At this point every iommu driver will cause a default_domain to be
selected, so we can finally remove this gap from the core code.
The following table explains what each driver supports and what the
resulting default_domain will be:
ops->defaut_domain
IDENTITY DMA PLATFORM v ARM32 dma-iommu ARCH
amd/iommu.c Y Y N/A either
apple-dart.c Y Y N/A either
arm-smmu.c Y Y IDENTITY either
qcom_iommu.c G Y IDENTITY either
arm-smmu-v3.c Y Y N/A either
exynos-iommu.c G Y IDENTITY either
fsl_pamu_domain.c Y Y N/A N/A PLATFORM
intel/iommu.c Y Y N/A either
ipmmu-vmsa.c G Y IDENTITY either
msm_iommu.c G IDENTITY N/A
mtk_iommu.c G Y IDENTITY either
mtk_iommu_v1.c G IDENTITY N/A
omap-iommu.c G IDENTITY N/A
rockchip-iommu.c G Y IDENTITY either
s390-iommu.c Y Y N/A N/A PLATFORM
sprd-iommu.c Y N/A DMA
sun50i-iommu.c G Y IDENTITY either
tegra-smmu.c G Y IDENTITY IDENTITY
virtio-iommu.c Y Y N/A either
spapr Y Y N/A N/A PLATFORM
* G means ops->identity_domain is used
* N/A means the driver will not compile in this configuration
ARM32 drivers select an IDENTITY default domain through either the
ops->identity_domain or directly requesting an IDENTIY domain through
alloc_domain().
In ARM64 mode tegra-smmu will still block the use of dma-iommu.c and
forces an IDENTITY domain.
S390 uses a PLATFORM domain to represent when the dma_ops are set to the
s390 iommu code.
fsl_pamu uses an PLATFORM domain.
POWER SPAPR uses PLATFORM and blocking to enable its weird VFIO mode.
The x86 drivers continue unchanged.
After this patch group->default_domain is only NULL for a short period
during bus iommu probing while all the groups are constituted. Otherwise
it is always !NULL.
This completes changing the iommu subsystem driver contract to a system
where the current iommu_domain always represents some form of translation
and the driver is continuously asserting a definable translation mode.
It resolves the confusion that the original ops->detach_dev() caused
around what translation, exactly, is the IOMMU performing after
detach. There were at least three different answers to that question in
the tree, they are all now clearly named with domain types.
Tested-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Niklas Schnelle <schnelle@linux.ibm.com>
Tested-by: Steven Price <steven.price@arm.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Prior to commit 1b932ceddd ("iommu: Remove detach_dev callbacks") the
sun50i_iommu_detach_device() function was being called by
ops->detach_dev().
This is an IDENTITY domain so convert sun50i_iommu_detach_device() into
sun50i_iommu_identity_attach() and a full IDENTITY domain and thus hook it
back up the same was as the old ops->detach_dev().
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/19-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
What msm does during msm_iommu_set_platform_dma() is actually putting the
iommu into identity mode.
Move to the new core support for ARM_DMA_USE_IOMMU by defining
ops->identity_domain.
This driver does not support IOMMU_DOMAIN_DMA, however it cannot be
compiled on ARM64 either. Most likely it is fine to support dma-iommu.c
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/14-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
What omap does during omap_iommu_set_platform_dma() is actually putting
the iommu into identity mode.
Move to the new core support for ARM_DMA_USE_IOMMU by defining
ops->identity_domain.
This driver does not support IOMMU_DOMAIN_DMA, however it cannot be
compiled on ARM64 either. Most likely it is fine to support dma-iommu.c
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/13-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
All ARM64 iommu drivers should support IOMMU_DOMAIN_DMA to enable
dma-iommu.c.
tegra is blocking dma-iommu usage, and also default_domain's, because it
wants an identity translation. This is needed for some device quirk. The
correct way to do this is to support IDENTITY domains and use
ops->def_domain_type() to return IOMMU_DOMAIN_IDENTITY for only the quirky
devices.
Add support for IOMMU_DOMAIN_DMA and force IOMMU_DOMAIN_IDENTITY mode for
everything so no behavior changes.
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/12-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Even though dma-iommu.c and CONFIG_ARM_DMA_USE_IOMMU do approximately the
same stuff, the way they relate to the IOMMU core is quiet different.
dma-iommu.c expects the core code to setup an UNMANAGED domain (of type
IOMMU_DOMAIN_DMA) and then configures itself to use that domain. This
becomes the default_domain for the group.
