platform/kernel/linux-starfive.git
3 years agoiommu/vt-d: Add present bit check in pasid entry setup helpers
Liu Yi L [Wed, 18 Aug 2021 13:48:52 +0000 (21:48 +0800)]
iommu/vt-d: Add present bit check in pasid entry setup helpers

The helper functions should not modify the pasid entries which are still
in use. Add a check against present bit.

Signed-off-by: Liu Yi L <yi.l.liu@intel.com>
Link: https://lore.kernel.org/r/20210817042425.1784279-1-yi.l.liu@intel.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210818134852.1847070-10-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Use pasid_pte_is_present() helper function
Liu Yi L [Wed, 18 Aug 2021 13:48:51 +0000 (21:48 +0800)]
iommu/vt-d: Use pasid_pte_is_present() helper function

Use the pasid_pte_is_present() helper for present bit check in the
intel_pasid_tear_down_entry().

Signed-off-by: Liu Yi L <yi.l.liu@intel.com>
Link: https://lore.kernel.org/r/20210817042425.1784279-1-yi.l.liu@intel.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210818134852.1847070-9-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Drop the kernel doc annotation
Andy Shevchenko [Wed, 18 Aug 2021 13:48:50 +0000 (21:48 +0800)]
iommu/vt-d: Drop the kernel doc annotation

Kernel doc validator is unhappy with the following

.../perf.c:16: warning: Function parameter or member 'latency_lock' not described in 'DEFINE_SPINLOCK'
.../perf.c:16: warning: expecting prototype for perf.c(). Prototype was for DEFINE_SPINLOCK() instead

Drop kernel doc annotation since the top comment is not in the required format.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20210729163538.40101-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210818134852.1847070-8-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Allow devices to have more than 32 outstanding PRs
Lu Baolu [Wed, 18 Aug 2021 13:48:49 +0000 (21:48 +0800)]
iommu/vt-d: Allow devices to have more than 32 outstanding PRs

The minimum per-IOMMU PRQ queue size is one 4K page, this is more entries
than the hardcoded limit of 32 in the current VT-d code. Some devices can
support up to 512 outstanding PRQs but underutilized by this limit of 32.
Although, 32 gives some rough fairness when multiple devices share the same
IOMMU PRQ queue, but far from optimal for customized use case. This extends
the per-IOMMU PRQ queue size to four 4K pages and let the devices have as
many outstanding page requests as they can.

Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720013856.4143880-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210818134852.1847070-7-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Preset A/D bits for user space DMA usage
Lu Baolu [Wed, 18 Aug 2021 13:48:48 +0000 (21:48 +0800)]
iommu/vt-d: Preset A/D bits for user space DMA usage

We preset the access and dirty bits for IOVA over first level usage only
for the kernel DMA (i.e., when domain type is IOMMU_DOMAIN_DMA). We should
also preset the FL A/D for user space DMA usage. The idea is that even the
user space A/D bit memory write is unnecessary. We should avoid it to
minimize the overhead.

Suggested-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720013856.4143880-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210818134852.1847070-6-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Enable Intel IOMMU scalable mode by default
Lu Baolu [Wed, 18 Aug 2021 13:48:47 +0000 (21:48 +0800)]
iommu/vt-d: Enable Intel IOMMU scalable mode by default

The commit 8950dcd83ae7d ("iommu/vt-d: Leave scalable mode default off")
leaves the scalable mode default off and end users could turn it on with
"intel_iommu=sm_on". Using the Intel IOMMU scalable mode for kernel DMA,
user-level device access and Shared Virtual Address have been enabled.
This enables the scalable mode by default if the hardware advertises the
support and adds kernel options of "intel_iommu=sm_on/sm_off" for end
users to configure it through the kernel parameters.

Suggested-by: Ashok Raj <ashok.raj@intel.com>
Suggested-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20210720013856.4143880-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210818134852.1847070-5-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Refactor Kconfig a bit
Lu Baolu [Wed, 18 Aug 2021 13:48:46 +0000 (21:48 +0800)]
iommu/vt-d: Refactor Kconfig a bit

Put all sub-options inside a "if INTEL_IOMMU" so that they don't need to
always depend on INTEL_IOMMU. Use IS_ENABLED() instead of #ifdef as well.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720013856.4143880-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210818134852.1847070-4-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Remove unnecessary oom message
Zhen Lei [Wed, 18 Aug 2021 13:48:45 +0000 (21:48 +0800)]
iommu/vt-d: Remove unnecessary oom message

Fixes scripts/checkpatch.pl warning:
WARNING: Possible unnecessary 'out of memory' message

Remove it can help us save a bit of memory.

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Link: https://lore.kernel.org/r/20210609124937.14260-1-thunder.leizhen@huawei.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210818134852.1847070-3-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Update the virtual command related registers
Lu Baolu [Wed, 18 Aug 2021 13:48:44 +0000 (21:48 +0800)]
iommu/vt-d: Update the virtual command related registers

The VT-d spec Revision 3.3 updated the virtual command registers, virtual
command opcode B register, virtual command response register and virtual
command capability register (Section 10.4.43, 10.4.44, 10.4.45, 10.4.46).
This updates the virtual command interface implementation in the Intel
IOMMU driver accordingly.

Fixes: 24f27d32ab6b7 ("iommu/vt-d: Enlightened PASID allocation")
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Sanjay Kumar <sanjay.k.kumar@intel.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20210713042649.3547403-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210818134852.1847070-2-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Allow enabling non-strict mode dynamically
Robin Murphy [Wed, 11 Aug 2021 12:21:38 +0000 (13:21 +0100)]
iommu: Allow enabling non-strict mode dynamically

Allocating and enabling a flush queue is in fact something we can
reasonably do while a DMA domain is active, without having to rebuild it
from scratch. Thus we can allow a strict -> non-strict transition from
sysfs without requiring to unbind the device's driver, which is of
particular interest to users who want to make selective relaxations to
critical devices like the one serving their root filesystem.

Disabling and draining a queue also seems technically possible to
achieve without rebuilding the whole domain, but would certainly be more
involved. Furthermore there's not such a clear use-case for tightening
up security *after* the device may already have done whatever it is that
you don't trust it not to do, so we only consider the relaxation case.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/d652966348c78457c38bf18daf369272a4ebc2c9.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Merge strictness and domain type configs
Robin Murphy [Wed, 11 Aug 2021 12:21:37 +0000 (13:21 +0100)]
iommu: Merge strictness and domain type configs

To parallel the sysfs behaviour, merge the new build-time option
for DMA domain strictness into the default domain type choice.

Suggested-by: Joerg Roedel <joro@8bytes.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: John Garry <john.garry@huawei.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/d04af35b9c0f2a1d39605d7a9b451f5e1f0c7736.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Only log strictness for DMA domains
Robin Murphy [Wed, 11 Aug 2021 12:21:36 +0000 (13:21 +0100)]
iommu: Only log strictness for DMA domains

When passthrough is enabled, the default strictness policy becomes
irrelevant, since any subsequent runtime override to a DMA domain type
now embodies an explicit choice of strictness as well. Save on noise by
only logging the default policy when it is meaningfully in effect.

Reviewed-by: John Garry <john.garry@huawei.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/9d2bcba880c6d517d0751ed8bd4960853030b4d7.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Expose DMA domain strictness via sysfs
Robin Murphy [Wed, 11 Aug 2021 12:21:35 +0000 (13:21 +0100)]
iommu: Expose DMA domain strictness via sysfs

The sysfs interface for default domain types exists primarily so users
can choose the performance/security tradeoff relevant to their own
workload. As such, the choice between the policies for DMA domains fits
perfectly as an additional point on that scale - downgrading a
particular device from a strict default to non-strict may be enough to
let it reach the desired level of performance, while still retaining
more peace of mind than with a wide-open identity domain. Now that we've
abstracted non-strict mode as a distinct type of DMA domain, allow it to
be chosen through the user interface as well.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/0e08da5ed4069fd3473cfbadda758ca983becdbf.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Express DMA strictness via the domain type
Robin Murphy [Wed, 11 Aug 2021 12:21:34 +0000 (13:21 +0100)]
iommu: Express DMA strictness via the domain type

Eliminate the iommu_get_dma_strict() indirection and pipe the
information through the domain type from the beginning. Besides
the flow simplification this also has several nice side-effects:

 - Automatically implies strict mode for untrusted devices by
   virtue of their IOMMU_DOMAIN_DMA override.
 - Ensures that we only end up using flush queues for drivers
   which are aware of them and can actually benefit.
 - Allows us to handle flush queue init failure by falling back
   to strict mode instead of leaving it to possibly blow up later.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/47083d69155577f1367877b1594921948c366eb3.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Prepare for multiple DMA domain types
Robin Murphy [Wed, 11 Aug 2021 12:21:33 +0000 (13:21 +0100)]
iommu/vt-d: Prepare for multiple DMA domain types

In preparation for the strict vs. non-strict decision for DMA domains
to be expressed in the domain type, make sure we expose our flush queue
awareness by accepting the new domain type, and test the specific
feature flag where we want to identify DMA domains in general. The DMA
ops reset/setup can simply be made unconditional, since iommu-dma
already knows only to touch DMA domains.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/31a8ef868d593a2f3826a6a120edee81815375a7.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu: Prepare for multiple DMA domain types
Robin Murphy [Wed, 11 Aug 2021 12:21:32 +0000 (13:21 +0100)]
iommu/arm-smmu: Prepare for multiple DMA domain types

In preparation for the strict vs. non-strict decision for DMA domains to
be expressed in the domain type, make sure we expose our flush queue
awareness by accepting the new domain type.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/8f217ef285bd0bb9456c27ef622d2efdbbca1ad8.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/amd: Prepare for multiple DMA domain types
Robin Murphy [Wed, 11 Aug 2021 12:21:31 +0000 (13:21 +0100)]
iommu/amd: Prepare for multiple DMA domain types

