Allow users to connect XRC QPs through the rdma_cm.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow the user to indicate the QP type separately from the port space
when allocating an rdma_cm_id. With RDMA_PS_IB, there is no longer a
1:1 relationship between the QP type and port space, so we need to
switch on the QP type to select between UD and connected QPs.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Add RDMA_PS_IB. XRC QP types will use the IB port space when operating
over the RDMA CM. For the 'IP protocol' field value, we select 0x3F,
which is listed as being for 'any local network'.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The XRC annex was updated to have XRC behave more like RD. Specifically,
the XRC TGT QPN moves from the local QPN to local EECN field. Lookup of
SRQN is done using the REQ/REP protocol.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Update the REQ and REP messages to support XRC connection setup
according to the XRC Annex. Several existing fields must be set to 0 or
1 when connecting XRC QPs, and a reserved field is changed to an
extended transport type.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow user space to operate on XRC TGT QPs the same way as other types
of QPs, with one notable exception: since XRC TGT QPs may be shared
among multiple processes, the XRC TGT QP is allowed to exist beyond the
lifetime of the creating process.
The process that creates the QP is allowed to destroy it, but if the
process exits without destroying the QP, then the QP will be left bound
to the lifetime of the XRCD.
TGT QPs are not associated with CQs or a PD.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC INI QPs are similar to send only RC QPs. Allow user space to create
INI QPs. Note that INI QPs do not require receive CQs.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
We require additional information to create XRC SRQs than we can
exchange using the existing create SRQ ABI. Provide an enhanced create
ABI for extended SRQ types.
Based on patches by Jack Morgenstein <jackm@dev.mellanox.co.il>
and Roland Dreier <roland@purestorage.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow user space to create XRC domains. Because XRCDs are expected to
be shared among multiple processes, we use inodes to identify an XRCD.
Based on patches by Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC TGT QPs are intended to be shared among multiple users and
processes. Allow the destruction of an XRC TGT QP to be done explicitly
through ib_destroy_qp() or when the XRCD is destroyed.
To support destroying an XRC TGT QP, we need to track TGT QPs with the
XRCD. When the XRCD is destroyed, all tracked XRC TGT QPs are also
cleaned up.
To avoid stale reference issues, if a user is holding a reference on a
TGT QP, we increment a reference count on the QP. The user releases the
reference by calling ib_release_qp. This releases any access to the QP
from a user above verbs, but allows the QP to continue to exist until
destroyed by the XRCD.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC ("eXtended reliable connected") is an IB transport that provides
better scalability by allowing senders to specify which shared receive
queue (SRQ) should be used to receive a message, which essentially
allows one transport context (QP connection) to serve multiple
destinations (as long as they share an adapter, of course).
XRC communication is between an initiator (INI) QP and a target (TGT)
QP. Target QPs are associated with SRQs through an XRCD. An XRC TGT QP
behaves like a receive-only RD QP. XRC INI QPs behave similarly to RC
QPs, except that work requests posted to an XRC INI QP must specify the
remote SRQ that is the target of the work request.
We define two new QP types for XRC, to distinguish between INI and TGT
QPs, and update the core layer to support XRC QPs.
This patch is derived from work by Jack Morgenstein
<jackm@dev.mellanox.co.il>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC ("eXtended reliable connected") is an IB transport that provides
better scalability by allowing senders to specify which shared receive
queue (SRQ) should be used to receive a message, which essentially
allows one transport context (QP connection) to serve multiple
destinations (as long as they share an adapter, of course).
XRC defines SRQs that are specifically used by XRC connections. Expand
the SRQ code to support XRC SRQs. An XRC SRQ is currently restricted to
only XRC use according to the IB XRC Annex.
Portions of this patch were derived from work by
Jack Morgenstein <jackm@dev.mellanox.co.il>.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Currently, there is only a single ("basic") type of SRQ, but with XRC
support we will add a second. Prepare for this by defining an SRQ type
and setting all current users to IB_SRQT_BASIC.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC ("eXtended reliable connected") is an IB transport that provides
better scalability by allowing senders to specify which shared receive
queue (SRQ) should be used to receive a message, which essentially
allows one transport context (QP connection) to serve multiple
destinations (as long as they share an adapter, of course).
A few new concepts are introduced to support this. This patch adds:
- A new device capability flag, IB_DEVICE_XRC, which low-level
drivers set to indicate that a device supports XRC.
- A new object type, XRC domains (struct ib_xrcd), and new verbs
ib_alloc_xrcd()/ib_dealloc_xrcd(). XRCDs are used to limit which
XRC SRQs an incoming message can target.
