| Age | Commit message (Collapse) | Author |
|---|
|
Add bit definitions for Intel PT MSRs to support trace output
directed to the memeory subsystem and holds a count if packet
bytes that have been sent out.
These are required by the upcoming PT support in KVM guests
for MSRs read/write emulation.
Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
intel_pt_validate_hw_cap() validates whether a given PT capability is
supported by the hardware. It checks the PT capability array which
reflects the capabilities of the hardware on which the code is executed.
For setting up PT for KVM guests this is not correct as the capability
array for the guest can be different from the host array.
Provide a new function to check against a given capability array.
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
pt_cap_get() is required by the upcoming PT support in KVM guests.
Export it and move the capabilites enum to a global header.
As a global functions, "pt_*" is already used for ptrace and
other things, so it makes sense to use "intel_pt_*" as a prefix.
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The Intel Processor Trace (PT) MSR bit defines are in a private
header. The upcoming support for PT virtualization requires these defines
to be accessible from KVM code.
Move them to the global MSR header file.
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When KVM has KVM_CAP_ARM_VM_IPA_SIZE we can test with > 40-bit IPAs by
using the 'type' field of KVM_CREATE_VM.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
We need to reset the offset for each mode as it will change
depending on the number of guest physical address bits.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
There's no reason not to always test the topmost physical
addresses, and if the user wants to try lower addresses
then '-p' (used to be '-o before this patch) can be used.
Let's remove the '-t' option and just always do what it did.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
It isn't necessary and can even cause problems when testing high
guest physical addresses. This patch leaves the test memory id-
mapped by default, but when using '-t' the test memory virtual
addresses stay the same even though the physical addresses switch
to the topmost valid addresses.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
[Preserved the iff and a probably intentional weird bracket notation.
Also dropped the style change to make a single-purpose patch. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Update the verbose license text with the matching SPDX
license identifier.
Signed-off-by: Peng Hao <peng.hao2@zte.com.cn>
[Changed deprecated GPL-2.0+ to GPL-2.0-or-later. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Although well-intentioned, keeping the KF() definition as a hint for
handling scattered CPUID features may be counter-productive. Simply
redefining the bit position only works for directly manipulating the
guest's CPUID leafs, e.g. it doesn't make guest_cpuid_has() magically
work. Taking an alternative approach, e.g. ensuring the bit position
is identical between the Linux-defined and hardware-defined features,
may be a simpler and/or more effective method of exposing scattered
features to the guest.
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Let the guest read the IA32_TSC MSR with the generic RDMSR instruction
as well as the specific RDTSC(P) instructions. Note that the hardware
applies the TSC multiplier and offset (when applicable) to the result of
RDMSR(IA32_TSC), just as it does to the result of RDTSC(P).
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Reviewed-by: Marc Orr <marcorr@google.com>
Reviewed-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
According to the SDM, "NMI-window exiting" VM-exits wake a logical
processor from the same inactive states as would an NMI and
"interrupt-window exiting" VM-exits wake a logical processor from the
same inactive states as would an external interrupt. Specifically, they
wake a logical processor from the shutdown state and from the states
entered using the HLT and MWAIT instructions.
Fixes: 6dfacadd5858 ("KVM: nVMX: Add support for activity state HLT")
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Suggested-by: Sean Christopherson <sean.j.christopherson@intel.com>
[Squashed comments of two Jim's patches and used the simplified code
hunk provided by Sean. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Currently, the nested guest's PAUSE intercept intentions are not being
honored. Instead, since the L0 hypervisor's pause_filter_count and
pause_filter_thresh values are still in place, these values are used
instead of those programmed in the VMCB by the L1 hypervisor.
To honor the desired PAUSE intercept support of the L1 hypervisor, the L0
hypervisor must use the PAUSE filtering fields of the L1 hypervisor. This
requires saving and restoring of both the L0 and L1 hypervisor's PAUSE
filtering fields.
Signed-off-by: William Tambe <william.tambe@amd.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Since the offset is added directly to the hva from the
gfn_to_hva_cache, a negative offset could result in an out of bounds
write. The existing BUG_ON only checks for addresses beyond the end of
the gfn_to_hva_cache, not for addresses before the start of the
gfn_to_hva_cache.
Note that all current call sites have non-negative offsets.
Fixes: 4ec6e8636256 ("kvm: Introduce kvm_write_guest_offset_cached()")
Reported-by: Cfir Cohen <cfir@google.com>
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Cfir Cohen <cfir@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Previously, in the case where (gpa + len) wrapped around, the entire
region was not validated, as the comment claimed. It doesn't actually
seem that wraparound should be allowed here at all.