The DMA ops reset/setup can simply be unconditional, since
iommu-dma already knows only to touch DMA domains.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/6450b4f39a5a086d505297b4a53ff1e4a7a0fe7c.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Introduce explicit type for non-strict DMA domains
Robin Murphy [Wed, 11 Aug 2021 12:21:30 +0000 (13:21 +0100)]
iommu: Introduce explicit type for non-strict DMA domains

Promote the difference between strict and non-strict DMA domains from an
internal detail to a distinct domain feature and type, to pave the road
for exposing it through the sysfs default domain interface.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/08cd2afaf6b63c58ad49acec3517c9b32c2bb946.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable: Remove non-strict quirk
Robin Murphy [Wed, 11 Aug 2021 12:21:29 +0000 (13:21 +0100)]
iommu/io-pgtable: Remove non-strict quirk

IO_PGTABLE_QUIRK_NON_STRICT was never a very comfortable fit, since it's
not a quirk of the pagetable format itself. Now that we have a more
appropriate way to convey non-strict unmaps, though, this last of the
non-quirk quirks can also go, and with the flush queue code also now
enforcing its own ordering we can have a lovely cleanup all round.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/155b5c621cd8936472e273a8b07a182f62c6c20d.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Indicate queued flushes via gather data
Robin Murphy [Wed, 11 Aug 2021 12:21:28 +0000 (13:21 +0100)]
iommu: Indicate queued flushes via gather data

Since iommu_iotlb_gather exists to help drivers optimise flushing for a
given unmap request, it is also the logical place to indicate whether
the unmap is strict or not, and thus help them further optimise for
whether to expect a sync or a flush_all subsequently. As part of that,
it also seems fair to make the flush queue code take responsibility for
enforcing the really subtle ordering requirement it brings, so that we
don't need to worry about forgetting that if new drivers want to add
flush queue support, and can consolidate the existing versions.

While we're adding to the kerneldoc, also fill in some info for
@freelist which was overlooked previously.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/bf5f8e2ad84e48c712ccbf80fa8c610594c7595f.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/dma: Remove redundant "!dev" checks
Robin Murphy [Wed, 11 Aug 2021 12:21:27 +0000 (13:21 +0100)]
iommu/dma: Remove redundant "!dev" checks

iommu_dma_init_domain() is now only called from iommu_setup_dma_ops(),
which has already assumed dev to be non-NULL.

Reviewed-by: John Garry <john.garry@huawei.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/06024523c080364390016550065e3cfe8031367e.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/virtio: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:25 +0000 (13:21 +0100)]
iommu/virtio: Drop IOVA cookie management

The core code bakes its own cookies now.

Reviewed-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/f05cd2d0a0f414de3180e2536c7656faf1e52418.1628682049.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/sun50i: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:24 +0000 (13:21 +0100)]
iommu/sun50i: Drop IOVA cookie management

The core code bakes its own cookies now.

CC: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/147edb0ba59be563df19cec3e63e621aa65b7b68.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/sprd: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:23 +0000 (13:21 +0100)]
iommu/sprd: Drop IOVA cookie management

The core code bakes its own cookies now.

Acked-by: Chunyan Zhang <zhang.lyra@gmail.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/4e7fc6e523cb4b63fb13f5be10041eb24c0dcb1e.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/rockchip: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:22 +0000 (13:21 +0100)]
iommu/rockchip: Drop IOVA cookie management

The core code bakes its own cookies now.

Tested-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/aff51e2da1e431987ae5fdafa62a6a7c4bd042dc.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/mtk: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:21 +0000 (13:21 +0100)]
iommu/mtk: Drop IOVA cookie management

The core code bakes its own cookies now.

CC: Yong Wu <yong.wu@mediatek.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/b856648e7ee2b1017e7c7c02e2ddd50eaf72cbf7.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/ipmmu-vmsa: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:20 +0000 (13:21 +0100)]
iommu/ipmmu-vmsa: Drop IOVA cookie management

The core code bakes its own cookies now.

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/dc5513293942d81f84edf61b354b236e5ac51dc2.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/exynos: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:19 +0000 (13:21 +0100)]
iommu/exynos: Drop IOVA cookie management

The core code bakes its own cookies now.

Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/12d88cbf44e57faa4f0512760e7ed3a9cba05ca8.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:18 +0000 (13:21 +0100)]
iommu/vt-d: Drop IOVA cookie management

The core code bakes its own cookies now.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/e9dbe3b6108f8538e17e0c5f59f8feeb714f51a4.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:17 +0000 (13:21 +0100)]
iommu/arm-smmu: Drop IOVA cookie management

The core code bakes its own cookies now.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/7ae3680dad9735cc69c3618866666896bd11e031.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/amd: Drop IOVA cookie management
Robin Murphy [Wed, 11 Aug 2021 12:21:16 +0000 (13:21 +0100)]
iommu/amd: Drop IOVA cookie management

The core code bakes its own cookies now.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/648e74e7422caa6a7db7fb0c36813c7bd2007af8.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Pull IOVA cookie management into the core
Robin Murphy [Wed, 11 Aug 2021 12:21:15 +0000 (13:21 +0100)]
iommu: Pull IOVA cookie management into the core

Now that everyone has converged on iommu-dma for IOMMU_DOMAIN_DMA
support, we can abandon the notion of drivers being responsible for the
cookie type, and consolidate all the management into the core code.

CC: Yong Wu <yong.wu@mediatek.com>
CC: Chunyan Zhang <chunyan.zhang@unisoc.com>
CC: Maxime Ripard <mripard@kernel.org>
Tested-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/46a2c0e7419c7d1d931762dc7b6a69fa082d199a.1628682048.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu-v3: Implement the map_pages() IOMMU driver callback
Xiang Chen [Sat, 31 Jul 2021 02:17:11 +0000 (10:17 +0800)]
iommu/arm-smmu-v3: Implement the map_pages() IOMMU driver callback

Implement the map_pages() callback for ARM SMMUV3 driver to allow calls
from iommu_map to map multiple pages of the same size in one call.
Also remove the map() callback for the ARM SMMUV3 driver as it will no
longer be used.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/1627697831-158822-3-git-send-email-chenxiang66@hisilicon.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu-v3: Implement the unmap_pages() IOMMU driver callback
Xiang Chen [Sat, 31 Jul 2021 02:17:10 +0000 (10:17 +0800)]
iommu/arm-smmu-v3: Implement the unmap_pages() IOMMU driver callback

Implement the unmap_pages() callback for ARM SMMUV3 driver to allow calls
from iommu_unmap to unmap multiple pages of the same size in one call.
Also remove the unmap() callback for the ARM SMMUV3 driver as it will
no longer be used.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/1627697831-158822-2-git-send-email-chenxiang66@hisilicon.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Move clflush'es from iotlb_sync_map() to map_pages()
Lu Baolu [Tue, 20 Jul 2021 02:06:15 +0000 (10:06 +0800)]
iommu/vt-d: Move clflush'es from iotlb_sync_map() to map_pages()

As the Intel VT-d driver has switched to use the iommu_ops.map_pages()
callback, multiple pages of the same size will be mapped in a call.
There's no need to put the clflush'es in iotlb_sync_map() callback.
Move them back into __domain_mapping() to simplify the code.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720020615.4144323-4-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Implement map/unmap_pages() iommu_ops callback
Lu Baolu [Tue, 20 Jul 2021 02:06:14 +0000 (10:06 +0800)]
iommu/vt-d: Implement map/unmap_pages() iommu_ops callback

Implement the map_pages() and unmap_pages() callback for the Intel IOMMU
driver to allow calls from iommu core to map and unmap multiple pages of
the same size in one call. With map/unmap_pages() implemented, the prior
map/unmap callbacks are deprecated.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720020615.4144323-3-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Report real pgsize bitmap to iommu core
Lu Baolu [Tue, 20 Jul 2021 02:06:13 +0000 (10:06 +0800)]
iommu/vt-d: Report real pgsize bitmap to iommu core

The pgsize bitmap is used to advertise the page sizes our hardware supports
to the IOMMU core, which will then use this information to split physically
contiguous memory regions it is mapping into page sizes that we support.

Traditionally the IOMMU core just handed us the mappings directly, after
making sure the size is an order of a 4KiB page and that the mapping has
natural alignment. To retain this behavior, we currently advertise that we
support all page sizes that are an order of 4KiB.

We are about to utilize the new IOMMU map/unmap_pages APIs. We could change
this to advertise the real page sizes we support.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210720020615.4144323-2-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Streamline iommu_iova_to_phys()
Robin Murphy [Thu, 15 Jul 2021 13:04:24 +0000 (14:04 +0100)]
iommu: Streamline iommu_iova_to_phys()

If people are going to insist on calling iommu_iova_to_phys()
pointlessly and expecting it to work, we can at least do ourselves a
favour by handling those cases in the core code, rather than repeatedly
across an inconsistent handful of drivers.

Since all the existing drivers implement the internal callback, and any
future ones are likely to want to work with iommu-dma which relies on
iova_to_phys a fair bit, we may as well remove that currently-redundant
check as well and consider it mandatory.

Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/f564f3f6ff731b898ff7a898919bf871c2c7745a.1626354264.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Remove mode argument from iommu_set_dma_strict()
John Garry [Mon, 12 Jul 2021 11:12:20 +0000 (19:12 +0800)]
iommu: Remove mode argument from iommu_set_dma_strict()

We only ever now set strict mode enabled in iommu_set_dma_strict(), so
just remove the argument.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-7-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/amd: Add support for IOMMU default DMA mode build options
Zhen Lei [Mon, 12 Jul 2021 11:12:19 +0000 (19:12 +0800)]
iommu/amd: Add support for IOMMU default DMA mode build options

Make IOMMU_DEFAULT_LAZY default for when AMD_IOMMU config is set, which
matches current behaviour.

