IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
To avoid adding copy and pasted strcmp codes in the future,
this patch adds an array "rcar_gen3_slave_whitelist" to check
whether the device can work with the IPMMU or not.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Some R-Car Gen3 SoCs has hardware restrictions on the IPMMU. So,
to check whether this R-Car Gen3 SoC can use the IPMMU correctly,
this patch modifies the ipmmu_slave_whitelist().
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
With the switch to dev_err for reporting errors from the
iommu log there was an unwanted newline introduced. The
reason was that the reporting was done in multiple dev_err()
calls, and dev_err adds a newline after every call.
Fix it by printing the log messages with only one dev_err()
call.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Because we already have the DEFINE_SHOW_ATTRIBUTE,there is no need
to define such a macro.So remove DEBUG_SEQ_FOPS_RO.
Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The Intel IOMMU driver opportunistically skips a few top level page
tables from the domain paging directory while programming the IOMMU
context entry. However there is an implicit assumption in the code that
domain's adjusted guest address width (agaw) would always be greater
than IOMMU's agaw.
The IOMMU capabilities in an upcoming platform cause the domain's agaw
to be lower than IOMMU's agaw. The issue is seen when the IOMMU supports
both 4-level and 5-level paging. The domain builds a 4-level page table
based on agaw of 2. However the IOMMU's agaw is set as 3 (5-level). In
this case the code incorrectly tries to skip page page table levels.
This causes the IOMMU driver to avoid programming the context entry. The
fix handles this case and programs the context entry accordingly.
Fixes: de24e55395 ("iommu/vt-d: Simplify domain_context_mapping_one")
Cc: <stable@vger.kernel.org>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Jacob Pan <jacob.jun.pan@linux.intel.com>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reported-by: Ramos Falcon, Ernesto R <ernesto.r.ramos.falcon@intel.com>
Tested-by: Ricardo Neri <ricardo.neri-calderon@linux.intel.com>
Signed-off-by: Sohil Mehta <sohil.mehta@intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
memunmap() should be used to free the return of memremap(), not
iounmap().
Fixes: dfddb969ed ('iommu/vt-d: Switch from ioremap_cache to memremap')
Signed-off-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Before this patch the iommu_map_page() function failed when
it tried to map a huge-page where smaller mappings existed
before.
With this change the page-table pages of the old mappings
are teared down, so that the huge-page can be mapped.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
This makes sure that __pte always contains the correct value
when the pointer to the next page-table level is derived.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Before this patch it was not possible the downgrade a
mapping established with page-mode 7 to a mapping using
smaller page-sizes, because the pte_level != level check
prevented that.
Treat page-mode 7 like a non-present mapping and allow to
overwrite it in alloc_pte().
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The function is a more generic version of free_pagetable()
and will be used to free only specific sub-trees of a
page-table.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Collect all pages that belong to a page-table in a list and
free them after the tree has been traversed. This allows to
implement safer page-table updates in subsequent patches.
Also move the functions for page-table freeing a bit upwards
in the file so that they are usable from the iommu_map() path.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
This register should have been programmed with the physical address
of the memory location containing the shadow tail pointer for
the guest virtual APIC log instead of the base address.
Fixes: 8bda0cfbdc ('iommu/amd: Detect and initialize guest vAPIC log')
Signed-off-by: Filippo Sironi <sironi@amazon.de>
Signed-off-by: Wei Wang <wawei@amazon.de>
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
If iommu_ops.add_device() fails, iommu_ops.domain_free() is still
called, leading to a crash, as the domain was only partially
initialized:
ipmmu-vmsa e67b0000.mmu: Cannot accommodate DMA translation for IOMMU page tables
sata_rcar ee300000.sata: Unable to initialize IPMMU context
iommu: Failed to add device ee300000.sata to group 0: -22
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000038
...
