[BUG] NULL-Pointer crash in mutex_can_spin_on_owner

[Frank Wunderlich]

Hi,

i got a crash in mutex_can_spin_on_owner while trying to get mPCIe card (ath10 driver) running in bananapi r64 (aarch64) using kernel 5.4

[   11.381653] DEBUG: Passed mtk_pcie_irq_domain_alloc 441 port:0x00000000bc0b7874 lock:0x0000000023f7fc2a
[   11.391943] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000260
[   11.402178] Mem abort info:
[   11.406356]   ESR = 0x96000005
[   11.410868]   EC = 0x25: DABT (current EL), IL = 32 bits
[   11.417530]   SET = 0, FnV = 0
[   11.421997]   EA = 0, S1PTW = 0
[   11.425220] Data abort info:
[   11.428196]   ISV = 0, ISS = 0x00000005
[   11.428199]   CM = 0, WnR = 0
[   11.428203] user pgtable: 4k pages, 39-bit VAs, pgdp=000000007c24b000
[   11.428206] [0000000000000260] pgd=0000000000000000, pud=0000000000000000
[   11.428214] Internal error: Oops: 96000005 [#1] PREEMPT SMP
[   11.428218] Modules linked in: ath10k_pci(+) ath10k_core mt7622(+) mt76 ath mac80211 libarc4 cfg80211 btmtkuart
034-ga5669c8a8027-dirty #32
[   11.428243] Hardware name: Bananapi BPI-R64 (DT)
[   11.428248] pstate: 00000005 (nzcv daif -PAN -UAO)
[   11.428260] pc : mutex_can_spin_on_owner+0x30/0x5c
[   11.428264] lr : mutex_can_spin_on_owner+0x24/0x5c
[   11.428268] sp : ffffffc010c3b580
[   11.428271] x29: ffffffc010c3b580 x28: 0000000000080000
[   11.428275] x27: ffffffc010c3b838 x26: 0000000000000001
[   11.428280] x25: 000000000000008d x24: 0000000000000002
[   11.428284] x23: ffffff803d4e6800 x22: ffffff803e155500
[   11.428288] x21: 0000000000000001 x20: ffffffc010808000
[   11.428293] x19: ffffff803e155500 x18: 0000000000000000
[   11.428297] x17: 0000000000000000 x16: 0000000000000000
[   11.428302] x15: 0000000000000000 x14: 0000000000000000
[   11.428306] x13: 0000000000000000 x12: 0000000000000001
[   11.428310] x11: 0000000000000001 x10: 00000000000007c0
[   11.428315] x9 : ffffffc010c3b3c0 x8 : ffffff803fdb07d0
[   11.428319] x7 : ffffff803e155500 x6 : ffffff800336a400
[   11.428324] x5 : 0000000000000220 x4 : 0000000000000220
[   11.428328] x3 : ffffffc0108b7000 x2 : ffffff800336a400
[   11.428332] x1 : ffffff800336a400 x0 : 0000000000000220
[   11.428337] Call trace:
[   11.428342]  mutex_can_spin_on_owner+0x30/0x5c
[   11.428349]  __mutex_lock.isra.9+0x58/0x2a4
[   11.428356]  __mutex_lock_slowpath+0x10/0x18
[   11.428360]  mutex_lock+0x44/0x68
[   11.428366]  mtk_pcie_irq_domain_alloc+0x60/0x120
[   11.428371]  irq_domain_alloc_irqs_hierarchy+0x14/0x1c
[   11.428375]  irq_domain_alloc_irqs_parent+0x14/0x24
[   11.428379]  msi_domain_alloc+0x90/0x130
[   11.428382]  irq_domain_alloc_irqs_hierarchy+0x14/0x1c
[   11.428386]  __irq_domain_alloc_irqs+0x140/0x2b4
[   11.428390]  msi_domain_alloc_irqs+0x134/0x2c4
[   11.428394]  pci_msi_setup_msi_irqs+0x28/0x38
[   11.428398]  __pci_enable_msi_range+0x208/0x30c
[   11.428401]  pci_enable_msi+0x18/0x28
[   11.428418]  ath10k_pci_probe+0x50c/0x6d8 [ath10k_pci]
[   11.428423]  pci_device_probe+0xb4/0x144
[   11.428430]  really_probe+0x238/0x3f8
[   11.428436]  driver_probe_device+0x114/0x124
[   11.428440]  device_driver_attach+0x40/0x68
[   11.428444]  __driver_attach+0x134/0x138
[   11.428448]  bus_for_each_dev+0x78/0xbc

(gdb) list *(mtk_pcie_irq_domain_alloc+0x60)
0xffffffc0102e68d0 is in mtk_pcie_irq_domain_alloc (drivers/pci/controller/pcie-mediatek.c:446).
441             printk(KERN_ALERT "DEBUG: Passed %s %d port:0x%p lock:0x%p\n",__FUNCTION__,__LINE__,(void *)port,(void *)&port->lock);
442
443             WARN_ON(nr_irqs != 1);
444             mutex_lock(&port->lock);
445
446             printk(KERN_ALERT "DEBUG: Passed %s %d\n",__FUNCTION__,__LINE__);
447             bit = find_first_zero_bit(port->msi_irq_in_use, MTK_MSI_IRQS_NUM);
448             printk(KERN_ALERT "DEBUG: Passed %s %d bit:%lu,max:%u\n",__FUNCTION__,__LINE__,bit,MTK_MSI_IRQS_NUM);
449
450             if (bit >= MTK_MSI_IRQS_NUM) {

(gdb) list *(mutex_can_spin_on_owner+0x30)
0xffffffc0100d96f8 is in mutex_can_spin_on_owner (kernel/locking/mutex.c:605).
600             /*
601              * As lock holder preemption issue, we both skip spinning if task is not
602              * on cpu or its cpu is preempted
603              */
604             if (owner)
605                     retval = owner->on_cpu && !vcpu_is_preempted(task_cpu(owner));
606             rcu_read_unlock();
607
608             /*
609              * If lock->owner is not set, the mutex has been released. Return true

mutex is initialized in mtk_pcie_allocate_msi_domains which is called before twice, so i guess the mutex itself is correctly initialized.

can you give me some hint to debug it?

regards Frank