We extend the KVM ISA extension ONE_REG interface to allow KVM
user space to detect and enable Zbb extension for Guest/VM.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
We add ONE_REG interface to enable/disable SBI extensions (just
like the ONE_REG interface for ISA extensions). This allows KVM
user-space to decide the set of SBI extension enabled for a Guest
and by default all SBI extensions are enabled.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
While alphabetized lists tend to become unalphabetized almost
as quickly as they get fixed up, it is preferred to keep select
lists in Kconfigs in order. Let's fix KVM's up.
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Anup Patel <anup@brainfault.org>
Read mmu_invalidate_seq before dropping the mmap_lock so that KVM can
detect if the results of vma_lookup() (e.g. vma_shift) become stale
before it acquires kvm->mmu_lock. This fixes a theoretical bug where a
VMA could be changed by userspace after vma_lookup() and before KVM
reads the mmu_invalidate_seq, causing KVM to install page table entries
based on a (possibly) no-longer-valid vma_shift.
Re-order the MMU cache top-up to earlier in user_mem_abort() so that it
is not done after KVM has read mmu_invalidate_seq (i.e. so as to avoid
inducing spurious fault retries).
It's unlikely that any sane userspace currently modifies VMAs in such a
way as to trigger this race. And even with directed testing I was unable
to reproduce it. But a sufficiently motivated host userspace might be
able to exploit this race.
Note KVM/ARM had the same bug and was fixed in a separate, near
identical patch (see Link).
Link: https://lore.kernel.org/kvm/20230313235454.2964067-1-dmatlack@google.com/
Fixes: 9955371cc0 ("RISC-V: KVM: Implement MMU notifiers")
Cc: stable@vger.kernel.org
Signed-off-by: David Matlack <dmatlack@google.com>
Tested-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Anup Patel <anup@brainfault.org>
Add missing clockticks and cas_count_* events for Intel SapphireRapids IMC
PMU. These events are useful to measure memory bandwidth.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Link: https://lore.kernel.org/r/20230419214241.2310385-1-eranian@google.com
Have local_clock() return sched_clock() if sched_clock_init() has not
yet run. sched_clock_cpu() has this check but it was not included in the
new noinstr implementation of local_clock().
The effect can be seen on x86 with CONFIG_PRINTK_TIME enabled, for
instance. scd->clock quickly reaches the value of TICK_NSEC and that
value is returned until sched_clock_init() runs.
dmesg without this patch:
[ 0.000000] kvm-clock: ...
[ 0.000002] kvm-clock: ...
[ 0.000672] clocksource: ...
[ 0.001000] tsc: ...
[ 0.001000] e820: ...
[ 0.001000] e820: ...
...
[ 0.001000] ..TIMER: ...
[ 0.001000] clocksource: ...
[ 0.378956] Calibrating delay loop ...
[ 0.379955] pid_max: ...
dmesg with this patch:
[ 0.000000] kvm-clock: ...
[ 0.000001] kvm-clock: ...
[ 0.000675] clocksource: ...
[ 0.002685] tsc: ...
[ 0.003331] e820: ...
[ 0.004190] e820: ...
...
[ 0.421939] ..TIMER: ...
[ 0.422842] clocksource: ...
[ 0.424582] Calibrating delay loop ...
[ 0.425580] pid_max: ...
Fixes: 776f22913b ("sched/clock: Make local_clock() noinstr")
Signed-off-by: Aaron Thompson <dev@aaront.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230413175012.2201-1-dev@aaront.org
Commit 95158a89dd ("sched,rt: Use the full cpumask for balancing")
allows find_lock_lowest_rq() to pick a task with migration disabled.
The purpose of the commit is to push the current running task on the
CPU that has the migrate_disable() task away.
However, there is a race which allows a migrate_disable() task to be
migrated. Consider:
CPU0 CPU1
push_rt_task
check is_migration_disabled(next_task)
task not running and
migration_disabled == 0
find_lock_lowest_rq(next_task, rq);
_double_lock_balance(this_rq, busiest);
raw_spin_rq_unlock(this_rq);
double_rq_lock(this_rq, busiest);
<<wait for busiest rq>>
<wakeup>
task become running
migrate_disable();
<context out>
deactivate_task(rq, next_task, 0);
set_task_cpu(next_task, lowest_rq->cpu);
WARN_ON_ONCE(is_migration_disabled(p));
Fixes: 95158a89dd ("sched,rt: Use the full cpumask for balancing")
Signed-off-by: Schspa Shi <schspa@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Tested-by: Dwaine Gonyier <dgonyier@redhat.com>
Introduce per-mm/cpu current concurrency id (mm_cid) to fix a PostgreSQL
sysbench regression reported by Aaron Lu.
