commit 1e35eba4055149c578baf0318d2f2f89ea3c44a0 upstream.
As spotted and explained in commit c12ab8dbc492 ("powerpc/8xx: Fix
Oops with STRICT_KERNEL_RWX without DEBUG_RODATA_TEST"), the selection
of STRICT_KERNEL_RWX without selecting DEBUG_RODATA_TEST has spotted
the lack of the DIRTY bit in the pinned kernel data TLBs.
This problem should have been detected a lot earlier if things had
been working as expected. But due to an incredible level of chance or
mishap, this went undetected because of a set of bugs: In fact the
DTLBs were not pinned, because instead of setting the reserve bit
in MD_CTR, it was set in MI_CTR that is the register for ITLBs.
But then, another huge bug was there: the physical address was
reset to 0 at the boundary between RO and RW areas, leading to the
same physical space being mapped at both 0xc0000000 and 0xc8000000.
This had by miracle no consequence until now because the entry was
not really pinned so it was overwritten soon enough to go undetected.
Of course, now that we really pin the DTLBs, it must be fixed as well.
Fixes: f76c8f6d25 ("powerpc/8xx: Add function to set pinned TLBs")
Cc: stable@vger.kernel.org # v5.8+
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Depends-on: c12ab8dbc492 ("powerpc/8xx: Fix Oops with STRICT_KERNEL_RWX without DEBUG_RODATA_TEST")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/a21e9a057fe2d247a535aff0d157a54eefee017a.1636963688.git.christophe.leroy@csgroup.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 56537faf8821e361d739fc5ff58c9c40f54a1d4c ]
When check_kvm_guest() succeeds in looking up a /hypervisor OF node, it
returns without performing a matching put for the lookup, leaving the
node's reference count elevated.
Add the necessary call to of_node_put(), rearranging the code slightly to
avoid repetition or goto.
Fixes: 107c55005f ("powerpc/pseries: Add KVM guest doorbell restrictions")
Signed-off-by: Nathan Lynch <nathanl@linux.ibm.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Tyrel Datwyler <tyreld@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210928124550.132020-1-nathanl@linux.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 95839225639ba7c3d8d7231b542728dcf222bf2d ]
Commit a21d1becaa3f ("powerpc: Reintroduce is_kvm_guest() as a fast-path
check") added is_kvm_guest() and changed kvm_para_available() to use it.
is_kvm_guest() checks a static key, kvm_guest, and that static key is
set in check_kvm_guest().
The problem is check_kvm_guest() is only called on pseries, and even
then only in some configurations. That means is_kvm_guest() always
returns false on all non-pseries and some pseries depending on
configuration. That's a bug.
For PR KVM guests this is noticable because they no longer do live
patching of themselves, which can be detected by the omission of a
message in dmesg such as:
KVM: Live patching for a fast VM worked
To fix it make check_kvm_guest() an initcall, to ensure it's always
called at boot. It needs to be core so that it runs before
kvm_guest_init() which is postcore. To be an initcall it needs to return
int, where 0 means success, so update that.
We still call it manually in pSeries_smp_probe(), because that runs
before init calls are run.
Fixes: a21d1becaa3f ("powerpc: Reintroduce is_kvm_guest() as a fast-path check")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210623130514.2543232-1-mpe@ellerman.id.au
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 496c5fe25c377ddb7815c4ce8ecfb676f051e9b6 upstream.
In isa206_idle_insn_mayloss() we store various registers into the stack
red zone, which is allowed.
However inside the IDLE_STATE_ENTER_SEQ_NORET macro we save r2 again,
to 0(r1), which corrupts the stack back chain.
We used to do the same in isa206_idle_insn_mayloss() itself, but we
fixed that in 73287caa9210 ("powerpc64/idle: Fix SP offsets when saving
GPRs"), however we missed that the macro also corrupts the back chain.
Corrupting the back chain is bad for debuggability but doesn't
necessarily cause a bug.
However we recently changed the stack handling in some KVM code, and it
now relies on the stack back chain being valid when it returns. The
corruption causes that code to return with r1 pointing somewhere in
kernel data, at some point LR is restored from the stack and we branch
to NULL or somewhere else invalid.
