Fixes the following:
1. POLL should not enable IRQ when work is not completed
2. No locking between TX descriptor cleaning and XMIT descriptor handling
3. No locking between RX POLL and XMIT modifying control register
4. Since TX cleaning (called from POLL) is running in parallel with XMIT
unnecessary locking is needed.
5. IRQ handler looks at RX frame status solely, this is wrong when IRQ is
temporarily disabled (in POLL), and when IRQ is shared.
6. IRQ handler clears IRQ status, which is unnecessary
7. TX queue was stopped in preventing cause when not MAX_SKB_FRAGS+1
descriptors were available after a SKB been scheduled by XMIT. Instead
the TX queue is stopped first when not enough descriptors are available
upon entering XMIT.
It was hard to split up this patch in smaller pieces since all are tied
together somehow.
Note the RX flag used in the interrupt handler does not signal that
interrupt was asserted, but that a frame was received. Same goes for TX.
Also, IRQ is not asserted when the RX flag is set before enabling IRQ
enable until a new frame is received. So extra care must be taken to
avoid enabling IRQ and all descriptors are already used, hence dead lock
will upon us. See new POLL implementation that enableds IRQ then look at
the RX flag to determine if one or more IRQs may have been missed. TX/RX
flags are cleared before handling previously enabled descriptors, this
ensures that the RX/TX flags are valid when determining if IRQ should be
turned on again.
By moving TX cleaning from POLL to XMIT in the standard case, removes some
locking trouble. Enabling TX cleaning from poll only when not enough TX
descriptors are available is safe because the TX queue is at the same time
stopped, thus XMIT will not be called. The TX queue is woken up again when
enough descriptrs are available.
TX Frames are always enabled with IRQ, however the TX IRQ Enable flag will
not be enabled until XMIT must wait for free descriptors.
Locking RX and XMIT parts of the driver from each other is needed because
the RX/TX enable bits share the same register.
Signed-off-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Frame error interrupts must also be handled since the RX flag only indicates
successful reception, it is unlikely but the old code may lead to dead lock
if 128 error frames are recieved in a row.
Signed-off-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A new SKB buffer should not be allocated when the old SKB is reused.
Signed-off-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is safe to enable all fragments before enabling the first descriptor,
this way all descriptors don't have to be processed twice, added extra
memory barrier.
Signed-off-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When NAPI is disabled there is no point in having IRQs enabled, TX/RX
should be off before clearing the TX/RX descriptor rings.
Signed-off-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unfortunately, not all CONFIG_OF platforms provide
pci_device_to_OF_node().
Change the test to CONFIG_SPARC for now to deal with
the build regressions.
Signed-off-by: David S. Miller <davem@davemloft.net>
Some minor comment errors and whitespace issues discovered while looking
into this are also addressed.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
If hardware asserted an interrupt and driver is down,
then there is nothing to do so return IRQ_HANDLED
instead of IRQ_NONE. Returning IRQ_NONE in above
situation causes screaming IRQ on virtual machines.
CC: Andy Gospodarek <gospo@redhat.com>
Signed-off-by: Tushar Dave <tushar.n.dave@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The firmware agent is not available during resume. Loading the firmware
during open() (see eee3a96c63) is not
enough.
close() is run during resume through rtl8169_reset_task(), whence the
mildly natural release of firmware in the driver removal method instead.
It will help with http://bugs.debian.org/609538. It will not avoid
the 60 seconds delay when:
- there is no firmware
- the driver is loaded and the device is not up before a suspend/resume
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Tested-by: Jarek Kamiński <jarek@vilo.eu.org>
Cc: Hayes <hayeswang@realtek.com>
Cc: Ben Hutchings <benh@debian.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use is_unicast_ether_addr from linux/etherdevice.h instead of custom
macros.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
From a check for !is_multicast_ether_addr it is not always obvious that
we're checking for a unicast address. So add this helper function to
make those code paths easier to read.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Acked-by: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is completely untested as I don't have an ARM build environment.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some ethtool operations can only be implemented for the WAN port, and
not all such operations are allowed to return an error code such as
-EOPNOTSUPP. Therefore, define two separate ethtool_ops structures
for WAN and non-WAN ports; simplify and rename the WAN-only functions.
This is completely untested as I don't have an ARM build environment.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_clone() dynamically allocates memory and may fail. If it does it
returns NULL. This means we'll dereference a NULL pointer in
drivers/net/usb/cdc_ncm.c::cdc_ncm_rx_fixup().
As far as I can tell, the proper way to deal with this is simply to goto
the error label.
Furthermore gcc complains that 'skb' may be used uninitialized:
drivers/net/usb/cdc_ncm.c: In function ‘cdc_ncm_rx_fixup’:
drivers/net/usb/cdc_ncm.c:922:18: warning: ‘skb’ may be used uninitialized in this function
and I believe it is right. On the line where we
pr_debug("invalid frame detected (ignored)" ...
we are using the local variable 'skb' but nothing has ever been assigned
to that variable yet. I believe the correct fix for that is to use
'skb_in' instead.
Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Regardless of whether the firmware update being performed by
vxge_fw_upgrade() is a success or not we must still remember to always
release_firmware() before returning.
Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Acked-by: Ram Vepa <ram.vepa@exar.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
icv_truncbits is set to 256 for sha512, so update
MAX_AH_AUTH_LEN to 64.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After recent changes, (percpu stats on vlan/tunnels...), we dont need
anymore per struct netdev_queue tx_bytes/tx_packets/tx_dropped counters.
Only remaining users are ixgbe, sch_teql, gianfar & macvlan :
1) ixgbe can be converted to use existing tx_ring counters.