Keep track of the currently allocated mm_cid for each mm/cpu rather than
freeing them immediately on context switch. This eliminates most atomic
operations when context switching back and forth between threads
belonging to different memory spaces in multi-threaded scenarios (many
processes, each with many threads). The per-mm/per-cpu mm_cid values are
serialized by their respective runqueue locks.
Thread migration is handled by introducing invocation to
sched_mm_cid_migrate_to() (with destination runqueue lock held) in
activate_task() for migrating tasks. If the destination cpu's mm_cid is
unset, and if the source runqueue is not actively using its mm_cid, then
the source cpu's mm_cid is moved to the destination cpu on migration.
Introduce a task-work executed periodically, similarly to NUMA work,
which delays reclaim of cid values when they are unused for a period of
time.
Keep track of the allocation time for each per-cpu cid, and let the task
work clear them when they are observed to be older than
SCHED_MM_CID_PERIOD_NS and unused. This task work also clears all
mm_cids which are greater or equal to the Hamming weight of the mm
cidmask to keep concurrency ids compact.
Because we want to ensure the mm_cid converges towards the smaller
values as migrations happen, the prior optimization that was done when
context switching between threads belonging to the same mm is removed,
because it could delay the lazy release of the destination runqueue
mm_cid after it has been replaced by a migration. Removing this prior
optimization is not an issue performance-wise because the introduced
per-mm/per-cpu mm_cid tracking also covers this more specific case.
Fixes: af7f588d8f ("sched: Introduce per-memory-map concurrency ID")
Reported-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Aaron Lu <aaron.lu@intel.com>
Link: https://lore.kernel.org/lkml/20230327080502.GA570847@ziqianlu-desk2/
Packet sockets, like tap, can be used as the backend for kernel vhost.
In packet sockets, virtio net header size is currently hardcoded to be
the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
always the case: some virtio features, such as mrg_rxbuf, need virtio
net header to be 12-byte long.
Mergeable buffers, as a virtio feature, is worthy of supporting: packets
that are larger than one-mbuf size will be dropped in vhost worker's
handle_rx if mrg_rxbuf feature is not used, but large packets
cannot be avoided and increasing mbuf's size is not economical.
With this virtio feature enabled by virtio-user, packet sockets with
hardcoded 10-byte virtio net header will parse mac head incorrectly in
packet_snd by taking the last two bytes of virtio net header as part of
mac header.
This incorrect mac header parsing will cause packet to be dropped due to
invalid ether head checking in later under-layer device packet receiving.
By adding extra field vnet_hdr_sz with utilizing holes in struct
packet_sock to record currently used virtio net header size and supporting
extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
sockets can know the exact length of virtio net header that virtio user
gives.
In packet_snd, tpacket_snd and packet_recvmsg, instead of using
hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
corresponding packet_sock, and parse mac header correctly based on this
information to avoid the packets being mistakenly dropped.
Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Leon Romanovsky says:
====================
Fixes to mlx5 IPsec implementation
This small patchset includes various fixes and one refactoring patch
which I collected for the features sent in this cycle, with one exception -
first patch.
First patch fixes code which was introduced in previous cycle, however I
was able to trigger FW error only in custom debug code, so don't see a
need to send it to net-rc.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
ARP discovery code has same logic for RX and TX flows, but with
different source and destination fields. Instead of duplicating
same code in mlx5e_ipsec_init_macs, let's refactor.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix size argument in memcmp to compare whole IPv6 address.
Fixes: b3beba1fb4 ("net/mlx5e: Allow policies with reqid 0, to support IKE policy holes")
Reviewed-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Addition of new err_xfrm label caused to error messages be overwritten.
Fix it by using proper NL_SET_ERR_MSG_WEAK_MOD macro together with change
in a default message.
Fixes: aa8bd0c951 ("net/mlx5e: Support IPsec acquire default SA")
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When trying to set IPsec policy block action the following error is
generated:
mlx5_cmd_out_err:803:(pid 3426): SET_FLOW_TABLE_ENTRY(0x936) op_mod(0x0) failed,
status bad parameter(0x3), syndrome (0x8708c3), err(-22)
This error means that drop action is not allowed when modify action is
set, so update the code to skip modify header for XFRM_POLICY_BLOCK action.
Fixes: 6721239672 ("net/mlx5e: Skip IPsec encryption for TX path without matching policy")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The system hang because of dsa_tag_8021q_port_setup()->
stmmac_vlan_rx_add_vid().