Only affects Power8 hosts running KVM guests, with dynamic_mt_modes
enabled (which it is by default).
The fixes tag below points to the commit that changed the KVM stack
handling, exposing this bug. The actual corruption of the back chain has
always existed since 948cf67c47 ("powerpc: Add NAP mode support on
Power7 in HV mode").
Fixes: 9b4416c5095c ("KVM: PPC: Book3S HV: Fix stack handling in idle_kvm_start_guest()")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211020094826.3222052-1-mpe@ellerman.id.au
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 73287caa9210ded6066833195f4335f7f688a46b upstream.
The idle entry/exit code saves/restores GPRs in the stack "red zone"
(Protected Zone according to PowerPC64 ELF ABI v2). However, the offset
used for the first GPR is incorrect and overwrites the back chain - the
Protected Zone actually starts below the current SP. In practice this is
probably not an issue, but it's still incorrect so fix it.
Also expand the comments to explain why using the stack "red zone"
instead of creating a new stackframe is appropriate here.
Signed-off-by: Christopher M. Riedl <cmr@codefail.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210206072342.5067-1-cmr@codefail.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 787252a10d9422f3058df9a4821f389e5326c440 ]
With PREEMPT_COUNT=y, when a CPU is offlined and then onlined again, we
get:
BUG: scheduling while atomic: swapper/1/0/0x00000000
no locks held by swapper/1/0.
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.15.0-rc2+ #100
Call Trace:
dump_stack_lvl+0xac/0x108
__schedule_bug+0xac/0xe0
__schedule+0xcf8/0x10d0
schedule_idle+0x3c/0x70
do_idle+0x2d8/0x4a0
cpu_startup_entry+0x38/0x40
start_secondary+0x2ec/0x3a0
start_secondary_prolog+0x10/0x14
This is because powerpc's arch_cpu_idle_dead() decrements the idle task's
preempt count, for reasons explained in commit a7c2bb8279 ("powerpc:
Re-enable preemption before cpu_die()"), specifically "start_secondary()
expects a preempt_count() of 0."
However, since commit 2c669ef6979c ("powerpc/preempt: Don't touch the idle
task's preempt_count during hotplug") and commit f1a0a376ca0c ("sched/core:
Initialize the idle task with preemption disabled"), that justification no
longer holds.
The idle task isn't supposed to re-enable preemption, so remove the
vestigial preempt_enable() from the CPU offline path.
Tested with pseries and powernv in qemu, and pseries on PowerVM.
Fixes: 2c669ef6979c ("powerpc/preempt: Don't touch the idle task's preempt_count during hotplug")
Signed-off-by: Nathan Lynch <nathanl@linux.ibm.com>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211015173902.2278118-1-nathanl@linux.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 23c216b335d1fbd716076e8263b54a714ea3cf0e ]
According to dma-api.rst, the dma_get_required_mask() helper should return
"the mask that the platform requires to operate efficiently". Which in
the case of PPC64 means the bypass mask and not a mask from an IOMMU table
which is shorter and slower to use due to map/unmap operations (especially
expensive on "pseries").
However the existing implementation ignores the possibility of bypassing
and returns the IOMMU table mask on the pseries platform which makes some
drivers (mpt3sas is one example) choose 32bit DMA even though bypass is
supported. The powernv platform sort of handles it by having a bigger
default window with a mask >=40 but it only works as drivers choose
63/64bit if the required mask is >32 which is rather pointless.
This reintroduces the bypass capability check to let drivers make
a better choice of the DMA mask.
Fixes: f1565c24b5 ("powerpc: use the generic dma_ops_bypass mode")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210930034454.95794-1-aik@ozlabs.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b8b928030332a0ca16d42433eb2c3085600d8704 ]
lscpu() uses core_siblings to list the number of sockets in the
system. core_siblings is set using topology_core_cpumask.
While optimizing the powerpc bootup path, Commit 4ca234a9cb
("powerpc/smp: Stop updating cpu_core_mask"). it was found that
updating cpu_core_mask() ended up taking a lot of time. It was thought
that on Powerpc, cpu_core_mask() would always be same as
cpu_cpu_mask() i.e number of sockets will always be equal to number of
nodes. As an optimization, cpu_core_mask() was made a snapshot of
cpu_cpu_mask().
