[PATCH v3 01/15] iommu/vt-d: Handle race between registration and device probe

[Baolu Lu]

On 2022/7/6 01:08, Robin Murphy wrote:
> That also highlights an issue with intel_iommu_get_resv_regions() taking
> dmar_global_lock from within a section where intel_iommu_init() already
> holds it, which already exists via probe_acpi_namespace_devices() when
> an ANDD device is probed, but gets more obvious with the upcoming change
> to iommu_device_register(). Since they are both read locks it manages
> not to deadlock in practice, so I'm leaving it here for someone with
> more confidence to tackle a larger rework of the locking.

I am trying to reproduce this problem. Strangely, even if I selected
CONFIG_LOCKDEP=y, the kernel didn't complain anything. :-)

In fact the rmrr list in the Intel IOMMU driver is always static after
parsing the ACPI/DMAR tables. There's no need to protect it with a lock.
Hence we can safely remove below down/up_read().

4512 static void intel_iommu_get_resv_regions(struct device *device,
4513                                          struct list_head *head)
4514 {
4515         int prot = DMA_PTE_READ | DMA_PTE_WRITE;
4516         struct iommu_resv_region *reg;
4517         struct dmar_rmrr_unit *rmrr;
4518         struct device *i_dev;
4519         int i;
4520
4521         down_read(&dmar_global_lock);
4522         for_each_rmrr_units(rmrr) {
4523                 for_each_active_dev_scope(rmrr->devices, 
rmrr->devices_cnt,
4524                                           i, i_dev) {

Best regards,
baolu