Date:   Tue, 9 May 2023 16:49:23 +0800
From:   Zhu Yanjun <yanjun.zhu@...ux.dev>
To:     "leon@...nel.org" <leon@...nel.org>,
        Jason Gunthorpe <jgg@...pe.ca>,
        RDMA mailing list <linux-rdma@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        akpm@...ux-foundation.org
Subject: Re: [PATCH 1/1] kernel/fork: Fix the error "raw_local_irq_restore()

There is something wrong with this commit. Please ignore it.

Zhu Yanjun

在 2023/5/8 16:57, Zhu Yanjun 写道:
> The following can fix this problem. Not sure if this will introduce any 
> risks.
> 
> Subject: [PATCH 1/1] kernel/fork: Fix the error "raw_local_irq_restore()
>   called with IRQs enabled"
> 
> The call trace are as below.
> "
> [ 6020.092175] Call Trace:
> [ 6020.092178]  <TASK>
> [ 6020.092181]  _raw_spin_unlock_irqrestore+0x67/0x70
> [ 6020.092190]  ib_send_mad+0x3fa/0x7d0 [ib_core]
> [ 6020.092299]  ? __pfx_ib_send_mad+0x10/0x10 [ib_core]
> [ 6020.092429]  ib_post_send_mad+0x338/0x4c0 [ib_core]
> [ 6020.092519]  ib_send_cm_rep+0x4b0/0x790 [ib_cm]
> [ 6020.092553]  rdma_accept+0x432/0x550 [rdma_cm]
> [ 6020.092585]  ? __pfx_rdma_accept+0x10/0x10 [rdma_cm]
> [ 6020.092613]  ? do_raw_spin_unlock+0x97/0x130
> [ 6020.092623]  ? preempt_count_sub+0x14/0xc0
> [ 6020.092632]  ? _raw_spin_unlock+0x29/0x50
> [ 6020.092637]  ? ucma_get_ctx+0xab/0x180 [rdma_ucm]
> [ 6020.092650]  ? check_stack_object+0x22/0x70
> [ 6020.092659]  ucma_accept+0x1e7/0x2e0 [rdma_ucm]
> [ 6020.092671]  ? __pfx_ucma_accept+0x10/0x10 [rdma_ucm]
> [ 6020.092683]  ? 
> __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
> [ 6020.092720]  ? _find_next_bit+0x37/0xc0
> [ 6020.092741]  ucma_write+0x139/0x1f0 [rdma_ucm]
> [ 6020.092752]  ? __pfx_ucma_write+0x10/0x10 [rdma_ucm]
> [ 6020.092764]  ? security_file_permission+0x4b/0x2b0
> [ 6020.092774]  vfs_write+0x161/0x740
> [ 6020.092781]  ? ib_uverbs_ioctl+0x17f/0x210 [ib_uverbs]
> [ 6020.092815]  ? __pfx_vfs_write+0x10/0x10
> [ 6020.092822]  ? __fget_files+0x158/0x230
> [ 6020.092831]  ? __fget_light+0xe2/0x100
> [ 6020.092838]  ksys_write+0x13f/0x160
> [ 6020.092843]  ? __pfx_ksys_write+0x10/0x10
> [ 6020.092847]  ? ktime_get_coarse_real_ts64+0x107/0x120
> [ 6020.092854]  ? ktime_get_coarse_real_ts64+0xaf/0x120
> [ 6020.092863]  do_syscall_64+0x37/0x90
> [ 6020.092870]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
> [ 6020.092877] RIP: 0033:0x7f2151e11847
> [ 6020.092882] Code: c3 66 90 41 54 49 89 d4 55 48 89 f5 53 89 fb 48 83 
> ec 10 e8 1b fd ff ff 4c 89 e2 48 89 ee 89 df 41 89 c0 b8 01 00 00 00 0f 
> 05 <48> 3d 00 f0 ff ff 77 35 44 89 c7 48 89 44 24 08 e8 54 fd ff ff 48
> [ 6020.092886] RSP: 002b:00007ffc2439a0f0 EFLAGS: 00000293 ORIG_RAX: 
> 0000000000000001
> [ 6020.092892] RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 
> 00007f2151e11847
> [ 6020.092896] RDX: 0000000000000130 RSI: 00007ffc2439a120 RDI: 
> 0000000000000003
> [ 6020.092899] RBP: 00007ffc2439a120 R08: 0000000000000000 R09: 
> 0000000000000000
> [ 6020.092902] R10: 00000000ffffffff R11: 0000000000000293 R12: 
> 0000000000000130
> [ 6020.092905] R13: 0000000000000011 R14: 0000000000000000 R15: 
> 000055c0ea6ffa18
> [ 6020.092914]  </TASK>
> [ 6020.092917] irq event stamp: 0