Furthermore, since some callers don't check the return code from this
function, it seems prudent to clear ghc->memslot in the event of an
error.
Fixes: 8f964525a121f ("KVM: Allow cross page reads and writes from cached translations.")
Reported-by: Cfir Cohen <cfir@google.com>
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Cfir Cohen <cfir@google.com>
Reviewed-by: Marc Orr <marcorr@google.com>
Cc: Andrew Honig <ahonig@google.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
Remove duplicated include.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
When we get a report like
==== Test Assertion Failure ====
x86_64/state_test.c:157: run->exit_reason == KVM_EXIT_IO
pid=955 tid=955 - Success
1 0x0000000000401350: main at state_test.c:154
2 0x00007fc31c9e9412: ?? ??:0
3 0x000000000040159d: _start at ??:?
Unexpected exit reason: 8 (SHUTDOWN),
it is not obvious which particular stage failed. Add the info.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
AMD doesn't seem to implement MSR_IA32_MCG_EXT_CTL and svm code in kvm
knows nothing about it, however, this MSR is among emulated_msrs and
thus returned with KVM_GET_MSR_INDEX_LIST. The consequent KVM_GET_MSRS,
of course, fails.
Report the MSR as unsupported to not confuse userspace.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
|
|
The memory allocation in b666a4b69739 ("kvm: x86: Dynamically allocate
guest_fpu", 2018-11-06) is wrong, there are other members in struct fpu
before the fpregs_state union and the patch should be doing something
similar to the code in fpu__init_task_struct_size. It's enough to run
a guest and then rmmod kvm to see slub errors which are actually caused
by memory corruption.
For now let's revert it to sizeof(struct fpu), which is conservative.
I have plans to move fsave/fxsave/xsave directly in KVM, without using
the kernel FPU helpers, and once it's done, the size of the object in
the cache will be something like kvm_xstate_size.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
dmesg reports that "Your touchpad (PNP: SYN3052 SYN0100 SYN0002 PNP0f13)
says it can support a different bus."
I've tested the offered psmouse.synaptics_intertouch=1 with 4.18.x and
4.19.x and it seems to work well. No problems seen with suspend/resume.
Also, it appears that RMI/SMBus mode is actually required for 3-4 finger
multitouch gestures to work -- otherwise they are not reported at all.
Information from dmesg in both modes:
psmouse serio3: synaptics: Touchpad model: 1, fw: 8.2, id: 0x1e2b1,
caps: 0xf00123/0x840300/0x2e800/0x0, board id: 3139, fw id: 2000742
psmouse serio3: synaptics: Trying to set up SMBus access
rmi4_smbus 6-002c: registering SMbus-connected sensor
rmi4_f01 rmi4-00.fn01: found RMI device,
manufacturer: Synaptics, product: TM3139-001, fw id: 2000742
Signed-off-by: Mantas Mikulėnas <grawity@gmail.com>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
* pm-devfreq:
PM / devfreq: add devfreq_suspend/resume() functions
PM / devfreq: add support for suspend/resume of a devfreq device
PM / devfreq: refactor set_target frequency function
* pm-avs:
PM / AVS: SmartReflex: Switch to SPDX Licence ID
PM / AVS: SmartReflex: NULL check before some freeing functions is not needed
PM / AVS: SmartReflex: remove unused function
* pm-tools:
tools/power/x86/intel_pstate_tracer: Fix non root execution for post processing a trace file
tools/power turbostat: consolidate duplicate model numbers
tools/power turbostat: fix goldmont C-state limit decoding
cpupower : Auto-completion for cpupower tool
tools/power turbostat: reduce debug output
tools/power turbosat: fix AMD APIC-id output
|
|
* pm-core:
PM-runtime: Switch autosuspend over to using hrtimers
* pm-qos:
PM / QoS: Change to use DEFINE_SHOW_ATTRIBUTE macro
* pm-domains:
PM / Domains: remove define_genpd_open_function() and define_genpd_debugfs_fops()
* pm-sleep:
PM / sleep: convert to DEFINE_SHOW_ATTRIBUTE
|
|
* pm-opp:
PM / Domains: Propagate performance state updates
PM / Domains: Factorize dev_pm_genpd_set_performance_state()
PM / Domains: Save OPP table pointer in genpd
OPP: Don't return 0 on error from of_get_required_opp_performance_state()
OPP: Add dev_pm_opp_xlate_performance_state() helper
OPP: Improve _find_table_of_opp_np()