ARM_DMA_USE_IOMMU does not use the default_domain, instead it directly
allocates an UNMANAGED domain and operates it just like an external
driver. In this case group->default_domain is NULL.
If the driver provides a global static identity_domain then automatically
use it as the default_domain when in ARM_DMA_USE_IOMMU mode.
This allows drivers that implemented default_domain == NULL as an IDENTITY
translation to trivially get a properly labeled non-NULL default_domain on
ARM32 configs.
With this arrangment when ARM_DMA_USE_IOMMU wants to disconnect from the
device the normal detach_domain flow will restore the IDENTITY domain as
the default domain. Overall this makes attach_dev() of the IDENTITY domain
called in the same places as detach_dev().
This effectively migrates these drivers to default_domain mode. For
drivers that support ARM64 they will gain support for the IDENTITY
translation mode for the dma_api and behave in a uniform way.
Drivers use this by setting ops->identity_domain to a static singleton
iommu_domain that implements the identity attach. If the core detects
ARM_DMA_USE_IOMMU mode then it automatically attaches the IDENTITY domain
during probe.
Drivers can continue to prevent the use of DMA translation by returning
IOMMU_DOMAIN_IDENTITY from def_domain_type, this will completely prevent
IOMMU_DMA from running but will not impact ARM_DMA_USE_IOMMU.
This allows removing the set_platform_dma_ops() from every remaining
driver.
Remove the set_platform_dma_ops from rockchip and mkt_v1 as all it does
is set an existing global static identity domain. mkt_v1 does not support
IOMMU_DOMAIN_DMA and it does not compile on ARM64 so this transformation
is safe.
Tested-by: Steven Price <steven.price@arm.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/9-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Except for dart (which forces IOMMU_DOMAIN_DMA) every driver returns 0 or
IDENTITY from ops->def_domain_type().
The drivers that return IDENTITY have some kind of good reason, typically
that quirky hardware really can't support anything other than IDENTITY.
Arrange things so that if the driver says it needs IDENTITY then
iommu_get_default_domain_type() either fails or returns IDENTITY. It will
not ignore the driver's override to IDENTITY.
Split the function into two steps, reducing the group device list to the
driver's def_domain_type() and the untrusted flag.
Then compute the result based on those two reduced variables. Fully reject
combining untrusted with IDENTITY.
Remove the debugging print on the iommu_group_store_type() failure path,
userspace should not be able to trigger kernel prints.
This makes the next patch cleaner that wants to force IDENTITY always for
ARM_IOMMU because there is no support for DMA.
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/8-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
What mtk does during mtk_iommu_v1_set_platform_dma() is actually putting
the iommu into identity mode. Make this available as a proper IDENTITY
domain.
The mtk_iommu_v1_def_domain_type() from
commit 8bbe13f52c ("iommu/mediatek-v1: Add def_domain_type") explains
this was needed to allow probe_finalize() to be called, but now the
IDENTITY domain will do the same job so change the returned
def_domain_type.
mkt_v1 is the only driver that returns IOMMU_DOMAIN_UNMANAGED from
def_domain_type(). This allows the next patch to enforce an IDENTITY
domain policy for this driver.
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/7-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Thierry says this is not used anymore, and doesn't think it makes sense as
an iommu driver. The HW it supports is about 10 years old now and newer HW
uses different IOMMU drivers.
As this is the only driver with a GART approach, and it doesn't really
meet the driver expectations from the IOMMU core, let's just remove it
so we don't have to think about how to make it fit in.
It has a number of identified problems:
- The assignment of iommu_groups doesn't match the HW behavior
- It claims to have an UNMANAGED domain but it is really an IDENTITY
domain with a translation aperture. This is inconsistent with the core
expectation for security sensitive operations
- It doesn't implement a SW page table under struct iommu_domain so
* It can't accept a map until the domain is attached
* It forgets about all maps after the domain is detached
* It doesn't clear the HW of maps once the domain is detached
(made worse by having the wrong groups)
Cc: Thierry Reding <treding@nvidia.com>
Cc: Dmitry Osipenko <digetx@gmail.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/6-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The PLATFORM domain will be set as the default domain and attached as
normal during probe. The driver will ignore the initial attach from a NULL
domain to the PLATFORM domain.