However that was found to be false with PowerPc KVM guests, where each
node could have more than one socket. So with Commit c47f892d7aa6
("powerpc/smp: Reintroduce cpu_core_mask"), cpu_core_mask was updated
based on chip_id but in an optimized way using some mask manipulations
and chip_id caching.
However on non-PowerNV and non-pseries KVM guests (i.e not
implementing cpu_to_chip_id(), continued to use a copy of
cpu_cpu_mask().
There are two issues that were noticed on such systems
1. lscpu would report one extra socket.
On a IBM,9009-42A (aka zz system) which has only 2 chips/ sockets/
nodes, lscpu would report
Architecture: ppc64le
Byte Order: Little Endian
CPU(s): 160
On-line CPU(s) list: 0-159
Thread(s) per core: 8
Core(s) per socket: 6
Socket(s): 3 <--------------
NUMA node(s): 2
Model: 2.2 (pvr 004e 0202)
Model name: POWER9 (architected), altivec supported
Hypervisor vendor: pHyp
Virtualization type: para
L1d cache: 32K
L1i cache: 32K
L2 cache: 512K
L3 cache: 10240K
NUMA node0 CPU(s): 0-79
NUMA node1 CPU(s): 80-159
2. Currently cpu_cpu_mask is updated when a core is
added/removed. However its not updated when smt mode switching or on
CPUs are explicitly offlined. However all other percpu masks are
updated to ensure only active/online CPUs are in the masks.
This results in build_sched_domain traces since there will be CPUs in
cpu_cpu_mask() but those CPUs are not present in SMT / CACHE / MC /
NUMA domains. A loop of threads running smt mode switching and core
add/remove will soon show this trace.
Hence cpu_cpu_mask has to be update at smt mode switch.
This will have impact on cpu_core_mask(). cpu_core_mask() is a
snapshot of cpu_cpu_mask. Different CPUs within the same socket will
end up having different cpu_core_masks since they are snapshots at
different points of time. This means when lscpu will start reporting
many more sockets than the actual number of sockets/ nodes / chips.
Different ways to handle this problem:
A. Update the snapshot aka cpu_core_mask for all CPUs whenever
cpu_cpu_mask is updated. This would a non-optimal solution.
B. Instead of a cpumask_var_t, make cpu_core_map a cpumask pointer
pointing to cpu_cpu_mask. However percpu cpumask pointer is frowned
upon and we need a clean way to handle PowerPc KVM guest which is
not a snapshot.
C. Update cpu_core_masks all PowerPc systems like in PowerPc KVM
guests using mask manipulations. This approach is relatively simple
and unifies with the existing code.
D. On top of 3, we could also resurrect get_physical_package_id which
could return a nid for the said CPU. However this is not needed at this
time.
Option C is the preferred approach for now.
While this is somewhat a revert of Commit 4ca234a9cb ("powerpc/smp:
Stop updating cpu_core_mask").
1. Plain revert has some conflicts
2. For chip_id == -1, the cpu_core_mask is made identical to
cpu_cpu_mask, unlike previously where cpu_core_mask was set to a core
if chip_id doesn't exist.
This goes by the principle that if chip_id is not exposed, then
sockets / chip / node share the same set of CPUs.
With the fix, lscpu o/p would be
Architecture: ppc64le
Byte Order: Little Endian
CPU(s): 160
On-line CPU(s) list: 0-159
Thread(s) per core: 8
Core(s) per socket: 6
Socket(s): 2 <--------------
NUMA node(s): 2
Model: 2.2 (pvr 004e 0202)
Model name: POWER9 (architected), altivec supported
Hypervisor vendor: pHyp
Virtualization type: para
L1d cache: 32K
L1i cache: 32K
L2 cache: 512K
L3 cache: 10240K
NUMA node0 CPU(s): 0-79
NUMA node1 CPU(s): 80-159
Fixes: 4ca234a9cb ("powerpc/smp: Stop updating cpu_core_mask")
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100401.412519-3-srikar@linux.vnet.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit a6cae77f1bc89368a4e2822afcddc45c3062d499 ]
commit 7c6986ade69e ("powerpc/stacktrace: Fix spurious "stale" traces in raise_backtrace_ipi()")
introduces udelay() call without including the linux/delay.h header.