Call trace:
ipmmu_domain_free+0x1c/0xa0
iommu_group_release+0x48/0x68
kobject_put+0x74/0xe8
kobject_del.part.0+0x3c/0x50
kobject_put+0x60/0xe8
iommu_group_get_for_dev+0xa8/0x1f0
ipmmu_add_device+0x1c/0x40
of_iommu_configure+0x118/0x190
Fix this by checking if the domain's context already exists, before
trying to destroy it.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Fixes: d25a2a16f0 ('iommu: Add driver for Renesas VMSA-compatible IPMMU')
Signed-off-by: Joerg Roedel <jroedel@suse.de>
iommu_ops are not supposed to change at runtime.
Functions 'iommu_device_set_ops' and 'bus_set_iommu' working with
const iommu_ops provided by <linux/iommu.h>. So mark the non-const
structs as const.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Yong Wu <yong.wu@mediatek.com>
(Change the title to iommu/mediatek: xx)
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The original motivation for iommu_map_sg() was to give IOMMU drivers the
chance to map an IOVA-contiguous scatterlist as efficiently as they
could. It turns out that there isn't really much driver-specific
business involved there, so now that the default implementation is
mandatory let's just improve that - the main thing we're after is to use
larger pages wherever possible, and as long as domain->pgsize_bitmap
reflects reality, iommu_map() can already do that in a generic way. All
we need to do is detect physically-contiguous segments and batch them
into a single map operation, since whatever we do here is transparent to
our caller and not bound by any segment-length restrictions on the list
itself.
Speaking of efficiency, there's really very little point in duplicating
the checks that iommu_map() is going to do anyway, so those get cleared
up in the process.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Merge updates from Andrew Morton:
- a few misc things
- ocfs2 updates
- most of MM
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (132 commits)
hugetlbfs: dirty pages as they are added to pagecache
mm: export add_swap_extent()
mm: split SWP_FILE into SWP_ACTIVATED and SWP_FS
tools/testing/selftests/vm/map_fixed_noreplace.c: add test for MAP_FIXED_NOREPLACE
mm: thp: relocate flush_cache_range() in migrate_misplaced_transhuge_page()
mm: thp: fix mmu_notifier in migrate_misplaced_transhuge_page()
mm: thp: fix MADV_DONTNEED vs migrate_misplaced_transhuge_page race condition
mm/kasan/quarantine.c: make quarantine_lock a raw_spinlock_t
mm/gup: cache dev_pagemap while pinning pages
Revert "x86/e820: put !E820_TYPE_RAM regions into memblock.reserved"
mm: return zero_resv_unavail optimization
mm: zero remaining unavailable struct pages
tools/testing/selftests/vm/gup_benchmark.c: add MAP_HUGETLB option
tools/testing/selftests/vm/gup_benchmark.c: add MAP_SHARED option
tools/testing/selftests/vm/gup_benchmark.c: allow user specified file
tools/testing/selftests/vm/gup_benchmark.c: fix 'write' flag usage
mm/gup_benchmark.c: add additional pinning methods
mm/gup_benchmark.c: time put_page()
mm: don't raise MEMCG_OOM event due to failed high-order allocation
mm/page-writeback.c: fix range_cyclic writeback vs writepages deadlock
...
Pull Devicetree updates from Rob Herring:
"A bit bigger than normal as I've been busy this cycle.
There's a few things with dependencies and a few things subsystem
maintainers didn't pick up, so I'm taking them thru my tree.
The fixes from Johan didn't get into linux-next, but they've been
waiting for some time now and they are what's left of what subsystem
maintainers didn't pick up.
Summary:
- Sync dtc with upstream version v1.4.7-14-gc86da84d30e4
- Work to get rid of direct accesses to struct device_node name and
type pointers in preparation for removing them. New helpers for
parsing DT cpu nodes and conversions to use the helpers. printk
conversions to %pOFn for printing DT node names. Most went thru
subystem trees, so this is the remainder.
- Fixes to DT child node lookups to actually be restricted to child
nodes instead of treewide.
- Refactoring of dtb targets out of arch code. This makes the support
more uniform and enables building all dtbs on c6x, microblaze, and
powerpc.