I found in stmmac_drv_probe() that cailing pm_runtime_put()
disabled the clock.
First, when the kernel is compiled with CONFIG_PM=y,The stmmac's
resume/suspend is active.
Secondly,stmmac as DSA master,the dsa_tag_8021q_port_setup() function
will callback stmmac_vlan_rx_add_vid when DSA dirver starts. However,
The system is hanged for the stmmac_vlan_rx_add_vid() accesses its
registers after stmmac's clock is closed.
I would suggest adding the pm_runtime_resume_and_get() to the
stmmac_vlan_rx_add_vid().This guarantees that resuming clock output
while in use.
Fixes: b3dcb31277 ("net: stmmac: correct clocks enabled in stmmac_vlan_rx_kill_vid()")
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Yan Wang <rk.code@outlook.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using the bitmap API is less verbose than hand writing it.
It also improves the semantic.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Link: https://lore.kernel.org/r/a51e9f32c19a007f4922943282cb12c89064440d.1681671848.git.christophe.jaillet@wanadoo.fr
Signed-off-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
The auto-silencer supports two modes: "thresholded" to fill up "just
enough", and "top-up" to fill up "as much as possible". The two modes
used rather distinct code paths, which this patch unifies. The only
remaining distinction is how much we actually want to fill.
This fixes a bug in thresholded mode, where we failed to use new_hw_ptr,
resulting in under-fill.
Top-up mode is now more well-behaved and much easier to understand in
corner cases.
This also updates comments in the proximity of silencing-related data
structures.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230420113324.877164-1-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
As we are revamping the way the pgtable walker evaluates some of the
data, make it clear that we rely on somew of the fields to be constant
across the lifetime of a walk.
For this, flag the start, end and phys fields of the walk data as
'const', which will generate an error if we were to accidentally
update these fields again.
Reviewed-by: Oliver Upton <oliver.upton@linux.dev>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Similar to the recently fixed stage-2 walker, the hyp map walker
increments the PA and VA of a walk separately. Unlike stage-2, there is
no bug here as the map walker has exclusive access to the stage-1 page
tables.
Nonetheless, in the interest of continuity throughout the page table
code, tweak the hyp map walker to avoid incrementing the PA and instead
use the VA as the authoritative source of how far along a table walk has
gotten. Calculate the PA to use for a leaf PTE by adding the offset of
the VA from the start of the walk to the starting PA.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230421071606.1603916-3-oliver.upton@linux.dev
Until now, the page table walker counted increments to the PA and IPA
of a walk in two separate places. While the PA is incremented as soon as
a leaf PTE is installed in stage2_map_walker_try_leaf(), the IPA is
actually bumped in the generic table walker context. Critically,
__kvm_pgtable_visit() rereads the PTE after the LEAF callback returns
to work out if a table or leaf was installed, and only bumps the IPA for
a leaf PTE.
This arrangement worked fine when we handled faults behind the write lock,
as the walker had exclusive access to the stage-2 page tables. However,
commit 1577cb5823 ("KVM: arm64: Handle stage-2 faults in parallel")
started handling all stage-2 faults behind the read lock, opening up a
race where a walker could increment the PA but not the IPA of a walk.
Nothing good ensues, as the walker starts mapping with the incorrect
IPA -> PA relationship.
For example, assume that two vCPUs took a data abort on the same IPA.
One observes that dirty logging is disabled, and the other observed that
it is enabled:
vCPU attempting PMD mapping vCPU attempting PTE mapping
====================================== =====================================
/* install PMD */
stage2_make_pte(ctx, leaf);
data->phys += granule;
/* replace PMD with a table */
stage2_try_break_pte(ctx, data->mmu);
stage2_make_pte(ctx, table);
/* table is observed */
ctx.old = READ_ONCE(*ptep);
table = kvm_pte_table(ctx.old, level);
/*
* map walk continues w/o incrementing
* IPA.
*/
__kvm_pgtable_walk(..., level + 1);
Bring an end to the whole mess by using the IPA as the single source of
truth for how far along a walk has gotten. Work out the correct PA to
map by calculating the IPA offset from the beginning of the walk and add
that to the starting physical address.
Cc: stable@vger.kernel.org
Fixes: 1577cb5823 ("KVM: arm64: Handle stage-2 faults in parallel")
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230421071606.1603916-2-oliver.upton@linux.dev
In the past setting the pin direction called pinctrl_gpio_direction()
which uses a mutex to serialize this. That was changed to set the
direction directly in the pin controller driver, but that lost the
serialization mechanism. Since the direction of multiple pins are in
the same register you can have a race condition, something that was
in fact observed with the cec-gpio driver.
