Commit 468808bd35 ("KVM: PPC: Book3S HV: Set process table for HPT
guests on POWER9", 2017-01-30) added a call to kvmppc_update_lpcr()
which doesn't hold the kvm->lock mutex around the call, as required.
This adds the lock/unlock pair, and for good measure, includes
the kvmppc_setup_partition_table() call in the locked region, since
it is altering global state of the VM.
This error appears not to have any fatal consequences for the host;
the consequences would be that the VCPUs could end up running with
different LPCR values, or an update to the LPCR value by userspace
using the one_reg interface could get overwritten, or the update
done by kvmhv_configure_mmu() could get overwritten.
Cc: stable@vger.kernel.org # v4.10+
Fixes: 468808bd35 ("KVM: PPC: Book3S HV: Set process table for HPT guests on POWER9")
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
The XIVE interrupt controller on POWER9 machines doesn't support byte
accesses to any register in the thread management area other than the
CPPR (current processor priority register). In particular, when
reading the PIPR (pending interrupt priority register), we need to
do a 32-bit or 64-bit load.
Cc: stable@vger.kernel.org # v4.13
Fixes: 2c4fb78f78 ("KVM: PPC: Book3S HV: Workaround POWER9 DD1.0 bug causing IPB bit loss")
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
This fix was intended for 4.13, but didn't get in because both
maintainers were on vacation.
Paul Mackerras:
"It adds mutual exclusion between list_add_rcu and list_del_rcu calls
on the kvm->arch.spapr_tce_tables list. Without this, userspace could
potentially trigger corruption of the list and cause a host crash or
worse."
We do ctx = kzalloc(sizeof(*ctx), GFP_KERNEL) and then later on call
anon_inode_getfd(), but if that fails we don't free ctx, so that
memory gets leaked. To fix it, this adds kfree(ctx) in the failure
path.
Signed-off-by: nixiaoming <nixiaoming@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
This merges in the 'ppc-kvm' topic branch from the powerpc tree in
order to bring in some fixes which touch both powerpc and KVM code.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
This adds information about storage keys to the struct returned by
the KVM_PPC_GET_SMMU_INFO ioctl. The new fields replace a pad field,
which was zeroed by previous kernel versions. Thus userspace that
knows about the new fields will see zeroes when running on an older
kernel, indicating that storage keys are not supported. The size of
the structure has not changed.
The number of keys is hard-coded for the CPUs supported by HV KVM,
which is just POWER7, POWER8 and POWER9.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Commit 2f2724630f ("KVM: PPC: Book3S HV: Cope with host using large
decrementer mode", 2017-05-22) added code to treat the hypervisor
decrementer (HDEC) as a 64-bit value on POWER9 rather than 32-bit.
Unfortunately, that commit missed one place where HDEC is treated
as a 32-bit value. This fixes it.
This bug should not have any user-visible consequences that I can
think of, beyond an occasional unnecessary exit to the host kernel.
If the hypervisor decrementer has gone negative, then the bottom
32 bits will be negative for about 4 seconds after that, so as
long as we get out of the guest within those 4 seconds we won't
conclude that the HDEC interrupt is spurious.
Reported-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Fixes: 2f2724630f ("KVM: PPC: Book3S HV: Cope with host using large decrementer mode")
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
binutils >= 2.26 now warns about misuse of register expressions in
assembler operands that are actually literals. In this instance r0 is
being used where a literal 0 should be used.
Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
[mpe: Split into separate KVM patch, tweak change log]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
KVM currently validates the size of the VPA registered by the client
against sizeof(struct lppaca), however we align (and therefore size)
that struct to 1kB to avoid crossing a 4kB boundary in the client.
PAPR calls for sizes >= 640 bytes to be accepted. Hard code this with
a comment.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
In handling a H_ENTER hypercall, the code in kvmppc_do_h_enter
clobbers the high-order two bits of the storage key, which is stored
in a split field in the second doubleword of the HPTE. Any storage
key number above 7 hence fails to operate correctly.
This makes sure we preserve all the bits of the storage key.
Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Ram Pai <linuxram@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
We should set "err = -ENOMEM;", otherwise it means we're returning
ERR_PTR(0) which is NULL. It results in a NULL pointer dereference in
the caller.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
There are some error paths in kvmppc_core_vcpu_create_e500() where we
forget to set the error code. It means that we return ERR_PTR(0) which
is NULL and it results in a NULL pointer dereference in the caller.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Al Viro pointed out that while one thread of a process is executing
in kvm_vm_ioctl_create_spapr_tce(), another thread could guess the
file descriptor returned by anon_inode_getfd() and close() it before
the first thread has added it to the kvm->arch.spapr_tce_tables list.