For "fullflush" param, just call iommu_set_dma_strict(true) directly.

Since we get a strict vs lazy mode print already in iommu_subsys_init(),
and maintain a deprecation print when "fullflush" param is passed, drop the
prints in amd_iommu_init_dma_ops().

Finally drop global flag amd_iommu_unmap_flush, as it has no longer has any
purpose.

[jpg: Rebase for relocated file and drop amd_iommu_unmap_flush]

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/1626088340-5838-6-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/vt-d: Add support for IOMMU default DMA mode build options
Zhen Lei [Mon, 12 Jul 2021 11:12:18 +0000 (19:12 +0800)]
iommu/vt-d: Add support for IOMMU default DMA mode build options

Make IOMMU_DEFAULT_LAZY default for when INTEL_IOMMU config is set,
as is current behaviour.

Also delete global flag intel_iommu_strict:
- In intel_iommu_setup(), call iommu_set_dma_strict(true) directly. Also
  remove the print, as iommu_subsys_init() prints the mode and we have
  already marked this param as deprecated.

- For cap_caching_mode() check in intel_iommu_setup(), call
  iommu_set_dma_strict(true) directly; also reword the accompanying print
  with a level downgrade and also add the missing '\n'.

- For Ironlake GPU, again call iommu_set_dma_strict(true) directly and
  keep the accompanying print.

[jpg: Remove intel_iommu_strict]

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-5-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Enhance IOMMU default DMA mode build options
Zhen Lei [Mon, 12 Jul 2021 11:12:17 +0000 (19:12 +0800)]
iommu: Enhance IOMMU default DMA mode build options

First, add build options IOMMU_DEFAULT_{LAZY|STRICT}, so that we have the
opportunity to set {lazy|strict} mode as default at build time. Then put
the two config options in an choice, as they are mutually exclusive.

[jpg: Make choice between strict and lazy only (and not passthrough)]

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-4-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Print strict or lazy mode at init time
John Garry [Mon, 12 Jul 2021 11:12:16 +0000 (19:12 +0800)]
iommu: Print strict or lazy mode at init time

As well as the default domain type, it's useful to know whether strict
or lazy for DMA domains, so add this info in a separate print.

The (stict/lazy) mode may be also set via iommu.strict earlyparm, but
this will be processed prior to iommu_subsys_init(), so that print will be
accurate for drivers which don't set the mode via custom means.

For the drivers which set the mode via custom means - AMD and Intel drivers
- they maintain prints to inform a change in policy or that custom cmdline
methods to change policy are deprecated.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-3-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Deprecate Intel and AMD cmdline methods to enable strict mode
John Garry [Mon, 12 Jul 2021 11:12:15 +0000 (19:12 +0800)]
iommu: Deprecate Intel and AMD cmdline methods to enable strict mode

Now that the x86 drivers support iommu.strict, deprecate the custom
methods.

Signed-off-by: John Garry <john.garry@huawei.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-2-git-send-email-john.garry@huawei.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu: Implement the map_pages() IOMMU driver callback
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:56 +0000 (06:38 -0700)]
iommu/arm-smmu: Implement the map_pages() IOMMU driver callback

Implement the map_pages() callback for the ARM SMMU driver
to allow calls from iommu_map to map multiple pages of
the same size in one call. Also, remove the map() callback
for the ARM SMMU driver, as it will no longer be used.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-16-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/arm-smmu: Implement the unmap_pages() IOMMU driver callback
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:55 +0000 (06:38 -0700)]
iommu/arm-smmu: Implement the unmap_pages() IOMMU driver callback

Implement the unmap_pages() callback for the ARM SMMU driver
to allow calls from iommu_unmap to unmap multiple pages of
the same size in one call. Also, remove the unmap() callback
for the SMMU driver, as it will no longer be used.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-15-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable-arm-v7s: Implement arm_v7s_map_pages()
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:54 +0000 (06:38 -0700)]
iommu/io-pgtable-arm-v7s: Implement arm_v7s_map_pages()

Implement the map_pages() callback for the ARM v7s io-pgtable
format.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-14-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable-arm-v7s: Implement arm_v7s_unmap_pages()
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:53 +0000 (06:38 -0700)]
iommu/io-pgtable-arm-v7s: Implement arm_v7s_unmap_pages()

Implement the unmap_pages() callback for the ARM v7s io-pgtable
format.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-13-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable-arm: Implement arm_lpae_map_pages()
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:52 +0000 (06:38 -0700)]
iommu/io-pgtable-arm: Implement arm_lpae_map_pages()

Implement the map_pages() callback for the ARM LPAE io-pgtable
format.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-12-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable-arm: Implement arm_lpae_unmap_pages()
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:51 +0000 (06:38 -0700)]
iommu/io-pgtable-arm: Implement arm_lpae_unmap_pages()

Implement the unmap_pages() callback for the ARM LPAE io-pgtable
format.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-11-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable-arm: Prepare PTE methods for handling multiple entries
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:50 +0000 (06:38 -0700)]
iommu/io-pgtable-arm: Prepare PTE methods for handling multiple entries

The PTE methods currently operate on a single entry. In preparation
for manipulating multiple PTEs in one map or unmap call, allow them
to handle multiple PTEs.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-10-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Add support for the map_pages() callback
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:49 +0000 (06:38 -0700)]
iommu: Add support for the map_pages() callback

Since iommu_pgsize can calculate how many pages of the
same size can be mapped/unmapped before the next largest
page size boundary, add support for invoking an IOMMU
driver's map_pages() callback, if it provides one.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-9-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Hook up '->unmap_pages' driver callback
Will Deacon [Wed, 16 Jun 2021 13:38:48 +0000 (06:38 -0700)]
iommu: Hook up '->unmap_pages' driver callback

Extend iommu_pgsize() to populate an optional 'count' parameter so that
we can direct unmapping operation to the ->unmap_pages callback if it
has been provided by the driver.

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-8-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Split 'addr_merge' argument to iommu_pgsize() into separate parts
Will Deacon [Wed, 16 Jun 2021 13:38:47 +0000 (06:38 -0700)]
iommu: Split 'addr_merge' argument to iommu_pgsize() into separate parts

The 'addr_merge' parameter to iommu_pgsize() is a fabricated address
intended to describe the alignment requirements to consider when
choosing an appropriate page size. On the iommu_map() path, this address
is the logical OR of the virtual and physical addresses.

Subsequent improvements to iommu_pgsize() will need to check the
alignment of the virtual and physical components of 'addr_merge'
independently, so pass them in as separate parameters and reconstruct
'addr_merge' locally.

No functional change.

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-7-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Use bitmap to calculate page size in iommu_pgsize()
Will Deacon [Wed, 16 Jun 2021 13:38:46 +0000 (06:38 -0700)]
iommu: Use bitmap to calculate page size in iommu_pgsize()

Avoid the potential for shifting values by amounts greater than the
width of their type by using a bitmap to compute page size in
iommu_pgsize().

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-6-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Add a map_pages() op for IOMMU drivers
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:45 +0000 (06:38 -0700)]
iommu: Add a map_pages() op for IOMMU drivers

Add a callback for IOMMU drivers to provide a path for the
IOMMU framework to call into an IOMMU driver, which can
call into the io-pgtable code, to map a physically contiguous
rnage of pages of the same size.

For IOMMU drivers that do not specify a map_pages() callback,
the existing logic of mapping memory one page block at a time
will be used.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Acked-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-5-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable: Introduce map_pages() as a page table op
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:44 +0000 (06:38 -0700)]
iommu/io-pgtable: Introduce map_pages() as a page table op

Mapping memory into io-pgtables follows the same semantics
that unmapping memory used to follow (i.e. a buffer will be
mapped one page block per call to the io-pgtable code). This
means that it can be optimized in the same way that unmapping
memory was, so add a map_pages() callback to the io-pgtable
ops structure, so that a range of pages of the same size
can be mapped within the same call.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-4-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu: Add an unmap_pages() op for IOMMU drivers
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:43 +0000 (06:38 -0700)]
iommu: Add an unmap_pages() op for IOMMU drivers

Add a callback for IOMMU drivers to provide a path for the
IOMMU framework to call into an IOMMU driver, which can call
into the io-pgtable code, to unmap a virtually contiguous
range of pages of the same size.

For IOMMU drivers that do not specify an unmap_pages() callback,
the existing logic of unmapping memory one page block at a time
will be used.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
Acked-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-3-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoiommu/io-pgtable: Introduce unmap_pages() as a page table op
Isaac J. Manjarres [Wed, 16 Jun 2021 13:38:42 +0000 (06:38 -0700)]
iommu/io-pgtable: Introduce unmap_pages() as a page table op

The io-pgtable code expects to operate on a single block or
granule of memory that is supported by the IOMMU hardware when
unmapping memory.

This means that when a large buffer that consists of multiple
such blocks is unmapped, the io-pgtable code will walk the page
tables to the correct level to unmap each block, even for blocks
that are virtually contiguous and at the same level, which can
incur an overhead in performance.

Introduce the unmap_pages() page table op to express to the
io-pgtable code that it should unmap a number of blocks of
the same size, instead of a single block. Doing so allows
multiple blocks to be unmapped in one call to the io-pgtable
code, reducing the number of page table walks, and indirect
calls.

Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-2-git-send-email-quic_c_gdjako@quicinc.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
3 years agoLinux 5.14-rc3
Linus Torvalds [Sun, 25 Jul 2021 22:35:14 +0000 (15:35 -0700)]
Linux 5.14-rc3

3 years agosmpboot: fix duplicate and misplaced inlining directive
Linus Torvalds [Sun, 25 Jul 2021 18:06:37 +0000 (11:06 -0700)]
smpboot: fix duplicate and misplaced inlining directive

gcc doesn't care, but clang quite reasonably pointed out that the recent
commit e9ba16e68cce ("smpboot: Mark idle_init() as __always_inlined to
work around aggressive compiler un-inlining") did some really odd
things:

    kernel/smpboot.c:50:20: warning: duplicate 'inline' declaration specifier [-Wduplicate-decl-specifier]
    static inline void __always_inline idle_init(unsigned int cpu)
                       ^

which not only has that duplicate inlining specifier, but the new
__always_inline was put in the wrong place of the function definition.

We put the storage class specifiers (ie things like "static" and
"extern") first, and the type information after that.  And while the
compiler may not care, we put the inline specifier before the types.

So it should be just

    static __always_inline void idle_init(unsigned int cpu)

instead.

Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'powerpc-5.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 25 Jul 2021 17:33:48 +0000 (10:33 -0700)]
Merge tag 'powerpc-5.14-3' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix guest to host memory corruption in H_RTAS due to missing nargs
   check.

 - Fix guest triggerable host crashes due to bad handling of nested
   guest TM state.

 - Fix possible crashes due to incorrect reference counting in
   kvm_arch_vcpu_ioctl().

 - Two commits fixing some regressions in KVM transactional memory
   handling introduced by the recent rework of the KVM code.

Thanks to Nicholas Piggin, Alexey Kardashevskiy, and Michael Neuling.

* tag 'powerpc-5.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  KVM: PPC: Book3S HV Nested: Sanitise H_ENTER_NESTED TM state
  KVM: PPC: Book3S: Fix H_RTAS rets buffer overflow
  KVM: PPC: Fix kvm_arch_vcpu_ioctl vcpu_load leak
  KVM: PPC: Book3S: Fix CONFIG_TRANSACTIONAL_MEM=n crash
  KVM: PPC: Book3S HV P9: Fix guest TM support

3 years agoMerge tag 'timers-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 25 Jul 2021 17:27:44 +0000 (10:27 -0700)]
Merge tag 'timers-urgent-2021-07-25' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Thomas Gleixner:
 "A small set of timer related fixes:

   - Plug a race between rearm and process tick in the posix CPU timers
     code

   - Make the optimization to avoid recalculation of the next timer
     interrupt work correctly when there are no timers pending"

* tag 'timers-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timers: Fix get_next_timer_interrupt() with no timers pending
  posix-cpu-timers: Fix rearm racing against process tick

3 years agoMerge tag 'locking-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 25 Jul 2021 17:21:19 +0000 (10:21 -0700)]
Merge tag 'locking-urgent-2021-07-25' of git://git./linux/kernel/git/tip/tip

Pull x86 jump label fix from Thomas Gleixner:
 "A single fix for jump labels to prevent the compiler from agressive
  un-inlining which results in a section mismatch"

* tag 'locking-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  jump_labels: Mark __jump_label_transform() as __always_inlined to work around aggressive compiler un-inlining

3 years agoMerge tag 'efi-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 25 Jul 2021 17:04:27 +0000 (10:04 -0700)]
Merge tag 'efi-urgent-2021-07-25' of git://git./linux/kernel/git/tip/tip

Pull EFI fixes from Thomas Gleixner:
 "A set of EFI fixes:

   - Prevent memblock and I/O reserved resources to get out of sync when
     EFI memreserve is in use.

   - Don't claim a non-existing table is invalid

   - Don't warn when firmware memory is already reserved correctly"

* tag 'efi-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  efi/mokvar: Reserve the table only if it is in boot services data
  efi/libstub: Fix the efi_load_initrd function description
  firmware/efi: Tell memblock about EFI iomem reservations
  efi/tpm: Differentiate missing and invalid final event log table.

3 years agoMerge tag 'core-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 25 Jul 2021 16:52:48 +0000 (09:52 -0700)]
Merge tag 'core-urgent-2021-07-25' of git://git./linux/kernel/git/tip/tip

Pull core fix from Thomas Gleixner:
 "A single update for the boot code to prevent aggressive un-inlining
  which causes a section mismatch"

* tag 'core-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  smpboot: Mark idle_init() as __always_inlined to work around aggressive compiler un-inlining

3 years agoMerge tag 'dma-mapping-5.14-1' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Sun, 25 Jul 2021 16:46:17 +0000 (09:46 -0700)]
Merge tag 'dma-mapping-5.14-1' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:

 - handle vmalloc addresses in dma_common_{mmap,get_sgtable} (Roman
   Skakun)

* tag 'dma-mapping-5.14-1' of git://git.infradead.org/users/hch/dma-mapping:
  dma-mapping: handle vmalloc addresses in dma_common_{mmap,get_sgtable}

3 years agoMerge tag '5.14-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 25 Jul 2021 00:26:47 +0000 (17:26 -0700)]
Merge tag '5.14-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Five cifs/smb3 fixes, including a DFS failover fix, two fallocate
  fixes, and two trivial coverity cleanups"

* tag '5.14-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix fallocate when trying to allocate a hole.
  CIFS: Clarify SMB1 code for POSIX delete file
  CIFS: Clarify SMB1 code for POSIX Create
  cifs: support share failover when remounting
  cifs: only write 64kb at a time when fallocating a small region of a file

3 years agoMerge tag 'riscv-for-linus-5.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 24 Jul 2021 22:34:04 +0000 (15:34 -0700)]
Merge tag 'riscv-for-linus-5.14-rc3' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:

 - properly set the memory size, which fixes 32-bit systems

 - allow initrd to load anywhere in memory, rather that restricting it
   to the first 256MiB

 - fix the 'mem=' parameter on 64-bit systems to properly account for
   the maximum supported memory now that the kernel is outside the
   linear map

 - avoid installing mappings into the last 4KiB of memory, which
   conflicts with error values

 - avoid the stack from being freed while it is being walked

 - a handful of fixes to the new copy to/from user routines

* tag 'riscv-for-linus-5.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: __asm_copy_to-from_user: Fix: Typos in comments
  riscv: __asm_copy_to-from_user: Remove unnecessary size check
  riscv: __asm_copy_to-from_user: Fix: fail on RV32
  riscv: __asm_copy_to-from_user: Fix: overrun copy
  riscv: stacktrace: pin the task's stack in get_wchan
  riscv: Make sure the kernel mapping does not overlap with IS_ERR_VALUE
  riscv: Make sure the linear mapping does not use the kernel mapping
  riscv: Fix memory_limit for 64-bit kernel
  RISC-V: load initrd wherever it fits into memory
  riscv: Fix 32-bit RISC-V boot failure

3 years agoACPI: fix NULL pointer dereference
Linus Torvalds [Sat, 24 Jul 2021 22:25:54 +0000 (15:25 -0700)]
ACPI: fix NULL pointer dereference

Commit 71f642833284 ("ACPI: utils: Fix reference counting in
for_each_acpi_dev_match()") started doing "acpi_dev_put()" on a pointer
that was possibly NULL.  That fails miserably, because that helper
inline function is not set up to handle that case.

Just make acpi_dev_put() silently accept a NULL pointer, rather than
calling down to put_device() with an invalid offset off that NULL
pointer.

Link: https://lore.kernel.org/lkml/a607c149-6bf6-0fd0-0e31-100378504da2@kernel.dk/
Reported-and-tested-by: Jens Axboe <axboe@kernel.dk>
Tested-by: Daniel Scally <djrscally@gmail.com>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 24 Jul 2021 20:08:31 +0000 (13:08 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Four fixes, all in drivers, all of which can lead to user visible
  problems in certain situations"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: target: Fix NULL dereference on XCOPY completion
  scsi: mpt3sas: Transition IOC to Ready state during shutdown
  scsi: target: Fix protect handling in WRITE SAME(32)
  scsi: iscsi: Fix iface sysfs attr detection

3 years agoMerge tag 'io_uring-5.14-2021-07-24' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 24 Jul 2021 20:03:40 +0000 (13:03 -0700)]
Merge tag 'io_uring-5.14-2021-07-24' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:

 - Fix a memory leak due to a race condition in io_init_wq_offload
   (Yang)

 - Poll error handling fixes (Pavel)

 - Fix early fdput() regression (me)

 - Don't reissue iopoll requests off release path (me)

 - Add a safety check for io-wq queue off wrong path (me)

* tag 'io_uring-5.14-2021-07-24' of git://git.kernel.dk/linux-block:
  io_uring: explicitly catch any illegal async queue attempt
  io_uring: never attempt iopoll reissue from release path
  io_uring: fix early fdput() of file
  io_uring: fix memleak in io_init_wq_offload()
  io_uring: remove double poll entry on arm failure
  io_uring: explicitly count entries for poll reqs

3 years agoMerge tag 'block-5.14-2021-07-24' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 24 Jul 2021 19:57:06 +0000 (12:57 -0700)]
Merge tag 'block-5.14-2021-07-24' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - NVMe pull request (Christoph):
    - tracing fix (Keith Busch)
    - fix multipath head refcounting (Hannes Reinecke)
    - Write Zeroes vs PI fix (me)
    - drop a bogus WARN_ON (Zhihao Cheng)

 - Increase max blk-cgroup policy size, now that mq-deadline
   uses it too (Oleksandr)

* tag 'block-5.14-2021-07-24' of git://git.kernel.dk/linux-block:
  nvme: set the PRACT bit when using Write Zeroes with T10 PI
  nvme: fix nvme_setup_command metadata trace event
  nvme: fix refcounting imbalance when all paths are down
  nvme-pci: don't WARN_ON in nvme_reset_work if ctrl.state is not RESETTING
  block: increase BLKCG_MAX_POLS

3 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 24 Jul 2021 19:55:06 +0000 (12:55 -0700)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "Two bugfixes for the I2C subsystem"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: mpc: Poll for MCF
  misc: eeprom: at24: Always append device id even if label property is set.

