commit 2c8a7c66c9 upstream.
On the Lenovo ThinkPad X201, when Intel VT-d is enabled in the BIOS, the
kernel boots with errors related to DMAR, the graphical interface appeared
quite choppy, and the system resets erratically within a minute after it
booted:
DMAR: DRHD: handling fault status reg 3
DMAR: [DMA Write NO_PASID] Request device [00:02.0] fault addr 0xb97ff000
[fault reason 0x05] PTE Write access is not set
Upon comparing boot logs with VT-d on/off, I found that the Intel Calpella
quirk (`quirk_calpella_no_shadow_gtt()') correctly applied the igfx IOMMU
disable/quirk correctly:
pci 0000:00:00.0: DMAR: BIOS has allocated no shadow GTT; disabling IOMMU
for graphics
Whereas with VT-d on, it went into the "else" branch, which then
triggered the DMAR handling fault above:
... else if (!disable_igfx_iommu) {
/* we have to ensure the gfx device is idle before we flush */
pci_info(dev, "Disabling batched IOTLB flush on Ironlake\n");
iommu_set_dma_strict();
}
Now, this is not exactly scientific, but moving 0x0044 to quirk_iommu_igfx
seems to have fixed the aforementioned issue. Running a few `git blame'
runs on the function, I have found that the quirk was originally
introduced as a fix specific to ThinkPad X201:
commit 9eecabcb9a ("intel-iommu: Abort IOMMU setup for igfx if BIOS gave
no shadow GTT space")
Which was later revised twice to the "else" branch we saw above:
- 2011: commit 6fbcfb3e46 ("intel-iommu: Workaround IOTLB hang on
Ironlake GPU")
- 2024: commit ba00196ca4 ("iommu/vt-d: Decouple igfx_off from graphic
identity mapping")
I'm uncertain whether further testings on this particular laptops were
done in 2011 and (honestly I'm not sure) 2024, but I would be happy to do
some distro-specific testing if that's what would be required to verify
this patch.
P.S., I also see IDs 0x0040, 0x0062, and 0x006a listed under the same
`quirk_calpella_no_shadow_gtt()' quirk, but I'm not sure how similar these
chipsets are (if they share the same issue with VT-d or even, indeed, if
this issue is specific to a bug in the Lenovo BIOS). With regards to
0x0062, it seems to be a Centrino wireless card, but not a chipset?
I have also listed a couple (distro and kernel) bug reports below as
references (some of them are from 7-8 years ago!), as they seem to be
similar issue found on different Westmere/Ironlake, Haswell, and Broadwell
hardware setups.
Cc: stable@vger.kernel.org
Fixes: 6fbcfb3e46 ("intel-iommu: Workaround IOTLB hang on Ironlake GPU")
Fixes: ba00196ca4 ("iommu/vt-d: Decouple igfx_off from graphic identity mapping")
Link: https://groups.google.com/g/qubes-users/c/4NP4goUds2c?pli=1
Link: https://bugs.archlinux.org/task/65362
Link: https://bbs.archlinux.org/viewtopic.php?id=230323
Reported-by: Wenhao Sun <weiguangtwk@outlook.com>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=197029
Signed-off-by: Mingcong Bai <jeffbai@aosc.io>
Link: https://lore.kernel.org/r/20250415133330.12528-1-jeffbai@aosc.io
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 12f7802197 upstream.
UBSan caught a bug with IOMMU SVA domains, where the reported exponent
value in __arm_smmu_tlb_inv_range() was >= 64.
__arm_smmu_tlb_inv_range() uses the domain's pgsize_bitmap to compute
the number of pages to invalidate and the invalidation range. Currently
arm_smmu_sva_domain_alloc() does not setup the iommu domain's
pgsize_bitmap. This leads to __ffs() on the value returning 64 and that
leads to undefined behaviour w.r.t. shift operations
Fix this by initializing the iommu_domain's pgsize_bitmap to PAGE_SIZE.
Effectively the code needs to use the smallest page size for
invalidation
Cc: stable@vger.kernel.org
Fixes: eb6c97647b ("iommu/arm-smmu-v3: Avoid constructing invalid range commands")
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Balbir Singh <balbirs@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jean-Philippe Brucker <jean-philippe@linaro.org>
Cc: Will Deacon <will@kernel.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20250412002354.3071449-1-balbirs@nvidia.com
Signed-off-by: Will Deacon <will@kernel.org>
commit b00d24997a upstream.
ASPEED VGA card has two built-in devices:
0008:06:00.0 PCI bridge: ASPEED Technology, Inc. AST1150 PCI-to-PCI Bridge (rev 06)
0008:07:00.0 VGA compatible controller: ASPEED Technology, Inc. ASPEED Graphics Family (rev 52)
Its toplogy looks like this:
+-[0008:00]---00.0-[01-09]--+-00.0-[02-09]--+-00.0-[03]----00.0 Sandisk Corp Device 5017
| +-01.0-[04]--
| +-02.0-[05]----00.0 NVIDIA Corporation Device
| +-03.0-[06-07]----00.0-[07]----00.0 ASPEED Technology, Inc. ASPEED Graphics Family
| +-04.0-[08]----00.0 Renesas Technology Corp. uPD720201 USB 3.0 Host Controller
| \-05.0-[09]----00.0 Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
\-00.1 PMC-Sierra Inc. Device 4028
The IORT logic populaties two identical IDs into the fwspec->ids array via
DMA aliasing in iort_pci_iommu_init() called by pci_for_each_dma_alias().
Though the SMMU driver had been able to handle this situation since commit
563b5cbe33 ("iommu/arm-smmu-v3: Cope with duplicated Stream IDs"), that
got broken by the later commit cdf315f907 ("iommu/arm-smmu-v3: Maintain
a SID->device structure"), which ended up with allocating separate streams
with the same stuffing.
On a kernel prior to v6.15-rc1, there has been an overlooked warning:
pci 0008:07:00.0: vgaarb: setting as boot VGA device
pci 0008:07:00.0: vgaarb: bridge control possible
pci 0008:07:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none
pcieport 0008:06:00.0: Adding to iommu group 14
ast 0008:07:00.0: stream 67328 already in tree <===== WARNING
ast 0008:07:00.0: enabling device (0002 -> 0003)
ast 0008:07:00.0: Using default configuration
ast 0008:07:00.0: AST 2600 detected
ast 0008:07:00.0: [drm] Using analog VGA
ast 0008:07:00.0: [drm] dram MCLK=396 Mhz type=1 bus_width=16
[drm] Initialized ast 0.1.0 for 0008:07:00.0 on minor 0
ast 0008:07:00.0: [drm] fb0: astdrmfb frame buffer device
With v6.15-rc, since the commit bcb81ac6ae ("iommu: Get DT/ACPI parsing
into the proper probe path"), the error returned with the warning is moved
to the SMMU device probe flow:
arm_smmu_probe_device+0x15c/0x4c0
__iommu_probe_device+0x150/0x4f8
probe_iommu_group+0x44/0x80
bus_for_each_dev+0x7c/0x100
bus_iommu_probe+0x48/0x1a8
iommu_device_register+0xb8/0x178
arm_smmu_device_probe+0x1350/0x1db0
which then fails the entire SMMU driver probe:
pci 0008:06:00.0: Adding to iommu group 21
pci 0008:07:00.0: stream 67328 already in tree
arm-smmu-v3 arm-smmu-v3.9.auto: Failed to register iommu
arm-smmu-v3 arm-smmu-v3.9.auto: probe with driver arm-smmu-v3 failed with error -22
Since SMMU driver had been already expecting a potential duplicated Stream
ID in arm_smmu_install_ste_for_dev(), change the arm_smmu_insert_master()
routine to ignore a duplicated ID from the fwspec->sids array as well.