- Various DT binding updates for Renesas r8a7744 SoC
- Vendor prefixes for Facebook, OLPC
- Restructuring of some ARM binding docs moving some peripheral
bindings out of board/SoC binding files
- New "secure-chosen" binding for secure world settings on ARM
- Dual licensing of 2 DT IRQ binding headers"
* tag 'devicetree-for-4.20' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux: (78 commits)
ARM: dt: relicense two DT binding IRQ headers
power: supply: twl4030-charger: fix OF sibling-node lookup
NFC: nfcmrvl_uart: fix OF child-node lookup
net: stmmac: dwmac-sun8i: fix OF child-node lookup
net: bcmgenet: fix OF child-node lookup
drm/msm: fix OF child-node lookup
drm/mediatek: fix OF sibling-node lookup
of: Add missing exports of node name compare functions
dt-bindings: Add OLPC vendor prefix
dt-bindings: misc: bk4: Add device tree binding for Liebherr's BK4 SPI bus
dt-bindings: thermal: samsung: Add SPDX license identifier
dt-bindings: clock: samsung: Add SPDX license identifiers
dt-bindings: timer: ostm: Add R7S9210 support
dt-bindings: phy: rcar-gen2: Add r8a7744 support
dt-bindings: can: rcar_can: Add r8a7744 support
dt-bindings: timer: renesas, cmt: Document r8a7744 CMT support
dt-bindings: watchdog: renesas-wdt: Document r8a7744 support
dt-bindings: thermal: rcar: Add device tree support for r8a7744
Documentation: dt: Add binding for /secure-chosen/stdout-path
dt-bindings: arm: zte: Move sysctrl bindings to their own doc
...
Pull IOMMU updates from Joerg Roedel:
- Debugfs support for the Intel VT-d driver.
When enabled, it now also exposes some of its internal data
structures to user-space for debugging purposes.
- ARM-SMMU driver now uses the generic deferred flushing and fast-path
iova allocation code.
This is expected to be a major performance improvement, as this
allocation path scales a lot better.
- Support for r8a7744 in the Renesas iommu driver
- Couple of minor fixes and improvements all over the place
* tag 'iommu-updates-v4.20' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu: (39 commits)
iommu/arm-smmu-v3: Remove unnecessary wrapper function
iommu/arm-smmu-v3: Add SPDX header
iommu/amd: Add default branch in amd_iommu_capable()
dt-bindings: iommu: ipmmu-vmsa: Add r8a7744 support
iommu/amd: Move iommu_init_pci() to .init section
iommu/arm-smmu: Support non-strict mode
iommu/io-pgtable-arm-v7s: Add support for non-strict mode
iommu/arm-smmu-v3: Add support for non-strict mode
iommu/io-pgtable-arm: Add support for non-strict mode
iommu: Add "iommu.strict" command line option
iommu/dma: Add support for non-strict mode
iommu/arm-smmu: Ensure that page-table updates are visible before TLBI
iommu/arm-smmu-v3: Implement flush_iotlb_all hook
iommu/arm-smmu-v3: Avoid back-to-back CMD_SYNC operations
iommu/arm-smmu-v3: Fix unexpected CMD_SYNC timeout
iommu/io-pgtable-arm: Fix race handling in split_blk_unmap()
iommu/arm-smmu-v3: Fix a couple of minor comment typos
iommu: Fix a typo
iommu: Remove .domain_{get,set}_windows
iommu: Tidy up window attributes
...
Pull KVM updates from Radim Krčmář:
"ARM:
- Improved guest IPA space support (32 to 52 bits)
- RAS event delivery for 32bit
- PMU fixes
- Guest entry hardening
- Various cleanups
- Port of dirty_log_test selftest
PPC:
- Nested HV KVM support for radix guests on POWER9. The performance
is much better than with PR KVM. Migration and arbitrary level of
nesting is supported.