That highlights a more general problem: there is no mutual exclusion
between writers to the spapr_tce_tables list, leading to the
possibility of the list becoming corrupted, which could cause a
host kernel crash.
To fix the mutual exclusion problem, we add a mutex_lock/unlock
pair around the list_del_rce in kvm_spapr_tce_release(). Also,
this moves the call to anon_inode_getfd() inside the region
protected by the kvm->lock mutex, after we have done the check for
a duplicate LIOBN. This means that if another thread does guess the
file descriptor and closes it, its call to kvm_spapr_tce_release()
will not do any harm because it will have to wait until the first
thread has released kvm->lock. With this, there are no failure
points in kvm_vm_ioctl_create_spapr_tce() after the call to
anon_inode_getfd().
The other things that the second thread could do with the guessed
file descriptor are to mmap it or to pass it as a parameter to a
KVM_DEV_VFIO_GROUP_SET_SPAPR_TCE ioctl on a KVM device fd. An mmap
call won't cause any harm because kvm_spapr_tce_mmap() and
kvm_spapr_tce_fault() don't access the spapr_tce_tables list or
the kvmppc_spapr_tce_table.list field, and the fields that they do use
have been properly initialized by the time of the anon_inode_getfd()
call.
The KVM_DEV_VFIO_GROUP_SET_SPAPR_TCE ioctl calls
kvm_spapr_tce_attach_iommu_group(), which scans the spapr_tce_tables
list looking for the kvmppc_spapr_tce_table struct corresponding to
the fd given as the parameter. Either it will find the new entry
or it won't; if it doesn't, it just returns an error, and if it
does, it will function normally. So, in each case there is no
harmful effect.
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
POWER9 CPUs have independent MMU contexts per thread, so KVM does not
need to quiesce secondary threads, so the hwthread_req/hwthread_state
protocol does not have to be used. So patch it away on POWER9, and patch
away the branch from the Linux idle wakeup to kvm_start_guest that is
never used.
Add a warning and error out of kvmppc_grab_hwthread in case it is ever
called on POWER9.
This avoids a hwsync in the idle wakeup path on POWER9.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
[mpe: Use WARN(...) instead of WARN_ON()/pr_err(...)]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Nixiaoming pointed out that there is a memory leak in
kvm_vm_ioctl_create_spapr_tce() if the call to anon_inode_getfd()
fails; the memory allocated for the kvmppc_spapr_tce_table struct
is not freed, and nor are the pages allocated for the iommu
tables. In addition, we have already incremented the process's
count of locked memory pages, and this doesn't get restored on
error.
David Hildenbrand pointed out that there is a race in that the
function checks early on that there is not already an entry in the
stt->iommu_tables list with the same LIOBN, but an entry with the
same LIOBN could get added between then and when the new entry is
added to the list.
This fixes all three problems. To simplify things, we now call
anon_inode_getfd() before placing the new entry in the list. The
check for an existing entry is done while holding the kvm->lock
mutex, immediately before adding the new entry to the list.
Finally, on failure we now call kvmppc_account_memlimit to
decrement the process's count of locked memory pages.
Reported-by: Nixiaoming <nixiaoming@huawei.com>
Reported-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This adds missing memory barriers to order updates/tests of
the virtual CPPR and MFRR, thus fixing a lost IPI problem.
While at it also document all barriers in this file.
This fixes a bug causing guest IPIs to occasionally get lost. The
symptom then is hangs or stalls in the guest.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Tested-by: Guilherme G. Piccoli <gpiccoli@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
This adds a workaround for a bug in POWER9 DD1 chips where changing
the CPPR (Current Processor Priority Register) can cause bits in the
IPB (Interrupt Pending Buffer) to get lost. Thankfully it only
happens when manually manipulating CPPR which is quite rare. When it
does happen it can cause interrupts to be delayed or lost.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
When msgsnd is used for IPIs to other cores, msgsync must be executed by
the target to order stores performed on the source before its msgsnd
(provided the source executes the appropriate sync).
Fixes: 1704a81cce ("KVM: PPC: Book3S HV: Use msgsnd for IPIs to other cores on POWER9")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Commit 05a4a95279 ("kernel/watchdog: split up config options") lost
the perf-based hardlockup detector's dependency on PERF_EVENTS, which
can result in broken builds with some powerpc configurations.