After this, the PLATFORM domain's attach_dev will be called whenever we
detach from an UNMANAGED domain (eg for VFIO). This is the same time the
original design would have called op->detach_dev().
This is temporary until the S390 dma-iommu.c conversion is merged.
Tested-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/4-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
POWER is using the set_platform_dma_ops() callback to hook up its private
dma_ops, but this is buired under some indirection and is weirdly
happening for a BLOCKED domain as well.
For better documentation create a PLATFORM domain to manage the dma_ops,
since that is what it is for, and make the BLOCKED domain an alias for
it. BLOCKED is required for VFIO.
Also removes the leaky allocation of the BLOCKED domain by using a global
static.
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/3-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
This is used when the iommu driver is taking control of the dma_ops,
currently only on S390 and power spapr. It is designed to preserve the
original ops->detach_dev() semantic that these S390 was built around.
Provide an opaque domain type and a 'default_domain' ops value that allows
the driver to trivially force any single domain as the default domain.
Update iommufd selftest to use this instead of set_platform_dma_ops
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/2-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
This allows a driver to set a global static to an IDENTITY domain and
the core code will automatically use it whenever an IDENTITY domain
is requested.
By making it always available it means the IDENTITY can be used in error
handling paths to force the iommu driver into a known state. Devices
implementing global static identity domains should avoid failing their
attach_dev ops.
To make global static domains simpler allow drivers to omit their free
function and update the iommufd selftest.
Convert rockchip to use the new mechanism.
Tested-by: Steven Price <steven.price@arm.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/1-v8-81230027b2fa+9d-iommu_all_defdom_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Pull kvm fixes from Paolo Bonzini:
"ARM:
- Fix EL2 Stage-1 MMIO mappings where a random address was used
- Fix SMCCC function number comparison when the SVE hint is set
RISC-V:
- Fix KVM_GET_REG_LIST API for ISA_EXT registers
- Fix reading ISA_EXT register of a missing extension
- Fix ISA_EXT register handling in get-reg-list test
- Fix filtering of AIA registers in get-reg-list test
x86:
- Fixes for TSC_AUX virtualization
- Stop zapping page tables asynchronously, since we don't zap them as
often as before"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: SVM: Do not use user return MSR support for virtualized TSC_AUX
KVM: SVM: Fix TSC_AUX virtualization setup
KVM: SVM: INTERCEPT_RDTSCP is never intercepted anyway
KVM: x86/mmu: Stop zapping invalidated TDP MMU roots asynchronously
KVM: x86/mmu: Do not filter address spaces in for_each_tdp_mmu_root_yield_safe()
KVM: x86/mmu: Open code leaf invalidation from mmu_notifier
KVM: riscv: selftests: Selectively filter-out AIA registers
KVM: riscv: selftests: Fix ISA_EXT register handling in get-reg-list
RISC-V: KVM: Fix riscv_vcpu_get_isa_ext_single() for missing extensions
RISC-V: KVM: Fix KVM_GET_REG_LIST API for ISA_EXT registers
KVM: selftests: Assert that vasprintf() is successful
KVM: arm64: nvhe: Ignore SVE hint in SMCCC function ID
KVM: arm64: Properly return allocated EL2 VA from hyp_alloc_private_va_range()
Pull tracing fixes from Steven Rostedt:
- Fix the "bytes" output of the per_cpu stat file
The tracefs/per_cpu/cpu*/stats "bytes" was giving bogus values as the
accounting was not accurate. It is suppose to show how many used
bytes are still in the ring buffer, but even when the ring buffer was
empty it would still show there were bytes used.
- Fix a bug in eventfs where reading a dynamic event directory (open)
and then creating a dynamic event that goes into that diretory screws
up the accounting.
On close, the newly created event dentry will get a "dput" without
ever having a "dget" done for it. The fix is to allocate an array on
dir open to save what dentries were actually "dget" on, and what ones
to "dput" on close.
* tag 'trace-v6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
eventfs: Remember what dentries were created on dir open
ring-buffer: Fix bytes info in per_cpu buffer stats
Pull cxl fixes from Dan Williams:
"A collection of regression fixes, bug fixes, and some small cleanups
to the Compute Express Link code.
