Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 11 11月, 2022 4 次提交
    • O
      !231 ROH: Support ROH basic functions and adapt ROH mode for RDMA/hns driver · 1789f7b7
      由 openeuler-ci-bot 提交于 
      Merge Pull Request from: @chenke1978 
 
[Description]
The ROH module driver consists of the ROH Core and ROH DRV
modules, which work with hardware to implement communication
between nodes through HCCS packets.
ROH Core is a protocol stack of the ROH architecture. It provides
related services for upper layers by invoking operation interfaces
provided by the ROH DRV.
The ROH DRV implements the lower layer functions of the ROH
featureand provides a series of interfaces for operating hardware
for the ROH Core.

This patch supports basic ROH functions, such as: sysfs file node
query, abnormal interrupt handling, reset capability and dfx
information query.

RDMA/hns supports ROH mode, mainly adapted to the device id of ROH,
and the different capabilities and features of RDMA/hns in ROH mode.

[Testing]
kernel options:
CONFIG_ROH=m
CONFIG_ROH_HNS=m

Test passed with below step:
1. Using a hardware environment that supports ROH, insmod net/hns,
 RDMA/hns and ROH related drivers:
 insmod hnae3.ko
 insmod hclge.ko
 insmod hns3.ko
 insmod roh_core.ko
 insmod hns-roh-v1.ko
 insmod hns-roce-hw-v2.ko
2. Check whether ROH generates the corresponding sysfs node:
 ls /sys/class/roh/hns3_0/
3. Check whether the abnormal interrupt information of roh is correct.
 The down or up of the network device link corresponding to roh will
 cause the roh abnormal interrupt count to increase.
 cat /proc/interrupts | grep roh
4. Configure the network port ip and check whether the ip/mac has the
 correct mapping relationship.
5. Query eid to check whether it complies with ip conversion rules:
 cat /sys/class/roh/hns3_0/node_eid
6. Use ping to check Ethernet communication is normal.
7. Use perftest to check RDMA communication is normal.
8. Use the "ethtool --reset eth1 all" command to trigger a reset. 
 
Link:https://gitee.com/openeuler/kernel/pulls/231 
Reviewed-by: Ling Mingqiang <lingmingqiang@huawei.com> 
Reviewed-by: Xie XiuQi <xiexiuqi@huawei.com> 
Reviewed-by: Zheng Zengkai <zhengzengkai@huawei.com> 
Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com>
      1789f7b7
    • O
      !217 AMD: Support HSMP feature for AMD EPYC platforms · e29772f7
      由 openeuler-ci-bot 提交于 
      Merge Pull Request from: @haochengxie 
 
HSMP is short of Host System Management Port. This is a feature of recent AMD EPYC platforms.The Host System Management Port (HSMP) is an interface to provide OS-level software with access to system management functions via a set of mailbox registers. 
 
Link:https://gitee.com/openeuler/kernel/pulls/217 
Reviewed-by: Chen Wei <chenwei@xfusion.com> 
Reviewed-by: Zheng Zengkai <zhengzengkai@huawei.com> 
Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com>
      e29772f7
    • O
      !226 Intel SPR: Adding core PMU support for OLK-5.10 · 50c172c5
      由 openeuler-ci-bot 提交于 
      Merge Pull Request from: @allen-shi 
 
This is a cherry-pick of [PR62](https://gitee.com/openeuler/kernel/pulls/62) from openEuler-22.09 branch.


This patchset is to enable Intel SPR core PMU support.
The first 5 patches enable Intel SPR core PMU support.
The last 3 patches fix bugs introduced by the above 5 patches.

 **Intel Kernel Issue** 
[#I596BF](https://gitee.com/openeuler/intel-kernel/issues/I596BF)

 **Test:** 
With this patch set:

```
        \# perf stat -e L1-dcache-load-misses -- sleep 1
        
         Performance counter stats for 'sleep 1':
        
                    10,134      L1-dcache-load-misses
        
               1.000727264 seconds time elapsed
        
               0.000584000 seconds user
               0.000000000 seconds sys
```

Without this patch set, it returns event "not supported":

```
        # perf stat -e L1-dcache-load-misses -- sleep 1
        
         Performance counter stats for 'sleep 1':
        
           <not supported>      L1-dcache-load-misses
        
               1.001434664 seconds time elapsed
        
               0.000000000 seconds user
               0.001318000 seconds sys
```