This patch is derived from work by Jack Morgenstein <jackm@dev.mellanox.co.il>.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This oops was reported recently:
d:mon> e
cpu 0xd: Vector: 300 (Data Access) at [c0000000fd4c7120]
pc: d00000000076f194: .t3_l2t_get+0x44/0x524 [cxgb3]
lr: d000000000b02108: .init_act_open+0x150/0x3d4 [cxgb3i]
sp: c0000000fd4c73a0
msr: 8000000000009032
dar: 0
dsisr: 40000000
current = 0xc0000000fd640d40
paca = 0xc00000000054ff80
pid = 5085, comm = iscsid
d:mon> t
[c0000000fd4c7450] d000000000b02108 .init_act_open+0x150/0x3d4 [cxgb3i]
[c0000000fd4c7500] d000000000e45378 .cxgbi_ep_connect+0x784/0x8e8 [libcxgbi]
[c0000000fd4c7650] d000000000db33f0 .iscsi_if_rx+0x71c/0xb18
[scsi_transport_iscsi2]
[c0000000fd4c7740] c000000000370c9c .netlink_data_ready+0x40/0xa4
[c0000000fd4c77c0] c00000000036f010 .netlink_sendskb+0x4c/0x9c
[c0000000fd4c7850] c000000000370c18 .netlink_sendmsg+0x358/0x39c
[c0000000fd4c7950] c00000000033be24 .sock_sendmsg+0x114/0x1b8
[c0000000fd4c7b50] c00000000033d208 .sys_sendmsg+0x218/0x2ac
[c0000000fd4c7d70] c00000000033f55c .sys_socketcall+0x228/0x27c
[c0000000fd4c7e30] c0000000000086a4 syscall_exit+0x0/0x40
--- Exception: c01 (System Call) at 00000080da560cfc
The root cause was an EEH error, which sent us down the offload_close path in
the cxgb3 driver, which in turn sets cdev->l2opt to NULL, without regard for
upper layer driver (like the cxgbi drivers) which might have execution contexts
in the middle of its use. The result is the oops above, when t3_l2t_get attempts
to dereference L2DATA(cdev)->nentries in arp_hash right after the EEH error handler sets it to NULL.
The fix is to prevent the setting of the NULL pointer until after there are no
further users of it. The t3cdev->l2opt pointer is now converted to be an rcu
pointer and the L2DATA macro is now called under the protection of the
rcu_read_lock(). When the EEH error path:
t3_adapter_error->offload_close->cxgb3_offload_deactivate
Is exectured, setting of that l2opt pointer to NULL, is now gated on an rcu
quiescence point, preventing, allowing L2DATA callers to safely check for a NULL
pointer without concern that the underlying data will be freeded before the
pointer is dereferenced.
This has been tested by the reporter and shown to fix the reproted oops
[nhorman: fix up unitinialised variable reported by Dan Carpenter]
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reviewed-by: Karen Xie <kxie@chelsio.com>
Cc: stable@kernel.org
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
RFC3270 mandates that iSCSI PDUs are padded to the closest integer
number of four byte words. Fix the iser code to support that on both
the TX/RX flows.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The code that prepares the SG associated with SCSI command for FMR was
buggy for systems with DMA addresses that don't fit in unsigned long,
e.g under the 32-bit based XenServer dom0 sizeof(dma_addr_t) is 8.
Fix that by casting to unsigned long long a masking constant used by
the code. This resolves a crash in iser_sg_to_page_vec on this system.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
target: Convert to DIV_ROUND_UP_SECTOR_T usage for sectors / dev_max_sectors
kernel.h: Add DIV_ROUND_UP_ULL and DIV_ROUND_UP_SECTOR_T macro usage
iscsi-target: Add iSCSI fabric support for target v4.1
iscsi: Add Serial Number Arithmetic LT and GT into iscsi_proto.h
iscsi: Use struct scsi_lun in iscsi structs instead of u8[8]
iscsi: Resolve iscsi_proto.h naming conflicts with drivers/target/iscsi
This allows us to move duplicated code in <asm/atomic.h>
(atomic_inc_not_zero() for now) to <linux/atomic.h>
Signed-off-by: Arun Sharma <asharma@fb.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch renames the following iscsi_proto.h structures to avoid
namespace issues with drivers/target/iscsi/iscsi_target_core.h:
*) struct iscsi_cmd -> struct iscsi_scsi_req
*) struct iscsi_cmd_rsp -> struct iscsi_scsi_rsp
*) struct iscsi_login -> struct iscsi_login_req
This patch includes useful ISCSI_FLAG_LOGIN_[CURRENT,NEXT]_STAGE*,
and ISCSI_FLAG_SNACK_TYPE_* definitions used by iscsi_target_mod, and
fixes the incorrect definition of struct iscsi_snack to following
RFC-3720 Section 10.16. SNACK Request.