The regressions arrived in the v6.5 dev cycle and missed the v6.6
merge window due to my personal absences this cycle. The most
important fixes are for scenarios where the CXL subsystem fails to
parse valid region configurations established by platform firmware.
This is important because agreement between OS and BIOS on the CXL
configuration is fundamental to implementing "OS native" error
handling, i.e. address translation and component failure
identification.
Other important fixes are a driver load error when the BIOS lets the
Linux PCI core handle AER events, but not CXL memory errors.
The other fixex might have end user impact, but for now are only known
to trigger in our test/emulation environment.
Summary:
- Fix multiple scenarios where platform firmware defined regions fail
to be assembled by the CXL core.
- Fix a spurious driver-load failure on platforms that enable OS
native AER, but not OS native CXL error handling.
- Fix a regression detecting "poison" commands when "security"
commands are also defined.
- Fix a cxl_test regression with the move to centralize CXL port
register enumeration in the CXL core.
- Miscellaneous small fixes and cleanups"
* tag 'cxl-fixes-6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl:
cxl/acpi: Annotate struct cxl_cxims_data with __counted_by
cxl/port: Fix cxl_test register enumeration regression
cxl/region: Refactor granularity select in cxl_port_setup_targets()
cxl/region: Match auto-discovered region decoders by HPA range
cxl/mbox: Fix CEL logic for poison and security commands
cxl/pci: Replace host_bridge->native_aer with pcie_aer_is_native()
PCI/AER: Export pcie_aer_is_native()
cxl/pci: Fix appropriate checking for _OSC while handling CXL RAS registers
Pull gpio fixes from Bartosz Golaszewski:
- fix an invalid usage of __free(kfree) leading to kfreeing an
ERR_PTR()
- fix an irq domain leak in gpio-tb10x
- MAINTAINERS update
* tag 'gpio-fixes-for-v6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
gpio: sim: fix an invalid __free() usage
gpio: tb10x: Fix an error handling path in tb10x_gpio_probe()
MAINTAINERS: gpio-regmap: make myself a maintainer of it
Pull misc fixes from Andrew Morton:
"13 hotfixes, 10 of which pertain to post-6.5 issues. The other three
are cc:stable"
* tag 'mm-hotfixes-stable-2023-09-23-10-31' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
proc: nommu: fix empty /proc/<pid>/maps
filemap: add filemap_map_order0_folio() to handle order0 folio
proc: nommu: /proc/<pid>/maps: release mmap read lock
mm: memcontrol: fix GFP_NOFS recursion in memory.high enforcement
pidfd: prevent a kernel-doc warning
argv_split: fix kernel-doc warnings
scatterlist: add missing function params to kernel-doc
selftests/proc: fixup proc-empty-vm test after KSM changes
revert "scripts/gdb/symbols: add specific ko module load command"
selftests: link libasan statically for tests with -fsanitize=address
task_work: add kerneldoc annotation for 'data' argument
mm: page_alloc: fix CMA and HIGHATOMIC landing on the wrong buddy list
sh: mm: re-add lost __ref to ioremap_prot() to fix modpost warning
Pull smb client fixes from Steve French:
"Six smb3 client fixes, including three for stable, from the SMB
plugfest (testing event) this week:
- Reparse point handling fix (found when investigating dir
enumeration when fifo in dir)
- Fix excessive thread creation for dir lease cleanup
- UAF fix in negotiate path
- remove duplicate error message mapping and fix confusing warning
message
- add dynamic trace point to improve debugging RDMA connection
attempts"
* tag '6.6-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
smb3: fix confusing debug message
smb: client: handle STATUS_IO_REPARSE_TAG_NOT_HANDLED
smb3: remove duplicate error mapping
cifs: Fix UAF in cifs_demultiplex_thread()
smb3: do not start laundromat thread when dir leases disabled
smb3: Add dynamic trace points for RDMA (smbdirect) reconnect
Pull i2c fixes from Wolfram Sang:
"A set of I2C driver fixes. Mostly fixing resource leaks or sanity
checks"
* tag 'i2c-for-6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
i2c: xiic: Correct return value check for xiic_reinit()
i2c: mux: gpio: Add missing fwnode_handle_put()
i2c: mux: demux-pinctrl: check the return value of devm_kstrdup()
i2c: designware: fix __i2c_dw_disable() in case master is holding SCL low
i2c: i801: unregister tco_pdev in i801_probe() error path
Pull LoongArch fixes from Huacai Chen:
"Fix lockdep, fix a boot failure, fix some build warnings, fix document
links, and some cleanups"
* tag 'loongarch-fixes-6.6-1' of git://git.kernel.org/pub/scm/linux/kernel/git/chenhuacai/linux-loongson:
docs/zh_CN/LoongArch: Update the links of ABI
docs/LoongArch: Update the links of ABI
LoongArch: Don't inline kasan_mem_to_shadow()/kasan_shadow_to_mem()
kasan: Cleanup the __HAVE_ARCH_SHADOW_MAP usage
LoongArch: Set all reserved memblocks on Node#0 at initialization
LoongArch: Remove dead code in relocate_new_kernel
LoongArch: Use _UL() and _ULL()
LoongArch: Fix some build warnings with W=1
LoongArch: Fix lockdep static memory detection
Pull iomap fixes from Darrick Wong:
- Return EIO on bad inputs to iomap_to_bh instead of BUGging, to deal
less poorly with block device io racing with block device resizing
- Fix a stale page data exposure bug introduced in 6.6-rc1 when
unsharing a file range that is not in the page cache
* tag 'iomap-6.6-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
iomap: convert iomap_unshare_iter to use large folios
iomap: don't skip reading in !uptodate folios when unsharing a range
iomap: handle error conditions more gracefully in iomap_to_bh
KVM/riscv fixes for 6.6, take #1
- Fix KVM_GET_REG_LIST API for ISA_EXT registers
- Fix reading ISA_EXT register of a missing extension
- Fix ISA_EXT register handling in get-reg-list test
- Fix filtering of AIA registers in get-reg-list test
When the TSC_AUX MSR is virtualized, the TSC_AUX value is swap type "B"
within the VMSA. This means that the guest value is loaded on VMRUN and
the host value is restored from the host save area on #VMEXIT.
Since the value is restored on #VMEXIT, the KVM user return MSR support
for TSC_AUX can be replaced by populating the host save area with the
current host value of TSC_AUX. And, since TSC_AUX is not changed by Linux
post-boot, the host save area can be set once in svm_hardware_enable().
This eliminates the two WRMSR instructions associated with the user return
MSR support.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Message-Id: <d381de38eb0ab6c9c93dda8503b72b72546053d7.1694811272.git.thomas.lendacky@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The checks for virtualizing TSC_AUX occur during the vCPU reset processing
path. However, at the time of initial vCPU reset processing, when the vCPU
is first created, not all of the guest CPUID information has been set. In
this case the RDTSCP and RDPID feature support for the guest is not in
place and so TSC_AUX virtualization is not established.
This continues for each vCPU created for the guest. On the first boot of
an AP, vCPU reset processing is executed as a result of an APIC INIT
event, this time with all of the guest CPUID information set, resulting
in TSC_AUX virtualization being enabled, but only for the APs. The BSP
always sees a TSC_AUX value of 0 which probably went unnoticed because,
at least for Linux, the BSP TSC_AUX value is 0.
Move the TSC_AUX virtualization enablement out of the init_vmcb() path and
into the vcpu_after_set_cpuid() path to allow for proper initialization of
the support after the guest CPUID information has been set.
With the TSC_AUX virtualization support now in the vcpu_set_after_cpuid()
path, the intercepts must be either cleared or set based on the guest
CPUID input.
Fixes: 296d5a17e7 ("KVM: SEV-ES: Use V_TSC_AUX if available instead of RDTSC/MSR_TSC_AUX intercepts")
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Message-Id: <4137fbcb9008951ab5f0befa74a0399d2cce809a.1694811272.git.thomas.lendacky@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
svm_recalc_instruction_intercepts() is always called at least once
before the vCPU is started, so the setting or clearing of the RDTSCP
intercept can be dropped from the TSC_AUX virtualization support.
Extracted from a patch by Tom Lendacky.
Cc: stable@vger.kernel.org
Fixes: 296d5a17e7 ("KVM: SEV-ES: Use V_TSC_AUX if available instead of RDTSC/MSR_TSC_AUX intercepts")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Stop zapping invalidate TDP MMU roots via work queue now that KVM
preserves TDP MMU roots until they are explicitly invalidated. Zapping
roots asynchronously was effectively a workaround to avoid stalling a vCPU
for an extended during if a vCPU unloaded a root, which at the time
happened whenever the guest toggled CR0.WP (a frequent operation for some
guest kernels).