3 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 24 Jul 2021 19:27:16 +0000 (12:27 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc mm fixes from Andrew Morton:
 "15 patches.

  VM subsystems affected by this patch series: userfaultfd, kfence,
  highmem, pagealloc, memblock, pagecache, secretmem, pagemap, and
  hugetlbfs"

* akpm:
  hugetlbfs: fix mount mode command line processing
  mm: fix the deadlock in finish_fault()
  mm: mmap_lock: fix disabling preemption directly
  mm/secretmem: wire up ->set_page_dirty
  writeback, cgroup: do not reparent dax inodes
  writeback, cgroup: remove wb from offline list before releasing refcnt
  memblock: make for_each_mem_range() traverse MEMBLOCK_HOTPLUG regions
  mm: page_alloc: fix page_poison=1 / INIT_ON_ALLOC_DEFAULT_ON interaction
  mm: use kmap_local_page in memzero_page
  mm: call flush_dcache_page() in memcpy_to_page() and memzero_page()
  kfence: skip all GFP_ZONEMASK allocations
  kfence: move the size check to the beginning of __kfence_alloc()
  kfence: defer kfence_test_init to ensure that kunit debugfs is created
  selftest: use mmap instead of posix_memalign to allocate memory
  userfaultfd: do not untag user pointers

3 years agoriscv: __asm_copy_to-from_user: Fix: Typos in comments
Akira Tsukamoto [Tue, 20 Jul 2021 08:53:23 +0000 (17:53 +0900)]
riscv: __asm_copy_to-from_user: Fix: Typos in comments

Fixing typos and grammar mistakes and using more intuitive label
name.

Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa210de ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoriscv: __asm_copy_to-from_user: Remove unnecessary size check
Akira Tsukamoto [Tue, 20 Jul 2021 08:52:36 +0000 (17:52 +0900)]
riscv: __asm_copy_to-from_user: Remove unnecessary size check

Clean up:

The size of 0 will be evaluated in the next step. Not
required here.

Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa210de ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoriscv: __asm_copy_to-from_user: Fix: fail on RV32
Akira Tsukamoto [Tue, 20 Jul 2021 08:51:45 +0000 (17:51 +0900)]
riscv: __asm_copy_to-from_user: Fix: fail on RV32

Had a bug when converting bytes to bits when the cpu was rv32.

The a3 contains the number of bytes and multiple of 8
would be the bits. The LGREG is holding 2 for RV32 and 3 for
RV32, so to achieve multiple of 8 it must always be constant 3.
The 2 was mistakenly used for rv32.

Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa210de ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoriscv: __asm_copy_to-from_user: Fix: overrun copy
Akira Tsukamoto [Tue, 20 Jul 2021 08:50:52 +0000 (17:50 +0900)]
riscv: __asm_copy_to-from_user: Fix: overrun copy

There were two causes for the overrun memory access.

The threshold size was too small.
The aligning dst require one SZREG and unrolling word copy requires
8*SZREG, total have to be at least 9*SZREG.

Inside the unrolling copy, the subtracting -(8*SZREG-1) would make
iteration happening one extra loop. Proper value is -(8*SZREG).

Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa210de ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agohugetlbfs: fix mount mode command line processing
Mike Kravetz [Fri, 23 Jul 2021 22:50:44 +0000 (15:50 -0700)]
hugetlbfs: fix mount mode command line processing

In commit 32021982a324 ("hugetlbfs: Convert to fs_context") processing
of the mount mode string was changed from match_octal() to fsparam_u32.

This changed existing behavior as match_octal does not require octal
values to have a '0' prefix, but fsparam_u32 does.

Use fsparam_u32oct which provides the same behavior as match_octal.

Link: https://lkml.kernel.org/r/20210721183326.102716-1-mike.kravetz@oracle.com
Fixes: 32021982a324 ("hugetlbfs: Convert to fs_context")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reported-by: Dennis Camera <bugs+kernel.org@dtnr.ch>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: fix the deadlock in finish_fault()
Qi Zheng [Fri, 23 Jul 2021 22:50:41 +0000 (15:50 -0700)]
mm: fix the deadlock in finish_fault()

Commit 63f3655f9501 ("mm, memcg: fix reclaim deadlock with writeback")
fix the following ABBA deadlock by pre-allocating the pte page table
without holding the page lock.

                                lock_page(A)
                                        SetPageWriteback(A)
                                        unlock_page(A)
  lock_page(B)
                                        lock_page(B)
  pte_alloc_one
    shrink_page_list
      wait_on_page_writeback(A)
                                        SetPageWriteback(B)
                                        unlock_page(B)

                                        # flush A, B to clear the writeback

Commit f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault()
codepaths") reworked the relevant code but ignored this race.  This will
cause the deadlock above to appear again, so fix it.

Link: https://lkml.kernel.org/r/20210721074849.57004-1-zhengqi.arch@bytedance.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: mmap_lock: fix disabling preemption directly
Muchun Song [Fri, 23 Jul 2021 22:50:38 +0000 (15:50 -0700)]
mm: mmap_lock: fix disabling preemption directly

Commit 832b50725373 ("mm: mmap_lock: use local locks instead of
disabling preemption") fixed a bug by using local locks.

But commit d01079f3d0c0 ("mm/mmap_lock: remove dead code for
!CONFIG_TRACING configurations") changed those lines back to the
original version.

I guess it was introduced by fixing conflicts.

Link: https://lkml.kernel.org/r/20210720074228.76342-1-songmuchun@bytedance.com
Fixes: d01079f3d0c0 ("mm/mmap_lock: remove dead code for !CONFIG_TRACING configurations")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/secretmem: wire up ->set_page_dirty
Mike Rapoport [Fri, 23 Jul 2021 22:50:35 +0000 (15:50 -0700)]
mm/secretmem: wire up ->set_page_dirty

Make secretmem up to date with the changes done in commit 0af573780b0b
("mm: require ->set_page_dirty to be explicitly wired up") so that
unconditional call to this method won't cause crashes.

Link: https://lkml.kernel.org/r/20210716063933.31633-1-rppt@kernel.org
Fixes: 0af573780b0b ("mm: require ->set_page_dirty to be explicitly wired up")
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agowriteback, cgroup: do not reparent dax inodes
Roman Gushchin [Fri, 23 Jul 2021 22:50:32 +0000 (15:50 -0700)]
writeback, cgroup: do not reparent dax inodes

The inode switching code is not suited for dax inodes.  An attempt to
switch a dax inode to a parent writeback structure (as a part of a
writeback cleanup procedure) results in a panic like this:

  run fstests generic/270 at 2021-07-15 05:54:02
  XFS (pmem0p2): EXPERIMENTAL big timestamp feature in use.  Use at your own risk!
  XFS (pmem0p2): DAX enabled. Warning: EXPERIMENTAL, use at your own risk
  XFS (pmem0p2): EXPERIMENTAL inode btree counters feature in use. Use at your own risk!
  XFS (pmem0p2): Mounting V5 Filesystem
  XFS (pmem0p2): Ending clean mount
  XFS (pmem0p2): Quotacheck needed: Please wait.
  XFS (pmem0p2): Quotacheck: Done.
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  BUG: unable to handle page fault for address: 0000000005b0f669
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 0 P4D 0
  Oops: 0000 [#1] SMP PTI
  CPU: 13 PID: 10479 Comm: kworker/13:16 Not tainted 5.14.0-rc1-master-8096acd7442e+ #8
  Hardware name: HP ProLiant DL360 Gen9/ProLiant DL360 Gen9, BIOS P89 09/13/2016
  Workqueue: inode_switch_wbs inode_switch_wbs_work_fn
  RIP: 0010:inode_do_switch_wbs+0xaf/0x470
  Code: 00 30 0f 85 c1 03 00 00 0f 1f 44 00 00 31 d2 48 c7 c6 ff ff ff ff 48 8d 7c 24 08 e8 eb 49 1a 00 48 85 c0 74 4a bb ff ff ff ff <48> 8b 50 08 48 8d 4a ff 83 e2 01 48 0f 45 c1 48 8b 00 a8 08 0f 85
  RSP: 0018:ffff9c66691abdc8 EFLAGS: 00010002
  RAX: 0000000005b0f661 RBX: 00000000ffffffff RCX: ffff89e6a21382b0
  RDX: 0000000000000001 RSI: ffff89e350230248 RDI: ffffffffffffffff
  RBP: ffff89e681d19400 R08: 0000000000000000 R09: 0000000000000228
  R10: ffffffffffffffff R11: ffffffffffffffc0 R12: ffff89e6a2138130
  R13: ffff89e316af7400 R14: ffff89e316af6e78 R15: ffff89e6a21382b0
  FS:  0000000000000000(0000) GS:ffff89ee5fb40000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000005b0f669 CR3: 0000000cb2410004 CR4: 00000000001706e0
  Call Trace:
   inode_switch_wbs_work_fn+0xb6/0x2a0
   process_one_work+0x1e6/0x380
   worker_thread+0x53/0x3d0
   kthread+0x10f/0x130
   ret_from_fork+0x22/0x30
  Modules linked in: xt_CHECKSUM xt_MASQUERADE xt_conntrack ipt_REJECT nf_reject_ipv4 nft_compat nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 nft_counter nf_tables nfnetlink bridge stp llc rfkill sunrpc intel_rapl_msr intel_rapl_common sb_edac x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel ipmi_ssif kvm mgag200 i2c_algo_bit iTCO_wdt irqbypass drm_kms_helper iTCO_vendor_support acpi_ipmi rapl syscopyarea sysfillrect intel_cstate ipmi_si sysimgblt ioatdma dax_pmem_compat fb_sys_fops ipmi_devintf device_dax i2c_i801 pcspkr intel_uncore hpilo nd_pmem cec dax_pmem_core dca i2c_smbus acpi_tad lpc_ich ipmi_msghandler acpi_power_meter drm fuse xfs libcrc32c sd_mod t10_pi crct10dif_pclmul crc32_pclmul crc32c_intel tg3 ghash_clmulni_intel serio_raw hpsa hpwdt scsi_transport_sas wmi dm_mirror dm_region_hash dm_log dm_mod
  CR2: 0000000005b0f669
  ---[ end trace ed2105faff8384f3 ]---
  RIP: 0010:inode_do_switch_wbs+0xaf/0x470
  Code: 00 30 0f 85 c1 03 00 00 0f 1f 44 00 00 31 d2 48 c7 c6 ff ff ff ff 48 8d 7c 24 08 e8 eb 49 1a 00 48 85 c0 74 4a bb ff ff ff ff <48> 8b 50 08 48 8d 4a ff 83 e2 01 48 0f 45 c1 48 8b 00 a8 08 0f 85
  RSP: 0018:ffff9c66691abdc8 EFLAGS: 00010002
  RAX: 0000000005b0f661 RBX: 00000000ffffffff RCX: ffff89e6a21382b0
  RDX: 0000000000000001 RSI: ffff89e350230248 RDI: ffffffffffffffff
  RBP: ffff89e681d19400 R08: 0000000000000000 R09: 0000000000000228
  R10: ffffffffffffffff R11: ffffffffffffffc0 R12: ffff89e6a2138130
  R13: ffff89e316af7400 R14: ffff89e316af6e78 R15: ffff89e6a21382b0
  FS:  0000000000000000(0000) GS:ffff89ee5fb40000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000005b0f669 CR3: 0000000cb2410004 CR4: 00000000001706e0
  Kernel panic - not syncing: Fatal exception
  Kernel Offset: 0x15200000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
  ---[ end Kernel panic - not syncing: Fatal exception ]---

