- 22 Jun, 2021 5 commits
-
-
Jack Wang authored
With fast memory registration on write request, rnbd-clt can do bigger IO without split. rnbd-clt now can query rtrs-clt to get the max_segments, instead of using BMAX_SEGMENTS. BMAX_SEGMENTS is not longer needed, so remove it. Link: https://lore.kernel.org/r/20210621055340.11789-6-jinpu.wang@ionos.com Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by:
Jack Wang <jinpu.wang@cloud.ionos.com> Reviewed-by:
Md Haris Iqbal <haris.iqbal@cloud.ionos.com> Signed-off-by:
Jason Gunthorpe <jgg@nvidia.com>
-
Jack Wang authored
As we can do fast memory registration on write, we can increase the max_segments, default to 512K. Link: https://lore.kernel.org/r/20210621055340.11789-5-jinpu.wang@ionos.comSigned-off-by:
-
Jack Wang authored
With write path fast memory registration, we need less memory for each request. With fast memory registration, we can reduce max_send_sge to save memory usage. Also convert the kmalloc_array to kcalloc. Link: https://lore.kernel.org/r/20210621055340.11789-4-jinpu.wang@ionos.comSigned-off-by:
-
Jack Wang authored
With fast memory registration in write path, we can reduce the memory consumption by using less max_send_sge, support IO bigger than 116 KB (29 segments * 4 KB) without splitting, and it also make the IO path more symmetric. To avoid some times MR reg failed, waiting for the invalidation to finish before the new mr reg. Introduce a refcount, only finish the request when both local invalidation and io reply are there. Link: https://lore.kernel.org/r/20210621055340.11789-3-jinpu.wang@ionos.comSigned-off-by:
Md Haris Iqbal <haris.iqbal@ionos.com> Signed-off-by:
Dima Stepanov <dmitrii.stepanov@ionos.com> Signed-off-by:
-
Jack Wang authored
Introduce tail wr, we can send as the last wr, we want to send the local invalidate wr after rdma wr in later patch. While at it, also fix coding style issue. Link: https://lore.kernel.org/r/20210621055340.11789-2-jinpu.wang@ionos.comSigned-off-by:
-
- 21 Jun, 2021 21 commits
-
-
Devesh Sharma authored
Changing ucontext ABI response structure to pass wqe_mode to user library. A flag in comp_mask has been set to indicate presence of wqe_mode. Moved wqe-mode ABI to uapi/rdma/bnxt_re-abi.h Link: https://lore.kernel.org/r/20210616202817.1185276-1-devesh.sharma@broadcom.comSigned-off-by:
Devesh Sharma <devesh.sharma@broadcom.com> Reviewed-by:
Leon Romanovsky <leonro@nvidia.com> Signed-off-by:
-
Anand Khoje authored
pahole shows two 4-byte holes in struct ib_port_data after pkey_list_lock and netdev_lock respectively. Shuffling the netdev_lock to be after pkey_list_lock, this shaves off eight bytes from the struct. Link: https://lore.kernel.org/r/20210616154509.1047-3-anand.a.khoje@oracle.comSuggested-by:
Haakon Bugge <haakon.bugge@oracle.com> Signed-off-by:
Anand Khoje <anand.a.khoje@oracle.com> Reviewed-by:
-
Anand Khoje authored
Removed port validity check from ib_get_cached_subnet_prefix() as this check is not needed because "port_num" is valid. Link: https://lore.kernel.org/r/20210616154509.1047-2-anand.a.khoje@oracle.comSuggested-by:
-
Leon Romanovsky authored
Compilation with W=1 produces warnings similar to the below. drivers/infiniband/ulp/ipoib/ipoib_main.c:320: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst All such occurrences were found with the following one line git grep -A 1 "\/\*\*" drivers/infiniband/ Link: https://lore.kernel.org/r/e57d5f4ddd08b7a19934635b44d6d632841b9ba7.1623823612.git.leonro@nvidia.com Reviewed-by: Jack Wang <jinpu.wang@ionos.com> #rtrs Reviewed-by:
Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Signed-off-by:
-
Yangyang Li authored