While a clever hack, zapping roots via an unbound worker had subtle,
unintended consequences on host scheduling, especially when zapping
multiple roots, e.g. as part of a memslot. Because the work of zapping a
root is no longer bound to the task that initiated the zap, things like
the CPU affinity and priority of the original task get lost. Losing the
affinity and priority can be especially problematic if unbound workqueues
aren't affined to a small number of CPUs, as zapping multiple roots can
cause KVM to heavily utilize the majority of CPUs in the system, *beyond*
the CPUs KVM is already using to run vCPUs.
When deleting a memslot via KVM_SET_USER_MEMORY_REGION, the async root
zap can result in KVM occupying all logical CPUs for ~8ms, and result in
high priority tasks not being scheduled in in a timely manner. In v5.15,
which doesn't preserve unloaded roots, the issues were even more noticeable
as KVM would zap roots more frequently and could occupy all CPUs for 50ms+.
Consuming all CPUs for an extended duration can lead to significant jitter
throughout the system, e.g. on ChromeOS with virtio-gpu, deleting memslots
is a semi-frequent operation as memslots are deleted and recreated with
different host virtual addresses to react to host GPU drivers allocating
and freeing GPU blobs. On ChromeOS, the jitter manifests as audio blips
during games due to the audio server's tasks not getting scheduled in
promptly, despite the tasks having a high realtime priority.
Deleting memslots isn't exactly a fast path and should be avoided when
possible, and ChromeOS is working towards utilizing MAP_FIXED to avoid the
memslot shenanigans, but KVM is squarely in the wrong. Not to mention
that removing the async zapping eliminates a non-trivial amount of
complexity.
Note, one of the subtle behaviors hidden behind the async zapping is that
KVM would zap invalidated roots only once (ignoring partial zaps from
things like mmu_notifier events). Preserve this behavior by adding a flag
to identify roots that are scheduled to be zapped versus roots that have
already been zapped but not yet freed.
Add a comment calling out why kvm_tdp_mmu_invalidate_all_roots() can
encounter invalid roots, as it's not at all obvious why zapping
invalidated roots shouldn't simply zap all invalid roots.
Reported-by: Pattara Teerapong <pteerapong@google.com>
Cc: David Stevens <stevensd@google.com>
Cc: Yiwei Zhang<zzyiwei@google.com>
Cc: Paul Hsia <paulhsia@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230916003916.2545000-4-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
All callers except the MMU notifier want to process all address spaces.
Remove the address space ID argument of for_each_tdp_mmu_root_yield_safe()
and switch the MMU notifier to use __for_each_tdp_mmu_root_yield_safe().
Extracted out of a patch by Sean Christopherson <seanjc@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Pull xfs fixes from Chandan Babu:
- Fix an integer overflow bug when processing an fsmap call
- Fix crash due to CPU hot remove event racing with filesystem mount
operation
- During read-only mount, XFS does not allow the contents of the log to
be recovered when there are one or more unrecognized rcompat features
in the primary superblock, since the log might have intent items
which the kernel does not know how to process
- During recovery of log intent items, XFS now reserves log space
sufficient for one cycle of a permanent transaction to execute.
Otherwise, this could lead to livelocks due to non-availability of
log space
- On an fs which has an ondisk unlinked inode list, trying to delete a
file or allocating an O_TMPFILE file can cause the fs to the shutdown
if the first inode in the ondisk inode list is not present in the
inode cache. The bug is solved by explicitly loading the first inode
in the ondisk unlinked inode list into the inode cache if it is not
already cached
A similar problem arises when the uncached inode is present in the
middle of the ondisk unlinked inode list. This second bug is
triggered when executing operations like quotacheck and bulkstat. In
this case, XFS now reads in the entire ondisk unlinked inode list
- Enable LARP mode only on recent v5 filesystems
- Fix a out of bounds memory access in scrub
- Fix a performance bug when locating the tail of the log during
mounting a filesystem
* tag 'xfs-6.6-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
xfs: use roundup_pow_of_two instead of ffs during xlog_find_tail
xfs: only call xchk_stats_merge after validating scrub inputs
xfs: require a relatively recent V5 filesystem for LARP mode
xfs: make inode unlinked bucket recovery work with quotacheck
xfs: load uncached unlinked inodes into memory on demand
xfs: reserve less log space when recovering log intent items
xfs: fix log recovery when unknown rocompat bits are set
xfs: reload entire unlinked bucket lists
xfs: allow inode inactivation during a ro mount log recovery
xfs: use i_prev_unlinked to distinguish inodes that are not on the unlinked list
xfs: remove CPU hotplug infrastructure
xfs: remove the all-mounts list
xfs: use per-mount cpumask to track nonempty percpu inodegc lists
xfs: fix an agbno overflow in __xfs_getfsmap_datadev
xfs: fix per-cpu CIL structure aggregation racing with dying cpus
xfs: fix select in config XFS_ONLINE_SCRUB_STATS
The cxl_test unit test environment models a CXL topology for
sysfs/user-ABI regression testing. It uses interface mocking via the
"--wrap=" linker option to redirect cxl_core routines that parse
hardware registers with versions that just publish objects, like
devm_cxl_enumerate_decoders().
