IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Pass-through devices to VM guest can get updated IRQ affinity
information via irq_set_affinity() when not running in guest mode.
Currently, AMD IOMMU driver in GA mode ignores the updated information
if the pass-through device is setup to use vAPIC regardless of guest_mode.
This could cause invalid interrupt remapping.
Also, the guest_mode bit should be set and cleared only when
SVM updates posted-interrupt interrupt remapping information.
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Cc: Joerg Roedel <jroedel@suse.de>
Fixes: d98de49a53e48 ('iommu/amd: Enable vAPIC interrupt remapping mode by default')
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Most dma_map_ops structures are never modified. Constify these
structures such that these can be write-protected.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
This callback should never return NULL. Print a warning if
that happens so that we notice and can fix it.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The generic device_group call-backs in iommu.c return NULL
in case of error. Since they are getting ERR_PTR values from
iommu_group_alloc(), just pass them up instead.
Reported-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The iommu_group_get_for_dev() function also attaches the
device to its group, so this code doesn't need to be in the
iommu driver.
Further by using this function the driver can make use of
default domains in the future.
Reviewed-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
get_cpu() disables preemption and returns the current CPU number. The
CPU number is only used once while retrieving the address of the local's
CPU deferred_flush pointer.
We can instead use raw_cpu_ptr() while we remain preemptible. The worst
thing that can happen is that flush_unmaps_timeout() is invoked multiple
times: once by taskA after seeing HIGH_WATER_MARK and then preempted to
another CPU and then by taskB which saw HIGH_WATER_MARK on the same CPU
as taskA. It is also likely that ->size got from HIGH_WATER_MARK to 0
right after its read because another CPU invoked flush_unmaps_timeout()
for this CPU.
The access to flush_data is protected by a spinlock so even if we get
migrated to another CPU or preempted - the data structure is protected.
While at it, I marked deferred_flush static since I can't find a
reference to it outside of this file.
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: iommu@lists.linux-foundation.org
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Commit 583248e6620a ("iommu/iova: Disable preemption around use of
this_cpu_ptr()") disables preemption while accessing a per-CPU variable.
This does keep lockdep quiet. However I don't see the point why it is
bad if we get migrated after its access to another CPU.
__iova_rcache_insert() and __iova_rcache_get() immediately locks the
variable after obtaining it - before accessing its members.
_If_ we get migrated away after retrieving the address of cpu_rcache
before taking the lock then the *other* task on the same CPU will
retrieve the same address of cpu_rcache and will spin on the lock.
alloc_iova_fast() disables preemption while invoking
free_cpu_cached_iovas() on each CPU. The function itself uses
per_cpu_ptr() which does not trigger a warning (like this_cpu_ptr()
does). It _could_ make sense to use get_online_cpus() instead but the we
have a hotplug notifier for CPU down (and none for up) so we are good.
Cc: Joerg Roedel <joro@8bytes.org>
Cc: iommu@lists.linux-foundation.org
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Cavium ThunderX2 SMMU doesn't support MSI and also doesn't have unique irq
lines for gerror, eventq and cmdq-sync.
New named irq "combined" is set as a errata workaround, which allows to
share the irq line by register single irq handler for all the interrupts.
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Geetha sowjanya <gakula@caviumnetworks.com>
[will: reworked irq equality checking and added SPI check]
Signed-off-by: Will Deacon <will.deacon@arm.com>
HiSilicon SMMUv3 on Hip06/Hip07 platforms doesn't support CMD_PREFETCH
command. The dt based support for this quirk is already present in the
driver(hisilicon,broken-prefetch-cmd). This adds ACPI support for the
quirk using the IORT smmu model number.
Signed-off-by: shameer <shameerali.kolothum.thodi@huawei.com>
Signed-off-by: hanjun <guohanjun@huawei.com>
[will: rewrote patch]
Signed-off-by: Will Deacon <will.deacon@arm.com>
Cavium ThunderX2 SMMU implementation doesn't support page 1 register space
and PAGE0_REGS_ONLY option is enabled as an errata workaround.
This option when turned on, replaces all page 1 offsets used for
EVTQ_PROD/CONS, PRIQ_PROD/CONS register access with page 0 offsets.
SMMU resource size checks are now based on SMMU option PAGE0_REGS_ONLY,
since resource size can be either 64k/128k.
For this, arm_smmu_device_dt_probe/acpi_probe has been moved before
platform_get_resource call, so that SMMU options are set beforehand.
Signed-off-by: Linu Cherian <linu.cherian@cavium.com>
Signed-off-by: Geetha Sowjanya <geethasowjanya.akula@cavium.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The model number is already defined in acpica and we are actually
waiting for the acpi maintainers to include it:
https://github.com/acpica/acpica/commit/d00a4eb86e64
Adding those temporary definitions until the change makes it into
include/acpi/actbl2.h. Once that is done this patch can be reverted.