Add a new spinlock to serialize writing to the FSEL registers.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Fixes: 1a4541b68e ("pinctrl-bcm2835: don't call pinctrl_gpio_direction()")
Link: https://lore.kernel.org/r/4302b66b-ca20-0f19-d2aa-ee8661118863@xs4all.nl
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The enum value "atest_usb2" appears twice. Remove the duplicate. The
meta-schema normally catches these, but schemas under "$defs" was not
getting checked. A fix for that is pending.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230418150613.1528233-1-robh@kernel.org
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
* kvm-arm64/spec-ptw:
: .
: On taking an exception from EL1&0 to EL2(&0), the page table walker is
: allowed to carry on with speculative walks started from EL1&0 while
: running at EL2 (see R_LFHQG). Given that the PTW may be actively using
: the EL1&0 system registers, the only safe way to deal with it is to
: issue a DSB before changing any of it.
:
: We already did the right thing for SPE and TRBE, but ignored the PTW
: for unknown reasons (probably because the architecture wasn't crystal
: clear at the time).
:
: This requires a bit of surgery in the nvhe code, though most of these
: patches are comments so that my future self can understand the purpose
: of these barriers. The VHE code is largely unaffected, thanks to the
: DSB in the context switch.
: .
KVM: arm64: vhe: Drop extra isb() on guest exit
KVM: arm64: vhe: Synchronise with page table walker on MMU update
KVM: arm64: pkvm: Document the side effects of kvm_flush_dcache_to_poc()
KVM: arm64: nvhe: Synchronise with page table walker on TLBI
KVM: arm64: nvhe: Synchronise with page table walker on vcpu run
Signed-off-by: Marc Zyngier <maz@kernel.org>
* kvm-arm64/smccc-filtering:
: .
: SMCCC call filtering and forwarding to userspace, courtesy of
: Oliver Upton. From the cover letter:
:
: "The Arm SMCCC is rather prescriptive in regards to the allocation of
: SMCCC function ID ranges. Many of the hypercall ranges have an
: associated specification from Arm (FF-A, PSCI, SDEI, etc.) with some
: room for vendor-specific implementations.
:
: The ever-expanding SMCCC surface leaves a lot of work within KVM for
: providing new features. Furthermore, KVM implements its own
: vendor-specific ABI, with little room for other implementations (like
: Hyper-V, for example). Rather than cramming it all into the kernel we
: should provide a way for userspace to handle hypercalls."
: .
KVM: selftests: Fix spelling mistake "KVM_HYPERCAL_EXIT_SMC" -> "KVM_HYPERCALL_EXIT_SMC"
KVM: arm64: Test that SMC64 arch calls are reserved
KVM: arm64: Prevent userspace from handling SMC64 arch range
KVM: arm64: Expose SMC/HVC width to userspace
KVM: selftests: Add test for SMCCC filter
KVM: selftests: Add a helper for SMCCC calls with SMC instruction
KVM: arm64: Let errors from SMCCC emulation to reach userspace
KVM: arm64: Return NOT_SUPPORTED to guest for unknown PSCI version
KVM: arm64: Introduce support for userspace SMCCC filtering
KVM: arm64: Add support for KVM_EXIT_HYPERCALL
KVM: arm64: Use a maple tree to represent the SMCCC filter
KVM: arm64: Refactor hvc filtering to support different actions
KVM: arm64: Start handling SMCs from EL1
KVM: arm64: Rename SMC/HVC call handler to reflect reality
KVM: arm64: Add vm fd device attribute accessors
KVM: arm64: Add a helper to check if a VM has ran once
KVM: x86: Redefine 'longmode' as a flag for KVM_EXIT_HYPERCALL
Signed-off-by: Marc Zyngier <maz@kernel.org>
* kvm-arm64/timer-vm-offsets: (21 commits)
: .
: This series aims at satisfying multiple goals:
:
: - allow a VMM to atomically restore a timer offset for a whole VM
: instead of updating the offset each time a vcpu get its counter
: written
:
: - allow a VMM to save/restore the physical timer context, something
: that we cannot do at the moment due to the lack of offsetting
:
: - provide a framework that is suitable for NV support, where we get
: both global and per timer, per vcpu offsetting, and manage
: interrupts in a less braindead way.
:
: Conflict resolution involves using the new per-vcpu config lock instead
: of the home-grown timer lock.
: .