Restore the dependency. Add it in for x86 too, despite x86 always
selecting PERF_EVENTS it seems reasonable to make the dependency
explicit.
Link: http://lkml.kernel.org/r/20170810114452.6673-1-npiggin@gmail.com
Fixes: 05a4a95279 ("kernel/watchdog: split up config options")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A bug in the VSX register saving that could cause userspace FP/VMX register
corruption. Never seen to happen (that we know of), was found by code
inspection, but still tagged for stable given the consequences.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=jIDJ
-----END PGP SIGNATURE-----
Merge tag 'powerpc-4.13-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"A bug in the VSX register saving that could cause userspace FP/VMX
register corruption.
Never seen to happen (that we know of), was found by code inspection,
but still tagged for stable given the consequences"
* tag 'powerpc-4.13-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc: Fix VSX enabling/flushing to also test MSR_FP and MSR_VEC
Add newer helpers to make the function usage simpler. It is always
recommended to use find_current_mm_pte() for walking the page table.
If we cannot use find_current_mm_pte(), it should be documented why
the said usage of __find_linux_pte() is safe against a parallel THP
split.
For now we have KVM code using __find_linux_pte(). This is because kvm
code ends up calling __find_linux_pte() in real mode with MSR_EE=0 but
with PACA soft_enabled = 1. We may want to fix that later and make
sure we keep the MSR_EE and PACA soft_enabled in sync. When we do that
we can switch kvm to use find_linux_pte().
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
VSX uses a combination of the old vector registers, the old FP
registers and new "second halves" of the FP registers.
Thus when we need to see the VSX state in the thread struct
(flush_vsx_to_thread()) or when we'll use the VSX in the kernel
(enable_kernel_vsx()) we need to ensure they are all flushed into
the thread struct if either of them is individually enabled.
Unfortunately we only tested if the whole VSX was enabled, not if they
were individually enabled.
Fixes: 72cd7b44bc ("powerpc: Uncomment and make enable_kernel_vsx() routine available")
Cc: stable@vger.kernel.org # v4.3+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
All fixes for code that went in this cycle.
- A revert of an optimisation to the syscall exit path, which could lead to an
oops on either older machines or machines with > 1T of memory.
- Disable some deep idle states if the firmware configuration for them fails.
- Re-enable HARD/SOFT lockup detectors in defconfigs after a Kconfig change.
- Six fairly small patches fixing bugs in our new watchdog code.
Thanks to:
Gautham R. Shenoy, Nicholas Piggin.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=Vy1K
-----END PGP SIGNATURE-----
Merge tag 'powerpc-4.13-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"All fixes for code that went in this cycle.
- a revert of an optimisation to the syscall exit path, which could
lead to an oops on either older machines or machines with > 1TB of
memory
- disable some deep idle states if the firmware configuration for
them fails
- re-enable HARD/SOFT lockup detectors in defconfigs after a Kconfig
change
- six fairly small patches fixing bugs in our new watchdog code
Thanks to: Gautham R Shenoy, Nicholas Piggin"
* tag 'powerpc-4.13-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/watchdog: add locking around init/exit functions
powerpc/watchdog: Fix marking of stuck CPUs
powerpc/watchdog: Fix final-check recovered case
powerpc/watchdog: Moderate touch_nmi_watchdog overhead
powerpc/watchdog: Improve watchdog lock primitive
powerpc: NMI IPI improve lock primitive
powerpc/configs: Re-enable HARD/SOFT lockup detectors
powerpc/powernv/idle: Disable LOSE_FULL_CONTEXT states when stop-api fails
Revert "powerpc/64: Avoid restore_math call if possible in syscall exit"
When CPUs start and stop the watchdog, they manipulate shared data
that is normally protected by the lock. Other CPUs can be running
concurrently at this time, so it's a good idea to use locking here
to be on the safe side.
Remove the barrier which is undocumented and didn't do anything.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When the SMP detector finds other CPUs stuck, it iterates over
them and marks them as stuck. This pulls them out of the pending
mask and allows the detector to continue with remaining good
CPUs (if nmi_watchdog=panic is not enabled).
The code to dothat was buggy because when setting a CPU stuck,
if the pending mask became empty, it resets it to keep the
watchdog running. However the iterator will continue to run
over the new pending mask and mark remaining good CPUs sas stuck.
Fix this by doing it with cpumask bitwise operations.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When the watchdog decides to panic, it takes the lock and double
checks everything (to avoid races with the CPU being unstuck or
panic()ed by something else).
