[PATCH v4 0/2] iommu/arm-smmu-v3: Make the queue depths tunable, and shrink them in a kdump kernel

Kiryl Shutsemau (Meta) kas at kernel.org
Wed Sep 2 05:17:22 PDT 2026


The queues are sized from the IDR1 maxima and allocated at probe, costing
megabytes per queue per SMMU instance. A kdump capture kernel pays that out
of a small crashkernel reservation, for queues it barely uses and two of
which it switches off anyway.

v3 clamped every queue to a page under is_kdump_kernel(). Review wanted a
general mechanism instead, so patch 1 adds a cmdq_entries module parameter
at the max_n_shift initialisation in arm_smmu_device_hw_probe() (Robin),
command queue only (Will), floored at one page (Jason). Patch 2 then has a
kdump kernel default all three depths to one page through the same helper,
with no command line needed (Jason). An explicit cmdq_entries still wins.

The v3 Reviewed-by from Breno, Pranjal, Jason and Nicolin, and Nicolin's
Tested-by, are dropped: this is not the patch they were given to.

Measured per instance under QEMU on -M virt,iommu=smmuv3, through a real
panic and kexec into a capture kernel:

                        4K page          64K page
      cmdq         1 MB -> 4 KB      8 MB -> 64 KB
      evtq         1 MB -> 4 KB     16 MB -> 64 KB

Every clamped queue lands on exactly one page. cmdq_entries moves the
command queue alone and beats the kdump default; the capture kernel
attached four devices with no CMD_SYNC timeout, GERROR or context fault, so
batching survives the floor. QEMU exposes no PRI queue, which takes the
same path. Also build-tested across 4K/16K/64K, TEGRA241_CMDQV=n,
CRASH_DUMP=n and =m, every commit warning-free.

The kdump stream table adoption series drops the event and PRI allocations
outright as of its v10. This sits on the same base, applies on top with no
conflict, and sizes the command queue that series leaves alone.

v3: https://lore.kernel.org/all/20260706084708.8072-1-kas@kernel.org/

Kiryl Shutsemau (Meta) (2):
  iommu/arm-smmu-v3: Add a cmdq_entries module parameter
  iommu/arm-smmu-v3: Default queue depths to one page in a kdump kernel

 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c   | 61 +++++++++++++++++--
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h   |  1 +
 .../iommu/arm/arm-smmu-v3/tegra241-cmdqv.c    |  5 +-
 3 files changed, 59 insertions(+), 8 deletions(-)


base-commit: cee9395acd8043be0644b25c34bfa86623f2b935
-- 
2.54.0




More information about the linux-arm-kernel mailing list