Note: this has been failing the iommu_device_probe() since 2021, although a
recent iommu commit in v6.15-rc1 that moves iommu_device_probe() started to
fail the SMMU driver probe. Since nobody has cared about DMA Alias support,
leave that as it was but fix the fundamental iommu_device_probe() breakage.
Fixes: cdf315f907 ("iommu/arm-smmu-v3: Maintain a SID->device structure")
Cc: stable@vger.kernel.org
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Link: https://lore.kernel.org/r/20250415185620.504299-1-nicolinc@nvidia.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 8dee308e4c upstream.
There is a string parsing logic error which can lead to an overflow of hid
or uid buffers. Comparing ACPIID_LEN against a total string length doesn't
take into account the lengths of individual hid and uid buffers so the
check is insufficient in some cases. For example if the length of hid
string is 4 and the length of the uid string is 260, the length of str
will be equal to ACPIID_LEN + 1 but uid string will overflow uid buffer
which size is 256.
The same applies to the hid string with length 13 and uid string with
length 250.
Check the length of hid and uid strings separately to prevent
buffer overflow.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: ca3bf5d47c ("iommu/amd: Introduces ivrs_acpihid kernel parameter")
Cc: stable@vger.kernel.org
Signed-off-by: Pavel Paklov <Pavel.Paklov@cyberprotect.ru>
Link: https://lore.kernel.org/r/20250325092259.392844-1-Pavel.Paklov@cyberprotect.ru
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b46064a188 upstream.
It turns out that deferred default domain creation leaves a subtle
race window during iommu_device_register() wherein a client driver may
asynchronously probe in parallel and get as far as performing DMA API
operations with dma-direct, only to be switched to iommu-dma underfoot
once the default domain attachment finally happens, with obviously
disastrous consequences. Even the wonky of_iommu_configure() path is at
risk, since iommu_fwspec_init() will no longer defer client probe as the
instance ops are (necessarily) already registered, and the "replay"
iommu_probe_device() call can see dev->iommu_group already set and so
think there's nothing to do either.
Fortunately we already have the right tool in the right place in the
form of iommu_device_use_default_domain(), which just needs to ensure
that said default domain is actually ready to *be* used. Deferring the
client probe shouldn't have too much impact, given that this only
happens while the IOMMU driver is probing, and thus due to kick the
deferred probe list again once it finishes.
Reported-by: Charan Teja Kalla <quic_charante@quicinc.com>
Fixes: 98ac73f99b ("iommu: Require a default_domain for all iommu drivers")
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/e88b94c9b575034a2c98a48b3d383654cbda7902.1740753261.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 280e5a3010 ]
If iommu_device_register() encounters an error, it can end up tearing
down already-configured groups and default domains, however this
currently still leaves devices hooked up to iommu-dma (and even
historically the behaviour in this area was at best inconsistent across
architectures/drivers...) Although in the case that an IOMMU is present
whose driver has failed to probe, users cannot necessarily expect DMA to
work anyway, it's still arguable that we should do our best to put
things back as if the IOMMU driver was never there at all, and certainly
the potential for crashing in iommu-dma itself is undesirable. Make sure
we clean up the dev->dma_iommu flag along with everything else.
Reported-by: Chen-Yu Tsai <wenst@chromium.org>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Closes: https://lore.kernel.org/all/CAGXv+5HJpTYmQ2h-GD7GjyeYT7bL9EBCvu0mz5LgpzJZtzfW0w@mail.gmail.com/
Tested-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/e788aa927f6d827dd4ea1ed608fada79f2bab030.1744284228.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 07172206a2 ]
Return -EINVAL instead of success if amd_ir_set_vcpu_affinity() is
invoked without use_vapic; lying to KVM about whether or not the IRTE was
configured to post IRQs is all kinds of bad.
Fixes: d98de49a53 ("iommu/amd: Enable vAPIC interrupt remapping mode by default")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20250404193923.1413163-6-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 55c85fa757 upstream.
The current implementation of iommufd_device_do_replace() implicitly
assumes that the input device has already been attached. However, there
is no explicit check to verify this assumption. If another device within
the same group has been attached, the replace operation might succeed,
but the input device itself may not have been attached yet.
As a result, the input device might not be tracked in the
igroup->device_list, and its reserved IOVA might not be added. Despite
this, the caller might incorrectly assume that the device has been
successfully replaced, which could lead to unexpected behavior or errors.
To address this issue, add a check to ensure that the input device has
been attached before proceeding with the replace operation. This check
will help maintain the integrity of the device tracking system and prevent
potential issues arising from incorrect assumptions about the device's
attachment status.
Fixes: e88d4ec154 ("iommufd: Add iommufd_device_replace()")
Link: https://patch.msgid.link/r/20250306034842.5950-1-yi.l.liu@intel.com
Cc: stable@vger.kernel.org
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit fb21b1568a upstream.
"attach_handle" was added exclusively for the iommufd_fault_iopf_handler()
used by IOPF/PRI use cases. Now, both the MSI and PASID series require to
reuse the attach_handle for non-fault cases.
Add a set of new attach/detach/replace helpers that does the attach_handle
allocation/releasing/replacement in the common path and also handles those
fault specific routines such as iopf enabling/disabling and auto response.
This covers both non-fault and fault cases in a clean way, replacing those
inline helpers in the header. The following patch will clean up those old
helpers in the fault.c file.
Link: https://patch.msgid.link/r/32687df01c02291d89986a9fca897bbbe2b10987.1738645017.git.nicolinc@nvidia.com
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 548183ea38 upstream.
Set the posted MSI irq_chip's irq_ack() hook to irq_move_irq() instead of
a dummy/empty callback so that posted MSIs process pending changes to the
IRQ's SMP affinity. Failure to honor a pending set-affinity results in
userspace being unable to change the effective affinity of the IRQ, as
IRQD_SETAFFINITY_PENDING is never cleared and so irq_set_affinity_locked()
always defers moving the IRQ.
