Bug 197071 - ACPI object counter overrun on frequent nvme resets
Summary: ACPI object counter overrun on frequent nvme resets
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: ACPI
Classification: Unclassified
Component: ACPICA-Core (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Robert Moore
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-09-29 04:39 UTC by Lv Zheng
Modified: 2018-01-22 07:59 UTC (History)
2 users (show)

See Also:
Kernel Version: 4.13
Subsystem:
Regression: No
Bisected commit-id:


Attachments

Description Lv Zheng 2017-09-29 04:39:16 UTC
From: Christoph Hellwig <hch@infradead.org>

Hi all,

when doing an error injection test with the nvme-pci driver on Linux 4.14
almost rc2 (820bf5c419e4b85298e5c3001bd1b5be46d60765 plus a few nvme
patches) that does a lot of nvme resets which shut down and re-enable the
PCI device I see a flood of warnings like:

[   78.237286] ACPI Warning: Large Reference Count (0x1001) in object ffff88007d702c68, Type=0x1D (20170728/utdelete-473)

Throwing in a dump_stack into the path it comes from:

[   78.238387] CPU: 0 PID: 286 Comm: kworker/u8:3 Not tainted 4.14.0-rc1+ #134
[   78.238998] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
[   78.239720] Workqueue: nvme-wq nvme_reset_work
[   78.239995] Call Trace:
[   78.239995]  dump_stack+0x63/0x83
[   78.239995]  acpi_ut_update_ref_count+0x2f7/0x2fe
[   78.239995]  acpi_ut_update_object_reference+0x114/0x182
[   78.239995]  acpi_ut_add_reference+0x1b/0x1e
[   78.239995]  acpi_ds_init_package_element+0x149/0x16e
[   78.239995]  ? acpi_ds_build_internal_object+0xca/0x12c
[   78.239995]  acpi_ds_build_internal_package_obj+0x1b2/0x258
[   78.239995]  ? acpi_ds_build_internal_package_obj+0x1b2/0x258
[   78.239995]  acpi_ds_eval_data_object_operands+0xd3/0x123
[   78.239995]  acpi_ds_exec_end_op+0x288/0x3e1
[   78.239995]  acpi_ps_parse_loop+0x519/0x57b
[   78.239995]  acpi_ps_parse_aml+0x93/0x29c
[   78.239995]  acpi_ps_execute_method+0x148/0x17f
[   78.239995]  acpi_ns_evaluate+0x1c1/0x24d
[   78.239995]  acpi_ut_evaluate_object+0x72/0x197
[   78.239995]  acpi_rs_get_prt_method_data+0x21/0x45
[   78.239995]  acpi_get_irq_routing_table+0x2c/0x30
[   78.239995]  acpi_pci_irq_find_prt_entry+0x8e/0x2c0
[   78.239995]  ? irq_get_irq_data+0x9/0x20
[   78.239995]  ? mp_unmap_irq+0xf/0x70
[   78.239995]  acpi_pci_irq_lookup+0x26/0x1a0
[   78.239995]  acpi_pci_irq_enable+0x5b/0x1b0
[   78.239995]  ? pci_read_config_word+0x22/0x30
[   78.239995]  pcibios_enable_device+0x28/0x30
[   78.239995]  do_pci_enable_device+0x5f/0xe0
[   78.239995]  pci_enable_device_flags+0xc3/0x110
[   78.239995]  pci_enable_device_mem+0xe/0x10
[   78.239995]  nvme_reset_work+0x4c/0x1610
[   78.239995]  ? sched_clock+0x9/0x10
[   78.239995]  ? sched_clock_local+0x17/0x90
[   78.239995]  ? _raw_spin_lock+0x9/0x10
[   78.239995]  ? pick_next_task_fair+0x420/0x690
[   78.239995]  ? _raw_spin_unlock_irq+0x9/0x20
[   78.239995]  ? finish_task_switch+0x7b/0x1f0
[   78.239995]  ? __schedule+0x2d7/0x800
[   78.239995]  process_one_work+0x1db/0x3e0
[   78.239995]  worker_thread+0x43/0x3f0
[   78.239995]  kthread+0x103/0x140
[   78.239995]  ? process_one_work+0x3e0/0x3e0
[   78.239995]  ? kthread_create_on_node+0x40/0x40
[   78.239995]  ret_from_fork+0x25/0x30
Comment 1 Robert Moore 2017-09-29 14:03:38 UTC
Please attach the acpidump for the machine, thanks.
Comment 2 Lv Zheng 2017-10-11 07:08:40 UTC
Please also re-configure the kernel with CONFIG_ACPI_DEBUG=y.
And provide a dmesg output by booting the new kernel with "acpi.trace_state=enable".

Thanks in advance.
Comment 3 Lv Zheng 2017-10-11 07:08:54 UTC
Please upload acpidump here.
Comment 4 Zhang Rui 2017-12-18 03:13:10 UTC
Christoph Hellwig <hch@infradead.org>, the original bug reporter is not in this thread, so I don't think he can get these messages.
I've sent an email to Hellwig...
Comment 5 Zhang Rui 2018-01-22 07:59:36 UTC
I didn't get response from Hellwig.
Bug closed.
Bob, if you think this reveals an ACPICA issue, please create an ACPICA bug report instead. :)

Note You need to log in before you can comment on or make changes to this bug.