KVM: arm64: Handle 32bit CNTPCTSS traps
KVM: arm64: selftests: Augment existing timer test to handle variable offset
KVM: arm64: selftests: Deal with spurious timer interrupts
KVM: arm64: selftests: Add physical timer registers to the sysreg list
KVM: arm64: nv: timers: Support hyp timer emulation
KVM: arm64: nv: timers: Add a per-timer, per-vcpu offset
KVM: arm64: Document KVM_ARM_SET_CNT_OFFSETS and co
KVM: arm64: timers: Abstract the number of valid timers per vcpu
KVM: arm64: timers: Fast-track CNTPCT_EL0 trap handling
KVM: arm64: Elide kern_hyp_va() in VHE-specific parts of the hypervisor
KVM: arm64: timers: Move the timer IRQs into arch_timer_vm_data
KVM: arm64: timers: Abstract per-timer IRQ access
KVM: arm64: timers: Rationalise per-vcpu timer init
KVM: arm64: timers: Allow save/restoring of the physical timer
KVM: arm64: timers: Allow userspace to set the global counter offset
KVM: arm64: Expose {un,}lock_all_vcpus() to the rest of KVM
KVM: arm64: timers: Allow physical offset without CNTPOFF_EL2
KVM: arm64: timers: Use CNTPOFF_EL2 to offset the physical timer
arm64: Add HAS_ECV_CNTPOFF capability
arm64: Add CNTPOFF_EL2 register definition
...
Signed-off-by: Marc Zyngier <maz@kernel.org>
* kvm-arm64/lock-inversion:
: .
: vm/vcpu lock inversion fixes, courtesy of Oliver Upton, plus a few
: extra fixes from both Oliver and Reiji Watanabe.
:
: From the initial cover letter:
:
: As it so happens, lock ordering in KVM/arm64 is completely backwards.
: There's a significant amount of VM-wide state that needs to be accessed
: from the context of a vCPU. Until now, this was accomplished by
: acquiring the kvm->lock, but that cannot be nested within vcpu->mutex.
:
: This series fixes the issue with some fine-grained locking for MP state
: and a new, dedicated mutex that can nest with both kvm->lock and
: vcpu->mutex.
: .
KVM: arm64: Have kvm_psci_vcpu_on() use WRITE_ONCE() to update mp_state
KVM: arm64: Acquire mp_state_lock in kvm_arch_vcpu_ioctl_vcpu_init()
KVM: arm64: vgic: Don't acquire its_lock before config_lock
KVM: arm64: Use config_lock to protect vgic state
KVM: arm64: Use config_lock to protect data ordered against KVM_RUN
KVM: arm64: Avoid lock inversion when setting the VM register width
KVM: arm64: Avoid vcpu->mutex v. kvm->lock inversion in CPU_ON
Signed-off-by: Marc Zyngier <maz@kernel.org>
Add support for the 8 GPIOs found on PMI632.
Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20230414-pmi632-v2-2-98bafa909c36@z3ntu.xyz
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The pinctrl-wpcm450 driver relies on MFD_SYSCON functionality in order
to find some of its MMIO registers. Select MFD_SYSCON from
PINCTRL_WPCM450 to ensure that it's enabled.
Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Link: https://lore.kernel.org/r/20230412185049.3782842-1-j.neuschaefer@gmx.net
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The fbdev test of IGT may write after EOF, which lead to out-of-bound
access for drm drivers with fbdev-generic. For example, run fbdev test
on a x86+ast2400 platform, with 1680x1050 resolution, will cause the
linux kernel hang with the following call trace:
Oops: 0000 [#1] PREEMPT SMP PTI
[IGT] fbdev: starting subtest eof
Workqueue: events drm_fb_helper_damage_work [drm_kms_helper]
[IGT] fbdev: starting subtest nullptr
RIP: 0010:memcpy_erms+0xa/0x20
RSP: 0018:ffffa17d40167d98 EFLAGS: 00010246
RAX: ffffa17d4eb7fa80 RBX: ffffa17d40e0aa80 RCX: 00000000000014c0
RDX: 0000000000001a40 RSI: ffffa17d40e0b000 RDI: ffffa17d4eb80000
RBP: ffffa17d40167e20 R08: 0000000000000000 R09: ffff89522ecff8c0
R10: ffffa17d4e4c5000 R11: 0000000000000000 R12: ffffa17d4eb7fa80
R13: 0000000000001a40 R14: 000000000000041a R15: ffffa17d40167e30
FS: 0000000000000000(0000) GS:ffff895257380000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffa17d40e0b000 CR3: 00000001eaeca006 CR4: 00000000001706e0
Call Trace:
<TASK>
? drm_fbdev_generic_helper_fb_dirty+0x207/0x330 [drm_kms_helper]
drm_fb_helper_damage_work+0x8f/0x170 [drm_kms_helper]
process_one_work+0x21f/0x430
worker_thread+0x4e/0x3c0
? __pfx_worker_thread+0x10/0x10
kthread+0xf4/0x120
? __pfx_kthread+0x10/0x10
ret_from_fork+0x2c/0x50
</TASK>
CR2: ffffa17d40e0b000
---[ end trace 0000000000000000 ]---
The is because damage rectangles computed by
drm_fb_helper_memory_range_to_clip() function is not guaranteed to be
bound in the screen's active display area. Possible reasons are:
1) Buffers are allocated in the granularity of page size, for mmap system
call support. The shadow screen buffer consumed by fbdev emulation may
also choosed be page size aligned.
