Using blue flame can improve latency by allowing the HW to more efficiently
access the WQE. This patch presents two functions that are used to allocate or
release HW resources for using blue flame; the caller need to supply a struct
mlx4_bf object when allocating resources. Consumers that make use of this API
should post doorbells to the UAR object pointed by the initialized struct
mlx4_bf;
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mlx4_en module now uses the new steering mechanism.
The RX packets are now steered through the MCG table instead
of Mac table for unicast, and default entry for multicast.
The feature is enabled through INIT_HCA
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
For Ethernet mode only,
When we want to register QP as promiscuous, it must be added to all the
existing steering entries and also to the default one.
The promiscuous QP might also be on of "real" QPs,
which means we need to monitor every entry to avoid duplicates and ensure
we close an entry when all it has is promiscuous QPs.
Same mechanism both for unicast and multicast.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
The same packet steering mechanism would be used both for IB and Ethernet,
Both multicasts and unicasts.
This commit prepares the general infrastructure for this.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
HW revision is derived from device ID and rev id.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver queries the FW for WOL support.
Ethtool get/set_wol is implemented accordingly.
Only magic packets are supported at the time.
Signed-off-by: Igor Yarovinsky <igory@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each RX ring will have its own interrupt vector, and TX rings will share one
(we mostly use polling for TX completions).
The vectors are assigned first time device is opened, and its name includes
the interface name and ring number.
Signed-off-by: Markuze Alex <markuze@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adding a pool of MSI-X vectors and EQs that can be used explicitly by mlx4_core
customers (mlx4_ib, mlx4_en). The consumers will assign their own names to the
interrupt vectors. Those vectors are not opened at mlx4 device initialization,
opened by demand.
Changed the max number of possible EQs according to the new scheme, no longer relies on
on number of cores.
The new functionality is exposed through mlx4_assign_eq() and mlx4_release_eq().
Customers that do not use the new API will get completion vectors as before.
Signed-off-by: Markuze Alex <markuze@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of reseting the module parameters each ifup or mtu change,
they are being set once at device initialization
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some irq_set_type() callbacks need to change the chip and the handler
when the trigger mode changes. We have already a (misnomed) setter
function for the handler which can be called from irq_set_type().
Provide one which allows to set chip and name as well. Put the
misnomed function under the COMPAT switch and provide a replacement.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Some irq chips need to call genirq functions for nested chips from
their callbacks. That upsets lockdep. So they need to set a different
lock class for those nested chips. Provide a helper function to avoid
open access to irq_desc.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
commit 86271e460a introduced a
regression that caused mac80211 queues in stopped state.
ath_drain_all_txq is called in driver flush which would reset
the stopped flag and the mac80211 queues were never started
after that. iperf traffic is completely stalled due to this issue.
Restart the mac80211 queues in driver flush only if the txqs were
drained.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
With the recent tx status optimization in mac80211, we bail out as
and and when invalid rate index is found. So the behavior of resetting
rate idx to -1 and count to 0 has changed for the rate indexes that
were not part of the driver's retry series.
This has resulted in ath9k using incorrect rate table index which
caused the system to panic. Ideally ath9k need to loop only for the
indexes that were part of the retry series and so simply use hw->max_rates
as the loop counter.
Pasted the stack trace of the panic issue for reference.
