Dwc lockdep fixes rpi 3.16.y #671
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Os wrapper function for spinlock init causes lockdep to show this false positive splat during boot: [ 3.789851] ============================================= [ 3.796603] [ INFO: possible recursive locking detected ] [ 3.803320] 3.16.1+ raspberrypi#5 Not tainted [ 3.808015] --------------------------------------------- [ 3.814730] khubd/18 is trying to acquire lock: [ 3.820537] (&(sl)->rlock){-.-...}, at: [<c0357168>] DWC_SPINLOCK_IRQSAVE+0xc/0x14 [ 3.830932] [ 3.830932] but task is already holding lock: [ 3.839274] (&(sl)->rlock){-.-...}, at: [<c0357168>] DWC_SPINLOCK_IRQSAVE+0xc/0x14 [ 3.849704] [ 3.849704] other info that might help us debug this: [ 3.858826] Possible unsafe locking scenario: [ 3.858826] [ 3.867334] CPU0 [ 3.871052] ---- [ 3.874721] lock(&(sl)->rlock); [ 3.879302] lock(&(sl)->rlock); [ 3.883815] [ 3.883815] *** DEADLOCK *** [ 3.883815] [ 3.892869] May be due to missing lock nesting notation [ 3.892869] [ 3.901736] 4 locks held by khubd/18: [ 3.906438] #0: (&dev->mutex){......}, at: [<c0321da8>] hub_thread+0x98/0x1000 [ 3.916026] raspberrypi#1: (&port_dev->status_lock){+.+.+.}, at: [<c032234c>] hub_thread+0x63c/0x1000 [ 3.926847] raspberrypi#2: (&bus->usb_address0_mutex){+.+.+.}, at: [<c031f164>] hub_port_init+0x5c/0xb24 [ 3.938015] raspberrypi#3: (&(sl)->rlock){-.-...}, at: [<c0357168>] DWC_SPINLOCK_IRQSAVE+0xc/0x14 [ 3.948730] [ 3.948730] stack backtrace: [ 3.955457] CPU: 0 PID: 18 Comm: khubd Not tainted 3.16.1+ raspberrypi#5 [ 3.962541] [<c00137e4>] (unwind_backtrace) from [<c0011530>] (show_stack+0x10/0x14) [ 3.972998] [<c0011530>] (show_stack) from [<c005985c>] (__lock_acquire+0x1420/0x1ae0) [ 3.983910] [<c005985c>] (__lock_acquire) from [<c005a6c8>] (lock_acquire+0x6c/0x8c) [ 3.994908] [<c005a6c8>] (lock_acquire) from [<c04a872c>] (_raw_spin_lock_irqsave+0x50/0x64) [ 4.006756] [<c04a872c>] (_raw_spin_lock_irqsave) from [<c0357168>] (DWC_SPINLOCK_IRQSAVE+0xc/0x14) [ 4.019382] [<c0357168>] (DWC_SPINLOCK_IRQSAVE) from [<c034ba9c>] (dwc_otg_hcd_select_transactions+0x20c/0x368) [ 4.033064] [<c034ba9c>] (dwc_otg_hcd_select_transactions) from [<c034c0f8>] (dwc_otg_hcd_urb_enqueue+0x158/0x1ec) [ 4.047017] [<c034c0f8>] (dwc_otg_hcd_urb_enqueue) from [<c034d8a0>] (dwc_otg_urb_enqueue+0x1a8/0x2e0) [ 4.059889] [<c034d8a0>] (dwc_otg_urb_enqueue) from [<c03258e4>] (usb_hcd_submit_urb+0xb8/0x870) [ 4.072316] [<c03258e4>] (usb_hcd_submit_urb) from [<c03276a0>] (usb_start_wait_urb+0x44/0xbc) [ 4.084786] [<c03276a0>] (usb_start_wait_urb) from [<c03277cc>] (usb_control_msg+0xb4/0xec) [ 4.097045] [<c03277cc>] (usb_control_msg) from [<c031f528>] (hub_port_init+0x420/0xb24) [ 