PM / Domains: Make genpd performance states orthogonal to the idlestates
OPP: Fix missing debugfs supply directory for OPPs
OPP: Use opp_table->regulators to verify no regulator case
OPP: Remove of_dev_pm_opp_find_required_opp()
OPP: Rename and relocate of_genpd_opp_to_performance_state()
OPP: Configure all required OPPs
OPP: Add dev_pm_opp_{set|put}_genpd_virt_dev() helper
PM / Domains: Add genpd_opp_to_performance_state()
OPP: Populate OPPs from "required-opps" property
OPP: Populate required opp tables from "required-opps" property
OPP: Separate out custom OPP handler specific code
OPP: Identify and mark genpd OPP tables
PM / Domains: Rename genpd virtual devices as virt_dev
|
|
* pm-cpuidle:
cpuidle: Add 'above' and 'below' idle state metrics
cpuidle: big.LITTLE: fix refcount leak
cpuidle: Add cpuidle.governor= command line parameter
cpuidle: poll_state: Disregard disable idle states
Documentation: admin-guide: PM: Add cpuidle document
* pm-cpufreq:
cpufreq: qcom-hw: Add support for QCOM cpufreq HW driver
dt-bindings: cpufreq: Introduce QCOM cpufreq firmware bindings
cpufreq: nforce2: Remove meaningless return
cpufreq: ia64: Remove unused header files
cpufreq: imx6q: save one condition block for normal case of nvmem read
cpufreq: imx6q: remove unused code
cpufreq: pmac64: add of_node_put()
cpufreq: powernv: add of_node_put()
Documentation: intel_pstate: Clarify coordination of P-State limits
cpufreq: intel_pstate: Force HWP min perf before offline
cpufreq: s3c24xx: Change to use DEFINE_SHOW_ATTRIBUTE macro
* pm-cpufreq-sched:
sched/cpufreq: Add the SPDX tags
|
|
* acpi-pci:
ACPI: Make PCI slot detection driver depend on PCI
ACPI/IORT: Stub out ACS functions when CONFIG_PCI is not set
arm64: select ACPI PCI code only when both features are enabled
PCI/ACPI: Allow ACPI to be built without CONFIG_PCI set
ACPICA: Remove PCI bits from ACPICA when CONFIG_PCI is unset
ACPI: Allow CONFIG_PCI to be unset for reboot
ACPI: Move PCI reset to a separate function
|
|
* acpi-tables:
ACPI / tables: Add an ifdef around amlcode and dsdt_amlcode
ACPI / tables: add DSDT AmlCode new declaration name support
ACPI: SPCR: Consider baud rate 0 as preconfigured state
* acpi-soc:
ACPI / LPSS: Ignore acpi_device_fix_up_power() return value
ACPI / APD: Add clock frequency for Hisilicon Hip08 SPI controller
* acpi-apei:
ACPI/APEI: Clear GHES block_status before panic()
ACPI, APEI, EINJ: Change to use DEFINE_SHOW_ATTRIBUTE macro
* acpi-misc:
ACPI: fix acpi_find_child_device() invocation in acpi_preset_companion()
|
|
* acpica:
ACPICA: Update version to 20181213
ACPICA: change coding style to match ACPICA, no functional change
ACPICA: Debug output: Add option to display method/object evaluation
ACPICA: disassembler: disassemble OEMx tables as AML
ACPICA: Add "Windows 2018.2" string in the _OSI support
ACPICA: Expressions in package elements are not supported
ACPICA: Update buffer-to-string conversions
ACPICA: add comments, no functional change
ACPICA: Remove defines that use deprecated flag
ACPICA: Add "Windows 2018" string in the _OSI support
ACPICA: Update version to 20181031
ACPICA: iASL: Enhance error detection
ACPICA: iASL: adding definition and disassembly for TPM2 revision 3
ACPICA: Use %d for signed int print formatting instead of %u
ACPICA: Debugger: refactor to fix unused variable warning
|
|
The current implementation of elan_i2c is known to not support those
2 laptops.
A proper fix is to tweak both elantech and elan_i2c to transmit the
correct information from PS/2, which would make a bad candidate for
stable.
So to give us some time for fixing the root of the problem, disable
elan_i2c for the devices we know are not behaving properly.
Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1803600
Link: https://bugs.archlinux.org/task/59714
Fixes: df077237cf55 Input: elantech - detect new ICs and setup Host Notify for them
Cc: stable@vger.kernel.org # v4.18+
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Acked-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
The gpio IP on Armada 370 at offset 0x18180 has neither a clk nor pwm
registers. So there is no need for a clk as the pwm isn't used anyhow.
So only check for the clk in the presence of the pwm registers. This fixes
a failure to probe the gpio driver for the above mentioned gpio device.