[ 754.093192] BUG: unable to handle kernel paging request at ffff88046a9025b0
[ 754.093256] IP: [<ffffffffa02eac49>] ath_tx_status+0x209/0x2f0 [ath9k]
[ 754.094888] Call Trace:
[ 754.094903] <IRQ>
[ 754.094928] [<ffffffffa051f883>] ieee80211_tx_status+0x203/0x9e0 [mac80211]
[ 754.094975] [<ffffffffa053e305>] ? __ieee80211_wake_queue+0x125/0x140 [mac80211]
[ 754.095017] [<ffffffffa02e66c9>] ath_tx_complete_buf+0x1b9/0x370 [ath9k]
[ 754.095054] [<ffffffffa02e6fcf>] ath_tx_complete_aggr+0x51f/0xb50 [ath9k]
[ 754.095098] [<ffffffffa05382a3>] ? ieee80211_prepare_and_rx_handle+0x173/0xab0 [mac80211]
[ 754.095148] [<ffffffff81350e62>] ? _raw_spin_unlock_irqrestore+0x32/0x40
[ 754.095186] [<ffffffffa02e9735>] ath_tx_tasklet+0x365/0x4b0 [ath9k]
[ 754.095224] [<ffffffff8107a2a2>] ? clockevents_program_event+0x62/0xa0
[ 754.095261] [<ffffffffa02e2628>] ath9k_tasklet+0x168/0x1c0 [ath9k]
[ 754.095298] [<ffffffff8105599b>] tasklet_action+0x6b/0xe0
[ 754.095331] [<ffffffff81056278>] __do_softirq+0x98/0x120
[ 754.095361] [<ffffffff8100cd5c>] call_softirq+0x1c/0x30
[ 754.095393] [<ffffffff8100efb5>] do_softirq+0x65/0xa0
[ 754.095423] [<ffffffff810563fd>] irq_exit+0x8d/0x90
[ 754.095453] [<ffffffff8100ebc1>] do_IRQ+0x61/0xe0
[ 754.095482] [<ffffffff81351413>] ret_from_intr+0x0/0x15
[ 754.095513] <EOI>
[ 754.095531] [<ffffffff81014375>] ? native_sched_clock+0x15/0x70
[ 754.096475] [<ffffffffa02bcfa6>] ? acpi_idle_enter_bm+0x24d/0x285 [processor]
[ 754.096475] [<ffffffffa02bcf9f>] ? acpi_idle_enter_bm+0x246/0x285 [processor]
[ 754.096475] [<ffffffff8127fab2>] cpuidle_idle_call+0x82/0x100
[ 754.096475] [<ffffffff8100a236>] cpu_idle+0xa6/0xf0
[ 754.096475] [<ffffffff81339bc1>] rest_init+0x91/0xa0
[ 754.096475] [<ffffffff814efccd>] start_kernel+0x3fd/0x408
[ 754.096475] [<ffffffff814ef347>] x86_64_start_reservations+0x132/0x136
[ 754.096475] [<ffffffff814ef451>] x86_64_start_kernel+0x106/0x115
[ 754.096475] RIP [<ffffffffa02eac49>] ath_tx_status+0x209/0x2f0 [ath9k]
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
On hardware busy the scan request pointer should be cleared, as higher
levels will release. This avoids a crash when that pointer is
erroneously used later.
Signed-off-by: Joseph J. Gunn <armadefuego@yahoo.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Clearly a mistake, since pointers won't suddenly
change their value...
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
commit 2c8cec5c10 (Cache learned PMTU information in inetpeer) added
an extra inet_putpeer() call in ip_rt_update_pmtu().
This results in various problems, since we can free one inetpeer, while
it is still in use.
Ref: http://www.spinics.net/lists/netdev/msg159121.html
Reported-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit 9435eb1cf0
("ipv4: Implement __ip_dev_find using new interface address hash.")
we reimplemented __ip_dev_find() so that it doesn't have to
do a full FIB table lookup.
Instead, it consults a hash table of addresses configured to
interfaces.
This works identically to the old code in all except one case,
and that is for loopback subnets.
The old code would match the loopback device for any IP address
that falls within a subnet configured to the loopback device.
Handle this corner case by doing the FIB lookup.
We could implement this via inet_addr_onlink() but:
1) Someone could configure many addresses to loopback and
inet_addr_onlink() is a simple list traversal.
2) We know the old code works.
Reported-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some filesystems (such as ext4) can return the same cookie value for
multiple files. If we try to start a readdir with one of these cookies,
the server will return the first file found with a cookie of the same
value. This can cause the client to enter an infinite loop.
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
nfs_opendir() created a context that held much more information than we
need for a readdir. This patch introduces a slimmed-down
nfs_open_dir_context that contains only the cookie and the cred used for
RPC operations. The new context will eventually be used to help detect
readdir loops.
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If we're doing a search by readdir cookie, we need to ensure that the
resulting f_pos is updated. To do so, we need to update the
desc->current_index, in the same way that we do in the search by
file offset case.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This patch adds the TCM_Loop Linux/SCSI LLD fabric module for
accessing TCM device backstores as locally accessable SCSI LUNs in
virtual SAS, FC, and iSCSI Target ports using the generic fabric
TransportID and Target Port WWN naming handlers from TCM's
target_core_fabric_lib.c The TCM_Loop module uses the generic fabric
configfs infratructure provided by target_core_fabric_configfs.c and
adds a module dependent attribute for the creation/release of the
virtual I_T Nexus connected the TCM_Loop Target and Initiator Ports.
TCM_Loop can also be used with scsi-generic and BSG drivers so that
STGT userspace fabric modules, QEMU-KVM and other hypervisor SCSI
passthrough support can access TCM device backstore and control CDB
emulation.