4.109018] [<c031f528>] (hub_port_init) from [<c0322360>] (hub_thread+0x650/0x1000) [ 4.120667] [<c0322360>] (hub_thread) from [<c003ec3c>] (kthread+0xc8/0xe4) [ 4.129668] [<c003ec3c>] (kthread) from [<c000e128>] (ret_from_fork+0x14/0x2c) This is false positive because the lockdep uses the lock variable name to keep track of locks. To fix this, the spin_lock_init function can't be in a wrapper function for spinlock name to recorder correctly. I noticed similar fix was already made for mutex debugging so used similar approach and added extra macro to be used to spinlock allocation when spinlock debugging is on. Signed-off-by: Kari Suvanto <[email protected]>
Lockdep gives this splat during boot: [ 4.136748] ================================= [ 4.145487] [ INFO: inconsistent lock state ] [ 4.154157] 3.16.1+ raspberrypi#12 Not tainted [ 4.161852] --------------------------------- [ 4.170459] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage. [ 4.180776] khubd/18 [HC0[0]:SC0[0]:HE1:SE1] takes: [ 4.189971] (&((spinlock_t *)hcd->lock)->rlock){?.-...}, at: [<c0351490>] dwc_otg_hcd_qh_free+0x18/0xcc [ 4.204074] {IN-HARDIRQ-W} state was registered at: [ 4.213359] [<c04a8680>] _raw_spin_lock+0x40/0x50 [ 4.222743] [<c033f010>] dwc_otg_handle_common_intr+0x44/0xd90 [ 4.233133] [<c03333c8>] dwc_otg_common_irq+0xc/0x18 [ 4.242739] [<c00634d4>] handle_irq_event_percpu+0x34/0x150 [ 4.252955] [<c006362c>] handle_irq_event+0x3c/0x5c [ 4.262395] [<c0065984>] handle_level_irq+0x8c/0x130 [ 4.271972] [<c0062e40>] generic_handle_irq+0x28/0x40 [ 4.281489] [<c000e96c>] handle_IRQ+0x30/0x84 [ 4.290442] [<c04a9698>] __irq_svc+0x38/0xd0 [ 4.299164] [<c0357194>] DWC_MODIFY_REG32+0x64/0x84 [ 4.308613] [<c0333af4>] dwc_otg_driver_probe+0x720/0x7a8 [ 4.318523] [<c02daab0>] platform_drv_probe+0x18/0x48 [ 4.328141] [<c02d926c>] really_probe+0x68/0x200 [ 4.337334] [<c02d94f8>] __driver_attach+0xa0/0xa4 [ 4.346592] [<c02d7998>] bus_for_each_dev+0x60/0x94 [ 4.355998] [<c02d8b70>] bus_add_driver+0x140/0x1ec [ 4.365262] [<c02d9b18>] driver_register+0x78/0xf8 [ 4.374561] [<c065e0f4>] dwc_otg_driver_init+0x58/0x114 [ 4.384271] [<c00086a4>] do_one_initcall+0x80/0x1cc [ 4.393661] [<c0646c94>] kernel_init_freeable+0xf0/0x1b0 [ 4.403501] [<c049d83c>] kernel_init+0x8/0xec [ 4.412308] [<c000e128>] ret_from_fork+0x14/0x2c [ 4.421444] irq event stamp: 2351 [ 4.428928] hardirqs last enabled at (2351): [<c04a89b8>] _raw_spin_unlock_irqrestore+0x7c/0x94 [ 4.442100] hardirqs last disabled at (2350): [<c04a87a0>] _raw_spin_lock_irqsave+0x1c/0x64 [ 4.454802] softirqs last enabled at (2320): [<c0023b94>] __do_softirq+0x1ac/0x26c [ 4.466739] softirqs last disabled at (2297): [<c0023f80>] irq_exit+0xac/0x100 [ 4.478260] [ 4.478260] other info that might help us debug this: [ 4.493215] Possible unsafe locking scenario: [ 4.493215] [ 4.507451] CPU0 [ 4.513932] ---- [ 4.520505] lock(&((spinlock_t *)hcd->lock)->rlock); [ 4.529767] <Interrupt> [ 4.536515] lock(&((spinlock_t *)hcd->lock)->rlock); [ 4.545951] [ 4.545951] *** DEADLOCK *** [ 4.545951] [ 4.564132] 3 locks held by khubd/18: [ 4.571884] #0: (&dev->mutex){......}, at: [<c0321da8>] hub_thread+0x98/0x1000 [ 4.583536] raspberrypi#1: (&port_dev->status_lock){+.+.+.}, at: [<c032234c>] hub_thread+0x63c/0x1000 [ 4.596241] raspberrypi#2: (&bus->usb_address0_mutex){+.+.+.}, at: [<c031f164>] hub_port_init+0x5c/0xb24 [ 4.609307] [ 4.609307] stack backtrace: [ 4.621705] CPU: 0 PID: 18 Comm: khubd Not tainted 3.16.1+ raspberrypi#12 [ 4.631643] [<c00137e4>] (unwind_backtrace) from [<c0011530>] (show_stack+0x10/0x14) [ 4.643580] [<c0011530>] (show_stack) from [<c049f650>] (print_usage_bug+0x240/0x2b0) [ 4.655608] [<c049f650>] (print_usage_bug) from [<c0057a24>] (mark_lock+0x1d0/0x67c) [ 4.667527] [<c0057a24>] (mark_lock) from [<c0058a10>] (__lock_acquire+0x5d4/0x1ae0) [ 4.679551] [<c0058a10>] (__lock_acquire) from [<c005a6c8>] (lock_acquire+0x6c/0x8c) [ 4.691610] [<c005a6c8>] (lock_acquire) from [<c04a8680>] (_raw_spin_lock+0x40/0x50) [ 4.703584] [<c04a8680>] (_raw_spin_lock) from [<c0351490>] (dwc_otg_hcd_qh_free+0x18/0xcc) [ 4.716305] [<c0351490>] (dwc_otg_hcd_qh_free) from [<c034a60c>] (dwc_otg_hcd_endpoint_disable+0x9c/0xb0) [ 4.730246] [<c034a60c>] (dwc_otg_hcd_endpoint_disable) from [<c034d650>] (endpoint_disable+0x18/0x24) [ 4.743919] [<c034d650>] (endpoint_disable) from [<c031e0d0>] (usb_ep0_reinit+0x14/0x38) [ 4.756379] [<c031e0d0>] (usb_ep0_reinit) from [<c031f3ac>] (hub_port_init+0x2a4/0xb24) [ 4.768652] [<c031f3ac>] (hub_port_init) from [<c0322360>] (hub_thread+0x650/0x1000) [ 4.780824] [<c0322360>] (hub_thread) from [<c003ec3c>] (kthread+0xc8/0xe4) [ 4.792231] [<c003ec3c>] (kthread) from [<c000e128>] (ret_from_fork+0x14/0x2c) This splat shows that the hcd spinlock is used from hard irq context and also from process context with irqs on. To fix this, use spinlock_irqsave instead of spinlock in dwc_otg_hcd_qh_free. Signed-off-by: Kari Suvanto <[email protected]>
Collaborator
|
@P33M. Up to you. |
Contributor
|
@karis79 thanks for ferreting this out. |
pfpacket
pushed a commit
to pfpacket/linux-rpi-rust
that referenced
this pull request
Apr 7, 2023
doc: rust: document how to write comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
2 separate lockdep fixes to dwc. First one fixing false positive due to os wrapper function and second one fixing inconsistent lock state in hcd lock.