> [ 6020.092919] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
> [ 6020.092923] hardirqs last disabled at (0): [<ffffffffad94dae6>] 
> copy_process+0x13e6/0x3720
> [ 6020.092931] softirqs last  enabled at (0): [<ffffffffad94dae6>] 
> copy_process+0x13e6/0x3720
> [ 6020.092936] softirqs last disabled at (0): [<0000000000000000>] 0x0
> [ 6020.092939] ---[ end trace 0000000000000000 ]---
> 
> "
> 
> Signed-off-by: Zhu Yanjun <yanjun.zhu@...ux.dev>
> ---
>   kernel/fork.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/fork.c b/kernel/fork.c
> index ed4e01daccaa..802e11e55494 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -2250,6 +2250,7 @@ __latent_entropy struct task_struct *copy_process(
>                                          struct kernel_clone_args *args)
>   {
>          int pidfd = -1, retval;
> +       unsigned long flags;
>          struct task_struct *p;
>          struct multiprocess_signals delayed;
>          struct file *pidfile = NULL;
> @@ -2320,11 +2321,11 @@ __latent_entropy struct task_struct *copy_process(
>          sigemptyset(&delayed.signal);
>          INIT_HLIST_NODE(&delayed.node);
> 
> -       spin_lock_irq(&current->sighand->siglock);
> +       spin_lock_irqsave(&current->sighand->siglock, flags);
>          if (!(clone_flags & CLONE_THREAD))
>                  hlist_add_head(&delayed.node, 
> &current->signal->multiprocess);
>          recalc_sigpending();
> -       spin_unlock_irq(&current->sighand->siglock);
> +       spin_unlock_irqrestore(&current->sighand->siglock, flags);
>          retval = -ERESTARTNOINTR;
>          if (task_sigpending(current))
>                  goto fork_out;
> -- 
> 2.27.0
> 
> 
> 在 2023/5/8 16:53, Zhu Yanjun 写道:
>> Hi, all
>>
>> I confronted the following problem. The call traces are as below.
>>
>> kernel/fork.c:
>>
>> "
>>
>> 2323 spin_lock_irq(&current->sighand->siglock); <---disable
>> 2324         if (!(clone_flags & CLONE_THREAD))
>> 2325                 hlist_add_head(&delayed.node, 
>> &current->signal->multiprocess);
>> 2326         recalc_sigpending();
>> 2327 spin_unlock_irq(&current->sighand->siglock); <--enable
>> "
>>
>> conflicts with the following
>>
>> "
>>
>> 1031         spin_lock_irqsave(&qp_info->send_queue.lock, flags); 
>> <---save
>> 1032         if (qp_info->send_queue.count < 
>> qp_info->send_queue.max_active) {
>> 1033                 trace_ib_mad_ib_send_mad(mad_send_wr, qp_info);
>> 1034                 ret = ib_post_send(mad_agent->qp, 
>> &mad_send_wr->send_wr.wr,
>> 1035                                    NULL);
>> 1036                 list = &qp_info->send_queue.list;
>> 1037         } else {
>> 1038                 ret = 0;
>> 1039                 list = &qp_info->overflow_list;
>> 1040         }
>> 1041
>> 1042         if (!ret) {
>> 1043                 qp_info->send_queue.count++;
>> 1044 list_add_tail(&mad_send_wr->mad_list.list, list);
>> 1045         }
>> 1046 spin_unlock_irqrestore(&qp_info->send_queue.lock, flags); <--restore
>>
>> "
>>
>> If we change spin_lock_irq/spin_unlock_irq to 
>> spin_lock_irqsave/spin_unlock_irqrestore, it seems that this problem 
>> is resolved.
>>
>> But not sure if other problems are introduced.