The crash happens on an attempt to iterate over attached pagecache pages
and check the dirty flag: a dax inode's xarray contains pfn's instead of
generic struct page pointers.

This happens for DAX and not for other kinds of non-page entries in the
inodes because it's a tagged iteration, and shadow/swap entries are
never tagged; only DAX entries get tagged.

Fix the problem by bailing out (with the false return value) of
inode_prepare_sbs_switch() if a dax inode is passed.

[willy@infradead.org: changelog addition]

Link: https://lkml.kernel.org/r/20210719171350.3876830-1-guro@fb.com
Fixes: c22d70a162d3 ("writeback, cgroup: release dying cgwbs by switching attached inodes")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Murphy Zhou <jencce.kernel@gmail.com>
Reported-by: Darrick J. Wong <djwong@kernel.org>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Tested-by: Murphy Zhou <jencce.kernel@gmail.com>
Acked-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agowriteback, cgroup: remove wb from offline list before releasing refcnt
Roman Gushchin [Fri, 23 Jul 2021 22:50:29 +0000 (15:50 -0700)]
writeback, cgroup: remove wb from offline list before releasing refcnt

Boyang reported that the commit c22d70a162d3 ("writeback, cgroup:
release dying cgwbs by switching attached inodes") causes the kernel to
crash while running xfstests generic/256 on ext4 on aarch64 and ppc64le.

  run fstests generic/256 at 2021-07-12 05:41:40
  EXT4-fs (vda3): mounted filesystem with ordered data mode. Opts: . Quota mode: none.
  Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
  Mem abort info:
     ESR = 0x96000005
     EC = 0x25: DABT (current EL), IL = 32 bits
     SET = 0, FnV = 0
     EA = 0, S1PTW = 0
     FSC = 0x05: level 1 translation fault
  Data abort info:
     ISV = 0, ISS = 0x00000005
     CM = 0, WnR = 0
  user pgtable: 64k pages, 48-bit VAs, pgdp=00000000b0502000
  [0000000000000000] pgd=0000000000000000, p4d=0000000000000000, pud=0000000000000000
  Internal error: Oops: 96000005 [#1] SMP
  Modules linked in: dm_flakey dm_snapshot dm_bufio dm_zero dm_mod loop tls rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache netfs rfkill sunrpc ext4 vfat fat mbcache jbd2 drm fuse xfs libcrc32c crct10dif_ce ghash_ce sha2_ce sha256_arm64 sha1_ce virtio_blk virtio_net net_failover virtio_console failover virtio_mmio aes_neon_bs [last unloaded: scsi_debug]
  CPU: 0 PID: 408468 Comm: kworker/u8:5 Tainted: G X --------- ---  5.14.0-0.rc1.15.bx.el9.aarch64 #1
  Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
  Workqueue: events_unbound cleanup_offline_cgwbs_workfn
  pstate: 004000c5 (nzcv daIF +PAN -UAO -TCO BTYPE=--)
  pc : cleanup_offline_cgwbs_workfn+0x320/0x394
  lr : cleanup_offline_cgwbs_workfn+0xe0/0x394
  sp : ffff80001554fd10
  x29: ffff80001554fd10 x28: 0000000000000000 x27: 0000000000000001
  x26: 0000000000000000 x25: 00000000000000e0 x24: ffffd2a2fbe671a8
  x23: ffff80001554fd88 x22: ffffd2a2fbe67198 x21: ffffd2a2fc25a730
  x20: ffff210412bc3000 x19: ffff210412bc3280 x18: 0000000000000000
  x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
  x14: 0000000000000000 x13: 0000000000000030 x12: 0000000000000040
  x11: ffff210481572238 x10: ffff21048157223a x9 : ffffd2a2fa276c60
  x8 : ffff210484106b60 x7 : 0000000000000000 x6 : 000000000007d18a
  x5 : ffff210416a86400 x4 : ffff210412bc0280 x3 : 0000000000000000
  x2 : ffff80001554fd88 x1 : ffff210412bc0280 x0 : 0000000000000003
  Call trace:
     cleanup_offline_cgwbs_workfn+0x320/0x394
     process_one_work+0x1f4/0x4b0
     worker_thread+0x184/0x540
     kthread+0x114/0x120
     ret_from_fork+0x10/0x18
  Code: d63f0020 97f99963 17ffffa6 f8588263 (f9400061)
  ---[ end trace e250fe289272792a ]---
  Kernel panic - not syncing: Oops: Fatal exception
  SMP: stopping secondary CPUs
  SMP: failed to stop secondary CPUs 0-2
  Kernel Offset: 0x52a2e9fa0000 from 0xffff800010000000
  PHYS_OFFSET: 0xfff0defca0000000
  CPU features: 0x00200251,23200840
  Memory Limit: none
  ---[ end Kernel panic - not syncing: Oops: Fatal exception ]---

The problem happens when cgwb_release_workfn() races with
cleanup_offline_cgwbs_workfn(): wb_tryget() in
cleanup_offline_cgwbs_workfn() can be called after percpu_ref_exit() is
cgwb_release_workfn(), which is basically a use-after-free error.

Fix the problem by making removing the writeback structure from the
offline list before releasing the percpu reference counter.  It will
guarantee that cleanup_offline_cgwbs_workfn() will not see and not
access writeback structures which are about to be released.

Link: https://lkml.kernel.org/r/20210716201039.3762203-1-guro@fb.com
Fixes: c22d70a162d3 ("writeback, cgroup: release dying cgwbs by switching attached inodes")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Boyang Xue <bxue@redhat.com>
Suggested-by: Jan Kara <jack@suse.cz>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Murphy Zhou <jencce.kernel@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomemblock: make for_each_mem_range() traverse MEMBLOCK_HOTPLUG regions
Mike Rapoport [Fri, 23 Jul 2021 22:50:26 +0000 (15:50 -0700)]
memblock: make for_each_mem_range() traverse MEMBLOCK_HOTPLUG regions

Commit b10d6bca8720 ("arch, drivers: replace for_each_membock() with
for_each_mem_range()") didn't take into account that when there is
movable_node parameter in the kernel command line, for_each_mem_range()
would skip ranges marked with MEMBLOCK_HOTPLUG.

The page table setup code in POWER uses for_each_mem_range() to create
the linear mapping of the physical memory and since the regions marked
as MEMORY_HOTPLUG are skipped, they never make it to the linear map.

A later access to the memory in those ranges will fail:

  BUG: Unable to handle kernel data access on write at 0xc000000400000000
  Faulting instruction address: 0xc00000000008a3c0
  Oops: Kernel access of bad area, sig: 11 [#1]
  LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA pSeries
  Modules linked in:
  CPU: 0 PID: 53 Comm: kworker/u2:0 Not tainted 5.13.0 #7
  NIP:  c00000000008a3c0 LR: c0000000003c1ed8 CTR: 0000000000000040
  REGS: c000000008a57770 TRAP: 0300   Not tainted  (5.13.0)
  MSR:  8000000002009033 <SF,VEC,EE,ME,IR,DR,RI,LE>  CR: 84222202  XER: 20040000
  CFAR: c0000000003c1ed4 DAR: c000000400000000 DSISR: 42000000 IRQMASK: 0
  GPR00: c0000000003c1ed8 c000000008a57a10 c0000000019da700 c000000400000000
  GPR04: 0000000000000280 0000000000000180 0000000000000400 0000000000000200
  GPR08: 0000000000000100 0000000000000080 0000000000000040 0000000000000300
  GPR12: 0000000000000380 c000000001bc0000 c0000000001660c8 c000000006337e00
  GPR16: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
  GPR20: 0000000040000000 0000000020000000 c000000001a81990 c000000008c30000
  GPR24: c000000008c20000 c000000001a81998 000fffffffff0000 c000000001a819a0
  GPR28: c000000001a81908 c00c000001000000 c000000008c40000 c000000008a64680
  NIP clear_user_page+0x50/0x80
  LR __handle_mm_fault+0xc88/0x1910
  Call Trace:
    __handle_mm_fault+0xc44/0x1910 (unreliable)
    handle_mm_fault+0x130/0x2a0
    __get_user_pages+0x248/0x610
    __get_user_pages_remote+0x12c/0x3e0
    get_arg_page+0x54/0xf0
    copy_string_kernel+0x11c/0x210
    kernel_execve+0x16c/0x220
    call_usermodehelper_exec_async+0x1b0/0x2f0
    ret_from_kernel_thread+0x5c/0x70
  Instruction dump:
  79280fa4 79271764 79261f24 794ae8e2 7ca94214 7d683a14 7c893a14 7d893050
  7d4903a6 60000000 60000000 60000000 <7c001fec7c091fec 7c081fec 7c051fec
  ---[ end trace 490b8c67e6075e09 ]---