- Disable nested HV-KVM on early POWER9 chips that need a particular
hardware bug workaround
- One VM per core mode to prevent potential data leaks
- PCI pass-through optimization
- merge ppc-kvm topic branch and kvm-ppc-fixes to get a better base
s390:
- Initial version of AP crypto virtualization via vfio-mdev
- Improvement for vfio-ap
- Set the host program identifier
- Optimize page table locking
x86:
- Enable nested virtualization by default
- Implement Hyper-V IPI hypercalls
- Improve #PF and #DB handling
- Allow guests to use Enlightened VMCS
- Add migration selftests for VMCS and Enlightened VMCS
- Allow coalesced PIO accesses
- Add an option to perform nested VMCS host state consistency check
through hardware
- Automatic tuning of lapic_timer_advance_ns
- Many fixes, minor improvements, and cleanups"
* tag 'kvm-4.20-1' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (204 commits)
KVM/nVMX: Do not validate that posted_intr_desc_addr is page aligned
Revert "kvm: x86: optimize dr6 restore"
KVM: PPC: Optimize clearing TCEs for sparse tables
x86/kvm/nVMX: tweak shadow fields
selftests/kvm: add missing executables to .gitignore
KVM: arm64: Safety check PSTATE when entering guest and handle IL
KVM: PPC: Book3S HV: Don't use streamlined entry path on early POWER9 chips
arm/arm64: KVM: Enable 32 bits kvm vcpu events support
arm/arm64: KVM: Rename function kvm_arch_dev_ioctl_check_extension()
KVM: arm64: Fix caching of host MDCR_EL2 value
KVM: VMX: enable nested virtualization by default
KVM/x86: Use 32bit xor to clear registers in svm.c
kvm: x86: Introduce KVM_CAP_EXCEPTION_PAYLOAD
kvm: vmx: Defer setting of DR6 until #DB delivery
kvm: x86: Defer setting of CR2 until #PF delivery
kvm: x86: Add payload operands to kvm_multiple_exception
kvm: x86: Add exception payload fields to kvm_vcpu_events
kvm: x86: Add has_payload and payload to kvm_queued_exception
KVM: Documentation: Fix omission in struct kvm_vcpu_events
KVM: selftests: add Enlightened VMCS test
...
Pull x86 mm updates from Ingo Molnar:
"Lots of changes in this cycle:
- Lots of CPA (change page attribute) optimizations and related
cleanups (Thomas Gleixner, Peter Zijstra)
- Make lazy TLB mode even lazier (Rik van Riel)
- Fault handler cleanups and improvements (Dave Hansen)
- kdump, vmcore: Enable kdumping encrypted memory with AMD SME
enabled (Lianbo Jiang)
- Clean up VM layout documentation (Baoquan He, Ingo Molnar)
- ... plus misc other fixes and enhancements"
* 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (51 commits)
x86/stackprotector: Remove the call to boot_init_stack_canary() from cpu_startup_entry()
x86/mm: Kill stray kernel fault handling comment
x86/mm: Do not warn about PCI BIOS W+X mappings
resource: Clean it up a bit
resource: Fix find_next_iomem_res() iteration issue
resource: Include resource end in walk_*() interfaces
x86/kexec: Correct KEXEC_BACKUP_SRC_END off-by-one error
x86/mm: Remove spurious fault pkey check
x86/mm/vsyscall: Consider vsyscall page part of user address space
x86/mm: Add vsyscall address helper
x86/mm: Fix exception table comments
x86/mm: Add clarifying comments for user addr space
x86/mm: Break out user address space handling
x86/mm: Break out kernel address space handling
x86/mm: Clarify hardware vs. software "error_code"
x86/mm/tlb: Make lazy TLB mode lazier
x86/mm/tlb: Add freed_tables element to flush_tlb_info
x86/mm/tlb: Add freed_tables argument to flush_tlb_mm_range
smp,cpumask: introduce on_each_cpu_cond_mask
smp: use __cpumask_set_cpu in on_each_cpu_cond
...