This may happen to work on master but the header that declares the
functionshould be included nonetheless.
Fixes: 7c6986ade69e ("powerpc/stacktrace: Fix spurious "stale" traces in raise_backtrace_ipi()")
Signed-off-by: Michal Suchanek <msuchanek@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210729180103.15578-1-msuchanek@suse.de
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f35d2f249ef05b9671e7898f09ad89aa78f99122 ]
copy-paste contains implicit "copy buffer" state that can contain
arbitrary user data (if the user process executes a copy instruction).
This could be snooped by another process if a context switch hits while
the state is live. So cp_abort is executed on context switch to clear
out possible sensitive data and prevent the leak.
cp_abort is done after the low level _switch(), which means it is never
reached by newly created tasks, so they could snoop on this buffer
between their first and second context switch.
Fix this by doing the cp_abort before calling _switch. Add some
comments which should make the issue harder to miss.
Fixes: 07d2a628bc ("powerpc/64s: Avoid cpabort in context switch when possible")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210622053036.474678-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit bab26238bbd44d5a4687c0a64fd2c7f2755ea937 ]
printk_safe_flush_on_panic() has special lock breaking code for the case
where we panic()ed with the console lock held. It relies on panic IPI
causing other CPUs to mark themselves offline.
Do as most other architectures do.
This effectively reverts commit de6e5d3841 ("powerpc: smp_send_stop do
not offline stopped CPUs"), unfortunately it may result in some false
positive warnings, but the alternative is more situations where we can
crash without getting messages out.
Fixes: de6e5d3841 ("powerpc: smp_send_stop do not offline stopped CPUs")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210623041245.865134-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 3729e0ec59a20825bd4c8c70996b2df63915e1dd ]
POWER9 and POWER10 asynchronous machine checks due to stores have their
cause reported in SRR1 but SRR1[42] is set, which in other cases
indicates DSISR cause.
Check for these cases and clear SRR1[42], so the cause matching uses
the i-side (SRR1) table.
Fixes: 7b9f71f974 ("powerpc/64s: POWER9 machine check handler")
Fixes: 201220bb0e ("powerpc/powernv: Machine check handler for POWER10")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210517140355.2325406-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f1a0a376ca0c4ef1fc3d24e3e502acbb5b795674 ]
As pointed out by commit
de9b8f5dcb ("sched: Fix crash trying to dequeue/enqueue the idle thread")
init_idle() can and will be invoked more than once on the same idle
task. At boot time, it is invoked for the boot CPU thread by
sched_init(). Then smp_init() creates the threads for all the secondary
CPUs and invokes init_idle() on them.
As the hotplug machinery brings the secondaries to life, it will issue
calls to idle_thread_get(), which itself invokes init_idle() yet again.
In this case it's invoked twice more per secondary: at _cpu_up(), and at
bringup_cpu().
Given smp_init() already initializes the idle tasks for all *possible*
CPUs, no further initialization should be required. Now, removing
init_idle() from idle_thread_get() exposes some interesting expectations
with regards to the idle task's preempt_count: the secondary startup always
issues a preempt_disable(), requiring some reset of the preempt count to 0
between hot-unplug and hotplug, which is currently served by
idle_thread_get() -> idle_init().
Given the idle task is supposed to have preemption disabled once and never
see it re-enabled, it seems that what we actually want is to initialize its
preempt_count to PREEMPT_DISABLED and leave it there. Do that, and remove
init_idle() from idle_thread_get().
Secondary startups were patched via coccinelle:
@begone@
@@
-preempt_disable();
...
cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210512094636.2958515-1-valentin.schneider@arm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 7c6986ade69e3c81bac831645bc72109cd798a80 upstream.
In raise_backtrace_ipi() we iterate through the cpumask of CPUs, sending
each an IPI asking them to do a backtrace, but we don't wait for the
backtrace to happen.