Switch xrcd index allocation and release from hns own bitmap interface to IDA interface. Link: https://lore.kernel.org/r/1623325814-55737-7-git-send-email-liweihang@huawei.comSigned-off-by:
Yangyang Li <liyangyang20@huawei.com> Signed-off-by:
Weihang Li <liweihang@huawei.com> Signed-off-by:
-
Yangyang Li authored
Switch pd index allocation and release from hns own bitmap interface to IDA interface. Link: https://lore.kernel.org/r/1623325814-55737-6-git-send-email-liweihang@huawei.comSigned-off-by:
-
Yangyang Li authored
Switch mtpt index allocation and release from hns own bitmap interface to IDA interface. Link: https://lore.kernel.org/r/1623325814-55737-5-git-send-email-liweihang@huawei.comSigned-off-by:
-
Yangyang Li authored
Round-robin (RR) is no longer used in the allocation of the bitmap table, and all the function input parameters that use this mechanism are BITMAP_NO_RR. The code that defines and uses the RR needs to be deleted. Link: https://lore.kernel.org/r/1623325814-55737-4-git-send-email-liweihang@huawei.comSigned-off-by:
-
Yangyang Li authored
hns_roce_bitmap_free_range() is only called inside hns_roce_bitmap_free(), and the input parameter "cnt" is set to a constant 1. In addition, the driver does not use alloc_range scenarios, so free_range does not need to exist. Link: https://lore.kernel.org/r/1623325814-55737-3-git-send-email-liweihang@huawei.comSigned-off-by:
-
Yangyang Li authored
The function is no longer used. Link: https://lore.kernel.org/r/1623325814-55737-2-git-send-email-liweihang@huawei.comSigned-off-by:
-
Wenpeng Liang authored
There are some '%u' for 'int' and '%d' for 'unsigend int', they should be fixed. Link: https://lore.kernel.org/r/1623325232-30900-1-git-send-email-liweihang@huawei.comSigned-off-by:
Wenpeng Liang <liangwenpeng@huawei.com> Signed-off-by:
-
Xi Wang authored
Remove unused members in srq context structure. Link: https://lore.kernel.org/r/1624262443-24528-10-git-send-email-liweihang@huawei.comSigned-off-by:
Xi Wang <wangxi11@huawei.com> Signed-off-by:
-
Yixing Liu authored
Use hr_write_reg() instead of roce_set_field(). Link: https://lore.kernel.org/r/1624262443-24528-9-git-send-email-liweihang@huawei.comSigned-off-by:
Yixing Liu <liuyixing1@huawei.com> Signed-off-by:
-
Yixing Liu authored
Use "hr_reg_write" to replace "roce_set_filed". Link: https://lore.kernel.org/r/1624262443-24528-8-git-send-email-liweihang@huawei.comSigned-off-by:
-
Lang Cheng authored
WQE_INDEX and OPCODE and QPN of CQE use redundant masks. Just remove them. Link: https://lore.kernel.org/r/1624262443-24528-7-git-send-email-liweihang@huawei.comSigned-off-by:
Lang Cheng <chenglang@huawei.com> Signed-off-by:
-
Lang Cheng authored
Fill all QPC fileds with hr_reg_*() instead of roce_set_*(). SQPN is used for HIP08 ES only, it should be removed. Link: https://lore.kernel.org/r/1624262443-24528-6-git-send-email-liweihang@huawei.comSigned-off-by:
-
Yixing Liu authored
Use hr_reg_*() to write CQ context, it's simpler than roce_set_*(). Link: https://lore.kernel.org/r/1624262443-24528-5-git-send-email-liweihang@huawei.comSigned-off-by:
-
Lang Cheng authored
In order to avoid to do bitwise operations on a boolean value, add a new register interface to avoid sparse comlaint about "dubious: x & !y" when calling hr_reg_write(ctx, field, !!val). Fixes: dc504774 ("RDMA/hns: Use new interface to set MPT related fields") Fixes: 495c2480 ("RDMA/hns: Add XRC subtype in QPC and XRC type in SRQC") Link: https://lore.kernel.org/r/1624262443-24528-4-git-send-email-liweihang@huawei.comSigned-off-by:
-
Weihang Li authored