The issue is most easily reproducible by setting /proc/irq/xx/smp_affinity
multiple times in quick succession, as only the first update is likely to
be handled in process context.
Fixes: ed1e48ea43 ("iommu/vt-d: Enable posted mode for device MSIs")
Cc: Robert Lippert <rlippert@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Reported-by: Wentao Yang <wentaoyang@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20250321194249.1217961-1-seanjc@google.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 93ae6e68b6 upstream.
We have recently seen report of lockdep circular lock dependency warnings
on platforms like Skylake and Kabylake:
======================================================
WARNING: possible circular locking dependency detected
6.14.0-rc6-CI_DRM_16276-gca2c04fe76e8+ #1 Not tainted
------------------------------------------------------
swapper/0/1 is trying to acquire lock:
ffffffff8360ee48 (iommu_probe_device_lock){+.+.}-{3:3},
at: iommu_probe_device+0x1d/0x70
but task is already holding lock:
ffff888102c7efa8 (&device->physical_node_lock){+.+.}-{3:3},
at: intel_iommu_init+0xe75/0x11f0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #6 (&device->physical_node_lock){+.+.}-{3:3}:
__mutex_lock+0xb4/0xe40
mutex_lock_nested+0x1b/0x30
intel_iommu_init+0xe75/0x11f0
pci_iommu_init+0x13/0x70
do_one_initcall+0x62/0x3f0
kernel_init_freeable+0x3da/0x6a0
kernel_init+0x1b/0x200
ret_from_fork+0x44/0x70
ret_from_fork_asm+0x1a/0x30
-> #5 (dmar_global_lock){++++}-{3:3}:
down_read+0x43/0x1d0
enable_drhd_fault_handling+0x21/0x110
cpuhp_invoke_callback+0x4c6/0x870
cpuhp_issue_call+0xbf/0x1f0
__cpuhp_setup_state_cpuslocked+0x111/0x320
__cpuhp_setup_state+0xb0/0x220
irq_remap_enable_fault_handling+0x3f/0xa0
apic_intr_mode_init+0x5c/0x110
x86_late_time_init+0x24/0x40
start_kernel+0x895/0xbd0
x86_64_start_reservations+0x18/0x30
x86_64_start_kernel+0xbf/0x110
common_startup_64+0x13e/0x141
-> #4 (cpuhp_state_mutex){+.+.}-{3:3}:
__mutex_lock+0xb4/0xe40
mutex_lock_nested+0x1b/0x30
__cpuhp_setup_state_cpuslocked+0x67/0x320
__cpuhp_setup_state+0xb0/0x220
page_alloc_init_cpuhp+0x2d/0x60
mm_core_init+0x18/0x2c0
start_kernel+0x576/0xbd0
x86_64_start_reservations+0x18/0x30
x86_64_start_kernel+0xbf/0x110
common_startup_64+0x13e/0x141
-> #3 (cpu_hotplug_lock){++++}-{0:0}:
__cpuhp_state_add_instance+0x4f/0x220
iova_domain_init_rcaches+0x214/0x280
iommu_setup_dma_ops+0x1a4/0x710
iommu_device_register+0x17d/0x260
intel_iommu_init+0xda4/0x11f0
pci_iommu_init+0x13/0x70
do_one_initcall+0x62/0x3f0
kernel_init_freeable+0x3da/0x6a0
kernel_init+0x1b/0x200
ret_from_fork+0x44/0x70
ret_from_fork_asm+0x1a/0x30
-> #2 (&domain->iova_cookie->mutex){+.+.}-{3:3}:
__mutex_lock+0xb4/0xe40
mutex_lock_nested+0x1b/0x30
iommu_setup_dma_ops+0x16b/0x710
iommu_device_register+0x17d/0x260
intel_iommu_init+0xda4/0x11f0
pci_iommu_init+0x13/0x70
do_one_initcall+0x62/0x3f0
kernel_init_freeable+0x3da/0x6a0
kernel_init+0x1b/0x200
ret_from_fork+0x44/0x70
ret_from_fork_asm+0x1a/0x30
-> #1 (&group->mutex){+.+.}-{3:3}:
__mutex_lock+0xb4/0xe40
mutex_lock_nested+0x1b/0x30
__iommu_probe_device+0x24c/0x4e0
probe_iommu_group+0x2b/0x50
bus_for_each_dev+0x7d/0xe0
iommu_device_register+0xe1/0x260
intel_iommu_init+0xda4/0x11f0
pci_iommu_init+0x13/0x70
do_one_initcall+0x62/0x3f0
kernel_init_freeable+0x3da/0x6a0
kernel_init+0x1b/0x200
ret_from_fork+0x44/0x70
ret_from_fork_asm+0x1a/0x30
-> #0 (iommu_probe_device_lock){+.+.}-{3:3}:
__lock_acquire+0x1637/0x2810
lock_acquire+0xc9/0x300
__mutex_lock+0xb4/0xe40
mutex_lock_nested+0x1b/0x30
iommu_probe_device+0x1d/0x70
intel_iommu_init+0xe90/0x11f0
pci_iommu_init+0x13/0x70
do_one_initcall+0x62/0x3f0
kernel_init_freeable+0x3da/0x6a0
kernel_init+0x1b/0x200
ret_from_fork+0x44/0x70
ret_from_fork_asm+0x1a/0x30
other info that might help us debug this:
Chain exists of:
iommu_probe_device_lock --> dmar_global_lock -->
&device->physical_node_lock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&device->physical_node_lock);
lock(dmar_global_lock);
lock(&device->physical_node_lock);
lock(iommu_probe_device_lock);
*** DEADLOCK ***
This driver uses a global lock to protect the list of enumerated DMA
remapping units. It is necessary due to the driver's support for dynamic
addition and removal of remapping units at runtime.
Two distinct code paths require iteration over this remapping unit list:
- Device registration and probing: the driver iterates the list to
register each remapping unit with the upper layer IOMMU framework
and subsequently probe the devices managed by that unit.
- Global configuration: Upper layer components may also iterate the list
to apply configuration changes.
The lock acquisition order between these two code paths was reversed. This
caused lockdep warnings, indicating a risk of deadlock. Fix this warning
by releasing the global lock before invoking upper layer interfaces for
device registration.
Fixes: b150654f74 ("iommu/vt-d: Fix suspicious RCU usage")
Closes: https://lore.kernel.org/linux-iommu/SJ1PR11MB612953431F94F18C954C4A9CB9D32@SJ1PR11MB6129.namprd11.prod.outlook.com/
Tested-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20250317035714.1041549-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 688124cc54 upstream.
Don't overwrite an IRTE that is posting IRQs to a vCPU with a posted MSI
entry if the host IRQ affinity happens to change. If/when the IRTE is
reverted back to "host mode", it will be reconfigured as a posted MSI or
remapped entry as appropriate.