We then iterate through the CPU mask again, and if any CPU hasn't done
the backtrace and cleared itself from the mask, we print a trace on its
behalf, noting that the trace may be "stale".
This works well enough when a CPU is not responding, because in that
case it doesn't receive the IPI and the sending CPU is left to print the
trace. But when all CPUs are responding we are left with a race between
the sending and receiving CPUs, if the sending CPU wins the race then it
will erroneously print a trace.
This leads to spurious "stale" traces from the sending CPU, which can
then be interleaved messily with the receiving CPU, note the CPU
numbers, eg:
[ 1658.929157][ C7] rcu: Stack dump where RCU GP kthread last ran:
[ 1658.929223][ C7] Sending NMI from CPU 7 to CPUs 1:
[ 1658.929303][ C1] NMI backtrace for cpu 1
[ 1658.929303][ C7] CPU 1 didn't respond to backtrace IPI, inspecting paca.
[ 1658.929362][ C1] CPU: 1 PID: 325 Comm: kworker/1:1H Tainted: G W E 5.13.0-rc2+ #46
[ 1658.929405][ C7] irq_soft_mask: 0x01 in_mce: 0 in_nmi: 0 current: 325 (kworker/1:1H)
[ 1658.929465][ C1] Workqueue: events_highpri test_work_fn [test_lockup]
[ 1658.929549][ C7] Back trace of paca->saved_r1 (0xc0000000057fb400) (possibly stale):
[ 1658.929592][ C1] NIP: c00000000002cf50 LR: c008000000820178 CTR: c00000000002cfa0
To fix it, change the logic so that the sending CPU waits 5s for the
receiving CPU to print its trace. If the receiving CPU prints its trace
successfully then the sending CPU just continues, avoiding any spurious
"stale" trace.
This has the added benefit of allowing all CPUs to print their traces in
order and avoids any interleaving of their output.
Fixes: 5cc05910f2 ("powerpc/64s: Wire up arch_trigger_cpumask_backtrace()")
Cc: stable@vger.kernel.org # v4.18+
Reported-by: Nathan Lynch <nathanl@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210625140408.3351173-1-mpe@ellerman.id.au
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 82123a3d1d5a306fdf50c968a474cc60fe43a80f upstream.
When checking if the probed instruction is the suffix of a prefixed
instruction, we access the instruction at the previous word. If the
probed instruction is the very first word of a module, we can end up
trying to access an invalid page.
Fix this by skipping the check for all instructions at the beginning of
a page. Prefixed instructions cannot cross a 64-byte boundary and as
such, we don't expect to encounter a suffix as the very first word in a
page for kernel text. Even if there are prefixed instructions crossing
a page boundary (from a module, for instance), the instruction will be
illegal, so preventing probing on the suffix of such prefix instructions
isn't worthwhile.
Fixes: b4657f7650 ("powerpc/kprobes: Don't allow breakpoints on suffixes")
Cc: stable@vger.kernel.org # v5.8+
Reported-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/0df9a032a05576a2fa8e97d1b769af2ff0eafbd6.1621416666.git.naveen.n.rao@linux.vnet.ibm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit e2f5efd0f0e229bd110eab513e7c0331d61a4649 ]
The immediate problem is that after commit
0bd3f9e953bd ("powerpc/legacy_serial: Use early_ioremap()") the kernel
silently reboots on some systems.
The reason is that early_ioremap() returns broken addresses as it uses
slot_virt[] array which initialized with offsets from FIXADDR_TOP ==
IOREMAP_END+FIXADDR_SIZE == KERN_IO_END - FIXADDR_SIZ + FIXADDR_SIZE ==
__kernel_io_end which is 0 when early_ioremap_setup() is called.
__kernel_io_end is initialized little bit later in early_init_mmu().
This fixes the initialization by swapping early_ioremap_setup() and
early_init_mmu().
