Currently when there's some failure to receive a callback (because we
couldn't find a matching xid, for example), we exit svc_recv with
sk_tcplen still set but without any pages saved with the socket. This
will cause a crash later in svc_tcp_restore_pages.
Instead, make sure we reset that tcp information whether the callback
received failed or succeeded.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
* 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
NFS: Change initial mount authflavor only when server returns NFS4ERR_WRONGSEC
NFS: Fix a signed vs. unsigned secinfo bug
Revert "net/sunrpc: Use static const char arrays"
This is a regression acctually, caused by the first patch series for
creating a formal strcut l2cap_chan.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
This is part of "moving things to l2cap_chan". As one the first move it
triggered a big number of changes in the funcions parameters, basically
changing the struct sock param to struct l2cap_chan.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
As we use struct list_head to keep L2CAP channels list the workaround with
del_list is not needed anymore.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Use a well known Kernel API is always a good idea than implement your own
list.
In the future we might use RCU on this list.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
struct l2cap_chan cames to create a clear separation between what
properties and data belongs to the L2CAP channel and what belongs to the
socket. By now we just fold the struct sock * in struct l2cap_chan as all
the channel info is struct l2cap_pinfo today.
In the next commits we will see a move of channel stuff to struct
l2cap_chan.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Commit 1018b5c016 ("Set rt->rt_iif more
sanely on output routes.") breaks rt_is_{output,input}_route.
This became the cause to return "IP_PKTINFO's ->ipi_ifindex == 0".
To fix it, this does:
1) Add "int rt_route_iif;" to struct rtable
2) For input routes, always set rt_route_iif to same value as rt_iif
3) For output routes, always set rt_route_iif to zero. Set rt_iif
as it is done currently.
4) Change rt_is_{output,input}_route() to test rt_route_iif
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have no other option but to inform userspace that we
have queued up their regulatory hint request when we are
given one given that nl80211 operates atomically on user
requests. The best we can do is accept the request, and
add a delayed work item for processing failure and cancel it
if we succeeed. Upon failure we restore the regulatory
settings and ignore the user input.
This fixes this reported bug:
https://bugzilla.kernel.org/show_bug.cgi?id=28112
Reported-by: gregoryx.alagnou@intel.com
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When we restore regulatory settings its possible CRDA
will not reply because of a bogus user entry. In this
case the bogus entry will prevent any further processing
on cfg80211 for regulatory domains even if we restore
regulatory settings.
To prevent this we suck out all pending requests when
restoring regulatory settings and add them back into the
queue after we have queued up the reset work.
The impact of not having this applied is that a user
with privileges can issue a userspace regulatory hint
while we are disasocciating and this would prevent any
further processing of regulatory domains.
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This allows user-space monitoring of BSS parameters for the associated
station. This is useful for debugging and verifying that the paramaters
are as expected.
[Exactly the same as before but bundled into a single message]
Signed-off-by: Paul Stewart <pstew@chromium.org>
Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Allow the NFSv4 server to make use of TCP autotuning behaviour, which
was previously disabled by setting the sk_userlocks variable.
Set the receive buffers to be big enough to receive the whole RPC
request, and set this for the listening socket, not the accept socket.
Remove the code that readjusts the receive/send buffer sizes for the
accepted socket. Previously this code was used to influence the TCP
window management behaviour, which is no longer needed when autotuning
is enabled.
This can improve IO bandwidth on networks with high bandwidth-delay
products, where a large tcp window is required. It also simplifies
performance tuning, since getting adequate tcp buffers previously
required increasing the number of nfsd threads.
Signed-off-by: Olga Kornievskaia <aglo@citi.umich.edu>
Cc: Jim Rees <rees@umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Ensure that we immediately read and buffer data from the incoming TCP
stream so that we grow the receive window quickly, and don't deadlock on
large READ or WRITE requests.
Also do some minor exit cleanup.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
It's much simpler just to copy the cb reply data than to play tricks
with pages. Callback replies will typically be very small (at least
until we implement cb_getattr, in which case files with very long ACLs
could pose a problem), so there's no loss in efficiency.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
If the client sents a record too short to contain even the beginning of
the rpc header, then just close the connection.
The current code drops the record data and continues. I don't see the
point. It's a hopeless situation and simpler just to cut off the
connection completely.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Don't requeue the socket in some cases where we know it's unnecessary.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The mv88e6085 is identical to the mv88e6095, except that all ports are
10/100 Mb/s, so use the existing setup code except for the cpu/dsa speed
selection in _setup_port().
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Acked-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
properly record sk_rxhash in ipv6 sockets (v2)
Noticed while working on another project that flows to sockets which I had open
on a test systems weren't getting steered properly when I had RFS enabled.
Looking more closely I found that:
1) The affected sockets were all ipv6
2) They weren't getting steered because sk->sk_rxhash was never set from the
incomming skbs on that socket.
This was occuring because there are several points in the IPv4 tcp and udp code
which save the rxhash value when a new connection is established. Those calls
to sock_rps_save_rxhash were never added to the corresponding ipv6 code paths.
This patch adds those calls. Tested by myself to properly enable RFS
functionalty on ipv6.
Change notes:
v2:
Filtered UDP to only arm RFS on bound sockets (Eric Dumazet)
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes spin_locks at CAN socket creation time by using RCU.