Drop the "mode" field, which doesn't differentiate between posted MSIs and
posted vCPUs, in favor of a dedicated posted_vcpu flag. Note! The two
posted_{msi,vcpu} flags are intentionally not mutually exclusive; an IRTE
can transition between posted MSI and posted vCPU.
Fixes: ed1e48ea43 ("iommu/vt-d: Enable posted mode for device MSIs")
Cc: stable@vger.kernel.org
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jacob Pan <jacob.jun.pan@linux.intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Link: https://lore.kernel.org/r/20250315025135.2365846-3-seanjc@google.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 2454823e97 upstream.
Add a helper to take care of reconfiguring an IRTE to deliver IRQs to the
host, i.e. not to a vCPU, and use the helper when an IRTE's vCPU affinity
is nullified, i.e. when KVM puts an IRTE back into "host" mode. Because
posted MSIs use an ephemeral IRTE, using modify_irte() puts the IRTE into
full remapped mode, i.e. unintentionally disables posted MSIs on the IRQ.
Fixes: ed1e48ea43 ("iommu/vt-d: Enable posted mode for device MSIs")
Cc: stable@vger.kernel.org
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jacob Pan <jacob.jun.pan@linux.intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Link: https://lore.kernel.org/r/20250315025135.2365846-2-seanjc@google.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 767e22001d upstream.
Two WARNINGs are observed when SMMU driver rolls back upon failure:
arm-smmu-v3.9.auto: Failed to register iommu
arm-smmu-v3.9.auto: probe with driver arm-smmu-v3 failed with error -22
------------[ cut here ]------------
WARNING: CPU: 5 PID: 1 at kernel/dma/mapping.c:74 dmam_free_coherent+0xc0/0xd8
Call trace:
dmam_free_coherent+0xc0/0xd8 (P)
tegra241_vintf_free_lvcmdq+0x74/0x188
tegra241_cmdqv_remove_vintf+0x60/0x148
tegra241_cmdqv_remove+0x48/0xc8
arm_smmu_impl_remove+0x28/0x60
devm_action_release+0x1c/0x40
------------[ cut here ]------------
128 pages are still in use!
WARNING: CPU: 16 PID: 1 at mm/page_alloc.c:6902 free_contig_range+0x18c/0x1c8
Call trace:
free_contig_range+0x18c/0x1c8 (P)
cma_release+0x154/0x2f0
dma_free_contiguous+0x38/0xa0
dma_direct_free+0x10c/0x248
dma_free_attrs+0x100/0x290
dmam_free_coherent+0x78/0xd8
tegra241_vintf_free_lvcmdq+0x74/0x160
tegra241_cmdqv_remove+0x98/0x198
arm_smmu_impl_remove+0x28/0x60
devm_action_release+0x1c/0x40
This is because the LVCMDQ queue memory are managed by devres, while that
dmam_free_coherent() is called in the context of devm_action_release().
Jason pointed out that "arm_smmu_impl_probe() has mis-ordered the devres
callbacks if ops->device_remove() is going to be manually freeing things
that probe allocated":
https://lore.kernel.org/linux-iommu/20250407174408.GB1722458@nvidia.com/
In fact, tegra241_cmdqv_init_structures() only allocates memory resources
which means any failure that it generates would be similar to -ENOMEM, so
there is no point in having that "falling back to standard SMMU" routine,
as the standard SMMU would likely fail to allocate memory too.
Remove the unwind part in tegra241_cmdqv_init_structures(), and return a
proper error code to ask SMMU driver to call tegra241_cmdqv_remove() via
impl_ops->device_remove(). Then, drop tegra241_vintf_free_lvcmdq() since
devres will take care of that.
Fixes: 483e0bd888 ("iommu/tegra241-cmdqv: Do not allocate vcmdq until dma_set_mask_and_coherent")
Cc: stable@vger.kernel.org
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20250407201908.172225-1-nicolinc@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 38e8844005 ]
Currently, mtk_iommu calls during probe iommu_device_register before
the hw_list from driver data is initialized. Since iommu probing issue
fix, it leads to NULL pointer dereference in mtk_iommu_device_group when
hw_list is accessed with list_first_entry (not null safe).
So, change the call order to ensure iommu_device_register is called
after the driver data are initialized.
Fixes: 9e3a2a6436 ("iommu/mediatek: Adapt sharing and non-sharing pgtable case")
Fixes: bcb81ac6ae ("iommu: Get DT/ACPI parsing into the proper probe path")
Reviewed-by: Yong Wu <yong.wu@mediatek.com>
Tested-by: Chen-Yu Tsai <wenst@chromium.org> # MT8183 Juniper, MT8186 Tentacruel
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Louis-Alexis Eyraud <louisalexis.eyraud@collabora.com>
Link: https://lore.kernel.org/r/20250403-fix-mtk-iommu-error-v2-1-fe8b18f8b0a8@collabora.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 99deffc409 ]
Commit bcb81ac6ae ("iommu: Get DT/ACPI parsing into the proper probe
path") changed the sequence of probing the SYSMMU controller devices and
calls to arm_iommu_attach_device(), what results in resuming SYSMMU
controller earlier, when it is still set to IDENTITY mapping. Such change
revealed the bug in IDENTITY handling in the exynos-iommu driver. When
SYSMMU controller is set to IDENTITY mapping, data->domain is NULL, so
adjust checks in suspend & resume callbacks to handle this case
correctly.
Fixes: b3d14960e6 ("iommu/exynos: Implement an IDENTITY domain")
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Link: https://lore.kernel.org/r/20250401202731.2810474-1-m.szyprowski@samsung.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit b150654f74 upstream.
Commit <d74169ceb0> ("iommu/vt-d: Allocate DMAR fault interrupts
locally") moved the call to enable_drhd_fault_handling() to a code
path that does not hold any lock while traversing the drhd list. Fix
it by ensuring the dmar_global_lock lock is held when traversing the
drhd list.
Without this fix, the following warning is triggered:
=============================
WARNING: suspicious RCU usage
6.14.0-rc3 #55 Not tainted
-----------------------------
drivers/iommu/intel/dmar.c:2046 RCU-list traversed in non-reader section!!
other info that might help us debug this:
rcu_scheduler_active = 1, debug_locks = 1
2 locks held by cpuhp/1/23:
#0: ffffffff84a67c50 (cpu_hotplug_lock){++++}-{0:0}, at: cpuhp_thread_fun+0x87/0x2c0
#1: ffffffff84a6a380 (cpuhp_state-up){+.+.}-{0:0}, at: cpuhp_thread_fun+0x87/0x2c0
stack backtrace:
CPU: 1 UID: 0 PID: 23 Comm: cpuhp/1 Not tainted 6.14.0-rc3 #55
Call Trace:
<TASK>
dump_stack_lvl+0xb7/0xd0
lockdep_rcu_suspicious+0x159/0x1f0
? __pfx_enable_drhd_fault_handling+0x10/0x10
enable_drhd_fault_handling+0x151/0x180
cpuhp_invoke_callback+0x1df/0x990
cpuhp_thread_fun+0x1ea/0x2c0
smpboot_thread_fn+0x1f5/0x2e0
? __pfx_smpboot_thread_fn+0x10/0x10
kthread+0x12a/0x2d0
? __pfx_kthread+0x10/0x10
ret_from_fork+0x4a/0x60
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
Holding the lock in enable_drhd_fault_handling() triggers a lockdep splat
about a possible deadlock between dmar_global_lock and cpu_hotplug_lock.