The exit label was misplaced and would result in all-CPUs backtrace
and watchdog panic even in the case that the condition was found to be
resolved.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Some code can go into a tight loop calling touch_nmi_watchdog (e.g.,
stop_machine CPU hotplug code). This can cause contention on watchdog
locks particularly if all CPUs with watchdog enabled are spinning in
the loops.
Avoid this storm of activity by running the watchdog timer callback
from this path if we have exceeded the timer period since it was last
run.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
- Hard-disable interrupts before taking the lock, which prevents
soft-NMI re-entrancy and therefore can prevent deadlocks.
- Use raw_ variants of local_irq_disable to avoid irq debugging.
- When the lock is contended, spin at low SMT priority, using
loads only, and with interrupts enabled (where possible).
Some stalls have been noticed at high loads that go away with improved
locking. There should not be so much locking contention in the first
place (which is addressed in a subsequent patch), but locking should
still be improved.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When the NMI IPI lock is contended, spin at low SMT priority, using
loads only, and with interrupts enabled (where possible). This
improves behaviour under high contention (e.g., a system crash when
a number of CPUs are trying to enter the debugger).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In commit 05a4a95279 ("kernel/watchdog: split up config options"),
CONFIG_LOCKUP_DETECTOR was split into two separate config options,
CONFIG_HARDLOCKUP_DETECTOR and CONFIG_SOFTLOCKUP_DETECTOR.
Our defconfigs still have CONFIG_LOCKUP_DETECTOR=y, but that is no longer
user selectable, and we don't mention the new options, so we end up with
none of them enabled.
So update the defconfigs to turn on the new SOFT and HARD options, the
end result being the same as what we had previously.
Fixes: 05a4a95279 ("kernel/watchdog: split up config options")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently, we use the opal call opal_slw_set_reg() to inform the
Sleep-Winkle Engine (SLW) to restore the contents of some of the
Hypervisor state on wakeup from deep idle states that lose full
hypervisor context (characterized by the flag
OPAL_PM_LOSE_FULL_CONTEXT).
However, the current code has a bug in that if opal_slw_set_reg()
fails, we don't disable the use of these deep states (winkle on
POWER8, stop4 onwards on POWER9).
This patch fixes this bug by ensuring that if programing the
sleep-winkle engine to restore the hypervisor states in
pnv_save_sprs_for_deep_states() fails, then we exclude such states by
clearing the OPAL_PM_LOSE_FULL_CONTEXT flag from
supported_cpuidle_states. As a result POWER8 will be prevented from
using winkle for CPU-Hotplug, and POWER9 will put the offlined CPUs to
the default stop state when available.
Further, we ensure in the initialization of the cpuidle-powernv driver
to only include those states whose flags are present in
supported_cpuidle_states, thereby skipping OPAL_PM_LOSE_FULL_CONTEXT
states when they have been disabled due to stop-api failure.
Fixes: 1e1601b38e ("powerpc/powernv/idle: Restore SPRs for deep idle
states via stop API.")
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If a vcpu exits due to request a user mode spinlock, then
the spinlock-holder may be preempted in user mode or kernel mode.
(Note that not all architectures trap spin loops in user mode,
only AMD x86 and ARM/ARM64 currently do).
But if a vcpu exits in kernel mode, then the holder must be
preempted in kernel mode, so we should choose a vcpu in kernel mode
as a more likely candidate for the lock holder.
This introduces kvm_arch_vcpu_in_kernel() to decide whether the
vcpu is in kernel-mode when it's preempted. kvm_vcpu_on_spin's
new argument says the same of the spinning VCPU.
Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Fixes for recently merged code:
- a fix for the _PAGE_DEVMAP support, which was breaking KVM on Power9 radix
- avoid a (harmless) lockdep warning in the early SMP code
- return failure for some uses of dma_set_mask() rather than falling back to 32-bits
- fix stack setup in watchdog soft_nmi_common() to use emergency stack
- fix of_irq_to_resource() error check in of_fsl_spi_probe()
Two fixes going to stable:
- fix saving of Transactional Memory SPRs in core dump
- fix __check_irq_replay missing decrementer interrupt
And two misc:
- fix 64-bit boot wrapper build with non-biarch compiler
- work around a POWER9 PMU hang after state-loss idle
Thanks to:
Alistair Popple, Aneesh Kumar K.V, Cyril Bur, Gustavo Romero, Jose Ricardo
Ziviani, Laurent Vivier, Nicholas Piggin, Oliver O'Halloran, Sergei Shtylyov,
Suraj Jitindar Singh, Thomas Gleixner.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=Rdy8
-----END PGP SIGNATURE-----
Merge tag 'powerpc-4.13-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"Fixes for recently merged code:
- a fix for the _PAGE_DEVMAP support, which was breaking KVM on
Power9 radix
- avoid a (harmless) lockdep warning in the early SMP code
- return failure for some uses of dma_set_mask() rather than falling
back to 32-bits
- fix stack setup in watchdog soft_nmi_common() to use emergency
stack
- fix of_irq_to_resource() error check in of_fsl_spi_probe()
Two fixes going to stable:
- fix saving of Transactional Memory SPRs in core dump
- fix __check_irq_replay missing decrementer interrupt
And two misc:
- fix 64-bit boot wrapper build with non-biarch compiler
- work around a POWER9 PMU hang after state-loss idle
Thanks to: Alistair Popple, Aneesh Kumar K.V, Cyril Bur, Gustavo
Romero, Jose Ricardo Ziviani, Laurent Vivier, Nicholas Piggin, Oliver
O'Halloran, Sergei Shtylyov, Suraj Jitindar Singh, Thomas Gleixner"
* tag 'powerpc-4.13-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/64: Fix __check_irq_replay missing decrementer interrupt
powerpc/perf: POWER9 PMU stops after idle workaround
powerpc/83xx/mpc832x_rdb: fix of_irq_to_resource() error check
powerpc/64s: Fix stack setup in watchdog soft_nmi_common()
powerpc/powernv/pci: Return failure for some uses of dma_set_mask()
powerpc/boot: Fix 64-bit boot wrapper build with non-biarch compiler
powerpc/smp: Call smp_ops->setup_cpu() directly on the boot CPU
powerpc/tm: Fix saving of TM SPRs in core dump
powerpc/mm: Fix pmd/pte_devmap() on non-leaf entries
If the decrementer wraps again and de-asserts the decrementer
exception while hard-disabled, __check_irq_replay() has a test to
notice the wrap when interrupts are re-enabled.
The decrementer check must be done when clearing the PACA_IRQ_HARD_DIS
flag, not when the PACA_IRQ_DEC flag is tested. Previously this worked
because the decrementer interrupt was always the first one checked
after clearing the hard disable flag, but HMI check was moved ahead of
that, which introduced this bug.
This can cause a missed decrementer interrupt if we soft-disable
interrupts then take an HMI which is recorded in irq_happened, then
hard-disable interrupts for > 4s to wrap the decrementer.
Fixes: e0e0d6b739 ("powerpc/64: Replay hypervisor maintenance interrupt first")
Cc: stable@vger.kernel.org # v4.9+
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
POWER9 DD2 PMU can stop after a state-loss idle in some conditions.
A solution is to set then clear MMCRA[60] after wake from state-loss
idle. MMCRA[60] is a non-architected bit, see the user manual for
details.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Reviewed-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Acked-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
of_irq_to_resource() has recently been fixed to return negative error #'s
along with 0 in case of failure, however the Freescale MPC832x RDB board
code still only regards 0 as a failure indication -- fix it up.
Fixes: 7a4228bbff ("of: irq: use of_irq_get() in of_irq_to_resource()")
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Scott Wood <oss@buserror.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The watchdog soft-NMI exception stack setup loads a stack pointer
twice, which is an obvious error. It ends up using the system reset
interrupt (true-NMI) stack, which is also a bug because the watchdog
could be preempted by a system reset interrupt that overwrites the
NMI stack.
Change the soft-NMI to use the "emergency stack". The current kernel
stack is not used, because of the longer-term goal to prevent
asynchronous stack access using soft-disable.
Fixes: 2104180a53 ("powerpc/64s: implement arch-specific hardlockup watchdog")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJZapWhAAoJEHm+PkMAQRiGKb0IAJM6b7SbWaw69Og7+qiFB+zZ
xp29iXqbE9fPISC6a5BRQV1ONjeDM6opGixGHqGC8Hla6k2IYz25VDNoF8wd0MXN
cz/Ih20vd3C5afxXGe5cTT8lsPAlV0mWXxForlu6j8jPeL62FPfq6RhEkw7AcrYL
yfYy3k3qSdOrrvBdII0WAAUi46UfIs+we9BQgbsMbkHOiqV2K0MOrzKE84Xbgepq
RAy2xg6P4b4+hTx8xTrYc1MXwpnqjRc0oJ08gdmiwW3AOOU7LxYFn7zDkLPWi9Rr
g4x6r4YhBTGxT4wNvovLIiqd9QFs//dMCuPWYwEtTICG48umIqqq24beQ0mvCdg=
=08Ic
-----END PGP SIGNATURE-----
Merge tag 'v4.13-rc1' into fixes
The fixes branch is based off a random pre-rc1 commit, because we had
some fixes that needed to go in before rc1 was released.
