Re: [PATCH] PCI: vmd: Prevent recursive locking on interrupt allocation

From: Jonathan Derrick
Date: Tue Mar 01 2022 - 14:24:15 EST


This is how it used to be before I removed the member to better
cache align the vmd_irq_list struct, but then the scru_struct grew
to negate that benefit.

So this is good
Reviewed-by: Jon Derrick <jonathan.derrick@xxxxxxxxx>

On 2/13/2022 6:54 AM, Thomas Gleixner wrote:
Tejas reported the following recursive locking issue:

swapper/0/1 is trying to acquire lock:
ffff8881074fd0a0 (&md->mutex){+.+.}-{3:3}, at: msi_get_virq+0x30/0xc0
but task is already holding lock:
ffff8881017cd6a0 (&md->mutex){+.+.}-{3:3}, at: __pci_enable_msi_range+0xf2/0x290
stack backtrace:
__mutex_lock+0x9d/0x920
msi_get_virq+0x30/0xc0
pci_irq_vector+0x26/0x30
vmd_msi_init+0xcc/0x210
msi_domain_alloc+0xbf/0x150
msi_domain_alloc_irqs_descs_locked+0x3e/0xb0
__pci_enable_msi_range+0x155/0x290
pci_alloc_irq_vectors_affinity+0xba/0x100
pcie_port_device_register+0x307/0x550
pcie_portdrv_probe+0x3c/0xd0
pci_device_probe+0x95/0x110

This is caused by the VMD MSI code which does a lookup of the Linux
interrupt number for an VMD managed MSI[X] vector. The lookup function
tries to acquire the already held mutex.

Avoid that by caching the Linux interrupt number at initialization time
instead of looking it up over and over.

Fixes: 82ff8e6b78fc ("PCI/MSI: Use msi_get_virq() in pci_get_vector()")
Reported-by: "Surendrakumar Upadhyay, TejaskumarX" <tejaskumarx.surendrakumar.upadhyay@xxxxxxxxx>
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>

---
drivers/pci/controller/vmd.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)

--- a/drivers/pci/controller/vmd.c
+++ b/drivers/pci/controller/vmd.c
@@ -99,11 +99,13 @@ struct vmd_irq {
* @srcu: SRCU struct for local synchronization.
* @count: number of child IRQs assigned to this vector; used to track
* sharing.
+ * @virq: The underlying VMD Linux interrupt number
*/
struct vmd_irq_list {
struct list_head irq_list;
struct srcu_struct srcu;
unsigned int count;
+ unsigned int virq;
};
struct vmd_dev {
@@ -253,7 +255,6 @@ static int vmd_msi_init(struct irq_domai
struct msi_desc *desc = arg->desc;
struct vmd_dev *vmd = vmd_from_bus(msi_desc_to_pci_dev(desc)->bus);
struct vmd_irq *vmdirq = kzalloc(sizeof(*vmdirq), GFP_KERNEL);
- unsigned int index, vector;
if (!vmdirq)
return -ENOMEM;
@@ -261,10 +262,8 @@ static int vmd_msi_init(struct irq_domai
INIT_LIST_HEAD(&vmdirq->node);
vmdirq->irq = vmd_next_irq(vmd, desc);
vmdirq->virq = virq;
- index = index_from_irqs(vmd, vmdirq->irq);
- vector = pci_irq_vector(vmd->dev, index);
- irq_domain_set_info(domain, virq, vector, info->chip, vmdirq,
+ irq_domain_set_info(domain, virq, vmdirq->irq->virq, info->chip, vmdirq,
handle_untracked_irq, vmd, NULL);
return 0;
}
@@ -685,7 +684,8 @@ static int vmd_alloc_irqs(struct vmd_dev
return err;
INIT_LIST_HEAD(&vmd->irqs[i].irq_list);
- err = devm_request_irq(&dev->dev, pci_irq_vector(dev, i),
+ vmd->irqs[i].virq = pci_irq_vector(dev, i);
+ err = devm_request_irq(&dev->dev, vmd->irqs[i].virq,
vmd_irq, IRQF_NO_THREAD,
vmd->name, &vmd->irqs[i]);
if (err)
@@ -969,7 +969,7 @@ static int vmd_suspend(struct device *de
int i;
for (i = 0; i < vmd->msix_count; i++)
- devm_free_irq(dev, pci_irq_vector(pdev, i), &vmd->irqs[i]);
+ devm_free_irq(dev, vmd->irqs[i].virq, &vmd->irqs[i]);
return 0;
}
@@ -981,7 +981,7 @@ static int vmd_resume(struct device *dev
int err, i;
for (i = 0; i < vmd->msix_count; i++) {
- err = devm_request_irq(dev, pci_irq_vector(pdev, i),
+ err = devm_request_irq(dev, vmd->irqs[i].virq,
vmd_irq, IRQF_NO_THREAD,
vmd->name, &vmd->irqs[i]);
if (err)