This is avoided by not holding dmar_global_lock when calling
iommu_device_register(), which initiates the device probe process.
Fixes: d74169ceb0 ("iommu/vt-d: Allocate DMAR fault interrupts locally")
Reported-and-tested-by: Ido Schimmel <idosch@nvidia.com>
Closes: https://lore.kernel.org/linux-iommu/Zx9OwdLIc_VoQ0-a@shredder.mtl.com/
Tested-by: Breno Leitao <leitao@debian.org>
Cc: stable@vger.kernel.org
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20250218022422.2315082-1-baolu.lu@linux.intel.com
Tested-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 64f792981e upstream.
Remove the device comparison check in context_setup_pass_through_cb.
pci_for_each_dma_alias already makes a decision on whether the
callback function should be called for a device. With the check
in place it will fail to create context entries for aliases as
it walks up to the root bus.
Fixes: 2031c469f8 ("iommu/vt-d: Add support for static identity domain")
Closes: https://lore.kernel.org/linux-iommu/82499eb6-00b7-4f83-879a-e97b4144f576@linux.intel.com/
Cc: stable@vger.kernel.org
Signed-off-by: Jerry Snitselaar <jsnitsel@redhat.com>
Link: https://lore.kernel.org/r/20250224180316.140123-1-jsnitsel@redhat.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9759ae2cee upstream.
The iopf_queue_remove_device() helper removes a device from the per-iommu
iopf queue when PRI is disabled on the device. It responds to all
outstanding iopf's with an IOMMU_PAGE_RESP_INVALID code and detaches the
device from the queue.
However, it fails to release the group structure that represents a group
of iopf's awaiting for a response after responding to the hardware. This
can cause a memory leak if iopf_queue_remove_device() is called with
pending iopf's.
Fix it by calling iopf_free_group() after the iopf group is responded.
Fixes: 1991123271 ("iommu: Track iopf group instead of last fault")
Cc: stable@vger.kernel.org
Suggested-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20250117055800.782462-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit ef75966abf ]
With recent kernel, AMDGPU failed to resume after suspend on certain laptop.
Sample log:
-----------
Nov 14 11:52:19 Thinkbook kernel: iommu ivhd0: AMD-Vi: Event logged [ILLEGAL_DEV_TABLE_ENTRY device=0000:06:00.0 pasid=0x00000 address=0x135300000 flags=0x0080]
Nov 14 11:52:19 Thinkbook kernel: AMD-Vi: DTE[0]: 7d90000000000003
Nov 14 11:52:19 Thinkbook kernel: AMD-Vi: DTE[1]: 0000100103fc0009
Nov 14 11:52:19 Thinkbook kernel: AMD-Vi: DTE[2]: 2000000117840013
Nov 14 11:52:19 Thinkbook kernel: AMD-Vi: DTE[3]: 0000000000000000
This is because in resume path, CNTRL[EPHEn] is not set. Fix this by
setting CNTRL[EPHEn] to 1 in resume path if EFR[EPHSUP] is set.
Note
May be better approach is to save the control register in suspend path
and restore it in resume path instead of trying to set indivisual
bits. We will have separate patch for that.
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219499
Fixes: c4cb231111 ("iommu/amd: Add support for enable/disable IOPF")
Tested-by: Hamish McIntyre-Bhatty <kernel-bugzilla@regd.hamishmb.com>
Signed-off-by: Vasant Hegde <vasant.hegde@amd.com>
Link: https://lore.kernel.org/r/20250127094411.5931-1-vasant.hegde@amd.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 3d49020a32 upstream.
The fault->mutex serializes the fault read()/write() fops and the
iommufd_fault_auto_response_faults(), mainly for fault->response. Also, it
was conveniently used to fence the fault->deliver in poll() fop and
iommufd_fault_iopf_handler().
However, copy_from/to_user() may sleep if pagefaults are enabled. Thus,
they could take a long time to wait for user pages to swap in, blocking
iommufd_fault_iopf_handler() and its caller that is typically a shared IRQ
handler of an IOMMU driver, resulting in a potential global DOS.
Instead of reusing the mutex to protect the fault->deliver list, add a
separate spinlock, nested under the mutex, to do the job.
iommufd_fault_iopf_handler() would no longer be blocked by
copy_from/to_user().
Add a free_list in iommufd_auto_response_faults(), so the spinlock can
simply fence a fast list_for_each_entry_safe routine.
Provide two deliver list helpers for iommufd_fault_fops_read() to use:
- Fetch the first iopf_group out of the fault->deliver list
- Restore an iopf_group back to the head of the fault->deliver list
Lastly, move the mutex closer to the response in the fault structure,
and update its kdoc accordingly.
Fixes: 07838f7fd5 ("iommufd: Add iommufd fault object")
Link: https://patch.msgid.link/r/20250117192901.79491-1-nicolinc@nvidia.com
Cc: stable@vger.kernel.org
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e94dc6ddda upstream.
The hardware limitation "max=19" actually comes from SMMU Command Queue.
So, it'd be more natural for tegra241-cmdqv driver to read it out rather
than hardcoding it itself.
This is not an issue yet for a kernel on a baremetal system, but a guest
kernel setting the queue base/size in form of IPA/gPA might result in a
noncontiguous queue in the physical address space, if underlying physical
pages backing up the guest RAM aren't contiguous entirely: e.g. 2MB-page
backed guest RAM cannot guarantee a contiguous queue if it is 8MB (capped
to VCMDQ_LOG2SIZE_MAX=19). This might lead to command errors when HW does
linear-read from a noncontiguous queue memory.
Adding this extra IDR1.CMDQS cap (in the guest kernel) allows VMM to set
SMMU's IDR1.CMDQS=17 for the case mentioned above, so a guest-level queue
will be capped to maximum 2MB, ensuring a contiguous queue memory.
Fixes: a3799717b8 ("iommu/tegra241-cmdqv: Fix alignment failure at max_n_shift")
Reported-by: Ian Kalinowski <ikalinowski@nvidia.com>
Cc: stable@vger.kernel.org
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Link: https://lore.kernel.org/r/20241219051421.1850267-1-nicolinc@nvidia.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit fcbd621567 ]
kmemleak noticed that the iopf queue allocated deep down within
arm_smmu_init_structures() can be leaked by a subsequent error return
from arm_smmu_device_probe(). Furthermore, after arm_smmu_device_reset()
we will also leave the SMMU enabled with an empty Stream Table, silently
blocking all DMA. This proves rather annoying for debugging said probe
failure, so let's handle it a bit better by putting the SMMU back into
(more or less) the same state as if it hadn't probed at all.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/5137901958471cf67f2fad5c2229f8a8f1ae901a.1733406914.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4231473890 ]
Add the compatible for the separate IOMMU on SDM670 for the Adreno GPU.