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Robert Richter <rrichter@cavium.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
When writing a new table entry, we must ensure that the contents of the
table is made visible to the SMMU page table walker before the updated
table entry itself.
This is currently achieved using wmb(), which expands to an expensive and
unnecessary DSB instruction. Ideally, we'd just use cmpxchg64_release when
writing the table entry, but this doesn't have memory ordering semantics
on !SMP systems.
Instead, use dma_wmb(), which emits DMB OSHST. Strictly speaking, this
does more than we require (since it targets the outer-shareable domain),
but it's likely to be significantly faster than the DSB approach.
Reported-by: Linu Cherian <linu.cherian@cavium.com>
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The LPAE/ARMv8 page table format relies on the ability to read and write
64-bit page table entries in an atomic fashion. With the move to a lockless
implementation, we also need support for cmpxchg64 to resolve races when
installing table entries concurrently.
Unfortunately, not all architectures support cmpxchg64, so the code can
fail to compiler when building for these architectures using COMPILE_TEST.
Rather than disable COMPILE_TEST altogether, instead check that
GENERIC_ATOMIC64 is not selected, which is a reasonable indication that
the architecture has support for 64-bit cmpxchg.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
As for SMMUv2, take advantage of io-pgtable's newfound tolerance for
concurrency. Unfortunately in this case the command queue lock remains a
point of serialisation for the unmap path, but there may be a little
more we can do to ameliorate that in future.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
With the io-pgtable code now robust against (valid) races, we no longer
need to serialise all operations with a lock. This might make broken
callers who issue concurrent operations on overlapping addresses go even
more wrong than before, but hey, they already had little hope of useful
or deterministic results.
We do however still have to keep a lock around to serialise the ATS1*
translation ops, as parallel iova_to_phys() calls could lead to
unpredictable hardware behaviour otherwise.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Mirroring the LPAE implementation, rework the v7s code to be robust
against concurrent operations. The same two potential races exist, and
are solved in the same manner, with the fixed 2-level structure making
life ever so slightly simpler.
What complicates matters compared to LPAE, however, is large page
entries, since we can't update a block of 16 PTEs atomically, nor assume
available software bits to do clever things with. As most users are
never likely to do partial unmaps anyway (due to DMA API rules), it
doesn't seem unreasonable for this case to remain behind a serialising
lock; we just pull said lock down into the bowels of the implementation
so it's well out of the way of the normal call paths.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
For parallel I/O with multiple concurrent threads servicing the same
device (or devices, if several share a domain), serialising page table
updates becomes a massive bottleneck. On reflection, though, we don't
strictly need to do that - for valid IOMMU API usage, there are in fact
only two races that we need to guard against: multiple map requests for
different blocks within the same region, when the intermediate-level
table for that region does not yet exist; and multiple unmaps of
different parts of the same block entry. Both of those are fairly easily
solved by using a cmpxchg to install the new table, such that if we then
find that someone else's table got there first, we can simply free ours
and continue.
Make the requisite changes such that we can withstand being called
without the caller maintaining a lock. In theory, this opens up a few
corners in which wildly misbehaving callers making nonsensical
overlapping requests might lead to crashes instead of just unpredictable
results, but correct code really does not deserve to pay a significant
performance cost for the sake of masking bugs in theoretical broken code.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Once we remove the serialising spinlock, a potential race opens up for
non-coherent IOMMUs whereby a caller of .map() can be sure that cache
maintenance has been performed on their new PTE, but will have no
guarantee that such maintenance for table entries above it has actually
completed (e.g. if another CPU took an interrupt immediately after
writing the table entry, but before initiating the DMA sync).
Handling this race safely will add some potentially non-trivial overhead
to installing a table entry, which we would much rather avoid on
coherent systems where it will be unnecessary, and where we are stirivng
to minimise latency by removing the locking in the first place.
To that end, let's introduce an explicit notion of cache-coherency to
io-pgtable, such that we will be able to avoid penalising IOMMUs which
know enough to know when they are coherent.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Whilst the short-descriptor format's split_blk_unmap implementation has
no need to be recursive, it followed the pattern of the LPAE version
anyway for the sake of consistency. With the latter now reworked for
both efficiency and future scalability improvements, tweak the former
similarly, not least to make it less obtuse.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The current split_blk_unmap implementation suffers from some inscrutable
pointer trickery for creating the tables to replace the block entry, but
more than that it also suffers from hideous inefficiency. For example,
the most pathological case of unmapping a level 3 page from a level 1
block will allocate 513 lower-level tables to remap the entire block at
page granularity, when only 2 are actually needed (the rest can be
covered by level 2 block entries).
Also, we would like to be able to relax the spinlock requirement in
future, for which the roll-back-and-try-again logic for race resolution
would be pretty hideous under the current paradigm.