GCC may reports an running time assert error when a value calculated from ib_mtu_enum_to_int() is using as 'val' in FIELD_PREDP: include/linux/compiler_types.h:328:38: error: call to '__compiletime_assert_1524' declared with attribute error: FIELD_PREP: value too large for the field So a check is added about whether integer mtu from ib_mtu_enum_to_int() is negative to avoid this warning. Link: https://lore.kernel.org/r/1624262443-24528-3-git-send-email-liweihang@huawei.comReported-by:
kernel test robot <lkp@intel.com> Signed-off-by:
-
Weihang Li authored
There is no need to use "!!" before "eq->eqe_size == HNS_ROCE_V3_EQE_SIZE", or sparse will complain about "dubious: x & !y". Fixes: 782832f2 ("RDMA/hns: Simplify the function config_eqc()") Link: https://lore.kernel.org/r/1624262443-24528-2-git-send-email-liweihang@huawei.comSigned-off-by:
-
Avihai Horon authored
Relaxed Ordering is a capability that can only benefit users that support it. All kernel ULPs should support Relaxed Ordering, as they are designed to read data only after observing the CQE and use the DMA API correctly. Hence, implicitly enable Relaxed Ordering by default for MR transfers in kernel ULPs. Link: https://lore.kernel.org/r/b7e820aab7402b8efa63605f4ea465831b3b1e5e.1623236426.git.leonro@nvidia.comSigned-off-by:
Avihai Horon <avihaih@nvidia.com> Signed-off-by:
-
- 18 Jun, 2021 6 commits
-
-
Xi Wang authored
Both of HIP08 and HIP09 require the extended doorbell information to be cleared before being used. Fixes: 6b63597d ("RDMA/hns: Add TSQ link table support") Link: https://lore.kernel.org/r/1623392089-35639-1-git-send-email-liweihang@huawei.comSigned-off-by:
-
Jack Wang authored
Currently we only check device max_qp_wr limit for IO connection, but not for service connection. We should check for both. So save the max_qp_wr device limit in wr_limit, and use it for both IO connections and service connections. While at it, also remove an outdated comments. Link: https://lore.kernel.org/r/20210614090337.29557-6-jinpu.wang@ionos.comSuggested-by:
Jack Wang <jinpu.wang@ionos.com> Signed-off-by:
Gioh Kim <gi-oh.kim@ionos.com> Reviewed-by:
-
Guoqing Jiang authored
Those variables are passed to create_cq, create_qp, rtrs_iu_alloc and rtrs_iu_free, so these *_size means the num of unit. And cq_size also means number of cq element. Also move the setting of cq_num to common path. Link: https://lore.kernel.org/r/20210614090337.29557-5-jinpu.wang@ionos.comSigned-off-by:
Guoqing Jiang <guoqing.jiang@cloud.ionos.com> Reviewed-by:
-
Md Haris Iqbal authored
When using rdma_rxe, post_one_recv() returns ENOMEM error due to the full recv queue. This patch increase the number of WR for receive queue to support all devices. Link: https://lore.kernel.org/r/20210614090337.29557-4-jinpu.wang@ionos.comSigned-off-by:
-
Jack Wang authored
We use device limit max_send_sge, which is suboptimal for memory usage. We don't need that much for User Con, 1 is enough. And for IO con, sess->max_segments + 1 is enough Link: https://lore.kernel.org/r/20210614090337.29557-3-jinpu.wang@ionos.comSigned-off-by:
-
Jack Wang authored
Currently rtrs when create_qp use a coarse numbers (bigger in general), which leads to hardware create more resources which only waste memory with no benefits. For max_send_wr, we don't really need alway max_qp_wr size when creating qp, reduce it to cq_size. For max_recv_wr, cq_size is enough. With the patch when sess_queue_depth=128, per session (2 paths) memory consumption reduced from 188 MB to 65MB When always_invalidate is enabled, we need send more wr, so treat it special. Fixes: 9cb83748 ("RDMA/rtrs: server: main functionality") Link: https://lore.kernel.org/r/20210614090337.29557-2-jinpu.wang@ionos.comSigned-off-by:
-
- 16 Jun, 2021 8 commits