This IOMMU has the compatible strings:
"qcom,sdm670-smmu-v2", "qcom,adreno-smmu", "qcom,smmu-v2"
While the SMMU 500 doesn't need an entry for this specific SoC, the
SMMU v2 compatible should have its own entry, as the fallback entry in
arm-smmu.c handles "qcom,smmu-v2" without per-process page table support
unless there is an entry here. This entry can't be the
"qcom,adreno-smmu" compatible because dedicated GPU IOMMUs can also be
SMMU 500 with different handling.
Signed-off-by: Richard Acayan <mailingradian@gmail.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20241114004713.42404-6-mailingradian@gmail.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e24c155105 ]
Resolve a UBSAN shift-out-of-bounds issue in iova_bitmap_offset_to_index()
where shifting the constant "1" (of type int) by bitmap->mapped.pgshift
(an unsigned long value) could result in undefined behavior.
The constant "1" defaults to a 32-bit "int", and when "pgshift" exceeds
31 (e.g., pgshift = 63) the shift operation overflows, as the result
cannot be represented in a 32-bit type.
To resolve this, the constant is updated to "1UL", promoting it to an
unsigned long type to match the operand's type.
Fixes: 58ccf0190d ("vfio: Add an IOVA bitmap support")
Link: https://patch.msgid.link/r/20250113223820.10713-1-qasdev00@gmail.com
Reported-by: syzbot <syzbot+85992ace37d5b7b51635@syzkaller.appspotmail.com>
Closes: https://syzkaller.appspot.com/bug?extid=85992ace37d5b7b51635
Signed-off-by: Qasim Ijaz <qasdev00@gmail.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d9df72c6ac ]
Fix an issue detected by syzbot:
WARNING in iommufd_device_unbind iommufd: Time out waiting for iommufd object to become free
Resolve a warning in iommufd_device_unbind caused by a timeout while
waiting for the shortterm_users reference count to reach zero. The
existing 10-second timeout is insufficient in some scenarios, resulting in
failures the above warning.
Increase the timeout in iommufd_object_dec_wait_shortterm from 10 seconds
to 60 seconds to allow sufficient time for the reference count to drop to
zero. This change prevents premature timeouts and reduces the likelihood
of warnings during iommufd_device_unbind.
Fixes: 6f9c4d8c46 ("iommufd: Do not UAF during iommufd_put_object()")
Link: https://patch.msgid.link/r/20241123195900.3176-1-surajsonawane0215@gmail.com
Reported-by: syzbot+c92878e123785b1fa2db@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=c92878e123785b1fa2db
Tested-by: syzbot+c92878e123785b1fa2db@syzkaller.appspotmail.com
Signed-off-by: Suraj Sonawane <surajsonawane0215@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 9b640ae7fb ]
The SMMUv3 spec has a note that BYPASS and ATS don't work together under
the STE EATS field definition. However there is another section "13.6.4
Full ATS skipping stage 1" that explains under certain conditions BYPASS
and ATS do work together if the STE is using S1DSS to select BYPASS and
the CD table has the possibility for a substream.
When these comments were written the understanding was that all forms of
BYPASS just didn't work and this was to be a future problem to solve.
It turns out that ATS and IDENTITY will always work just fine:
- If STE.Config = BYPASS then the PCI ATS is disabled
- If a PASID domain is attached then S1DSS = BYPASS and ATS will be
enabled. This meets the requirements of 13.6.4 to automatically
generate 1:1 ATS replies on the RID.
Update the comments to reflect this.
Fixes: 7497f4211f ("iommu/arm-smmu-v3: Make changing domains be hitless for ATS")
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/0-v1-f27174f44f39+27a33-smmuv3_ats_note_jgg@nvidia.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 74536f9196 upstream.
The qi_batch is allocated when assigning cache tag for a domain. While
for nested parent domain, it is missed. Hence, when trying to map pages
to the nested parent, NULL dereference occurred. Also, there is potential
memleak since there is no lock around domain->qi_batch allocation.
To solve it, add a helper for qi_batch allocation, and call it in both
the __cache_tag_assign_domain() and __cache_tag_assign_parent_domain().
BUG: kernel NULL pointer dereference, address: 0000000000000200
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 8104795067 P4D 0
Oops: Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 223 UID: 0 PID: 4357 Comm: qemu-system-x86 Not tainted 6.13.0-rc1-00028-g4b50c3c3b998-dirty #2632
Call Trace:
? __die+0x24/0x70
? page_fault_oops+0x80/0x150
? do_user_addr_fault+0x63/0x7b0
? exc_page_fault+0x7c/0x220
? asm_exc_page_fault+0x26/0x30
? cache_tag_flush_range_np+0x13c/0x260
intel_iommu_iotlb_sync_map+0x1a/0x30
iommu_map+0x61/0xf0
batch_to_domain+0x188/0x250
iopt_area_fill_domains+0x125/0x320
? rcu_is_watching+0x11/0x50
iopt_map_pages+0x63/0x100
iopt_map_common.isra.0+0xa7/0x190
iopt_map_user_pages+0x6a/0x80
iommufd_ioas_map+0xcd/0x1d0
iommufd_fops_ioctl+0x118/0x1c0
__x64_sys_ioctl+0x93/0xc0
do_syscall_64+0x71/0x140
entry_SYSCALL_64_after_hwframe+0x76/0x7e
Fixes: 705c1cdf1e ("iommu/vt-d: Introduce batched cache invalidation")
Cc: stable@vger.kernel.org
Co-developed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20241210130322.17175-1-yi.l.liu@intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1f2557e08a upstream.
The current implementation removes cache tags after disabling ATS,
leading to potential memory leaks and kernel crashes. Specifically,
CACHE_TAG_DEVTLB type cache tags may still remain in the list even
after the domain is freed, causing a use-after-free condition.