Also, this patch updates libiscsi, iSER, be2iscsi, and bn2xi to
use the updated structure definitions in a handful of locations.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (26 commits)
IB/qib: Defer HCA error events to tasklet
mlx4_core: Bump the driver version to 1.0
RDMA/cxgb4: Use printk_ratelimited() instead of printk_ratelimit()
IB/mlx4: Support PMA counters for IBoE
IB/mlx4: Use flow counters on IBoE ports
IB/pma: Add include file for IBA performance counters definitions
mlx4_core: Add network flow counters
mlx4_core: Fix location of counter index in QP context struct
mlx4_core: Read extended capabilities into the flags field
mlx4_core: Extend capability flags to 64 bits
IB/mlx4: Generate GID change events in IBoE code
IB/core: Add GID change event
RDMA/cma: Don't allow IPoIB port space for IBoE
RDMA: Allow for NULL .modify_device() and .modify_port() methods
IB/qib: Update active link width
IB/qib: Fix potential deadlock with link down interrupt
IB/qib: Add sysfs interface to read free contexts
IB/mthca: Remove unnecessary read of PCI_CAP_ID_EXP
IB/qib: Remove double define
IB/qib: Remove unnecessary read of PCI_CAP_ID_EXP
...
With ib_qib options:
options ib_qib krcvqs=1 pcie_caps=0x51 rcvhdrcnt=4096 singleport=1 ibmtu=4
a run of ib_write_bw -a yields the following:
------------------------------------------------------------------
#bytes #iterations BW peak[MB/sec] BW average[MB/sec]
1048576 5000 2910.64 229.80
------------------------------------------------------------------
The top cpu use in a profile is:
CPU: Intel Architectural Perfmon, speed 2400.15 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask
of 0x00 (No unit mask) count 1002300
Counted LLC_MISSES events (Last level cache demand requests from this core that
missed the LLC) with a unit mask of 0x41 (No unit mask) count 10000
samples % samples % app name symbol name
15237 29.2642 964 17.1195 ib_qib.ko qib_7322intr
12320 23.6618 1040 18.4692 ib_qib.ko handle_7322_errors
4106 7.8860 0 0 vmlinux vsnprintf
Analysis of the stats, profile, the code, and the annotated profile indicate:
- All of the overflow interrupts (one per packet overflow) are
serviced on CPU0 with no mitigation on the frequency.
- All of the receive interrupts are being serviced by CPU0. (That is
the way truescale.cmds statically allocates the kctx IRQs to CPU)
- The code is spending all of its time servicing QIB_I_C_ERROR
RcvEgrFullErr interrupts on CPU0, starving the packet receive
processing.
- The decode_err routine is very inefficient, using a printf variant
to format a "%s" and continues to loop when the errs mask has been
cleared.
- Both qib_7322intr and handle_7322_errors read pci registers, which
is very inefficient.
The fix does the following:
- Adds a tasklet to service QIB_I_C_ERROR
- Replaces the very inefficient scnprintf() with a memcpy(). A field
is added to qib_hwerror_msgs to save the sizeof("string") at
compile time so that a strlen is not needed during err_decode().
- The most frequent errors (Overflows) are serviced first to exit the
loop as early as possible.
- The loop now exits as soon as the errs mask is clear rather than
fruitlessly looping through the msp array.
With this fix the performance changes to:
------------------------------------------------------------------
#bytes #iterations BW peak[MB/sec] BW average[MB/sec]
1048576 5000 2990.64 2941.35
------------------------------------------------------------------
During testing of the error handling overflow patch, it was determined
that some CPU's were slower when servicing both overflow and receive
interrupts on CPU0 with different MSI interrupt vectors.
This patch adds an option (krcvq01_no_msi) to not use a dedicated MSI
interrupt for kctx's < 2 and to service them on the default interrupt.
For some CPUs, the cost of the interrupt enter/exit is more costly
than then the additional PCI read in the default handler.
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
- unify vlan and nonvlan rx path
- kill nesvnic->vlan_grp and nes_netdev_vlan_rx_register
- allow to turn on/off rx/tx vlan accel via ethtool (set_features)
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since printk_ratelimit() shouldn't be used anymore (see comment in
include/linux/printk.h), replace it with printk_ratelimited().
Signed-off-by: Manuel Zerpies <manuel.f.zerpies@ww.stud.uni-erlangen.de>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Use the per port counter attached to all QPs created on that port to
implement port level packets/bytes performance counters a la IB.