Both issues can be resolved most neatly by turning things sideways:
instead of repeatedly recursing into __arm_lpae_map() map to build up an
entire new sub-table depth-first, we can directly replace the block
entry with a next-level table of block/page entries, then repeat by
unmapping at the next level if necessary. With a little refactoring of
some helper functions, the code ends up not much bigger than before, but
considerably easier to follow and to adapt in future.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Whilst we don't support the PXN bit at all, so should never encounter a
level 1 section or supersection PTE with it set, it would still be wise
to check both table type bits to resolve any theoretical ambiguity.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
iommu_device_register returns an error code and, although it currently
never fails, we should check its return value anyway.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
[will: adjusted to follow arm-smmu.c]
Signed-off-by: Will Deacon <will.deacon@arm.com>
of_device_ids are not supposed to change at runtime. All functions
working with of_device_ids provided by <linux/of.h> work with const
of_device_ids. So mark the non-const structs as const.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Revision C of IORT now allows us to identify ARM MMU-401 and the Cavium
ThunderX implementation. Wire them up so that we can probe these models
once firmware starts using the new codes in place of generic ones, and
so that the appropriate features and quirks get enabled when we do.
For the sake of backports and mitigating sychronisation problems with
the ACPICA headers, we'll carry a backup copy of the new definitions
locally for the short term to make life simpler.
CC: stable@vger.kernel.org # 4.10
Acked-by: Robert Richter <rrichter@cavium.com>
Tested-by: Robert Richter <rrichter@cavium.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
File size before:
text data bss dec hex filename
6146 56 9 6211 1843 drivers/iommu/io-pgtable-arm-v7s.o
File size After adding 'const':
text data bss dec hex filename
6170 24 9 6203 183b drivers/iommu/io-pgtable-arm-v7s.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Waiting for a CMD_SYNC to be processed involves waiting for the command
queue to drain, which can take an awful lot longer than waiting for a
single entry to become available. Consequently, the common timeout value
of 100us has been observed to be too short on some platforms when a
CMD_SYNC is issued into a queued full of TLBI commands.
This patch resolves the issue by using a different (1s) timeout when
waiting for the CMDQ to drain and using a simple back-off mechanism
when polling the cons pointer in the absence of WFE support.
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
[will: rewrote commit message and cosmetic changes]
Signed-off-by: Will Deacon <will.deacon@arm.com>
To benefit from IOTLB flushes on other CPUs we have to free
the already flushed IOVAs from the ring-buffer before we do
the queue_ring_full() check.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
After we made sure that all IOMMUs have been disabled we
need to make sure that all resources we allocated are
released again.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The function will also be used to free iommu resources when
amd_iommu=off was specified on the kernel command line. So
rename the function to reflect that.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When booting, make sure the IOMMUs are disabled. They could
be previously enabled if we boot into a kexec or kdump
kernel. So make sure they are off.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When booting into a kdump kernel, suppress IO_PAGE_FAULTs by
default for all devices. But allow the faults again when a
domain is assigned to a device.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Add a timer to each dma_ops domain so that we flush unused
IOTLB entries regularily, even if the queues don't get full
all the time.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The counters are increased every time the TLB for a given
domain is flushed. We also store the current value of that
counter into newly added entries of the flush-queue, so that
we can tell whether this entry is already flushed.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The queue flushing is pretty inefficient when it flushes the
queues for all cpus at once. Further it flushes all domains
from all IOMMUs for all CPUs, which is overkill as well.
Rip it out to make room for something more efficient.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Currently if there is no room to add a command to the command buffer, the
driver performs a "completion wait" which only returns when all commands
on the queue have been processed. There is no need to wait for the entire
command queue to be executed before adding the next command.
Update the driver to perform the same udelay() loop that the "completion
wait" performs, but instead re-read the head pointer to determine if
sufficient space is available. The very first time it is found that there
is no space available, the udelay() will be skipped to immediately perform
the opportunistic read of the head pointer. If it is still found that
there is not sufficient space, then the udelay() will be performed.
Signed-off-by: Leo Duran <leo.duran@amd.com>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
As newer, higher speed devices are developed, perf data shows that the
amount of MMIO that is performed when submitting commands to the IOMMU
causes performance issues. Currently, the command submission path reads
the command buffer head and tail pointers and then writes the tail
pointer once the command is ready.
The tail pointer is only ever updated by the driver so it can be tracked
by the driver without having to read it from the hardware.
The head pointer is updated by the hardware, but can be read
opportunistically. Reading the head pointer only when it appears that
there might not be room in the command buffer and then re-checking the
available space reduces the number of times the head pointer has to be
read.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
struct irq_domain_ops is not modified, so it can be made const.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Misbehaving devices can cause an endless chain of
io-page-faults, flooding dmesg and making the system-log
unusable or even prevent the system from booting.
So ratelimit the error messages about io-page-faults on a
per-device basis.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
struct irq_domain_ops is not modified, so it can be made const.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