This issue really shows up when multiple VFs from different PFs
passed through to a single user-space process via vfio-pci. In such
cases, the kernel may crash with kernel messages like:
BUG: kernel NULL pointer dereference, address: 0000000000000014
PGD 19036a067 P4D 1940a3067 PUD 136c9b067 PMD 0
Oops: Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 74 UID: 0 PID: 3183 Comm: testCli Not tainted 6.11.9 #2
RIP: 0010:cache_tag_flush_range+0x9b/0x250
Call Trace:
<TASK>
? __die+0x1f/0x60
? page_fault_oops+0x163/0x590
? exc_page_fault+0x72/0x190
? asm_exc_page_fault+0x22/0x30
? cache_tag_flush_range+0x9b/0x250
? cache_tag_flush_range+0x5d/0x250
intel_iommu_tlb_sync+0x29/0x40
intel_iommu_unmap_pages+0xfe/0x160
__iommu_unmap+0xd8/0x1a0
vfio_unmap_unpin+0x182/0x340 [vfio_iommu_type1]
vfio_remove_dma+0x2a/0xb0 [vfio_iommu_type1]
vfio_iommu_type1_ioctl+0xafa/0x18e0 [vfio_iommu_type1]
Move cache_tag_unassign_domain() before iommu_disable_pci_caps() to fix
it.
Fixes: 3b1d9e2b2d ("iommu/vt-d: Add cache tag assignment interface")
Cc: stable@vger.kernel.org
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20241129020506.576413-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1f80621816 upstream.
During boot some of the calls to tegra241_cmdqv_get_cmdq() will happen
in preemptible context. As this function calls smp_processor_id(), if
CONFIG_DEBUG_PREEMPT is enabled, these calls will trigger a series of
"BUG: using smp_processor_id() in preemptible" backtraces.
As tegra241_cmdqv_get_cmdq() only calls smp_processor_id() to use the
CPU number as a factor to balance out traffic on cmdq usage, it is safe
to use raw_smp_processor_id() here.
Cc: <stable@vger.kernel.org>
Fixes: 918eb5c856 ("iommu/arm-smmu-v3: Add in-kernel support for NVIDIA Tegra241 (Grace) CMDQV")
Signed-off-by: Luis Claudio R. Goncalves <lgoncalv@redhat.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Nicolin Chen <nicolinc@nvidia.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Link: https://lore.kernel.org/r/Z1L1mja3nXzsJ0Pk@uudg.org
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit e3a682eaf2 ]
If a page is mapped starting at 0 that is equal to or larger than can fit
in the current mode (number of table levels) it results in corrupting the
mapping as the following logic assumes the mode is correct for the page
size being requested.
There are two issues here, the check if the address fits within the table
uses the start address, it should use the last address to ensure that last
byte of the mapping fits within the current table mode.
The second is if the mapping is exactly the size of the full page table it
has to add another level to instead hold a single IOPTE for the large
size.
Since both corner cases require a 0 IOVA to be hit and doesn't start until
a page size of 2^48 it is unlikely to ever hit in a real system.
Reported-by: Alejandro Jimenez <alejandro.j.jimenez@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/0-v1-27ab08d646a1+29-amd_0map_jgg@nvidia.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit af7f478051 upstream.
As fput() calls the file->f_op->release op, where fault obj and ictx are
getting released, there is no need to release these two after fput() one
more time, which would result in imbalanced refcounts:
refcount_t: decrement hit 0; leaking memory.
WARNING: CPU: 48 PID: 2369 at lib/refcount.c:31 refcount_warn_saturate+0x60/0x230
Call trace:
refcount_warn_saturate+0x60/0x230 (P)
refcount_warn_saturate+0x60/0x230 (L)
iommufd_fault_fops_release+0x9c/0xe0 [iommufd]
...
VFS: Close: file count is 0 (f_op=iommufd_fops [iommufd])
WARNING: CPU: 48 PID: 2369 at fs/open.c:1507 filp_flush+0x3c/0xf0
Call trace:
filp_flush+0x3c/0xf0 (P)
filp_flush+0x3c/0xf0 (L)
__arm64_sys_close+0x34/0x98
...
imbalanced put on file reference count
WARNING: CPU: 48 PID: 2369 at fs/file.c:74 __file_ref_put+0x100/0x138
Call trace:
__file_ref_put+0x100/0x138 (P)
__file_ref_put+0x100/0x138 (L)
__fput_sync+0x4c/0xd0
Drop those two lines to fix the warnings above.
Cc: stable@vger.kernel.org
Fixes: 07838f7fd5 ("iommufd: Add iommufd fault object")
Link: https://patch.msgid.link/r/b5651beb3a6b1adeef26fffac24607353bf67ba1.1733212723.git.nicolinc@nvidia.com
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 229e6ee43d upstream.
Null pointer dereference occurs due to a race between smmu
driver probe and client driver probe, when of_dma_configure()
for client is called after the iommu_device_register() for smmu driver
probe has executed but before the driver_bound() for smmu driver
has been called.
Following is how the race occurs:
T1:Smmu device probe T2: Client device probe
really_probe()
arm_smmu_device_probe()
iommu_device_register()
really_probe()
platform_dma_configure()
of_dma_configure()
of_dma_configure_id()
of_iommu_configure()
iommu_probe_device()
iommu_init_device()
arm_smmu_probe_device()
arm_smmu_get_by_fwnode()
driver_find_device_by_fwnode()
driver_find_device()
next_device()
klist_next()
/* null ptr
assigned to smmu */
/* null ptr dereference
while smmu->streamid_mask */
driver_bound()
klist_add_tail()
When this null smmu pointer is dereferenced later in
arm_smmu_probe_device, the device crashes.
Fix this by deferring the probe of the client device
until the smmu device has bound to the arm smmu driver.
Fixes: 021bb8420d ("iommu/arm-smmu: Wire up generic configuration support")
Cc: stable@vger.kernel.org
Co-developed-by: Prakash Gupta <quic_guptap@quicinc.com>
Signed-off-by: Prakash Gupta <quic_guptap@quicinc.com>
Signed-off-by: Pratyush Brahma <quic_pbrahma@quicinc.com>
Link: https://lore.kernel.org/r/20241004090428.2035-1-quic_pbrahma@quicinc.com
[will: Add comment]
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit d71fa842d3 upstream.
ARM_LPAE_LVL_IDX() takes into account concatenated PGDs and can return
an index spanning multiple page-table pages given a sufficiently large
input address. However, when the resulting index is used to calculate
the number of remaining entries in the page, the possibility of
concatenation is ignored and we end up computing a negative upper bound:
max_entries = ARM_LPAE_PTES_PER_TABLE(data) - map_idx_start;
On the map path, this results in a negative 'mapped' value being
returned but on the unmap path we can leak child tables if they are
skipped in __arm_lpae_free_pgtable().
Introduce an arm_lpae_max_entries() helper to convert a table index into
the remaining number of entries within a single page-table page.
Cc: <stable@vger.kernel.org>
Signed-off-by: Mostafa Saleh <smostafa@google.com>
Link: https://lore.kernel.org/r/20241024162516.2005652-2-smostafa@google.com
[will: Tweaked comment and commit message]
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5492f0c408 upstream.
While testing some io-pgtable changes, I ran into a compiler warning
from the Tegra CMDQ driver:
drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c:803:23: warning: unused variable 'cmdqv_debugfs_dir' [-Wunused-variable]
803 | static struct dentry *cmdqv_debugfs_dir;
| ^~~~~~~~~~~~~~~~~
1 warning generated.