2) The DIV_ROUND_UP() used in drm_fb_helper_memory_range_to_clip()
will introduce off-by-one error.
For example, on a 16KB page size system, in order to store a 1920x1080
XRGB framebuffer, we need allocate 507 pages. Unfortunately, the size
1920*1080*4 can not be divided exactly by 16KB.
1920 * 1080 * 4 = 8294400 bytes
506 * 16 * 1024 = 8290304 bytes
507 * 16 * 1024 = 8306688 bytes
line_length = 1920*4 = 7680 bytes
507 * 16 * 1024 / 7680 = 1081.6
off / line_length = 507 * 16 * 1024 / 7680 = 1081
DIV_ROUND_UP(507 * 16 * 1024, 7680) will yeild 1082
memcpy_toio() typically issue the copy line by line, when copy the last
line, out-of-bound access will be happen. Because:
1082 * line_length = 1082 * 7680 = 8309760, and 8309760 > 8306688
Note that userspace may still write to the invisiable area if a larger
buffer than width x stride is exposed. But it is not a big issue as
long as there still have memory resolve the access if not drafting so
far.
- Also limit the y1 (Daniel)
- keep fix patch it to minimal (Daniel)
- screen_size is page size aligned because of it need mmap (Thomas)
- Adding fixes tag (Thomas)
Signed-off-by: Sui Jingfeng <suijingfeng@loongson.cn>
Fixes: aa15c677cc ("drm/fb-helper: Fix vertical damage clipping")
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/dri-devel/ad44df29-3241-0d9e-e708-b0338bf3c623@189.cn/
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20230420030500.1578756-1-suijingfeng@loongson.cn
Shannon Nelson says:
====================
pds_core driver
Summary:
--------
This patchset implements a new driver for use with the AMD/Pensando
Distributed Services Card (DSC), intended to provide core configuration
services through the auxiliary_bus and through a couple of EXPORTed
functions for use initially in VFio and vDPA feature specific drivers.
To keep this patchset to a manageable size, the pds_vdpa and pds_vfio
drivers have been split out into their own patchsets to be reviewed
separately.
Detail:
-------
AMD/Pensando is making available a new set of devices for supporting vDPA,
VFio, and potentially other features in the Distributed Services Card
(DSC). These features are implemented through a PF that serves as a Core
device for controlling and configuring its VF devices. These VF devices
have separate drivers that use the auxiliary_bus to work through the Core
device as the control path.
Currently, the DSC supports standard ethernet operations using the
ionic driver. This is not replaced by the Core-based devices - these
new devices are in addition to the existing Ethernet device. Typical DSC
configurations will include both PDS devices and Ionic Eth devices.
However, there is a potential future path for ethernet services to come
through this device as well.
The Core device is a new PCI PF/VF device managed by a new driver
'pds_core'. The PF device has access to an admin queue for configuring
the services used by the VFs, and sets up auxiliary_bus devices for each
vDPA VF for communicating with the drivers for the vDPA devices. The VFs
may be for VFio or vDPA, and other services in the future; these VF types
are selected as part of the DSC internal FW configurations, which is out
of the scope of this patchset.
When the vDPA support set is enabled in the core PF through its devlink
param, auxiliary_bus devices are created for each VF that supports the
feature. The vDPA driver then connects to and uses this auxiliary_device
to do control path configuration through the PF device. This can then be
used with the vdpa kernel module to provide devices for virtio_vdpa kernel
module for host interfaces, or vhost_vdpa kernel module for interfaces
exported into your favorite VM.
A cheap ASCII diagram of a vDPA instance looks something like this:
,----------.
| vdpa |
'----------'
| ||
ctl data
| ||
.----------. ||
| pds_vdpa | ||
'----------' ||
| ||
pds_core.vDPA.1 ||
| ||
.---------------. ||
| pds_core | ||
'---------------' ||
|| || ||
09:00.0 09:00.1
== PCI ============================================
|| ||
.----------. .----------.