Derived from a patch by Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allocate flow counter per Ethernet/IBoE port, and attach this counter
to all the QPs created on that port. Based on patch by Eli Cohen
<eli@mellanox.co.il>.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Move the various definitions and mad structures needed for software
implementation of IBA PM agent from the ipath and qib drivers into a
single include file, which in turn could be used by more consumers.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
IBoE doesn't use LIDs. Use the GID change event to update the IB core
cache for addition/deletion of GIDs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Add IB GID change event type. This is needed for IBoE when the HW
driver updates the GID (e.g when new VLANs are added/deleted) table
and the change should be reflected to the IB core cache.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This patch fixes a kernel crash in cma_set_qkey().
When the link layer is Ethernet, it is wrong to use IPoIB port space
since no IPoIB interface is available. Specifically, setting the
Q_Key when port space is RDMA_PS_IPOIB requires MGID calculation and
an SA query, which doesn't make sense over Ethernet.
Signed-off-by: Moni Shoua <monis@mellanox.co.il>
Acked-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
These methods don't make sense for iWARP devices, so rather than
forcing them to implement stubs, just return -ENOSYS in the core if
the hardware driver doesn't set .modify_device and/or .modify_port.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Update the active link width on QLE7220 chips when link goes down if
chip width does not match shadowed width.
Signed-off-by: Mitko Haralanov <mitko@qlogic.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
There is a possibility of a deadlock due to the way locks are
acquired and released in qib_set_uevent_bits(). The function
qib_set_uevent_bits() is called in process context and it uses
spin_lock() and spin_unlock(). This same lock is acquired/released
in interrupt context which can lead to a deadlock when running on
the same cpu.
The fix is to replace spin_lock() and spin_unlock() with
spin_lock_irqsave() and spin_unlock_irqrestore() respectively in
qib_set_uevent_bits().
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Indicate the number of free user contexts via the sysfs file
/sys/class/infiniband/qib0/nfreectxts as required for PSM.
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The PCIE capability offset is saved during PCI bus walking. It will
remove an unnecessary search in the PCI configuration space if this
value is referenced instead of reacquiring it. Also, pci_is_pcie is a
better way of determining if the device is PCIE or not (as it uses the
same saved PCIE capability offset).
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The PCIE capability offset is saved during PCI bus walking. It will
remove an unnecessary search in the PCI configuration space if this
value is referenced instead of reacquiring it.
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Acked-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Adapt to new api. We plan to remove old one later. Almost all
changes are trivial, but there is one real fix: the following code is
unsafe:
int ncpus = num_online_cpus()
for (i = 0; i < ncpus; i++) {
..
}
because 1) we don't guarantee last bit of online cpus is equal to
num_online_cpus(). some arch assign sparse cpu number. 2) cpu
hotplugging may change cpu_online_mask at same time. we need to pin
it by get_online_cpus().
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Adapt to use new APIs. We plan to remove old one later and plan to
change current->cpus_allowed implementation.
No functional change.
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Avoid assigning an IS_ERR value to the cm_id pointer. This fixes a
few anomalies in the error flow due to confusion about checking for
NULL vs IS_ERR, and eliminates the need to test for the IS_ERR value
every time we wish to determine if the cma_id object has a cm device
associated with it.
Also, eliminate the now-unnecessary procedure cma_has_cm_dev (we can
check directly for the existence of the device pointer -- for a
non-NULL check, makes no difference if it is the iwarp or the ib
pointer).
Finally, make a few code changes here to improve coding consistency.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
SCSI scanning of a channel🆔lun triplet in Linux works as follows
(function scsi_scan_target() in drivers/scsi/scsi_scan.c):
- If lun == SCAN_WILD_CARD, send a REPORT LUNS command to the target
and process the result.
- If lun != SCAN_WILD_CARD, send an INQUIRY command to the LUN
corresponding to the specified channel🆔lun triplet to verify
whether the LUN exists.
So a SCSI driver must either take the channel and target id values in
account in its quecommand() function or it should declare that it only
supports one channel and one target id.
Currently the ib_srp driver does neither. As a result scanning the
SCSI bus via e.g. rescan-scsi-bus.sh causes many duplicate SCSI
devices to be created. For each 0:0:L device, several duplicates are
created with the same LUN number and with (C:I) != (0:0). Fix this by
declaring that the ib_srp driver only supports one channel and one
target id.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Cc: <stable@kernel.org>
Acked-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Commits 71c29bd5c2 ("IB/uverbs: Add devnode method to set path/mode")
and c3af0980ce ("IB: Add devnode methods to cm_class and umad_class")
added devnode methods that set the mode.
However, these methods don't check for a NULL mode, and so we get a
crash when unloading modules because devtmpfs_delete_node() calls
device_get_devnode() with mode == NULL.
Add the missing checks.
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.de>
[ Also fix cm.c. - Roland ]
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
