- 26 May, 2016 29 commits
-
-
Christoph Lameter authored
In practice, each RDMA device has a unique set of counters that the hardware implements. Having a central set of counters that they must all adhere to is limiting and causes many useful counters to not be available. Therefore we create a dynamic counter registration infrastructure. The driver must implement a stats structure allocation routine, in which the driver must place the directory name it wants, a list of names for all of the counters, an array of u64 counters themselves, plus a few generic configuration options. We then implement a core routine to create a sysfs file for each of the named stats elements, and a core routine to retrieve the stats when any of the sysfs attribute files are read. To avoid excessive beating on the stats generation routine in the drivers, the core code also caches the stats for a short period of time so that someone attempting to read all of the stats in a given device's directory will not result in a stats generation call per file read. Future work will attempt to standardize just the shared stats elements, and possibly add a method to get the stats via netlink in addition to sysfs. Signed-off-by:
Christoph Lameter <cl@linux.com> Signed-off-by:
Mark Bloch <markb@mellanox.com> Reviewed-by:
Steve Wise <swise@opengridcomputing.com> Signed-off-by:
Doug Ledford <dledford@redhat.com> [ Add caching, make structure names more informative, add i40iw support, other significant rewrites from the original patch ]
-
Doug Ledford authored
-
Jubin John authored
The pio map initialization function is off by 1 causing the last kernel send context that is allocated to not get mapped into the pio map which leads to the last kernel send context not being used by any of the qps. The send context reserved for VL15 is taken care of by setting the scontext variable that is used as the index into the kernel send context array to 1 and does not need to be accounted for in the kernel send context counting loop as it is currently done. Fix the kernel send context counting loop to account for all the allocated send contexts and map all of them to the different VLs. Reviewed-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by:
Jianxin Xiong <jianxin.xiong@intel.com> Signed-off-by:
Jubin John <jubin.john@intel.com> Signed-off-by:
-
Dean Luick authored
Two 8051 link settings, external device config and tuning method, were written in the wrong location and the previous settings were not cleared. For both, clear the old value and write the new value. Fixes: 8ebd4cf1 ("staging/rdma/hfi1: Add active and optical cable support") Reviewed-by:
Dean Luick <dean.luick@intel.com> Signed-off-by:
-
Sebastian Sanchez authored
When FM is disabled, and the HFI port on the switch is changed from MgmtAllowed=YES to MgmtAllowed=NO and the link is bounced, FULL_MGMT_P_KEY doesn't get cleared from the pkey table. This also occurs when the QSFP cable is moved from a switch port with MgmtAllowed=YES to a MgmtAllowed=NO port. Clear pkey entry properly. Also, when the driver is loaded and the switch port is set to MgmtAllowed=NO, FULL_MGMT_P_KEY shouldn't be added to pkey table after FM is started. Only set FULL_MGMT_P_KEY in the pkey table if switch port is configured to MgmtAllowed=YES. Reviewed-by:
Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by:
-
Mike Marciniszyn authored
rdmavt allows the driver to specify the size of the ack queue, but only uses it for the modify QP limit testing for setting the atomic limit value. The driver dependent size is now used to size the s_ack_queue ring dynamicially. Since the driver knows its size, the driver will use its define for any ring size dependent code. Reviewed-by:
Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by:
-
Mike Marciniszyn authored
This matches the ib_qp_attr size and avoids a extremely large value when the lower level driver registers. As part of the patch, the u8 ordinals are moved to the end of the struct to reduce pahole noted excesses. Reviewed-by:
-
Mike Marciniszyn authored
Commit b9b06cb6 ("IB/hfi1: Fix missing lock/unlock in verbs drain callback") added a spin lock. Unfortunately, the new lock code can be called from a base level interrupt state, and an interrupt that can get stacked will attempt to get the same lock. Fix by using the flag save/restore spin lock variation. Cc: stable@vger.kernel.org # 4.6+ Reviewed-by:
-
Jianxin Xiong authored
Enable trace generation for packets with the "Send Last with Invalidate" and "Send Only with Invalidate" opcodes. Reviewed-by:
-
Jianxin Xiong authored
A new union member "ieth" (Invalidate Extended Transport Header) is added to the packet header definition in preparation of supporting the send with invalidate opcode. Reviewed-by:
-
Doug Ledford authored
-
Dennis Dalessandro authored
The TODO list for the hfi1 driver was completed during 4.6. In addition other objections raised (which are far beyond what was in the TODO list) have been addressed as well. It is now time to remove the driver from staging and into the drivers/infiniband sub-tree. Reviewed-by:
-
Dennis Dalessandro authored
The deletion of a cdev is not a fence for holding off references to the structure. The driver attempts to delete the cdev and then proceeds to free the parent structure, the hfi1_devdata, or dd. This can potentially lead to a kernel panic in situations where a user has an FD for the cdev open, and the pci device gets removed. If the user then closes the FD there will be a NULL dereference when trying to do put on the cdev's kobject. Fix this by pointing the cdev's kobject.parent at a new kobject embedded in its parent structure. Also take a reference when the device is opened and put it back when it is closed. Reviewed-by:
Ira Weiny <ira.weiny@intel.com> Signed-off-by:
-
Dennis Dalessandro authored
Add a trace message to HFI1s user IOCTL handling. This allows debugging of which IOCTLs are being handled by the driver. Reviewed-by:
-
Dennis Dalessandro authored
Remove the write() handler for user space commands now that ioctl handling is available. User apps will need to change to use ioctl from this point forward. Reviewed-by:
-
Dennis Dalessandro authored
IOCTL is more suited to what user space commands need to do than the write() interface. Add IOCTL definitions for all existing write commands and the handling for those. The write() interface will be removed in a follow on patch. Reviewed-by:
-
Dennis Dalessandro authored
The HFI1_CMD_SDMA_STATUS_UPD command was never implemented it has no reason to live in the driver. Remove it. Reviewed-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
-
Dennis Dalessandro authored
The snoop/diag interface is better served by an implementation which is more general and usable by other drivers perhaps. Go ahead and remove the code now and get rid of the char dev. We can put the feature back when we have a more agreeable solution. Reviewed-by:
-
Dennis Dalessandro authored
Remove EPROM handling from the cdev which is used for user application data traffic. Reviewed-by:
-
Dennis Dalessandro authored
Remove UI char device which exposes direct access to registers for user space. This was put in to aid in debugging the hardware. We are looking into alternatives means of providing the same functionality. This removes another char device from HFI1's footprint. Reviewed-by:
-
Dennis Dalessandro authored
hfi1 current exports a cdev that can be used to target all of the hfi's in the system. However there is a problem with this approach in that the devices could be on different subnets. This is a problem that user space can figure out and explicitly tell the driver on which device to create a context. Remove the multi-purpose cdev leaving a dedicated cdev for each port. Also remove the striping capability that is dependent upon the user choosing the multi-purpose cdev. It is now up to user space to determine how to stripe contexts. Reviewed-by:
-
Dennis Dalessandro authored
Remove the usage of an anti-pattern goto in hfi1_cdev_init to improve code readability. Suggested-by:
Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Reviewed-by:
-
Jianxin Xiong authored
During the processing of a user SDMA request, if there was an error before the request counter was increased, the state of the packet queue could be updated incorrectly, causing the counter to underflow. As the result, the process could get stuck later since the counter could never get back to 0. This patch adds a condition to guard the packet queue update so that the counter is only decreased if it has been increased before the error happens. Reviewed-by:
-
Jubin John authored
Building the qib driver with gcc version 6.1.0 raises the following build warning: drivers/infiniband/hw/qib/qib_iba7322.c:1311:39: warning: 'qib_7322_intr_msgs' defined but not used [-Wunused-const-variable=] static const struct qib_hwerror_msgs qib_7322_intr_msgs[] = { ^~~~~~~~~~~~~~~~~~ Remove the unused qib_7322_intr_msgs[] Reviewed-by: -
Ira Weiny authored
This comment was old, the MTU enums have been defined. Reviewed-by:
-
Jubin John authored
sdma_event_names[] is only used within CONFIG_SDMA_VERBOSITY ifdefs, so when CONFIG_SDMA_VERBOSITY is disabled, it results in the following 0-day build warning: >> drivers/infiniband/hw/hfi1/sdma.c:137:27: warning: 'sdma_event_names' >> defined but not used [-Wunused-const-variable=] static const char * const sdma_event_names[] = { ^~~~~~~~~~~~~~~~ This occurs on the following compiler: compiler: gcc-6 (Debian 6.1.1-1) 6.1.1 20160430 For more information check: https://lists.01.org/pipermail/kbuild-all/2016-May/020060.html Fix this warning by defining sdma_event_name[] only within the CONFIG_SDMA_VERBOSITY ifdefs. Reported-by:kbuild test robot <fengguang.wu@intel.com> Reviewed-by:
-
Jubin John authored
Use kzalloc_node instead of kzalloc for rdmavt memory region segment allocation to optimize for performance on NUMA platforms. Reviewed-by:
-
Mike Marciniszyn authored
The usage of the various vmalloc APIs do not consistently zero memory when allocating the swqe. Insure zeroing variants are used. Reviewed-by:
-
Mitko Haralanov authored
Commit e88c9271 ("IB/hfi1: Fix buffer cache corner case which may cause corruption") introduced a bug which may cause a reference count of a interval RB node to be leaked in the case where an SDMA transfer from that node completes at the same time as the node is being extended. If a node is being extended, it is first removed from the RB tree in order to be processed without the risk of an invalidation event removing the node at the same time. If a SDMA completion happens during that time, the completion handler will fail to find the node in the RB tree and, therefore, fail to correctly decrement its refcount. This leaves the node in the tree and its pages pinned for the duration of the user process. To prevent this from happening the io vector adds a reference to the RB node, which is used during the SDMA completion instead of looking up the node in the RB tree. This change adds a performance improvement as a side effect by avoiding the RB tree lookup. Fixes: e88c9271 ("IB/hfi1: Fix buffer cache corner case which may cause corruption") Reviewed-by:
Harish Chegondi <harish.chegondi@intel.com> Signed-off-by:
-
- 25 May, 2016 5 commits
-
-
Mark Bloch authored
In IB networks, and specifically in IPoIB/rdmacm traffic, the device address of an IPoIB interface is used as a means to exchange information between nodes needed for communication. Currently an IPoIB interface will always be created with a device address based on its node GUID without a way to change that. This change adds the ability to set the device address of an IPoIB interface by value. We use the set mac address ndo to do that. The flow should be broken down to two: 1) The GID value is already in the GID table, in this case the interface will be able to set carrier up. 2) The GID value is not yet in the GID table, in this case the interface won't try to join the multicast group and will wait (listen on GID_CHANGE event) until the GID is inserted. In order to track those changes, we add a new flag: * IPOIB_FLAG_DEV_ADDR_SET. When set, it means the dev_addr is a based on a value in the gid table. this bit will be cleared upon a dev_addr change triggered by the user and set after validation. Per IB spec the port GUID can't change if the module is loaded. port GUID is the basis for GID at index 0 which is the basis for the default device address of a ipoib interface. The issue is that there are devices that don't follow the spec, they change the port GUID while HCA is powered on, so in order not to break userspace applications. We need to check if the user wanted to control the device address and we assume that if he sets the device address back to be based on GID index 0, he no longer wishs to control it. In order to track this, we add an additional flag: * IPOIB_FLAG_DEV_ADDR_CTRL When setting the device address, there is no validation of the upper twelve bytes of the device address (flags, qpn, subnet prefix) as those bytes are not under the control of the user. Signed-off-by:
Leon Romanovsky <leonro@mellanox.com> Signed-off-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
-
Erez Shitrit authored
Check (via an SA query) if the SM supports the new option for SendOnly multicast joins. If the SM supports that option it will use the new join state to create such multicast group. If SendOnlyFullMember is supported, we wouldn't use faked FullMember state join for SendOnly MCG, use the correct state if supported. This check is performed at every invocation of mcast_restart task, to be sure that the driver stays in sync with the current state of the SM. Signed-off-by:
Erez Shitrit <erezsh@mellanox.com> Reviewed-by:
-
Erez Shitrit authored
There are four types for MCG, FullMember, NonMember, SendOnlyNonMember, and the new added type: SendOnlyFullMember. Add support for the new SendOnlyFullMember join state. The new type allows host to send join request as sendonly, it will cause the group to be created but without getting packets from this multicast back to the host. Signed-off-by:
-
Erez Shitrit authored
New SA query function to return the ClassPortInfo struct from the SA. If the SM supports FullMemberSendOnly mode for MCG's, it sets a capability bit in the capability_mask2 field of the response. Signed-off-by:
-
Erez Shitrit authored
Change struct ib_class_port_info to conform to IB Spec 1.3 That in order to get specific capability mask from ClassPortInfo mad. >From the IB Spec, ClassPortInfo section: "CapabilityMask2 Bits 0-26: Additional class-specific capabilities... RespTimeValue the rest 5 bits" The new struct now has one field for capabilitymask2 (previously was the reserved field) and the resp_time field. And it fixes up qib and srpt, use of the field repurposed to be used as capabilitymask2: IB/qib: Change pma_get_classportinfo IB/srpt: Adjust the use of ib_class_port_info Signed-off-by:Hal Rosenstock <hal@mellanox.com> Signed-off-by:
-
- 24 May, 2016 6 commits
-
-
Mark Bloch authored
There is an assumption that rdmacm is used only between nodes in the same IB subnet, this why ARP resolution can be used to turn IP to GID in rdmacm. When dealing with IB communication between subnets this assumption is no longer valid. ARP resolution will get us the next hop device address and not the peer node's device address. To solve this issue, we will check user space if it can provide the GID of the peer node, and fail if not. We add a sequence number to identify each request and fill in the GID upon answer from userspace. Signed-off-by:
-
Mark Bloch authored
Move SA ibnl client registration to ib_core module init. This will allow us to register a single client to handle all RDMA_NL_LS operations and make it SA independent. Signed-off-by:
-
Mark Bloch authored
This commits adds a new RDMA local service operation: - IP to GID resolution. The client request would include the ifindex of the outgoing interface and would place in an attribute (LS_NLA_TYPE_IPV4 or LS_NLA_TYPE_IPV6) the destnation IP. The local service would answer with a message that has the attribute: - LS_NLA_TYPE_DGID - The destination GID. Signed-off-by:
-
Mark Bloch authored
Consolidate ib_sa into ib_core, this commit eliminates ib_sa.ko and makes it part of ib_core.ko Signed-off-by:
-
Mark Bloch authored
Consolidate ib_mad into ib_core, this commit eliminates ib_mad.ko and makes it part of ib_core.ko Signed-off-by:
-
Leon Romanovsky authored
IB address resolution is declared as a module (ib_addr.ko) which loads itself before IB core module (ib_core.ko). It causes to the scenario where IB netlink which is initialized by IB core can't be used by ib_addr.ko. In order to solve it, we are converting ib_addr.ko to be part of IB core module. Signed-off-by:
-