,------| PF |---| VF |-------,
| '----------' '----------' |
| DSC |
| |
------------------------------------------
Changes:
v11:
- change strncpy to strscpy
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/oe-kbuild-all/202304181137.WaZTYyAa-lkp@intel.com/
v10:
Link: https://lore.kernel.org/netdev/20230418003228.28234-1-shannon.nelson@amd.com/
- remove CONFIG_DEBUG_FS guard static inline stuff
- remove unnecessary 0 and null initializations
- verify in driver load that PDS_CORE_DRV_NAME matches KBUILD_MODNAME
- remove debugfs irqs_show(), redundant with /proc
- return -ENOMEM if intr_info = kcalloc() fails
- move the status code enum into pds_core_if.h as part of API definition
- fix up one place in pdsc_devcmd_wait() we're using the status codes where we could use the errno
- remove redundant calls to flush_workqueue()
- grab config_lock before testing state bits in pdsc_fw_reporter_diagnose()
- change pdsc_color_match() to return bool
- remove useless VIF setup loop and just setup vDPA services for now
- remove pf pointer from struct padev and have clients use pci_physfn()
- drop use of "vf" in auxdev.c function names, make more generic
- remove last of client ops struct and simply export the functions
- drop drivers@pensando.io from MAINTAINERS and add new include dir
- include dynamic_debug.h in adminq.c to protect dynamic_hex_dump()
- fixed fw_slot type from u8 to int for handling error returns
- fixed comment spelling
- changed void arg in pdsc_adminq_post() to struct pdsc *
v9:
Link: https://lore.kernel.org/netdev/20230406234143.11318-1-shannon.nelson@amd.com/
- change pdsc field name id to uid to clarify the unique id used for aux device
- remove unnecessary pf->state and other checks in aux device creation
- hardcode fw slotnames for devlink info, don't use strings from FW
- handle errors from PDS_CORE_CMD_INIT devcmd call
- tighten up health thread use of config_lock
- remove pdsc_queue_health_check() layer over queuing health check
- start pds_core.rst file in first patch, add to it incrementally
- give more user interaction info in commit messages
- removed a few more extraneous includes
v8:
Link: https://lore.kernel.org/netdev/20230330234628.14627-1-shannon.nelson@amd.com/
- fixed deadlock problem, use devl_health_reporter_destroy() when devlink is locked
- don't clear client_id until after auxiliary_device_uninit()
v7:
Link: https://lore.kernel.org/netdev/20230330192313.62018-1-shannon.nelson@amd.com/
- use explicit devlink locking and devl_* APIs
- move some of devlink setup logic into probe and remove
- use debugfs_create_u{type}() for state and queue head and tail
- add include for linux/vmalloc.h
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/oe-kbuild-all/202303260420.Tgq0qobF-lkp@intel.com/
v6:
Link: https://lore.kernel.org/netdev/20230324190243.27722-1-shannon.nelson@amd.com/
- removed version.h include noticed by kernel test robot's version check
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/oe-kbuild-all/202303230742.pX3ply0t-lkp@intel.com/
- fixed up the more egregious checkpatch line length complaints
- make sure pdsc_auxbus_dev_register() checks padev pointer errcode
v5:
Link: https://lore.kernel.org/netdev/20230322185626.38758-1-shannon.nelson@amd.com/
- added devlink health reporter for FW issues
- removed asic_type, asic_rev, serial_num, fw_version from debugfs as
they are available through other means
- trimed OS info in pdsc_identify(), we don't need to send that much info to the FW
- removed reg/unreg from auxbus client API, they are now in the core when VF
is started
- removed need for pdsc definition in client by simplifying the padev to only carry
struct pci_dev pointers rather than full struct pdsc to the pf and vf
- removed the unused pdsc argument in pdsc_notify()
- moved include/linux/pds/pds_core.h to driver/../pds_core/core.h
- restored a few pds_core_if.h interface values and structs that are shared
with FW source
- moved final config_lock unlock to before tear down of timer and workqueue
to be sure there are no deadlocks while waiting for any stragglers
- changed use of PAGE_SIZE to local PDS_PAGE_SIZE to keep with FW layout needs
without regard to kernel PAGE_SIZE configuration
- removed the redundant *adminqcq argument from pdsc_adminq_post()
v4:
Link: https://lore.kernel.org/netdev/20230308051310.12544-1-shannon.nelson@amd.com/
- reworked to attach to both Core PF and vDPA VF PCI devices
- now creates auxiliary_device as part of each VF PCI probe, removes them on PCI remove
- auxiliary devices now use simple unique id rather than PCI address for identifier
- replaced home-grown event publishing with kernel-based notifier service
- dropped live_migration parameter, not needed when not creating aux device for it
- replaced devm_* functions with traditional interfaces
- added MAINTAINERS entry
- removed lingering traces of set/get_vf attribute adminq commands