Starting with:
Commit 19ab69a60e ("cxl/port: Store the port's Component Register mappings in struct cxl_port")
...port register enumeration is moved into devm_cxl_add_port(). This
conflicts with the "cxl_test avoids emulating registers stance" so
either the port code needs to be refactored (too violent), or modified
so that register enumeration is skipped on "fake" cxl_test ports
(annoying, but straightforward).
This conflict has happened previously and the "check for platform
device" workaround to avoid instrusive refactoring was deployed in those
scenarios. In general, refactoring should only benefit production code,
test code needs to remain minimally instrusive to the greatest extent
possible.
This was missed previously because it may sometimes just cause warning
messages to be emitted, but it can also cause test failures. The
backport to -stable is only nice to have for clean cxl_test runs.
Fixes: 19ab69a60e ("cxl/port: Store the port's Component Register mappings in struct cxl_port")
Cc: stable@vger.kernel.org
Reported-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Tested-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/169476525052.1013896.6235102957693675187.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Using the following code with libtracefs:
int dfd;
// create the directory events/kprobes/kp1
tracefs_kprobe_raw(NULL, "kp1", "schedule_timeout", "time=$arg1");
// Open the kprobes directory
dfd = tracefs_instance_file_open(NULL, "events/kprobes", O_RDONLY);
// Do a lookup of the kprobes/kp1 directory (by looking at enable)
tracefs_file_exists(NULL, "events/kprobes/kp1/enable");
// Now create a new entry in the kprobes directory
tracefs_kprobe_raw(NULL, "kp2", "schedule_hrtimeout", "expires=$arg1");
// Do another lookup to create the dentries
tracefs_file_exists(NULL, "events/kprobes/kp2/enable"))
// Close the directory
close(dfd);
What happened above, the first open (dfd) will call
dcache_dir_open_wrapper() that will create the dentries and up their ref
counts.
Now the creation of "kp2" will add another dentry within the kprobes
directory.
Upon the close of dfd, eventfs_release() will now do a dput for all the
entries in kprobes. But this is where the problem lies. The open only
upped the dentry of kp1 and not kp2. Now the close is decrementing both
kp1 and kp2, which causes kp2 to get a negative count.
Doing a "trace-cmd reset" which deletes all the kprobes cause the kernel
to crash! (due to the messed up accounting of the ref counts).
To solve this, save all the dentries that are opened in the
dcache_dir_open_wrapper() into an array, and use this array to know what
dentries to do a dput on in eventfs_release().
Since the dcache_dir_open_wrapper() calls dcache_dir_open() which uses the
file->private_data, we need to also add a wrapper around dcache_readdir()
that uses the cursor assigned to the file->private_data. This is because
the dentries need to also be saved in the file->private_data. To do this
create the structure:
struct dentry_list {
void *cursor;
struct dentry **dentries;
};
Which will hold both the cursor and the dentries. Some shuffling around is
needed to make sure that dcache_dir_open() and dcache_readdir() only see
the cursor.
Link: https://lore.kernel.org/linux-trace-kernel/20230919211804.230edf1e@gandalf.local.home/
Link: https://lore.kernel.org/linux-trace-kernel/20230922163446.1431d4fa@gandalf.local.home
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Ajay Kaher <akaher@vmware.com>
Fixes: 6394044955 ("eventfs: Implement eventfs lookup, read, open functions")
Reported-by: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The 'bytes' info in file 'per_cpu/cpu<X>/stats' means the number of
bytes in cpu buffer that have not been consumed. However, currently
after consuming data by reading file 'trace_pipe', the 'bytes' info
was not changed as expected.