Some other tests:

```
# perf mem record -- sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.030 MB perf.data (24 samples) ]
[root@spr4s01 linux]# perf script --fields comm,weight,ip,sym
       perf-exec                0 ffffffffaf80cb0b __intel_pmu_enable_all.constprop.0
       perf-exec              489 ffffffffafdfa612 acpi_os_read_memory
       perf-exec              359 ffffffffafa67daa perf_iterate_ctx
       perf-exec              435 ffffffffafa67daa perf_iterate_ctx
       perf-exec              123 ffffffffafa6d0eb perf_event_addr_filters_exec
       perf-exec              168 ffffffffafa67daa perf_iterate_ctx
           sleep              163 ffffffffafa67d72 perf_iterate_ctx
           sleep              174 ffffffffafa67d72 perf_iterate_ctx
           sleep               83 ffffffffafa67d92 perf_iterate_ctx
           sleep              159 ffffffffafa67d7c perf_iterate_ctx
           sleep              133 ffffffffafa67d7c perf_iterate_ctx
           sleep                0 ffffffffafa886aa filemap_map_pages
           sleep                0 ffffffffaf80cb0b __intel_pmu_enable_all.constprop.0
           sleep              184 ffffffffafa67d92 perf_iterate_ctx
           sleep                0 ffffffffb025029e sync_regs
           sleep                0 ffffffffafa886aa filemap_map_pages
           sleep                0     7ffb7f2b7ba0 [unknown]
           sleep                0     7ffb7f2b7cd1 [unknown]
           sleep              150 ffffffffafa67d92 perf_iterate_ctx
           sleep                0 ffffffffafa886aa filemap_map_pages
           sleep                0 ffffffffafae8a37 page_remove_file_rmap
           sleep                0 ffffffffafad1c2f zap_pte_range
           sleep                0 ffffffffafad1c2f zap_pte_range
           sleep                0 ffffffffafaec00c unlink_anon_vmas

# perf record -e cpu/event=0xcd,umask=0x1/ -e cpu/event=0x03,umask=0x82/ -- cat /proc/cpuinfo
    ……
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.029 MB perf.data (175 samples) ]
# perf script|head -5
             cat 254544 10879.930447:          1 cpu/event=0x03,umask=0x82/:  ffffffffa7653768 memcpy_erms+0x8 (vmlinux)
             cat 254544 10879.930511:          1 cpu/event=0x03,umask=0x82/:  ffffffffa6cda323 vm_area_dup+0x63 (vmlinux)
             cat 254544 10879.930513:          2 cpu/event=0x03,umask=0x82/:  ffffffffa6edf233 __split_vma+0xa3 (vmlinux)
             cat 254544 10879.930564:         32 cpu/event=0x03,umask=0x82/:  ffffffffa6e8b2a7 filemap_fault+0x2d7 (vmlinux)
             cat 254544 10879.930656:         48 cpu/event=0x03,umask=0x82/:  ffffffffa76504de sync_regs+0x1e (vmlinux)

        
# perf record -e instructions:ppp -a -- true
[ perf record: Woken up 0 times to write data ]
[ perf record: Captured and wrote 0.775 MB perf.data (3142 samples) ]
# perf script|head -5
            perf 254537 [000] 10812.672089:          1 instructions:ppp:  ffffffffa764f883 do_syscall_64+0x33 (vmlinux)
            perf 254537 [000] 10812.672101:          1 instructions:ppp:  ffffffffa6c262cd nmi_handle+0x6d (vmlinux)
            perf 254537 [000] 10812.672103:          4 instructions:ppp:  ffffffffa6c2bcb5 native_sched_clock+0x5 (vmlinux)
            perf 254537 [000] 10812.672104:         74 instructions:ppp:  ffffffffa6c2bcb5 native_sched_clock+0x5 (vmlinux)
            perf 254537 [000] 10812.672105:       1779 instructions:ppp:  ffffffffa6c2bcb5 native_sched_clock+0x5 (vmlinux)

# perf record -d --data-page-size -e major-faults -a -- sleep 1
Error:
Asking for the data page size isnt supported by this kernel.

# perf record -d -e major-faults -a -- sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 1.497 MB perf.data (3 samples) ]
# perf script
            perf 25239 [447]   660.473507:          1 major-faults:            5c6d74 perf_mmap__consume+0x24 (/usr/bin/perf)
            perf 25239 [447]   660.473876:          1 major-faults:            5c6d74 perf_mmap__consume+0x24 (/usr/bin/perf)
            perf 25239 [447]   660.474401:          1 major-faults:            5c6d74 perf_mmap__consume+0x24 (/usr/bin/perf)
# perf script --fields comm,addr
            perf     7fe0459a3408
            perf     7fe03ec0d408
            perf     7fe0333de408
```

 **Known Issue:** 
N/A

 **Default Config Change:** 
N/A 
 
Link:https://gitee.com/openeuler/kernel/pulls/226 
Reviewed-by: Chen Wei <chenwei@xfusion.com> 
Reviewed-by: Zheng Zengkai <zhengzengkai@huawei.com> 
Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com>
      50c172c5
    • O
      !230 Intel SPR: Add Sapphire Rapids server intel-uncore-freq support for OLK-5.10 · 188b2ee0
      由 openeuler-ci-bot 提交于 
      Merge Pull Request from: @allen-shi 
 
This is a cherry-pick of [PR125](https://gitee.com/openeuler/kernel/pulls/125) from openEuler-22.09 branch.

Sapphire Rapids uncore frequency control is the same as Skylake and
Ice Lake. Add the Sapphire Rapids CPU model number to the match array.

backport upstream patch without change.
 
 
Link:https://gitee.com/openeuler/kernel/pulls/230 
Reviewed-by: Chen Wei <chenwei@xfusion.com> 
Reviewed-by: Zheng Zengkai <zhengzengkai@huawei.com> 
Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com>
      188b2ee0
  3. 10 11月, 2022 36 次提交