- trimmed some include lists
- cleaned a kernel test robot complaint about a stray unused variable
Link: https://lore.kernel.org/oe-kbuild-all/202302181049.yeUQMeWY-lkp@intel.com/
v3:
Link: https://lore.kernel.org/netdev/20230217225558.19837-1-shannon.nelson@amd.com/
- changed names from "pensando" to "amd" and updated copyright strings
- dropped the DEVLINK_PARAM_GENERIC_ID_FW_BANK for future development
- changed the auxiliary device creation to be triggered by the
PCI bus event BOUND_DRIVER, and torn down at UNBIND_DRIVER in order
to properly handle users using the sysfs bind/unbind functions
- dropped some noisy log messages
- rebased to current net-next
RFC to v2:
Link: https://lore.kernel.org/netdev/20221207004443.33779-1-shannon.nelson@amd.com/
- added separate devlink param patches for DEVLINK_PARAM_GENERIC_ID_ENABLE_MIGRATION
and DEVLINK_PARAM_GENERIC_ID_FW_BANK, and dropped the driver specific implementations
- updated descriptions for the new devlink parameters
- dropped netdev support
- dropped vDPA patches, will followup later
- separated fw update and fw bank select into their own patches
RFC:
Link: https://lore.kernel.org/netdev/20221118225656.48309-1-snelson@pensando.io/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Remaining documentation and Kconfig hook for building the driver.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the Core device gets an event from the device, or notices
the device FW to be up or down, it needs to send those events
on to the clients that have an event handler. Add the code to
pass along the events to the clients.
The entry points pdsc_register_notify() and pdsc_unregister_notify()
are EXPORTed for other drivers that want to listen for these events.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the client API operations for running adminq commands.
The core registers the client with the FW, then the client
has a context for requesting adminq services. We expect
to add additional operations for other clients, including
requesting additional private adminqs and IRQs, but don't have
the need yet.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the devlink parameter switches so the user can enable
the features supported by the VFs. The only feature supported
at the moment is vDPA.
Example:
devlink dev param set pci/0000:2b:00.0 \
name enable_vnet cmode runtime value true
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
An auxiliary_bus device is created for each vDPA type VF at VF
probe and destroyed at VF remove. The aux device name comes
from the driver name + VIF type + the unique id assigned at PCI
probe. The VFs are always removed on PF remove, so there should
be no issues with VFs trying to access missing PF structures.
The auxiliary_device names will look like "pds_core.vDPA.nn"
where 'nn' is the VF's uid.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the initial VF PCI driver framework for the new
pds_vdpa VF device, which will work in conjunction with an
auxiliary_bus client of the pds_core driver. This does the
very basics of registering for the new VF device, setting
up debugfs entries, and registering with devlink.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Virtual Interfaces (VIFs) supported by the DSC's
configuration (vDPA, Eth, RDMA, etc) are reported in the
dev_ident struct and made visible in debugfs. At this point
only vDPA is supported in this driver so we only setup
devices for that feature.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add in the support for doing firmware updates. Of the two
main banks available, a and b, this updates the one not in
use and then selects it for the next boot.
Example:
devlink dev flash pci/0000:b2:00.0 \
file pensando/dsc_fw_1.63.0-22.tar
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the service routines for submitting and processing
the adminq messages and for handling notifyq events.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set up the basic adminq and notifyq queue structures. These are
used mostly by the client drivers for feature configuration.
These are essentially the same adminq and notifyq as in the
ionic driver.
Part of this includes querying for device identity and FW
information, so we can make that available to devlink dev info.
$ devlink dev info pci/0000:b5:00.0
pci/0000:b5:00.0:
driver pds_core
serial_number FLM18420073
versions:
fixed:
asic.id 0x0
asic.rev 0x0
running:
fw 1.51.0-73
stored:
fw.goldfw 1.15.9-C-22
fw.mainfwa 1.60.0-73
fw.mainfwb 1.60.0-57
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add devlink health reporting on top of our fw watchdog.
Example:
# devlink health show pci/0000:2b:00.0 reporter fw
pci/0000:2b:00.0:
reporter fw
state healthy error 0 recover 0
# devlink health diagnose pci/0000:2b:00.0 reporter fw
Status: healthy State: 1 Generation: 0 Recoveries: 0
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