Making for_each_mem_range() include MEMBLOCK_HOTPLUG regions in the
traversal fixes this issue.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=1976100
Link: https://lkml.kernel.org/r/20210712071132.20902-1-rppt@kernel.org
Fixes: b10d6bca8720 ("arch, drivers: replace for_each_membock() with for_each_mem_range()")
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Tested-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: <stable@vger.kernel.org> [5.10+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: page_alloc: fix page_poison=1 / INIT_ON_ALLOC_DEFAULT_ON interaction
Sergei Trofimovich [Fri, 23 Jul 2021 22:50:23 +0000 (15:50 -0700)]
mm: page_alloc: fix page_poison=1 / INIT_ON_ALLOC_DEFAULT_ON interaction

To reproduce the failure we need the following system:

 - kernel command: page_poison=1 init_on_free=0 init_on_alloc=0

 - kernel config:
    * CONFIG_INIT_ON_ALLOC_DEFAULT_ON=y
    * CONFIG_INIT_ON_FREE_DEFAULT_ON=y
    * CONFIG_PAGE_POISONING=y

Resulting in:

    0000000085629bdd: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    0000000022861832: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00000000c597f5b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    CPU: 11 PID: 15195 Comm: bash Kdump: loaded Tainted: G     U     O      5.13.1-gentoo-x86_64 #1
    Hardware name: System manufacturer System Product Name/PRIME Z370-A, BIOS 2801 01/13/2021
    Call Trace:
     dump_stack+0x64/0x7c
     __kernel_unpoison_pages.cold+0x48/0x84
     post_alloc_hook+0x60/0xa0
     get_page_from_freelist+0xdb8/0x1000
     __alloc_pages+0x163/0x2b0
     __get_free_pages+0xc/0x30
     pgd_alloc+0x2e/0x1a0
     mm_init+0x185/0x270
     dup_mm+0x6b/0x4f0
     copy_process+0x190d/0x1b10
     kernel_clone+0xba/0x3b0
     __do_sys_clone+0x8f/0xb0
     do_syscall_64+0x68/0x80
     entry_SYSCALL_64_after_hwframe+0x44/0xae

Before commit 51cba1ebc60d ("init_on_alloc: Optimize static branches")
init_on_alloc never enabled static branch by default.  It could only be
enabed explicitly by init_mem_debugging_and_hardening().

But after commit 51cba1ebc60d, a static branch could already be enabled
by default.  There was no code to ever disable it.  That caused
page_poison=1 / init_on_free=1 conflict.

This change extends init_mem_debugging_and_hardening() to also disable
static branch disabling.

Link: https://lkml.kernel.org/r/20210714031935.4094114-1-keescook@chromium.org
Link: https://lore.kernel.org/r/20210712215816.1512739-1-slyfox@gentoo.org
Fixes: 51cba1ebc60d ("init_on_alloc: Optimize static branches")
Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Co-developed-by: Kees Cook <keescook@chromium.org>
Reported-by: Mikhail Morfikov <mmorfikov@gmail.com>
Reported-by: <bowsingbetee@pm.me>
Tested-by: <bowsingbetee@protonmail.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: use kmap_local_page in memzero_page
Christoph Hellwig [Fri, 23 Jul 2021 22:50:20 +0000 (15:50 -0700)]
mm: use kmap_local_page in memzero_page

The commit message introducing the global memzero_page explicitly
mentions switching to kmap_local_page in the commit log but doesn't
actually do that.

Link: https://lkml.kernel.org/r/20210713055231.137602-3-hch@lst.de
Fixes: 28961998f858 ("iov_iter: lift memzero_page() to highmem.h")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: call flush_dcache_page() in memcpy_to_page() and memzero_page()
Christoph Hellwig [Fri, 23 Jul 2021 22:50:17 +0000 (15:50 -0700)]
mm: call flush_dcache_page() in memcpy_to_page() and memzero_page()

memcpy_to_page and memzero_page can write to arbitrary pages, which
could be in the page cache or in high memory, so call
flush_kernel_dcache_pages to flush the dcache.

This is a problem when using these helpers on dcache challeneged
architectures.  Right now there are just a few users, chances are no one
used the PC floppy driver, the aha1542 driver for an ISA SCSI HBA, and a
few advanced and optional btrfs and ext4 features on those platforms yet
since the conversion.

Link: https://lkml.kernel.org/r/20210713055231.137602-2-hch@lst.de
Fixes: bb90d4bc7b6a ("mm/highmem: Lift memcpy_[to|from]_page to core")
Fixes: 28961998f858 ("iov_iter: lift memzero_page() to highmem.h")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Cc: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokfence: skip all GFP_ZONEMASK allocations
Alexander Potapenko [Fri, 23 Jul 2021 22:50:14 +0000 (15:50 -0700)]
kfence: skip all GFP_ZONEMASK allocations

Allocation requests outside ZONE_NORMAL (MOVABLE, HIGHMEM or DMA) cannot
be fulfilled by KFENCE, because KFENCE memory pool is located in a zone
different from the requested one.

Because callers of kmem_cache_alloc() may actually rely on the
allocation to reside in the requested zone (e.g.  memory allocations
done with __GFP_DMA must be DMAable), skip all allocations done with
GFP_ZONEMASK and/or respective SLAB flags (SLAB_CACHE_DMA and
SLAB_CACHE_DMA32).

Link: https://lkml.kernel.org/r/20210714092222.1890268-2-glider@google.com
Fixes: 0ce20dd84089 ("mm: add Kernel Electric-Fence infrastructure")
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Marco Elver <elver@google.com>
Acked-by: Souptick Joarder <jrdr.linux@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Souptick Joarder <jrdr.linux@gmail.com>
Cc: <stable@vger.kernel.org> [5.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokfence: move the size check to the beginning of __kfence_alloc()
Alexander Potapenko [Fri, 23 Jul 2021 22:50:11 +0000 (15:50 -0700)]
kfence: move the size check to the beginning of __kfence_alloc()

Check the allocation size before toggling kfence_allocation_gate.

This way allocations that can't be served by KFENCE will not result in
waiting for another CONFIG_KFENCE_SAMPLE_INTERVAL without allocating
anything.

Link: https://lkml.kernel.org/r/20210714092222.1890268-1-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Suggested-by: Marco Elver <elver@google.com>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Marco Elver <elver@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: <stable@vger.kernel.org> [5.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokfence: defer kfence_test_init to ensure that kunit debugfs is created
Weizhao Ouyang [Fri, 23 Jul 2021 22:50:08 +0000 (15:50 -0700)]
kfence: defer kfence_test_init to ensure that kunit debugfs is created

kfence_test_init and kunit_init both use the same level late_initcall,
which means if kfence_test_init linked ahead of kunit_init,
kfence_test_init will get a NULL debugfs_rootdir as parent dentry, then
kfence_test_init and kfence_debugfs_init both create a debugfs node
named "kfence" under debugfs_mount->mnt_root, and it will throw out
"debugfs: Directory 'kfence' with parent '/' already present!" with
EEXIST.  So kfence_test_init should be deferred.

Link: https://lkml.kernel.org/r/20210714113140.2949995-1-o451686892@gmail.com
Signed-off-by: Weizhao Ouyang <o451686892@gmail.com>
Tested-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoselftest: use mmap instead of posix_memalign to allocate memory
Peter Collingbourne [Fri, 23 Jul 2021 22:50:04 +0000 (15:50 -0700)]
selftest: use mmap instead of posix_memalign to allocate memory

This test passes pointers obtained from anon_allocate_area to the
userfaultfd and mremap APIs.  This causes a problem if the system
allocator returns tagged pointers because with the tagged address ABI
the kernel rejects tagged addresses passed to these APIs, which would
end up causing the test to fail.  To make this test compatible with such
system allocators, stop using the system allocator to allocate memory in
anon_allocate_area, and instead just use mmap.

Link: https://lkml.kernel.org/r/20210714195437.118982-3-pcc@google.com
Link: https://linux-review.googlesource.com/id/Icac91064fcd923f77a83e8e133f8631c5b8fc241
Fixes: c47174fc362a ("userfaultfd: selftest")
Co-developed-by: Lokesh Gidra <lokeshgidra@google.com>
Signed-off-by: Lokesh Gidra <lokeshgidra@google.com>
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Dave Martin <Dave.Martin@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Alistair Delva <adelva@google.com>
Cc: William McVicker <willmcvicker@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Mitch Phillips <mitchp@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: <stable@vger.kernel.org> [5.4]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agouserfaultfd: do not untag user pointers
Peter Collingbourne [Fri, 23 Jul 2021 22:50:01 +0000 (15:50 -0700)]
userfaultfd: do not untag user pointers

Patch series "userfaultfd: do not untag user pointers", v5.

If a user program uses userfaultfd on ranges of heap memory, it may end
up passing a tagged pointer to the kernel in the range.start field of
the UFFDIO_REGISTER ioctl.  This can happen when using an MTE-capable
allocator, or on Android if using the Tagged Pointers feature for MTE
readiness [1].