Pull ia64 updates from Tony Luck:
"Miscellaneous ia64 fixes from Christoph"
* tag 'please-pull-next' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux:
intel-iommu: mark intel_dma_ops static
ia64: remove machvec_dma_sync_{single,sg}
ia64/sn2: remove no-ops dma sync methods
ia64: remove the unused iommu_dma_init function
ia64: remove the unused pci_iommu_shutdown function
ia64: remove the unused bad_dma_address symbol
ia64: remove iommu_dma_supported
ia64: remove the dead iommu_sac_force variable
ia64: remove the kern_mem_attribute export
Simplify the code by removing an unnecessary wrapper function.
This was left behind by commit 2f657add07
("iommu/arm-smmu-v3: Specialise CMD_SYNC handling")
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Recent gcc warns about switching on an enumeration, but not having
an explicit case statement for all members of the enumeration. To
show the compiler this is intentional, we simply add a default case
with nothing more than a break statement.
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Boris Ostrovsky reported a memory leak with device passthrough when SME
is active.
The VFIO driver uses iommu_iova_to_phys() to get the physical address for
an iova. This physical address is later passed into vfio_unmap_unpin() to
unpin the memory. The vfio_unmap_unpin() uses pfn_valid() before unpinning
the memory. The pfn_valid() check was failing because encryption mask was
part of the physical address returned. This resulted in the memory not
being unpinned and therefore leaked after the guest terminates.
The memory encryption mask must be cleared from the physical address in
iommu_iova_to_phys().
Fixes: 2543a786aa ("iommu/amd: Allow the AMD IOMMU to work with memory encryption")
Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: <iommu@lists.linux-foundation.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: kvm@vger.kernel.org
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: <stable@vger.kernel.org> # 4.14+
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
All we need is to wire up .flush_iotlb_all properly and implement the
domain attribute, and iommu-dma and io-pgtable will do the rest for us.
The only real subtlety is documenting the barrier semantics we're
introducing between io-pgtable and the drivers for non-strict flushes.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
As for LPAE, it's simply a case of skipping the leaf invalidation for a
regular unmap, and ensuring that the one in split_blk_unmap() is paired
with an explicit sync ASAP rather than relying on one which might only
eventually happen way down the line.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Now that io-pgtable knows how to dodge strict TLB maintenance, all
that's left to do is bridge the gap between the IOMMU core requesting
DOMAIN_ATTR_DMA_USE_FLUSH_QUEUE for default domains, and showing the
appropriate IO_PGTABLE_QUIRK_NON_STRICT flag to alloc_io_pgtable_ops().
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[rm: convert to domain attribute, tweak commit message]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Non-strict mode is simply a case of skipping 'regular' leaf TLBIs, since
the sync is already factored out into ops->iotlb_sync at the core API
level. Non-leaf invalidations where we change the page table structure
itself still have to be issued synchronously in order to maintain walk
caches correctly.
To save having to reason about it too much, make sure the invalidation
in arm_lpae_split_blk_unmap() just performs its own unconditional sync
to minimise the window in which we're technically violating the break-
before-make requirement on a live mapping. This might work out redundant
with an outer-level sync for strict unmaps, but we'll never be splitting
blocks on a DMA fastpath anyway.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[rm: tweak comment, commit message, split_blk_unmap logic and barriers]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add a generic command line option to enable lazy unmapping via IOVA
flush queues, which will initally be suuported by iommu-dma. This echoes
the semantics of "intel_iommu=strict" (albeit with the opposite default
value), but in the driver-agnostic fashion of "iommu.passthrough".
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[rm: move handling out of SMMUv3 driver, clean up documentation]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
[will: dropped broken printk when parsing command-line option]
Signed-off-by: Will Deacon <will.deacon@arm.com>
With the flush queue infrastructure already abstracted into IOVA
domains, hooking it up in iommu-dma is pretty simple. Since there is a
degree of dependency on the IOMMU driver knowing what to do to play
along, we key the whole thing off a domain attribute which will be set
on default DMA ops domains to request non-strict invalidation. That way,
drivers can indicate the appropriate support by acknowledging the
attribute, and we can easily fall back to strict invalidation otherwise.