2) macvlan incremented txq->tx_dropped, it can use the
dev->stats.tx_dropped counter.
3) sch_teql : almost revert ab35cd4b8f (Use net_device internal stats)
Now we have ndo_get_stats64(), use it, even for "unsigned long"
fields (No need to bring back a struct net_device_stats)
4) gianfar adds a stats structure per tx queue to hold
tx_bytes/tx_packets
This removes a lockdep warning (and possible lockup) in rndis gadget,
calling dev_get_stats() from hard IRQ context.
Ref: http://www.spinics.net/lists/netdev/msg149202.html
Reported-by: Neil Jones <neiljay@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jarek Poplawski <jarkao2@gmail.com>
CC: Alexander Duyck <alexander.h.duyck@intel.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Sandeep Gopalpet <sandeep.kumar@freescale.com>
CC: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a loop in ctnetlink_get_conntrack() that can be
triggered if you use the same socket to receive events and to
perform a GET operation. Under heavy load, netlink_unicast()
may return -EAGAIN, this error code is reserved in nfnetlink for
the module load-on-demand. Instead, we return -ENOBUFS which is
the appropriate error code that has to be propagated to
user-space.
Reported-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Fix new kernel-doc warning (copy-paste typo):
Warning(net/ethernet/eth.c:366): No description found for parameter 'rxqs'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RED statistics structure includes backlog field which is not
set or used by any code.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
kzalloc'd memory doesn't need a memset to 0.
Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Linux IPv6 forwards unicast packets, which are link layer multicasts...
The hole was present since day one. I was 100% this check is there, but it is not.
The problem shows itself, f.e. when Microsoft Network Load Balancer runs on a network.
This software resolves IPv6 unicast addresses to multicast MAC addresses.
Signed-off-by: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
The IPv6 tproxy patches split IPv6 defragmentation off of conntrack, but
failed to update the #ifdef stanzas guarding the defragmentation related
fields and code in skbuff and conntrack related code in nf_defrag_ipv6.c.
This patch adds the required #ifdefs so that IPv6 tproxy can truly be used
without connection tracking.
Original report:
http://marc.info/?l=linux-netdev&m=129010118516341&w=2
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: KOVACS Krisztian <hidden@balabit.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Commit fe10ae5338 adds a memset() to clear
the structure being sent back to userspace, but accidentally used the
wrong size.
Reported-by: Brad Spengler <spender@grsecurity.net>
Signed-off-by: Kees Cook <kees.cook@canonical.com>
Cc: stable@kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
If the Link Start fails in cxgb4vf_open(), we need to back out any state
that we've built up ...
Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
broute table init hook sets up the "br_should_route_hook" pointer,
which then gets called from br_input.
commit a386f99025
(bridge: add proper RCU annotation to should_route_hook)
introduced a typedef, and then changed this to:
br_should_route_hook_t *rhook;
[..]
rhook = rcu_dereference(br_should_route_hook);
if (*rhook(skb))
problem is that "br_should_route_hook" contains the address of the function,
so calling *rhook() results in kernel panic.
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The netlink interface to dump the connection tracking table has a race
when entries are deleted at the same time. A customer reported a crash
and the backtrace showed thatctnetlink_dump_table was running while a
conntrack entry was being destroyed.
(see https://bugzilla.vyatta.com/show_bug.cgi?id=6402).
According to RCU documentation, when using hlist_nulls the reader
must handle the case of seeing a deleted entry and not proceed
further down the linked list. The old code would continue
which caused the scan to walk into the free list.
This patch uses locking (rather than RCU) for this operation which
is guaranteed safe, and no longer requires getting reference while
doing dump operation.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
skb_cow_data() may allocate a new data buffer, so pointers on
skb should be set after this function.
Bug was introduced by commit dff3bb06 ("ah4: convert to ahash")
and 8631e9bd ("ah6: convert to ahash").
Signed-off-by: Wang Xuefu <xuefu.wang@6wind.com>
Acked-by: Krzysztof Witek <krzysztof.witek@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For SHA256, RFC4868 requires to truncate ICV length to 128 bits,
hence MAX_AH_AUTH_LEN should be updated to 16.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Maximum trunc length is defined by MAX_AH_AUTH_LEN (in bytes)
and need to be checked when this value is set (in bits) by
the user. In ah4.c and ah6.c a BUG_ON() checks this condiftion.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the skb array is not fully allocated, and the allocation
is done as it's requested, which is not the expected way.
This patch just allocate the full skb array at driver initialization.
Also, this patch increases ehea version to 107.
Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
pcnet_cs:
add another ID of "corega Ether CF-TD" 10Base-T PCMCIA card.
Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
inet_csk_bind_conflict() logic currently disallows a bind() if
it finds a friend socket (a socket bound on same address/port)
satisfying a set of conditions :
1) Current (to be bound) socket doesnt have sk_reuse set
OR
2) other socket doesnt have sk_reuse set
OR
3) other socket is in LISTEN state
We should add the CLOSE state in the 3) condition, in order to avoid two
REUSEADDR sockets in CLOSE state with same local address/port, since
this can deny further operations.
Note : a prior patch tried to address the problem in a different (and
buggy) way. (commit fda48a0d7a tcp: bind() fix when many ports
are bound).
Reported-by: Gaspar Chilingarov <gasparch@gmail.com>
Reported-by: Daniel Baluta <daniel.baluta@gmail.com>
Tested-by: Daniel Baluta <daniel.baluta@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update rtl_phy_write_fw function. The new function could
parse the complex firmware which is used by RTL8111E and later.
The new firmware may read data and do some operations, not just
do writing only.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
