Registration of a mmu_notifier requires the caller to hold a mmget() on
the mm as registration is not permitted to race with exit_mmap(). There is
a BUG_ON inside the mmu_notifier to guard against this.
Normally creating a umem is done against current which implicitly holds
the mmget(), however an implicit ODP child is created from a pagefault
work queue and is not guaranteed to have a mmget().
Call mmget() around this registration and abort faulting if the MM has
gone to exit_mmap().
Before the patch below the notifier was registered when the implicit ODP
parent was created, so there was no chance to register a notifier outside
of current.
Fixes: c571feca2dc9 ("RDMA/odp: use mmu_notifier_get/put for 'struct ib_ucontext_per_mm'")
Link: https://<email address hidden>
Signed-off-by: Leon Romanovsky <email address hidden>
Signed-off-by: Jason Gunthorpe <email address hidden>
(backported from commit a4e63bce1414df7ab6eb82ca9feb8494ce13e554)
[chengen - remove tgid-related code and adjust jump label from out_tgid
to out_mmput]
Signed-off-by: Chengen Du <email address hidden>
Acked-by: Tim Gardner <email address hidden>
Acked-by: Manuel Diewald <email address hidden>
Signed-off-by: Roxana Nicolescu <email address hidden>
This reverts commit 40613da52b13fb21c5566f10b287e0ca8c12c4e9 and the
subsequent fix to it:
cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus")
40613da52b13 fixed a problem where hot-adding a device with large BARs
failed if the bridge windows programmed by firmware were not large enough.
cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources()
only for non-root bus") fixed a problem with 40613da52b13: an ACPI hot-add
of a device on a PCI root bus (common in the virt world) or firmware
sending ACPI Bus Check to non-existent Root Ports (e.g., on Dell Inspiron
7352/0W6WV0) caused a NULL pointer dereference and suspend/resume hangs.
Unfortunately the combination of 40613da52b13 and cc22522fd55e caused other
problems:
- Fiona reported that hot-add of SCSI disks in QEMU virtual machine fails
sometimes.
- Dongli reported a similar problem with hot-add of SCSI disks.
- Jonathan reported a console freeze during boot on bare metal due to an
error in radeon GPU initialization.
Revert both patches to avoid adding these problems. This means we will
again see the problems with hot-adding devices with large BARs and the NULL
pointer dereferences and suspend/resume issues that 40613da52b13 and
cc22522fd55e were intended to fix.
Fixes: 40613da52b13 ("PCI: acpiphp: Reassign resources on bridge if necessary")
Fixes: cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus")
Reported-by: Fiona Ebner <email address hidden>
Closes: https://<email address hidden>
Reported-by: Dongli Zhang <email address hidden>
Closes: https://<email address hidden>
Reported-by: Jonathan Woithe <email address hidden>
Closes: https://<email address hidden>
Signed-off-by: Bjorn Helgaas <email address hidden>
Acked-by: Michael S. Tsirkin <email address hidden>
Acked-by: Igor Mammedov <email address hidden>
Cc: <email address hidden>
(cherry picked from commit 5df12742b7e3aae2594a30a9d14d5d6e9e7699f4)
Signed-off-by: Chengen Du <email address hidden>
Acked-by: Tim Gardner <email address hidden>
Acked-by: Manuel Diewald <email address hidden>
Signed-off-by: Roxana Nicolescu <email address hidden>