# cat per_cpu/cpu0/stats
entries: 0
overrun: 0
commit overrun: 0
bytes: 568 <--- 'bytes' is problematical !!!
oldest event ts: 8651.371479
now ts: 8653.912224
dropped events: 0
read events: 8
The root cause is incorrect stat on cpu_buffer->read_bytes. To fix it:
1. When stat 'read_bytes', account consumed event in rb_advance_reader();
2. When stat 'entries_bytes', exclude the discarded padding event which
is smaller than minimum size because it is invisible to reader. Then
use rb_page_commit() instead of BUF_PAGE_SIZE at where accounting for
page-based read/remove/overrun.
Also correct the comments of ring_buffer_bytes_cpu() in this patch.
Link: https://lore.kernel.org/linux-trace-kernel/20230921125425.1708423-1-zhengyejian1@huawei.com
Cc: stable@vger.kernel.org
Fixes: c64e148a3b ("trace: Add ring buffer stats to measure rate of events")
Signed-off-by: Zheng Yejian <zhengyejian1@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull thermal control fix from Rafael Wysocki:
"Unbreak the trip point update sysfs interface that has been broken
since the 6.3 cycle (Rafael Wysocki)"
* tag 'thermal-6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
thermal: sysfs: Fix trip_point_hyst_store()
Pull ACPI fixes from Rafael Wysocki:
"These fix a general ACPI processor driver regression and an ia64 build
issue, both introduced recently.
Specifics:
- Fix recently introduced uninitialized memory access issue in the
ACPI processor driver (Michal Wilczynski)
- Fix ia64 build inadvertently broken by recent ACPI processor driver
changes, which is prudent to do for 6.6 even though ia64 support is
slated for removal in 6.7 (Ard Biesheuvel)"
* tag 'acpi-6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI: processor: Fix uninitialized access of buf in acpi_set_pdc_bits()
acpi: Provide ia64 dummy implementation of acpi_proc_quirk_mwait_check()
Pull arm64 fixes from Will Deacon:
"Small crop of relatively boring arm64 fixes for -rc3.
That's not to say we don't have any juicy bugs, however, it's just
that fixes for those are likely to come via -mm and -tip for a hugetlb
and an atomics issue respectively. I get left with the
documentation...
- Fix detection of "ClearBHB" and "Hinted Conditional Branch" features
- Fix broken wildcarding for Arm PMU MAINTAINERS entry
- Add missing documentation for userspace-visible ID register fields"
* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
arm64: Document missing userspace visible fields in ID_AA64ISAR2_EL1
arm64/hbc: Document HWCAP2_HBC
arm64/sme: Include ID_AA64PFR1_EL1.SME in cpu-feature-registers.rst
arm64: cpufeature: Fix CLRBHB and BC detection
MAINTAINERS: Use wildcard pattern for ARM PMU headers
Pull x86 rethunk fixes from Borislav Petkov:
"Fix the patching ordering between static calls and return thunks"
* tag 'x86_urgent_for_v6.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86,static_call: Fix static-call vs return-thunk
x86/alternatives: Remove faulty optimization
Pull misc x86 fixes from Ingo Molnar:
- Fix a kexec bug
- Fix an UML build bug
- Fix a handful of SRSO related bugs
- Fix a shadow stacks handling bug & robustify related code
* tag 'x86-urgent-2023-09-22' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/shstk: Add warning for shadow stack double unmap
x86/shstk: Remove useless clone error handling
x86/shstk: Handle vfork clone failure correctly
x86/srso: Fix SBPB enablement for spec_rstack_overflow=off
x86/srso: Don't probe microcode in a guest
x86/srso: Set CPUID feature bits independently of bug or mitigation status
x86/srso: Fix srso_show_state() side effect
x86/asm: Fix build of UML with KASAN
x86/mm, kexec, ima: Use memblock_free_late() from ima_free_kexec_buffer()
Pull scheduler fix from Ingo Molnar:
"Fix a PF_IDLE initialization bug that generated warnings on tiny-RCU"
* tag 'sched-urgent-2023-09-22' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
kernel/sched: Modify initial boot task idle setup