Fixes: 757642f9a584 ("gpio: mvebu: Add limited PWM support")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Gregory CLEMENT <gregory.clement@bootlin.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
|
|
spi_read() and spi_write() require DMA-safe memory. When
CONFIG_VMAP_STACK is selected, those functions cannot be used
with buffers on stack.
This patch replaces calls to spi_read() and spi_write() by
spi_write_then_read() which doesn't require DMA-safe buffers.
Fixes: 0c36ec314735 ("gpio: gpio driver for max7301 SPI GPIO expander")
Cc: <stable@vger.kernel.org>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
|
|
Commit ec0daae685b2 ("gpio: omap: Add level wakeup handling for omap4
based SoCs") attempted to fix omap4 GPIO wakeup handling as it was
blocking deeper SoC idle states. However this caused a regression for
GPIOs during runtime having over second long latencies for Ethernet
GPIO interrupt as reportedy by Russell King <rmk+kernel@armlinux.org.uk>.
Let's fix this issue by doing a partial revert of the breaking commit.
We still want to keep the quirk handling around as it is also used for
OMAP_GPIO_QUIRK_IDLE_REMOVE_TRIGGER.
The real fix for omap4 GPIO wakeup handling involves fixes for
omap_set_gpio_trigger() and omap_gpio_unmask_irq() and will be posted
separately. And we must keep the wakeup bit enabled during runtime
because of module doing clock autogating with autoidle configured.
Reported-by: Russell King <rmk+kernel@armlinux.org.uk>
Fixes: ec0daae685b2 ("gpio: omap: Add level wakeup handling for omap4
based SoCs")
Cc: Aaro Koskinen <aaro.koskinen@iki.fi>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Cc: Keerthy <j-keerthy@ti.com>
Cc: Ladislav Michl <ladis@linux-mips.org>
Cc: Russell King <rmk+kernel@armlinux.org.uk>
Cc: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
|
|
Add ethtool private flag 'xdp_tx_mpwqe' to control the feature
from userspace.
Feature is set ON by default, if supported.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Add support for the HW feature of multi-packet WQE in XDP
xmit flow.
The conventional TX descriptor (WQE, Work Queue Element) serves
a single packet. Our HW has support for multi-packet WQE (MPWQE)
in which a single descriptor serves multiple TX packets.
This reduces both the PCI overhead and the CPU cycles wasted on
writing them.
In this patch we add support for the HW feature, which is supported
starting from ConnectX-5.
Performance:
Tested packet rate for UDP 64Byte multi-stream over ConnectX-5 NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
XDP_TX:
We see a huge gain on single port ConnectX-5, and reach the 100 Mpps
milestone.
* Single-port HCA:
Before: 70 Mpps
After: 100 Mpps (+42.8%)
* Dual-port HCA:
Before: 51.7 Mpps
After: 57.3 Mpps (+10.8%)
* In both cases we tested traffic on one port and for now On Dual-port HCAs
we see only small gain, we are working to overcome this bottleneck, but
for the moment only with experimental firmware on dual port HCAs we can
reach the wanted numbers as seen on Single-port HCAs.
XDP_REDIRECT:
Redirect from (A) ConnectX-5 to (B) ConnectX-5.
Due to a setup limitation, (A) and (B) are on different NUMA nodes,
so absolute performance numbers are not optimal.
Note:
Below is the transmit rate of (B), not the redirect rate of (A)
which is in some cases higher.
* (B) is single-port:
Before: 77 Mpps
After: 90 Mpps (+16.8%)
* (B) is dual-port:
Before: 61 Mpps
After: 72 Mpps (+18%)
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Each xdp_wqe_info instance describes the number of data-segments
and WQEBBs of the WQE.
This is useful for a downstream patch that adds support for
Multi-Packet TX WQE feature.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
This provides infrastructure to have multiple xdp_info instances
for the same consumer index.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Instead of calculating the control segment to be used upon an
XDP xmit doorbell, save it in SQ structure.
Nullify when no pending doorbell.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Do not ignore the CQE opcode.
This helps expose issues and debug them.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Do not maintain an SQ state bit to indicate whether an
XDP SQ serves redirect operations.
Instead, rely on the fact that such an XDP SQ doesn't reside
in an RQ instance, while the others do.
This info is not known to the XDP SQ functions themselves,
and they rely on their callers to distinguish between the cases.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
program check
At the end of the RQ polling loop, some XDP-related operations
might be required. Before checking them one by one, check if
an XDP program is even loaded.
Combine all the checks and operations in a single function in xdp files.
This saves unnecessary checks for non-XDP flows.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
The opcode indicates about the error reason.