When a fault subsequently occurs, the tag is stripped from the fault
address returned to the application in the fault.address field of struct
uffd_msg.  However, from the application's perspective, the tagged
address *is* the memory address, so if the application is unaware of
memory tags, it may get confused by receiving an address that is, from
its point of view, outside of the bounds of the allocation.  We observed
this behavior in the kselftest for userfaultfd [2] but other
applications could have the same problem.

Address this by not untagging pointers passed to the userfaultfd ioctls.
Instead, let the system call fail.  Also change the kselftest to use
mmap so that it doesn't encounter this problem.

[1] https://source.android.com/devices/tech/debug/tagged-pointers
[2] tools/testing/selftests/vm/userfaultfd.c

This patch (of 2):

Do not untag pointers passed to the userfaultfd ioctls.  Instead, let
the system call fail.  This will provide an early indication of problems
with tag-unaware userspace code instead of letting the code get confused
later, and is consistent with how we decided to handle brk/mmap/mremap
in commit dcde237319e6 ("mm: Avoid creating virtual address aliases in
brk()/mmap()/mremap()"), as well as being consistent with the existing
tagged address ABI documentation relating to how ioctl arguments are
handled.

The code change is a revert of commit 7d0325749a6c ("userfaultfd: untag
user pointers") plus some fixups to some additional calls to
validate_range that have appeared since then.

[1] https://source.android.com/devices/tech/debug/tagged-pointers
[2] tools/testing/selftests/vm/userfaultfd.c

Link: https://lkml.kernel.org/r/20210714195437.118982-1-pcc@google.com
Link: https://lkml.kernel.org/r/20210714195437.118982-2-pcc@google.com
Link: https://linux-review.googlesource.com/id/I761aa9f0344454c482b83fcfcce547db0a25501b
Fixes: 63f0c6037965 ("arm64: Introduce prctl() options to control the tagged user addresses ABI")
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Alistair Delva <adelva@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Martin <Dave.Martin@arm.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mitch Phillips <mitchp@google.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: William McVicker <willmcvicker@google.com>
Cc: <stable@vger.kernel.org> [5.4]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoriscv: stacktrace: pin the task's stack in get_wchan
Jisheng Zhang [Fri, 23 Jul 2021 00:22:26 +0000 (08:22 +0800)]
riscv: stacktrace: pin the task's stack in get_wchan

Pin the task's stack before calling walk_stackframe() in get_wchan().
This can fix the panic as reported by Andreas when CONFIG_VMAP_STACK=y:

[   65.609696] Unable to handle kernel paging request at virtual address ffffffd0003bbde8
[   65.610460] Oops [#1]
[   65.610626] Modules linked in: virtio_blk virtio_mmio rtc_goldfish btrfs blake2b_generic libcrc32c xor raid6_pq sg dm_multipath dm_mod scsi_dh_rdac scsi_dh_emc scsi_dh_alua efivarfs
[   65.611670] CPU: 2 PID: 1 Comm: systemd Not tainted 5.14.0-rc1-1.g34fe32a-default #1 openSUSE Tumbleweed (unreleased) c62f7109153e5a0897ee58ba52393ad99b070fd2
[   65.612334] Hardware name: riscv-virtio,qemu (DT)
[   65.613008] epc : get_wchan+0x5c/0x88
[   65.613334]  ra : get_wchan+0x42/0x88
[   65.613625] epc : ffffffff800048a4 ra : ffffffff8000488a sp : ffffffd00021bb90
[   65.614008]  gp : ffffffff817709f8 tp : ffffffe07fe91b80 t0 : 00000000000001f8
[   65.614411]  t1 : 0000000000020000 t2 : 0000000000000000 s0 : ffffffd00021bbd0
[   65.614818]  s1 : ffffffd0003bbdf0 a0 : 0000000000000001 a1 : 0000000000000002
[   65.615237]  a2 : ffffffff81618008 a3 : 0000000000000000 a4 : 0000000000000000
[   65.615637]  a5 : ffffffd0003bc000 a6 : 0000000000000002 a7 : ffffffe27d370000
[   65.616022]  s2 : ffffffd0003bbd90 s3 : ffffffff8071a81e s4 : 0000000000003fff
[   65.616407]  s5 : ffffffffffffc000 s6 : 0000000000000000 s7 : ffffffff81618008
[   65.616845]  s8 : 0000000000000001 s9 : 0000000180000040 s10: 0000000000000000
[   65.617248]  s11: 000000000000016b t3 : 000000ff00000000 t4 : 0c6aec92de5e3fd7
[   65.617672]  t5 : fff78f60608fcfff t6 : 0000000000000078
[   65.618088] status: 0000000000000120 badaddr: ffffffd0003bbde8 cause: 000000000000000d
[   65.618621] [<ffffffff800048a4>] get_wchan+0x5c/0x88
[   65.619008] [<ffffffff8022da88>] do_task_stat+0x7a2/0xa46
[   65.619325] [<ffffffff8022e87e>] proc_tgid_stat+0xe/0x16
[   65.619637] [<ffffffff80227dd6>] proc_single_show+0x46/0x96
[   65.619979] [<ffffffff801ccb1e>] seq_read_iter+0x190/0x31e
[   65.620341] [<ffffffff801ccd70>] seq_read+0xc4/0x104
[   65.620633] [<ffffffff801a6bfe>] vfs_read+0x6a/0x112
[   65.620922] [<ffffffff801a701c>] ksys_read+0x54/0xbe
[   65.621206] [<ffffffff801a7094>] sys_read+0xe/0x16
[   65.621474] [<ffffffff8000303e>] ret_from_syscall+0x0/0x2
[   65.622169] ---[ end trace f24856ed2b8789c5 ]---
[   65.622832] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b

Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoio_uring: explicitly catch any illegal async queue attempt
Jens Axboe [Fri, 23 Jul 2021 17:53:54 +0000 (11:53 -0600)]
io_uring: explicitly catch any illegal async queue attempt

Catch an illegal case to queue async from an unrelated task that got
the ring fd passed to it. This should not be possible to hit, but
better be proactive and catch it explicitly. io-wq is extended to
check for early IO_WQ_WORK_CANCEL being set on a work item as well,
so it can run the request through the normal cancelation path.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
3 years agoio_uring: never attempt iopoll reissue from release path
Jens Axboe [Fri, 23 Jul 2021 17:49:29 +0000 (11:49 -0600)]
io_uring: never attempt iopoll reissue from release path

There are two reasons why this shouldn't be done:

1) Ring is exiting, and we're canceling requests anyway. Any request
   should be canceled anyway. In theory, this could iterate for a
   number of times if someone else is also driving the target block
   queue into request starvation, however the likelihood of this
   happening is miniscule.

2) If the original task decided to pass the ring to another task, then
   we don't want to be reissuing from this context as it may be an
   unrelated task or context. No assumptions should be made about
   the context in which ->release() is run. This can only happen for pure
   read/write, and we'll get -EFAULT on them anyway.

Link: https://lore.kernel.org/io-uring/YPr4OaHv0iv0KTOc@zeniv-ca.linux.org.uk/
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
3 years agoMerge tag 'for-5.14-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Fri, 23 Jul 2021 19:49:07 +0000 (12:49 -0700)]
Merge tag 'for-5.14-rc2-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "A few fixes and one patch to help some block layer API cleanups:

   - skip missing device when running fstrim

   - fix unpersisted i_size on fsync after expanding truncate

   - fix lock inversion problem when doing qgroup extent tracing

   - replace bdgrab/bdput usage, replace gendisk by block_device"

* tag 'for-5.14-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: store a block_device in struct btrfs_ordered_extent
  btrfs: fix lock inversion problem when doing qgroup extent tracing
  btrfs: check for missing device in btrfs_trim_fs
  btrfs: fix unpersisted i_size on fsync after expanding truncate

3 years agoMerge tag 'ceph-for-5.14-rc3' of git://github.com/ceph/ceph-client
Linus Torvalds [Fri, 23 Jul 2021 18:30:12 +0000 (11:30 -0700)]
Merge tag 'ceph-for-5.14-rc3' of git://github.com/ceph/ceph-client

Pull ceph fixes from Ilya Dryomov:
 "A subtle deadlock on lock_rwsem (marked for stable) and rbd fixes for
  a -rc1 regression.

  Also included a rare WARN condition tweak"

* tag 'ceph-for-5.14-rc3' of git://github.com/ceph/ceph-client:
  rbd: resurrect setting of disk->private_data in rbd_init_disk()
  ceph: don't WARN if we're still opening a session to an MDS
  rbd: don't hold lock_rwsem while running_list is being drained
  rbd: always kick acquire on "acquired" and "released" notifications

3 years agoMerge tag 'trace-v5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Fri, 23 Jul 2021 18:25:21 +0000 (11:25 -0700)]
Merge tag 'trace-v5.14-rc2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:

 - Fix deadloop in ring buffer because of using stale "read" variable

 - Fix synthetic event use of field_pos as boolean and not an index

 - Fixed histogram special var "cpu" overriding event fields called
   "cpu"

 - Cleaned up error prone logic in alloc_synth_event()

 - Removed call to synchronize_rcu_tasks_rude() when not needed

 - Removed redundant initialization of a local variable "ret"

 - Fixed kernel crash when updating tracepoint callbacks of different
   priorities.

* tag 'trace-v5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracepoints: Update static_call before tp_funcs when adding a tracepoint
  ftrace: Remove redundant initialization of variable ret
  ftrace: Avoid synchronize_rcu_tasks_rude() call when not necessary
  tracing: Clean up alloc_synth_event()
  tracing/histogram: Rename "cpu" to "common_cpu"
  tracing: Synthetic event field_pos is an index not a boolean
  tracing: Fix bug in rb_per_cpu_empty() that might cause deadloop.