Inspired by the discussion with Kurt van Dijck and Eric Dumazet the RCU code
was partly derived from af_phonet.c
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 411b5e0561.
Olga Kornievskaia reports:
Problem: linux client mounting linux server using rc4-hmac-md5
enctype. gssd fails with create a context after receiving a reply from
the server.
Diagnose: putting printout statements in the server kernel and
kerberos libraries revealed that client and server derived different
integrity keys.
Server kernel code was at fault due the the commit
[aglo@skydive linux-pnfs]$ git show 411b5e0561
Trond: The problem is that since it relies on virt_to_page(), you cannot
call sg_set_buf() for data in the const section.
Reported-by: Olga Kornievskaia <aglo@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org [2.6.36+]
Fix the request transition from linger -> normal request. The key is to
preserve r_osd and requeue on the same OSD. Reregister as a normal request,
add the request to the proper queues, then unregister the linger. Fix the
unregister helper to avoid clearing r_osd (and also simplify the parallel
check in __unregister_request()).
Reported-by: Henry Chang <henry.cy.chang@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (27 commits)
ipv6: Don't pass invalid dst_entry pointer to dst_release().
mlx4: fix kfree on error path in new_steering_entry()
tcp: len check is unnecessarily devastating, change to WARN_ON
sctp: malloc enough room for asconf-ack chunk
sctp: fix auth_hmacs field's length of struct sctp_cookie
net: Fix dev dev_ethtool_get_rx_csum() for forced NETIF_F_RXCSUM
usbnet: use eth%d name for known ethernet devices
starfire: clean up dma_addr_t size test
iwlegacy: fix bugs in change_interface
carl9170: Fix tx aggregation problems with some clients
iwl3945: disable hw scan by default
wireless: rt2x00: rt2800usb.c add and identify ids
iwl3945: do not deprecate software scan
mac80211: fix aggregation frame release during timeout
cfg80211: fix BSS double-unlinking (continued)
cfg80211:: fix possible NULL pointer dereference
mac80211: fix possible NULL pointer dereference
mac80211: fix NULL pointer dereference in ieee80211_key_alloc()
ath9k: fix a chip wakeup related crash in ath9k_start
mac80211: fix a crash in minstrel_ht in HT mode with no supported MCS rates
...
The allocation in mgmt_control() code are in user context and not locked
by any spinlock, so it's not recommended the use of GFP_ATOMIC there.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
kernel_thread() is a low-level implementation detail and
EXPORT_SYMBOL(kernel_thread) is scheduled for removal.
Use the <linux/kthread.h> API instead.
Signed-off-by: Szymon Janc <szymon.janc@tieto.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
kernel_thread() is a low-level implementation detail and
EXPORT_SYMBOL(kernel_thread) is scheduled for removal.
Use the <linux/kthread.h> API instead.
Signed-off-by: Szymon Janc <szymon.janc@tieto.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
kernel_thread() is a low-level implementation detail and
EXPORT_SYMBOL(kernel_thread) is scheduled for removal.
Use the <linux/kthread.h> API instead.
Signed-off-by: Szymon Janc <szymon.janc@tieto.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
The ethtool ETHTOOL_PHYS_ID command runs for an arbitrarily long
period of time, holding the RTNL lock. This blocks routing updates,
device enumeration, and various important operations that one might
want to keep running while hunting for the flashing LED.
We need to drop the RTNL lock during this operation, but currently the
core implementation is a thin wrapper around a driver operation and
drivers may well depend upon holding the lock.
Define a new driver operation 'set_phys_id' with an argument that sets
the ID indicator on/off/inactive/active (the last optional, for any
driver or firmware that prefers to handle blinking asynchronously).
When this is defined, the ethtool core drops the lock while waiting
and only acquires it around calls to this operation.
Deprecate the 'phys_id' operation in favour of this. It can be
removed once all in-tree drivers are converted.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
This patch uses __copy_from_user_nocache on transmit to bypass data
cache for a performance improvement. skb_add_data_nocache and
skb_copy_to_page_nocache can be called by sendmsg functions to use
this feature, initial support is in tcp_sendmsg. This functionality is
configurable per device using ethtool.
Presumably, this feature would only be useful when the driver does
not touch the data. The feature is turned on by default if a device
indicates that it does some form of checksum offload; it is off by
default for devices that do no checksum offload or indicate no checksum
is necessary. For the former case copy-checksum is probably done
anyway, in the latter case the device is likely loopback in which case
the no cache copy is probably not beneficial.
This patch was tested using 200 instances of netperf TCP_RR with
1400 byte request and one byte reply. Platform is 16 core AMD x86.
No-cache copy disabled:
672703 tps, 97.13% utilization
50/90/99% latency:244.31 484.205 1028.41
No-cache copy enabled:
702113 tps, 96.16% utilization,
50/90/99% latency 238.56 467.56 956.955
Using 14000 byte request and response sizes demonstrate the
effects more dramatically:
No-cache copy disabled:
79571 tps, 34.34 %utlization
50/90/95% latency 1584.46 2319.59 5001.76
No-cache copy enabled:
83856 tps, 34.81% utilization
50/90/95% latency 2508.42 2622.62 2735.88
Note especially the effect on latency tail (95th percentile).
This seems to provide a nice performance improvement and is
consistent in the tests I ran. Presumably, this would provide
the greatest benfits in the presence of an application workload
stressing the cache and a lot of transmit data happening.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