Printing it helps in debug.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
POWER9 Witherspoon machines come with 4 or 6 V100 GPUs which are not
pluggable PCIe devices but still have PCIe links which are used
for config space and MMIO. In addition to that the GPUs have 6 NVLinks
which are connected to other GPUs and the POWER9 CPU. POWER9 chips
have a special unit on a die called an NPU which is an NVLink2 host bus
adapter with p2p connections to 2 to 3 GPUs, 3 or 2 NVLinks to each.
These systems also support ATS (address translation services) which is
a part of the NVLink2 protocol. Such GPUs also share on-board RAM
(16GB or 32GB) to the system via the same NVLink2 so a CPU has
cache-coherent access to a GPU RAM.
This exports GPU RAM to the userspace as a new VFIO device region. This
preregisters the new memory as device memory as it might be used for DMA.
This inserts pfns from the fault handler as the GPU memory is not onlined
until the vendor driver is loaded and trained the NVLinks so doing this
earlier causes low level errors which we fence in the firmware so
it does not hurt the host system but still better be avoided; for the same
reason this does not map GPU RAM into the host kernel (usual thing for
emulated access otherwise).
This exports an ATSD (Address Translation Shootdown) register of NPU which
allows TLB invalidations inside GPU for an operating system. The register
conveniently occupies a single 64k page. It is also presented to
the userspace as a new VFIO device region. One NPU has 8 ATSD registers,
each of them can be used for TLB invalidation in a GPU linked to this NPU.
This allocates one ATSD register per an NVLink bridge allowing passing
up to 6 registers. Due to the host firmware bug (just recently fixed),
only 1 ATSD register per NPU was actually advertised to the host system
so this passes that alone register via the first NVLink bridge device in
the group which is still enough as QEMU collects them all back and
presents to the guest via vPHB to mimic the emulated NPU PHB on the host.
In order to provide the userspace with the information about GPU-to-NVLink
connections, this exports an additional capability called "tgt"
(which is an abbreviated host system bus address). The "tgt" property
tells the GPU its own system address and allows the guest driver to
conglomerate the routing information so each GPU knows how to get directly
to the other GPUs.
For ATS to work, the nest MMU (an NVIDIA block in a P9 CPU) needs to
know LPID (a logical partition ID or a KVM guest hardware ID in other
words) and PID (a memory context ID of a userspace process, not to be
confused with a linux pid). This assigns a GPU to LPID in the NPU and
this is why this adds a listener for KVM on an IOMMU group. A PID comes
via NVLink from a GPU and NPU uses a PID wildcard to pass it through.
This requires coherent memory and ATSD to be available on the host as
the GPU vendor only supports configurations with both features enabled
and other configurations are known not to work. Because of this and
because of the ways the features are advertised to the host system
(which is a device tree with very platform specific properties),
this requires enabled POWERNV platform.
The V100 GPUs do not advertise any of these capabilities via the config
space and there are more than just one device ID so this relies on
the platform to tell whether these GPUs have special abilities such as
NVLinks.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
|
VFIO regions already support region capabilities with a limited set of
fields. However the subdriver might have to report to the userspace
additional bits.
This adds an add_capability() hook to vfio_pci_regops.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
|
So far we only allowed mapping of MMIO BARs to the userspace. However
there are GPUs with on-board coherent RAM accessible via side
channels which we also want to map to the userspace. The first client
for this is NVIDIA V100 GPU with NVLink2 direct links to a POWER9
NPU-enabled CPU; such GPUs have 16GB RAM which is coherently mapped
to the system address space, we are going to export these as an extra
PCI region.
We already support extra PCI regions and this adds support for mapping
them to the userspace.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
|
When a page fault happens in a GPU, the GPU signals the OS and the GPU
driver calls the fault handler which populated a page table; this allows
the GPU to complete an ATS request.
On the bare metal get_user_pages() is enough as it adds a pte to
the kernel page table but under KVM the partition scope tree does not get
updated so ATS will still fail.
This reads a byte from an effective address which causes HV storage
interrupt and KVM updates the partition scope tree.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
|
A broken device tree might contain more than 8 values and introduce hard
to debug memory corruption bug. This adds the boundary check.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
|
In order to make ATS work and translate addresses for arbitrary
LPID and PID, we need to program an NPU with LPID and allow PID wildcard
matching with a specific MSR mask.
This implements a helper to assign a GPU to LPAR and program the NPU
with a wildcard for PID and a helper to do clean-up. The helper takes
MSR (only DR/HV/PR/SF bits are allowed) to program them into NPU2 for
ATS checkout requests support.
This exports pnv_npu2_unmap_lpar_dev() as following patches will use it
from the VFIO driver.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
|