However we now need to fix some code that went in after that point, but
before rc1, so merge rc1 to get that code into fixes so we can fix it!
PPC: host crash fixes.
x86: bugfixes, including making nested posted interrupts really work.
Generic: tweaks to kvm_stat and to uevents
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJZe2EYAAoJEL/70l94x66D4JYH/AnvioKWTsplhUKt4Y4JlpJX
EXYjQd/CIZ+MHNNUH+U+XEj6tKQymKrz4TeZSs1o0nyxCeyparR3gK27OYVpPspN
GkPSit3hyRgW9r5uXp6pZCJuFCAMpMZ6z4sKbT1FxDhnWnpWayV9w8KA+yQT/UUX
dNQ9JJPUxApcM4NCaj2OCQ8K1koNIDCc52+jATf0iK/Heiaf6UGqCcHXUIy5I5wM
OWk05Qm32VBAYb6P6FfoyGdLMNAAkJtr1fyOJDkxX730CYgwpjIP0zifnJ1bt8V2
YRnjvPO5QciDHbZ8VynwAkKi0ZAd8psjwXh0KbyahPL/2/sA2xCztMH25qweriI=
=fsfr
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull KVM fixes from Paolo Bonzini:
"s390:
- SRCU fix
PPC:
- host crash fixes
x86:
- bugfixes, including making nested posted interrupts really work
Generic:
- tweaks to kvm_stat and to uevents"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: LAPIC: Fix reentrancy issues with preempt notifiers
tools/kvm_stat: add '-f help' to get the available event list
tools/kvm_stat: use variables instead of hard paths in help output
KVM: nVMX: Fix loss of L2's NMI blocking state
KVM: nVMX: Fix posted intr delivery when vcpu is in guest mode
x86: irq: Define a global vector for nested posted interrupts
KVM: x86: do mask out upper bits of PAE CR3
KVM: make pid available for uevents without debugfs
KVM: s390: take srcu lock when getting/setting storage keys
KVM: VMX: remove unused field
KVM: PPC: Book3S HV: Fix host crash on changing HPT size
KVM: PPC: Book3S HV: Enable TM before accessing TM registers
The highlight is Ben's patch to work around a host killing bug when running KVM
guests with the Radix MMU on Power9. See the long change log of that commit for
more detail.
And then three fairly minor fixes:
- Fix of_node_put() underflow during reconfig remove, using old DLPAR tools.
- Fix recently introduced ld version check with 64-bit LE-only toolchain.
- Free the subpage_prot_table correctly, avoiding a memory leak.
Thanks to:
Aneesh Kumar K.V, Benjamin Herrenschmidt, Laurent Vivier.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=Z6uF
-----END PGP SIGNATURE-----
Merge tag 'powerpc-4.13-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"The highlight is Ben's patch to work around a host killing bug when
running KVM guests with the Radix MMU on Power9. See the long change
log of that commit for more detail.
And then three fairly minor fixes:
- fix of_node_put() underflow during reconfig remove, using old DLPAR
tools.
- fix recently introduced ld version check with 64-bit LE-only
toolchain.
- free the subpage_prot_table correctly, avoiding a memory leak.
Thanks to: Aneesh Kumar K.V, Benjamin Herrenschmidt, Laurent Vivier"
* tag 'powerpc-4.13-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/mm/hash: Free the subpage_prot_table correctly
powerpc/Makefile: Fix ld version check with 64-bit LE-only toolchain
powerpc/pseries: Fix of_node_put() underflow during reconfig remove
powerpc/mm/radix: Workaround prefetch issue with KVM
Commit 8e3f1b1d82 ("powerpc/powernv/pci: Enable 64-bit devices to access
>4GB DMA space") introduced the ability for PCI device drivers to request a
DMA mask between 64 and 32 bits and actually get a mask greater than
32-bits. However currently if certain machine configuration dependent
conditions are not meet the code silently falls back to a 32-bit mask.