For more information please see:
http://linux-iscsi.org/wiki/Tcm_loop
[jejb: fixed up checkpatch stuff]
Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The map_bh() call will have already set the buffer_head to mapped.
Signed-off-by: Feng Tang <feng.tang@intel.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Disable Interrupt MBX completion will disable the interrupt on
successful completion. Fixed the bug where driver was waiting for
Interrupt to come in for its completion. Now driver will poll for
disable interrupt MBX completion.
Signed-off-by: Sarang Radke <sarang.radke@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
This reverts commit 24d720b726.
Previously we thought there was little possibility that devices would
crash with this, but some have been found.
Reported-by: Alan Stern <stern@rowland.harvard.edu>
Cc: Luben Tuikov <ltuikov@yahoo.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Driver does not detect a new CQE (completion queue entry) if a thread receives
the wakup when it is in TASK_RUNNING state. Fix is to set the state to
TASK_INTERRUPTIBLE while holding the fp_work_lock.
Also, Use __set_current_task() since it is now set inside a spinlock with
synchronization.
Two other related optimizations:
1. After we exit the while (!kthread_should_stop()) loop, use
__set_current_state() since synchronization is no longer needed.
2. Remove set_current_state(TASK_RUNNING) after schedule() since it
should always be TASK_RUNNING after schedule().
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Move relogin to all devices code from do_dpc to new
fuction qla4xxx_relogin_all_devices()
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
This provides the flexibility to modify the qdepth based on different
target devices to make the best use of system resources.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
We don't need to check ddb old state we can take action
based on ddb new state.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
remove ha flag AF_HBA_GOING_AWAY and added flag AF_HA_REMOVAL
to mark the other ISP-4xxx port to indicate that the driver is
being removed, so that the other port will not re-initialize
while in the process of removing the ha due to driver unload
or hba hotplug.
Signed-off-by: Karen Higgins <karen.higgins@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Driver has capability to add device dynamically and present
them to OS, driver no longer need to wait for DDBs to come
online during driver initialization.
Driver still issues a relogin for DDBs that are not online,
but no longer wait for DDB to come online.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Karen Higgins <karen.higgins@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
FW is not able to process mbox command if FW state is failed. This will cause
mbox command to timeout and adapter reset. We have separate function to detect
FW failed state and do adapter reset. So to avoid mbox command timeout, do not
process mbox command in case of FW state failed.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Prasanna Mumbai <prasanna.mumbai@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Remove process all aen code from qla4xxx_initialize_ddb_list()
as DPC activities should be done in DPC only.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
GCC 4.6 explicitly represents the MDR register. It may be accessed
via the "z" constraint. Perhaps more importantly, it tracks when
the MDR register is clobbered and uses the RETF instruction if the
incoming value is still valid.
Thus it is important to (at least) clobber the MDR register in
relevant inline assembly fragments, lest RETF be used incorrectly.
The only instances I could find are here. There are reads of the
MDR register in kernel/gdb-stub.c, but that's harmless. Although,
frankly, __builtin_return_address(0) might be a better thing in
those cases. Certainly MDR isn't going to contain anything else
that might be useful...
Signed-off-by: Richard Henderson <rth@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
There have been many complaints that an enclosure with subenclosures
isn't attached to by the ses driver. Until now, though, no-one had
been willing to provide access to one.
Subenclosures are added simply by flattening the tree (i.e. all
subenclosure devices show up under the one main device). This may have
consequences if the naming is only unique per subenclosure, but that's a
bug for another day. The tested array had no page 7, so no device
naming at all. It also only had the disk devices on one of its
subenclosures (all the others had power, fans, temperature and various
sensors), so testing of this is fairly rudimentary.
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The check for module state MODULE_STATE_LIVE is no longer required for LLDs, as
libfcoe transport takes care of it. Reference:
http://marc.info/?l=linux-scsi&m=129989565903046&w=2
Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
bnx2x now uses seperate MTUs for networking and FCoE. FCoE MTU is fixed to 2500
and bnx2fc now needs to match this logic by using FCOE_MTU instead of
netdev->mtu.
Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
bnx2fc only operates on bnx2x hardware devices and not master bonding
devices, so there is no need to check for bonding. Even if the bnx2x
device is "enslaved" into a bonding device, FCoE is unaffected as it
has its own MAC address and queues.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
tgt lock is needed during -
bnx2fc_return_rqe to protect the rq_prod_idx.
bnx2fc_get_next_rqe to protect rq_cons_idx
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
