different
forwarding method.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
During some debugging I needed to look into how /proc/net/ipv6_route
operated and in my digging I found its calling fib6_clean_all() which uses
"write_lock_bh(&table->tb6_lock)" before doing the walk of the table. I
found this on 2.6.32, but reading the code I believe the same basic idea
exists currently. Looking at the rtnetlink code they are only calling
"read_lock_bh(&table->tb6_lock);" via fib6_dump_table(). While I realize
reading from proc isn't the recommended way of fetching the ipv6 route
table; taking a write lock seems unnecessary and would probably cause
network performance issues.
To verify this I loaded up the ipv6 route table and then ran iperf in 3
cases:
* doing nothing
* reading ipv6 route table via proc
(while :; do cat /proc/net/ipv6_route > /dev/null; done)
* reading ipv6 route table via rtnetlink
(while :; do ip -6 route show table all > /dev/null; done)
* Load the ipv6 route table up with:
* for ((i = 0;i < 4000;i++)); do ip route add unreachable 2000::$i; done
* iperf commands:
* client: iperf -i 1 -V -c <ipv6 addr>
* server: iperf -V -s
* iperf results - 3 runs each (in Mbits/sec)
* nothing: client: 927,927,927 server: 927,927,927
* proc: client: 179,97,96,113 server: 142,112,133
* iproute: client: 928,927,928 server: 927,927,927
lock_stat shows taking the write lock is causing the slowdown. Using this
info I decided to write a version of fib6_clean_all() which replaces
write_lock_bh(&table->tb6_lock) with read_lock_bh(&table->tb6_lock). With
this new function I see the same results as with my rtnetlink iperf test.
Signed-off-by: Josh Hunt <joshhunt00@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|