This makes it hard for device drivers to detect which mask they actually
got. Instead we should return an error when the request could not be
fulfilled which allows drivers to either fallback or implement other
workarounds as documented in DMA-API-HOWTO.txt.
Signed-off-by: Alistair Popple <alistair@popple.id.au>
Acked-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Historically the boot wrapper was always built 32-bit big endian, even
for 64-bit kernels. That was because old firmwares didn't necessarily
support booting a 64-bit image. Because of that arch/powerpc/boot/Makefile
uses CROSS32CC for compilation.
However when we added 64-bit little endian support, we also added
support for building the boot wrapper 64-bit. However we kept using
CROSS32CC, because in most cases it is just CC and everything works.
However if the user doesn't specify CROSS32_COMPILE (which no one ever
does AFAIK), and CC is *not* biarch (32/64-bit capable), then CROSS32CC
becomes just "gcc". On native systems that is probably OK, but if we're
cross building it definitely isn't, leading to eg:
gcc ... -m64 -mlittle-endian -mabi=elfv2 ... arch/powerpc/boot/cpm-serial.c
gcc: error: unrecognized argument in option ‘-mabi=elfv2’
gcc: error: unrecognized command line option ‘-mlittle-endian’
make: *** [zImage] Error 2
To fix it, stop using CROSS32CC, because we may or may not be building
32-bit. Instead setup a BOOTCC, which defaults to CC, and only use
CROSS32_COMPILE if it's set and we're building for 32-bit.
Fixes: 147c05168f ("powerpc/boot: Add support for 64bit little endian wrapper")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Cyril Bur <cyrilbur@gmail.com>
In smp_cpus_done() we need to call smp_ops->setup_cpu() for the boot
CPU, which means it has to run *on* the boot CPU.
In the past we ensured it ran on the boot CPU by changing the CPU
affinity mask of current directly. That was removed in commit
6d11b87d55 ("powerpc/smp: Replace open coded task affinity logic"),
and replaced with a work queue call.
Unfortunately using a work queue leads to a lockdep warning, now that
the CPU hotplug lock is a regular semaphore:
======================================================
WARNING: possible circular locking dependency detected
...
kworker/0:1/971 is trying to acquire lock:
(cpu_hotplug_lock.rw_sem){++++++}, at: [<c000000000100974>] apply_workqueue_attrs+0x34/0xa0
but task is already holding lock:
((&wfc.work)){+.+.+.}, at: [<c0000000000fdb2c>] process_one_work+0x25c/0x800
...
CPU0 CPU1
---- ----
lock((&wfc.work));
lock(cpu_hotplug_lock.rw_sem);
lock((&wfc.work));
lock(cpu_hotplug_lock.rw_sem);
Although the deadlock can't happen in practice, because
smp_cpus_done() only runs in early boot before CPU hotplug is allowed,
lockdep can't tell that.
Luckily in commit 8fb12156b8 ("init: Pin init task to the boot CPU,
initially") tglx changed the generic code to pin init to the boot CPU
to begin with. The unpinning of init from the boot CPU happens in
sched_init_smp(), which is called after smp_cpus_done().
So smp_cpus_done() is always called on the boot CPU, which means we
don't need the work queue call at all - and the lockdep warning goes
away.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Currently flush_tmregs_to_thread() does not save the TM SPRs (TFHAR,
TFIAR, TEXASR) to the thread struct, unless the process is currently
inside a suspended transaction.
If the process is core dumping, and the TM SPRs have changed since the
last time the process was context switched, then we will save stale
values of the TM SPRs to the core dump.
Fix it by saving the live register state to the thread struct in that
case.
Fixes: 08e1c01d6a ("powerpc/ptrace: Enable support for TM SPR state")
Cc: stable@vger.kernel.org # v4.8+
Signed-off-by: Gustavo Romero <gromero@linux.vnet.ibm.com>
Reviewed-by: Cyril Bur <cyrilbur@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The Radix MMU translation tree as defined in ISA v3.0 contains two
different types of entry, directories and leaves. Leaves are
identified by _PAGE_PTE being set.
The formats of the two entries are different, with the directory
entries containing no spare bits for use by software. In particular
the bit we use for _PAGE_DEVMAP is not reserved for software, and is
part of the NLB (Next Level Base) field, essentially the address of
the next level in the tree.
Note that the Linux pte_t is not == _PAGE_PTE. A huge page pmd
entry (or devmap!) is also a leaf and so has _PAGE_PTE set, even
though we use a pmd_t for it in Linux.