The flush queue callback needs a handle on the iommu_domain which owns
our cookie, so we have to add a pointer back to that, but neatly, that's
also sufficient to indicate whether we're using a flush queue or not,
and thus which way to release IOVAs. The only slight subtlety is
switching __iommu_dma_unmap() from calling iommu_unmap() to explicit
iommu_unmap_fast()/iommu_tlb_sync() so that we can elide the sync
entirely in non-strict mode.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[rm: convert to domain attribute, tweak comments and commit message]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The IO-pgtable code relies on the driver TLB invalidation callbacks to
ensure that all page-table updates are visible to the IOMMU page-table
walker.
In the case that the page-table walker is cache-coherent, we cannot rely
on an implicit DSB from the DMA-mapping code, so we must ensure that we
execute a DSB in our tlb_add_flush() callback prior to triggering the
invalidation.
Cc: <stable@vger.kernel.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Fixes: 2df7a25ce4 ("iommu/arm-smmu: Clean up DMA API usage")
Signed-off-by: Will Deacon <will.deacon@arm.com>
.flush_iotlb_all is currently stubbed to arm_smmu_iotlb_sync() since the
only time it would ever need to actually do anything is for callers
doing their own explicit batching, e.g.:
iommu_unmap_fast(domain, ...);
iommu_unmap_fast(domain, ...);
iommu_iotlb_flush_all(domain, ...);
where since io-pgtable still issues the TLBI commands implicitly in the
unmap instead of implementing .iotlb_range_add, the "flush" only needs
to ensure completion of those already-in-flight invalidations.
However, we're about to start using it in anger with flush queues, so
let's get a proper implementation wired up.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
[rm: document why it wasn't a bug]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Putting adjacent CMD_SYNCs into the command queue is nonsensical, but
can happen when multiple CPUs are inserting commands. Rather than leave
the poor old hardware to chew through these operations, we can instead
drop the subsequent SYNCs and poll for completion of the first. This
has been shown to improve IO performance under pressure, where the
number of SYNC operations reduces by about a third:
CMD_SYNCs reduced: 19542181
CMD_SYNCs total: 58098548 (include reduced)
CMDs total: 116197099 (TLBI:SYNC about 1:1)
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The condition break condition of:
(int)(VAL - sync_idx) >= 0
in the __arm_smmu_sync_poll_msi() polling loop requires that sync_idx
must be increased monotonically according to the sequence of the CMDs in
the cmdq.
However, since the msidata is populated using atomic_inc_return_relaxed()
before taking the command-queue spinlock, then the following scenario
can occur:
CPU0 CPU1
msidata=0
msidata=1
insert cmd1
insert cmd0
smmu execute cmd1
smmu execute cmd0
poll timeout, because msidata=1 is overridden by
cmd0, that means VAL=0, sync_idx=1.
This is not a functional problem, since the caller will eventually either
timeout or exit due to another CMD_SYNC, however it's clearly not what
the code is supposed to be doing. Fix it, by incrementing the sequence
count with the command-queue lock held, allowing us to drop the atomic
operations altogether.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[will: dropped the specialised cmd building routine for now]
Signed-off-by: Will Deacon <will.deacon@arm.com>
In removing the pagetable-wide lock, we gained the possibility of the
vanishingly unlikely case where we have a race between two concurrent
unmappers splitting the same block entry. The logic to handle this is
fairly straightforward - whoever loses the race frees their partial
next-level table and instead dereferences the winner's newly-installed
entry in order to fall back to a regular unmap, which intentionally
echoes the pre-existing case of recursively splitting a 1GB block down
to 4KB pages by installing a full table of 2MB blocks first.
Unfortunately, the chump who implemented that logic failed to update the
condition check for that fallback, meaning that if said race occurs at
the last level (where the loser's unmap_idx is valid) then the unmap
won't actually happen. Fix that to properly account for both the race
and recursive cases.
Fixes: 2c3d273eab ("iommu/io-pgtable-arm: Support lockless operation")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
[will: re-jig control flow to avoid duplicate cmpxchg test]
Signed-off-by: Will Deacon <will.deacon@arm.com>