>>
>> [ 6020.091840] ------------[ cut here ]------------
>> [ 6020.091846] raw_local_irq_restore() called with IRQs enabled
>> [ 6020.091857] WARNING: CPU: 82 PID: 10603 at 
>> kernel/locking/irqflag-debug.c:10 warn_bogus_irq_restore+0x21/0x30
>> [ 6020.091872] Modules linked in: rdma_rxe ip6_udp_tunnel udp_tunnel 
>> xt_CHECKSUM xt_MASQUERADE xt_conntrack ipt_REJECT nf_reject_ipv4 
>> nft_compat nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6 
>> nf_defrag_ipv4 nf_tables nfnetlink tun bridge stp llc vfat fat ext4 
>> mbcache jbd2 rpcrdma sunrpc intel_rapl_msr intel_rapl_common rdma_ucm 
>> i10nm_edac ib_srpt nfit ib_isert libnvdimm iscsi_target_mod 
>> target_core_mod x86_pkg_temp_thermal ib_iser intel_powerclamp libiscsi 
>> coretemp scsi_transport_iscsi ib_umad rdma_cm ib_ipoib iw_cm kvm_intel 
>> ib_cm ipmi_ssif kvm mlx5_ib i40e irqbypass crct10dif_pclmul 
>> crc32_pclmul ghash_clmulni_intel ib_uverbs sha512_ssse3 iTCO_wdt rapl 
>> iTCO_vendor_support mei_me intel_cstate ib_core joydev acpi_ipmi 
>> intel_uncore mei ipmi_si pcspkr i2c_i801 dax_hmem isst_if_mmio 
>> isst_if_mbox_pci wmi_bmof isst_if_common intel_pch_thermal i2c_smbus 
>> ipmi_devintf ipmi_msghandler acpi_power_meter ip_tables xfs libcrc32c 
>> sd_mod t10_pi crc64_rocksoft crc64 sg mgag200 mlx5_core drm_kms_helper 
>> syscopyarea mlxfw sysfillrect
>> [ 6020.092063]  sysimgblt tls i2c_algo_bit ahci drm_shmem_helper 
>> libahci ice pci_hyperv_intf crc32c_intel drm psample megaraid_sas 
>> libata tg3 wmi dm_mirror dm_region_hash dm_log dm_mod fuse [last 
>> unloaded: irdma]
>> [ 6020.092108] CPU: 82 PID: 10603 Comm: rping Kdump: loaded Tainted: G 
>> S  B              6.4.0-rc1+ #30
>> [ 6020.092115] Hardware name: Dell Inc. PowerEdge R750/06V45N, BIOS 
>> 1.2.4 05/28/2021
>> [ 6020.092119] RIP: 0010:warn_bogus_irq_restore+0x21/0x30
>> [ 6020.092126] Code: 90 90 90 90 90 90 90 90 90 80 3d 35 50 79 01 00 
>> 74 05 c3 cc cc cc cc 48 c7 c7 a0 d7 e9 ae c6 05 20 50 79 01 01 e8 1f 
>> 15 da fe <0f> 0b c3 cc cc cc cc cc cc cc cc cc cc cc cc 90 90 90 90 90 
>> 90 90
>> [ 6020.092131] RSP: 0018:ff11000180e477e8 EFLAGS: 00010286
>> [ 6020.092137] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 
>> 0000000000000000
>> [ 6020.092141] RDX: 0000000000000004 RSI: dffffc0000000000 RDI: 
>> ffffffffb27b31c0
>> [ 6020.092144] RBP: ff110001175d69e8 R08: ffe21c02006d7f6a R09: 
>> ffe21c02006d7f6a
>> [ 6020.092148] R10: ff110010036bfb4b R11: ffe21c02006d7f69 R12: 
>> ff110001175d6a28
>> [ 6020.092152] R13: ff110001175d6a30 R14: ff110001175d69e8 R15: 
>> 0000000000000000
>> [ 6020.092155] FS:  00007f215281c740(0000) GS:ff11001003680000(0000) 
>> knlGS:0000000000000000
>> [ 6020.092160] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> [ 6020.092163] CR2: 00007f2147fffda0 CR3: 00000001a1b2c005 CR4: 
>> 0000000000771ee0
>> [ 6020.092167] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 
>> 0000000000000000
>> [ 6020.092170] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 
>> 0000000000000400
>> [ 6020.092173] PKRU: 55555554
>> [ 6020.092175] Call Trace:
>> [ 6020.092178]  <TASK>
>> [ 6020.092181]  _raw_spin_unlock_irqrestore+0x67/0x70
>> [ 6020.092190]  ib_send_mad+0x3fa/0x7d0 [ib_core]
>> [ 6020.092299]  ? __pfx_ib_send_mad+0x10/0x10 [ib_core]
>> [ 6020.092429]  ib_post_send_mad+0x338/0x4c0 [ib_core]
>> [ 6020.092519]  ib_send_cm_rep+0x4b0/0x790 [ib_cm]
>> [ 6020.092553]  rdma_accept+0x432/0x550 [rdma_cm]
>> [ 6020.092585]  ? __pfx_rdma_accept+0x10/0x10 [rdma_cm]
>> [ 6020.092613]  ? do_raw_spin_unlock+0x97/0x130
>> [ 6020.092623]  ? preempt_count_sub+0x14/0xc0
>> [ 6020.092632]  ? _raw_spin_unlock+0x29/0x50
>> [ 6020.092637]  ? ucma_get_ctx+0xab/0x180 [rdma_ucm]
>> [ 6020.092650]  ? check_stack_object+0x22/0x70
>> [ 6020.092659]  ucma_accept+0x1e7/0x2e0 [rdma_ucm]
>> [ 6020.092671]  ? __pfx_ucma_accept+0x10/0x10 [rdma_ucm]
>> [ 6020.092683]  ? 