The fix is to ensure that the pmd/pte_devmap() confirm they are
looking at a leaf entry (_PAGE_PTE) as well as checking _PAGE_DEVMAP.
Fixes: ebd3119793 ("powerpc/mm: Add devmap support for ppc64")
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Tested-by: Laurent Vivier <lvivier@redhat.com>
Tested-by: Jose Ricardo Ziviani <joserz@linux.vnet.ibm.com>
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
[mpe: Add a comment in the code and flesh out change log]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Fixes: dad6f37c26 ("powerpc: subpage_protect: Increase the array size to take care of 64TB")
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Tested-by: Ram Pai <linuxram@us.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In commit efe0160cfd ("powerpc/64: Linker on-demand sfpr functions
for modules"), we added an ld version check early in the powerpc
top-level Makefile.
Because the Makefile runs before the kernel config is setup, the
checks for CONFIG_CPU_LITTLE_ENDIAN etc. all take the default case. So
we end up configuring ld for 32-bit big endian.
That would be OK, except that for historical (or perhaps no) reason,
we use 'override LD' to add the endian flags to the LD variable
itself, rather than the normal approach of adding them to LDFLAGS.
The end result is that when we check the ld version we run it as:
$(CROSS_COMPILE)ld -EB -m elf32ppc --version
This often works, unless you are using a 64-bit only and/or little
endian only, toolchain. In which case you see something like:
$ make defconfig
powerpc64le-linux-ld: unrecognised emulation mode: elf32ppc
Supported emulations: elf64lppc elf32lppc elf32lppclinux elf32lppcsim
/bin/sh: 1: [: -ge: unexpected operator
The proper fix is to stop using 'override LD', but that will require a
fair bit of testing. Instead we can fix it for now just by reordering
the Makefile to do the version check earlier.
Fixes: efe0160cfd ("powerpc/64: Linker on-demand sfpr functions for modules")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As for commit 68baf692c4 ("powerpc/pseries: Fix of_node_put()
underflow during DLPAR remove"), the call to of_node_put() must be
removed from pSeries_reconfig_remove_node().
dlpar_detach_node() and pSeries_reconfig_remove_node() both call
of_detach_node(), and thus the node should not be released in both
cases.
Fixes: 0829f6d1f6 ("of: device_node kobject lifecycle fixes")
Cc: stable@vger.kernel.org # v3.15+
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There's a somewhat architectural issue with Radix MMU and KVM.
When coming out of a guest with AIL (Alternate Interrupt Location, ie,
MMU enabled), we start executing hypervisor code with the PID register
still containing whatever the guest has been using.
The problem is that the CPU can (and will) then start prefetching or
speculatively load from whatever host context has that same PID (if
any), thus bringing translations for that context into the TLB, which
Linux doesn't know about.
This can cause stale translations and subsequent crashes.
Fixing this in a way that is neither racy nor a huge performance
impact is difficult. We could just make the host invalidations always
use broadcast forms but that would hurt single threaded programs for
example.
We chose to fix it instead by partitioning the PID space between guest
and host. This is possible because today Linux only use 19 out of the
20 bits of PID space, so existing guests will work if we make the host
use the top half of the 20 bits space.
We additionally add support for a property to indicate to Linux the
size of the PID register which will be useful if we eventually have
processors with a larger PID space available.
There is still an issue with malicious guests purposefully setting the
PID register to a value in the hosts PID range. Hopefully future HW
can prevent that, but in the meantime, we handle it with a pair of
kludges:
- On the way out of a guest, before we clear the current VCPU in the
PACA, we check the PID and if it's outside of the permitted range
we flush the TLB for that PID.
- When context switching, if the mm is "new" on that CPU (the
corresponding bit was set for the first time in the mm cpumask), we
check if any sibling thread is in KVM (has a non-NULL VCPU pointer
in the PACA). If that is the case, we also flush the PID for that
CPU (core).
This second part is needed to handle the case where a process is
migrated (or starts a new pthread) on a sibling thread of the CPU
coming out of KVM, as there's a window where stale translations can
exist before we detect it and flush them out.
A future optimization could be added by keeping track of whether the
PID has ever been used and avoid doing that for completely fresh PIDs.
We could similarily mark PIDs that have been the subject of a global
invalidation as "fresh". But for now this will do.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Rework the asm to build with CONFIG_PPC_RADIX_MMU=n, drop
unneeded include of kvm_book3s_asm.h]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