Guard the variable declaration with CONFIG_IOMMU_DEBUGFS to silence the
warning.
Signed-off-by: Will Deacon <will@kernel.org>
Cc: Jiri Slaby <jirislaby@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a3799717b8 ]
When configuring a kernel with PAGE_SIZE=4KB, depending on its setting of
CONFIG_CMA_ALIGNMENT, VCMDQ_LOG2SIZE_MAX=19 could fail the alignment test
and trigger a WARN_ON:
WARNING: at drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c:3646
Call trace:
arm_smmu_init_one_queue+0x15c/0x210
tegra241_cmdqv_init_structures+0x114/0x338
arm_smmu_device_probe+0xb48/0x1d90
Fix it by capping max_n_shift to CMDQ_MAX_SZ_SHIFT as SMMUv3 CMDQ does.
Fixes: 918eb5c856 ("iommu/arm-smmu-v3: Add in-kernel support for NVIDIA Tegra241 (Grace) CMDQV")
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Link: https://lore.kernel.org/r/20241111030226.1940737-1-nicolinc@nvidia.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f1645676f2 ]
There are some issues in pgtable_walk():
1. Super page is dumped as non-present page
2. dma_pte_superpage() should not check against leaf page table entries
3. Pointer pte is never NULL so checking it is meaningless
4. When an entry is not present, it still makes sense to dump the entry
content.
Fix 1,2 by checking dma_pte_superpage()'s returned value after level check.
Fix 3 by removing pte check.
Fix 4 by checking present bit after printing.
By this chance, change to print "page table not present" instead of "PTE
not present" to be clearer.
Fixes: 914ff7719e ("iommu/vt-d: Dump DMAR translation structure when DMA fault occurs")
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Link: https://lore.kernel.org/r/20241024092146.715063-3-zhenzhong.duan@intel.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 6ceb93f952 ]
There are some issues in dmar_fault_dump_ptes():
1. return value of phys_to_virt() is used for checking if an entry is
present.
2. dump is confusing, e.g., "pasid table entry is not present", confusing
by unpresent pasid table vs. unpresent pasid table entry. Current code
means the former.
3. pgtable_walk() is called without checking if page table is present.
Fix 1 by checking present bit of an entry before dump a lower level entry.
Fix 2 by removing "entry" string, e.g., "pasid table is not present".
Fix 3 by checking page table present before walk.
Take issue 3 for example, before fix:
[ 442.240357] DMAR: pasid dir entry: 0x000000012c83e001
[ 442.246661] DMAR: pasid table entry[0]: 0x0000000000000000
[ 442.253429] DMAR: pasid table entry[1]: 0x0000000000000000
[ 442.260203] DMAR: pasid table entry[2]: 0x0000000000000000
[ 442.266969] DMAR: pasid table entry[3]: 0x0000000000000000
[ 442.273733] DMAR: pasid table entry[4]: 0x0000000000000000
[ 442.280479] DMAR: pasid table entry[5]: 0x0000000000000000
[ 442.287234] DMAR: pasid table entry[6]: 0x0000000000000000
[ 442.293989] DMAR: pasid table entry[7]: 0x0000000000000000
[ 442.300742] DMAR: PTE not present at level 2
After fix:
...
[ 357.241214] DMAR: pasid table entry[6]: 0x0000000000000000
[ 357.248022] DMAR: pasid table entry[7]: 0x0000000000000000
[ 357.254824] DMAR: scalable mode page table is not present
Fixes: 914ff7719e ("iommu/vt-d: Dump DMAR translation structure when DMA fault occurs")
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Link: https://lore.kernel.org/r/20241024092146.715063-2-zhenzhong.duan@intel.com
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit ecda483339 ]
This fixes a crash when surprise hot-unplugging a PCI device. This crash
happens because during hot-unplug __iommu_group_set_domain_nofail()
attaching the default domain fails when the platform no longer
recognizes the device as it has already been removed and we end up with
a NULL domain pointer and UAF. This is exactly the case referred to in
the second comment in __iommu_device_set_domain() and just as stated
there if we can instead attach the blocking domain the UAF is prevented
as this can handle the already removed device. Implement the blocking
domain to use this handling. With this change, the crash is fixed but
we still hit a warning attempting to change DMA ownership on a blocked
device.
Fixes: c76c067e48 ("s390/pci: Use dma-iommu layer")
Co-developed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20240910211516.137933-1-mjrosato@linux.ibm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Replace comma between expressions with semicolons.
Using a ',' in place of a ';' can have unintended side effects.
Although that is not the case here, it is seems best to use ';'
unless ',' is intended.
Found by inspection.
No functional change intended.
Compile tested only.
Fixes: e3b1be2e73 ("iommu/arm-smmu-v3: Reorganize struct arm_smmu_ctx_desc_cfg")
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20240923021557.3432068-1-nichen@iscas.ac.cn
Signed-off-by: Will Deacon <will@kernel.org>
The function arm_smmu_init_strtab_2lvl uses the expression
((1 << smmu->sid_bits) - 1)
to calculate the largest StreamID value. However, this fails for the
maximum allowed value of SMMU_IDR1.SIDSIZE which is 32. The C standard
states:
"If the value of the right operand is negative or is greater than or
equal to the width of the promoted left operand, the behavior is
undefined."
With smmu->sid_bits being 32, the prerequisites for undefined behavior
are met. We observed that the value of (1 << 32) is 1 and not 0 as we
initially expected.
Similar bit shift operations in arm_smmu_init_strtab_linear seem to not
be affected, because it appears to be unlikely for an SMMU to have
SMMU_IDR1.SIDSIZE set to 32 but then not support 2-level Stream tables
This issue was found by Ryan Huang <tzukui@google.com> on our team.
Fixes: ce410410f1 ("iommu/arm-smmu-v3: Add arm_smmu_strtab_l1/2_idx()")
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Link: https://lore.kernel.org/r/20241002015357.1766934-1-danielmentz@google.com
Signed-off-by: Will Deacon <will@kernel.org>
CPRE workarounds are implicated in at least 5 MMU-500 errata, some of
which remain unfixed. The comment and warning message have proven to be
unhelpfully misleading about this scope, so reword them to get the point
across with less risk of going out of date or confusing users.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/dfa82171b5248ad7cf1f25592101a6eec36b8c9a.1728400877.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
no_llseek had been defined to NULL two years ago, in commit 868941b144
("fs: remove no_llseek")
To quote that commit,
At -rc1 we'll need do a mechanical removal of no_llseek -
git grep -l -w no_llseek | grep -v porting.rst | while read i; do
sed -i '/\<no_llseek\>/d' $i
done
would do it.
Unfortunately, that hadn't been done. Linus, could you do that now, so
that we could finally put that thing to rest? All instances are of the
form
.llseek = no_llseek,
so it's obviously safe.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
