On Tue, Mar 26, 2019 at 07:43:17PM +0800, Dongdong Liu wrote: > Current we met another deadlock issue in hotplug driver. The calltrace is as below. > The deadlock triggered by a hotplug event during a sysfs "remove" operation. > Any suggestion to fix such deadlock ? That's a known problem, deadlocks may occur if hotplug ports are cascaded. I came up with a kludge to work around it but withdrew the patch: https://patchwork.ozlabs.org/patch/930403/ The real solution is to make the sections protected by pci_lock_rescan_remove() smaller or eliminate them as far as possible. So, no good solution available right now, sorry. Thanks, Lukas > > [ 4112.297250] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. > [ 4112.305069] bash D 0 6502 2207 0x00000200 > [ 4112.310544] Call trace: > [ 4112.312981] __switch_to+0x94/0xe8 > [ 4112.316373] __schedule+0x270/0x8b0 > [ 4112.319852] schedule+0x2c/0x88 > [ 4112.322981] schedule_timeout+0x224/0x448 > [ 4112.326979] wait_for_common+0x198/0x2a0 > [ 4112.330892] wait_for_completion+0x28/0x38 > [ 4112.334979] kthread_stop+0x60/0x190 > [ 4112.338544] __free_irq+0x1c0/0x348 > [ 4112.342022] free_irq+0x40/0x88 > [ 4112.345153] pcie_shutdown_notification+0x54/0x80 > [ 4112.349847] pciehp_remove+0x30/0x50 > [ 4112.353413] pcie_port_remove_service+0x3c/0x58 > [ 4112.357932] device_release_driver_internal+0x1b4/0x250 > [ 4112.363146] device_release_driver+0x28/0x38 > [ 4112.367406] bus_remove_device+0xd4/0x160 > [ 4112.371405] device_del+0x128/0x348 > [ 4112.374880] device_unregister+0x24/0x78 > [ 4112.378792] remove_iter+0x48/0x58 > [ 4112.382183] device_for_each_child+0x6c/0xb8 > [ 4112.386443] pcie_port_device_remove+0x2c/0x48 > [ 4112.390876] pcie_portdrv_remove+0x5c/0x68 > [ 4112.394963] pci_device_remove+0x48/0xd8 > [ 4112.398874] device_release_driver_internal+0x1b4/0x250 > [ 4112.404088] device_release_driver+0x28/0x38 > [ 4112.408348] pci_stop_bus_device+0x84/0xb8 > [ 4112.412434] pci_stop_and_remove_bus_device_locked+0x24/0x40 > [ 4112.418083] remove_store+0xa4/0xb8 > [ 4112.421560] dev_attr_store+0x44/0x60 > [ 4112.425213] sysfs_kf_write+0x58/0x80 > [ 4112.428864] kernfs_fop_write+0xe8/0x1f0 > [ 4112.432776] __vfs_write+0x60/0x190 > [ 4112.436255] vfs_write+0xac/0x1c0 > [ 4112.439560] ksys_write+0x6c/0xd8 > [ 4112.442861] __arm64_sys_write+0x24/0x30 > [ 4112.446773] el0_svc_common+0xa0/0x180 > [ 4112.450511] el0_svc_handler+0x38/0x78 > [ 4112.454249] el0_svc+0x8/0xc > [ 4112.457122] INFO: task irq/97-pciehp:17365 blocked for more than 120 seconds. > [ 4112.464248] Tainted: P W OE 4.19.25-vhulk1901.1.0.h111.aarch64+ #2 > [ 4112.471980] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. > [ 4112.479798] irq/97-pciehp D 0 17365 2 0x00000228 > [ 4112.485273] Call trace: > [ 4112.487710] __switch_to+0x94/0xe8 > [ 4112.491098] __schedule+0x270/0x8b0 > [ 4112.494575] schedule+0x2c/0x88 > [ 4112.497706] schedule_preempt_disabled+0x14/0x20 > [ 4112.502313] __mutex_lock.isra.1+0x1fc/0x540 > [ 4112.506572] __mutex_lock_slowpath+0x24/0x30 > [ 4112.510833] mutex_lock+0x80/0xa8 > [ 4112.514138] pci_lock_rescan_remove+0x20/0x28 > [ 4112.518485] pciehp_configure_device+0x30/0x140 > [ 4112.523005] pciehp_handle_presence_or_link_change+0x35c/0x4b0 > [ 4112.528826] pciehp_ist+0x1cc/0x1d0 > [ 4112.532305] irq_thread_fn+0x30/0x80 > [ 4112.535870] irq_thread+0x128/0x200 > [ 4112.539349] kthread+0x134/0x138 > [ 4112.542563] ret_from_fork+0x10/0x18