>> __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
>> [ 6020.092720]  ? _find_next_bit+0x37/0xc0
>> [ 6020.092741]  ucma_write+0x139/0x1f0 [rdma_ucm]
>> [ 6020.092752]  ? __pfx_ucma_write+0x10/0x10 [rdma_ucm]
>> [ 6020.092764]  ? security_file_permission+0x4b/0x2b0
>> [ 6020.092774]  vfs_write+0x161/0x740
>> [ 6020.092781]  ? ib_uverbs_ioctl+0x17f/0x210 [ib_uverbs]
>> [ 6020.092815]  ? __pfx_vfs_write+0x10/0x10
>> [ 6020.092822]  ? __fget_files+0x158/0x230
>> [ 6020.092831]  ? __fget_light+0xe2/0x100
>> [ 6020.092838]  ksys_write+0x13f/0x160
>> [ 6020.092843]  ? __pfx_ksys_write+0x10/0x10
>> [ 6020.092847]  ? ktime_get_coarse_real_ts64+0x107/0x120
>> [ 6020.092854]  ? ktime_get_coarse_real_ts64+0xaf/0x120
>> [ 6020.092863]  do_syscall_64+0x37/0x90
>> [ 6020.092870]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
>> [ 6020.092877] RIP: 0033:0x7f2151e11847
>> [ 6020.092882] Code: c3 66 90 41 54 49 89 d4 55 48 89 f5 53 89 fb 48 
>> 83 ec 10 e8 1b fd ff ff 4c 89 e2 48 89 ee 89 df 41 89 c0 b8 01 00 00 
>> 00 0f 05 <48> 3d 00 f0 ff ff 77 35 44 89 c7 48 89 44 24 08 e8 54 fd ff 
>> ff 48
>> [ 6020.092886] RSP: 002b:00007ffc2439a0f0 EFLAGS: 00000293 ORIG_RAX: 
>> 0000000000000001
>> [ 6020.092892] RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 
>> 00007f2151e11847
>> [ 6020.092896] RDX: 0000000000000130 RSI: 00007ffc2439a120 RDI: 
>> 0000000000000003
>> [ 6020.092899] RBP: 00007ffc2439a120 R08: 0000000000000000 R09: 
>> 0000000000000000
>> [ 6020.092902] R10: 00000000ffffffff R11: 0000000000000293 R12: 
>> 0000000000000130
>> [ 6020.092905] R13: 0000000000000011 R14: 0000000000000000 R15: 
>> 000055c0ea6ffa18
>> [ 6020.092914]  </TASK>
>> [ 6020.092917] irq event stamp: 0
>> [ 6020.092919] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
>> [ 6020.092923] hardirqs last disabled at (0): [<ffffffffad94dae6>] 
>> copy_process+0x13e6/0x3720
>> [ 6020.092931] softirqs last  enabled at (0): [<ffffffffad94dae6>] 
>> copy_process+0x13e6/0x3720
>> [ 6020.092936] softirqs last disabled at (0): [<0000000000000000>] 0x0
>> [ 6020.092939] ---[ end trace 0000000000000000 ]---
>>
>> Zhu Yanjun
>>

Powered by blists - more mailing lists