-
-
Jason Gunthorpe authored
The driver's device group can be specified as part of the ops structure like the device's port group. No need for the complicated API. Link: https://lore.kernel.org/r/8964785a34fd3a29ff5b6693493f575b717e594d.1623427137.git.leonro@nvidia.comSigned-off-by:
-
Jason Gunthorpe authored
Now that the port_groups data is being destroyed and managed by the core code this restriction is no longer needed. All the ib_port_attrs are compatible with the core's sysfs lifecycle. When the main device is destroyed and moved to another namespace the driver's port sysfs can be created/destroyed as well due to it now being a simple attribute list. Link: https://lore.kernel.org/r/afd8b676eace2821692d44489ff71856277c48d1.1623427137.git.leonro@nvidia.comSigned-off-by:
-
Jason Gunthorpe authored
init_port was only being used to register sysfs attributes against the port kobject. Now that all users are creating static attribute_group's we can simply set the attribute_group list in the ops and the core code can just handle it directly. This makes all the sysfs management quite straightforward and prevents any driver from abusing the naked port kobject in future because no driver code can access it. Link: https://lore.kernel.org/r/114f68f3d921460eafe14cea5a80ca65d81729c3.1623427137.git.leonro@nvidia.comSigned-off-by:
-
Jason Gunthorpe authored
hfi1 should not be creating a mess of kobjects to attach to the port kobject - this is all attributes. The proper API is to create an attribute_group list and create it against the port's kobject. Link: https://lore.kernel.org/r/cbe0ccb6175dd22274359b6ad803a37435a70e91.1623427137.git.leonro@nvidia.comTested-by:
Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com> Signed-off-by:
-
Jason Gunthorpe authored
qib should not be creating a mess of kobjects to attach to the port kobject - this is all attributes. The proper API is to create an attribute_group list and create it against the port's kobject. Link: https://lore.kernel.org/r/911e0031e1ed495b0006e8a6efec7b67a702cd5e.1623427137.git.leonro@nvidia.comTested-by:
-
Jason Gunthorpe authored
This code is trying to attach a list of counters grouped into 4 groups to the ib_port sysfs. Instead of creating a bunch of kobjects simply express everything naturally as an ib_port_attribute and add a single attribute_groups list. Remove all the naked kobject manipulations. Link: https://lore.kernel.org/r/0d5a7241ee0fe66622de04fcbaafaf6a791d5c7c.1623427137.git.leonro@nvidia.comSigned-off-by:
-
Jason Gunthorpe authored
Other things outside the core code are creating attributes against the port. This patch exposes the basic machinery to do this. The ib_port_attribute type allows creating groups of attributes attatched to the port and comes with the usual machinery to do this. Link: https://lore.kernel.org/r/5c4aeae57f6fa7c59a1d6d1c5506069516ae9bbf.1623427137.git.leonro@nvidia.comSigned-off-by:
-
Jason Gunthorpe authored
This call does nothing because the ib_port kobj is nested under a struct device kobject and the dev_uevent_filter() function of the struct device blocks uevents for any children kobj's that are not also struct devices. A uevent for the struct device will be triggered after ib_setup_port_attrs() returns which causes udev to pick up all the deep "attributes" which are implemented as kobjects nested under a struct device and assign them to the udev object for the struct device: $ udevadm info -a /sys/class/infiniband/ibp0s9 ATTR{ports/1/counters/excessive_buffer_overrun_errors}=="0" Link: https://lore.kernel.org/r/49231c92c7d4c60686de18f7e20932d0c82160ee.1623427137.git.leonro@nvidia.comSigned-off-by:
-
