[roce-1126]RDMA/hns: bugfix for slab-out-of-bounds when loading hip08 driver

[Fred Kimmy]

"[Bug Description]
KASAN: slab-out-of-bounds in hns_roce_table_mhop_put+0x584/0x828
    [hns_roce]
    Read of size 8 at addr ffff802185e08300 by task rmmod/270

    Call trace:
    dump_backtrace+0x0/0x1e8
    show_stack+0x14/0x20
    dump_stack+0xc4/0xfc
    print_address_description+0x60/0x270
    __kasan_report+0x164/0x1b8
    kasan_report+0xc/0x18
    __asan_load8+0x84/0xa8
    hns_roce_table_mhop_put+0x584/0x828 [hns_roce]
    hns_roce_table_put+0x174/0x1a0 [hns_roce]
    hns_roce_mr_free+0x124/0x210 [hns_roce]
    hns_roce_dereg_mr+0x90/0xb8 [hns_roce]
    ib_dealloc_pd_user+0x60/0xf0
    ib_mad_port_close+0x128/0x1d8
    ib_mad_remove_device+0x94/0x118
    remove_client_context+0xa0/0xe0
    disable_device+0xfc/0x1c0
    __ib_unregister_device+0x60/0xe0
    ib_unregister_device+0x24/0x38
    hns_roce_exit+0x3c/0x138 [hns_roce]
    __hns_roce_hw_v2_uninit_instance.isra.30+0x28/0x50 [hns_roce_hw_v2]
    hns_roce_hw_v2_uninit_instance+0x44/0x60 [hns_roce_hw_v2]
    hclge_uninit_client_instance+0x15c/0x238 [hclge]
    hnae3_uninit_client_instance+0x84/0xa8 [hnae3]
    hnae3_unregister_client+0x84/0x158 [hnae3]
    hns_roce_hw_v2_exit+0x14/0x20 [hns_roce_hw_v2]
    __arm64_sys_delete_module+0x20c/0x308
    el0_svc_handler+0xbc/0x210
    el0_svc+0x8/0xc

    Allocated by task 255:
    __kasan_kmalloc.isra.0+0xd0/0x180
    kasan_kmalloc+0xc/0x18
    __kmalloc+0x16c/0x328
    hns_roce_init_hem_table+0x20c/0x428 [hns_roce]
    hns_roce_init+0x214/0xfe0 [hns_roce]
    __hns_roce_hw_v2_init_instance+0x284/0x330 [hns_roce_hw_v2]
    hns_roce_hw_v2_init_instance+0xd0/0x1b8 [hns_roce_hw_v2]
    hclge_init_roce_client_instance+0x180/0x310 [hclge]
    hclge_init_client_instance+0xcc/0x508 [hclge]
    hnae3_init_client_instance.part.3+0x3c/0x80 [hnae3]
    hnae3_register_client+0x134/0x1a8 [hnae3]
    0xffff200009c00014
    do_one_initcall+0x9c/0x3e0
    do_init_module+0xd4/0x2d8
    load_module+0x3284/0x3690
    __se_sys_init_module+0x274/0x308
    __arm64_sys_init_module+0x40/0x50
    el0_svc_handler+0xbc/0x210
    el0_svc+0x8/0xc

    Freed by task 0:
    (stack is not available)

    The buggy address belongs to the object at ffff802185e06300
    which belongs to the cache kmalloc-8k of size 8192
    The buggy address is located 0 bytes to the right of
    8192-byte region [ffff802185e06300, ffff802185e08300)
    The buggy address belongs to the page:
    page:ffff7fe008617800 refcount:1 mapcount:0 mapping:ffff802340020e00 index:0x0
    compound_mapcount: 0
    flags: 0x5fffe00000010200(slab|head)
    raw: 5fffe00000010200 dead000000000100 dead000000000200 ffff802340020e00
    raw: 0000000000000000 00000000803e003e 00000001ffffffff 0000000000000000
    page dumped because: kasan: bad access detected
     Memory state around the buggy address:
    ffff802185e08200: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    ffff802185e08280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    >ffff802185e08300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
    ^
    ffff802185e08380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
    ffff802185e08400: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    ==================================================================
    Disabling lock debugging due to kernel taint

[Steps to Reproduce]
Enable KASAN and configure PAGE_SIZE to 64K, insmod hns roce driver and then rmmod it.

[Actual Results]
Call trace because of slab-out-of-bound.

[Expected Results]
Success

[Reproducibility]
Inevitably

[Additional information]
Hardware: D06 CS
Firmware: NA
Kernel: NA

[Resolution]
Not configure eq->next when number of eq_buf is 1 in eq_mhop_alloc()."

RDMA/hns: bugfix for slab-out-of-bounds when loading hip08 driver
RDMA/hns: Bugfix for slab-out-of-bounds when unloading hip08 driver