Fixes: 265c3491c4 ("powerpc: Add support for GENERIC_EARLY_IOREMAP")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
[mpe: Drop unrelated cleanup & cleanup change log]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210520032919.358935-1-aik@ozlabs.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit cc7130bf119add37f36238343a593b71ef6ecc1e ]
The IOMMU table is divided into pools for concurrent mappings and each
pool has a separate spinlock. When taking the ownership of an IOMMU group
to pass through a device to a VM, we lock these spinlocks which triggers
a false negative warning in lockdep (below).
This fixes it by annotating the large pool's spinlock as a nest lock
which makes lockdep not complaining when locking nested locks if
the nest lock is locked already.
===
WARNING: possible recursive locking detected
5.11.0-le_syzkaller_a+fstn1 #100 Not tainted
--------------------------------------------
qemu-system-ppc/4129 is trying to acquire lock:
c0000000119bddb0 (&(p->lock)/1){....}-{2:2}, at: iommu_take_ownership+0xac/0x1e0
but task is already holding lock:
c0000000119bdd30 (&(p->lock)/1){....}-{2:2}, at: iommu_take_ownership+0xac/0x1e0
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(p->lock)/1);
lock(&(p->lock)/1);
===
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210301063653.51003-1-aik@ozlabs.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c47f892d7aa62765bf0689073f75990b4517a4cf ]
Daniel reported that with Commit 4ca234a9cb ("powerpc/smp: Stop
updating cpu_core_mask") QEMU was unable to set single NUMA node SMP
topologies such as:
-smp 8,maxcpus=8,cores=2,threads=2,sockets=2
i.e he expected 2 sockets in one NUMA node.
The above commit helped to reduce boot time on Large Systems for
example 4096 vCPU single socket QEMU instance. PAPR is silent on
having more than one socket within a NUMA node.
cpu_core_mask and cpu_cpu_mask for any CPU would be same unless the
number of sockets is different from the number of NUMA nodes.
One option is to reintroduce cpu_core_mask but use a slightly
different method to arrive at the cpu_core_mask. Previously each CPU's
chip-id would be compared with all other CPU's chip-id to verify if
both the CPUs were related at the chip level. Now if a CPU 'A' is
found related / (unrelated) to another CPU 'B', all the thread
siblings of 'A' and thread siblings of 'B' are automatically marked as
related / (unrelated).
Also if a platform doesn't support ibm,chip-id property, i.e its
cpu_to_chip_id returns -1, cpu_core_map holds a copy of
cpu_cpu_mask().
Fixes: 4ca234a9cb ("powerpc/smp: Stop updating cpu_core_mask")
Reported-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210415120934.232271-2-srikar@linux.vnet.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1ef1dd9c7ed27b080445e1576e8a05957e0e4dfc ]
If identical_pvr_fixup() is not inlined, there are two modpost warnings:
WARNING: modpost: vmlinux.o(.text+0x54e8): Section mismatch in reference
from the function identical_pvr_fixup() to the function
.init.text:of_get_flat_dt_prop()
The function identical_pvr_fixup() references
the function __init of_get_flat_dt_prop().
This is often because identical_pvr_fixup lacks a __init
annotation or the annotation of of_get_flat_dt_prop is wrong.
WARNING: modpost: vmlinux.o(.text+0x551c): Section mismatch in reference
from the function identical_pvr_fixup() to the function
.init.text:identify_cpu()
The function identical_pvr_fixup() references
the function __init identify_cpu().
This is often because identical_pvr_fixup lacks a __init
annotation or the annotation of identify_cpu is wrong.
identical_pvr_fixup() calls two functions marked as __init and is only
called by a function marked as __init so it should be marked as __init
as well. At the same time, remove the inline keywork as it is not
necessary to inline this function. The compiler is still free to do so
if it feels it is worthwhile since commit 889b3c1245 ("compiler:
remove CONFIG_OPTIMIZE_INLINING entirely").
Fixes: 14b3d926a2 ("[POWERPC] 4xx: update 440EP(x)/440GR(x) identical PVR issue workaround")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://github.com/ClangBuiltLinux/linux/issues/1316
Link: https://lore.kernel.org/r/20210302200829.2680663-1-nathan@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fbced1546eaaab57a32e56c974ea8acf10c6abd8 ]
If fadump_calculate_reserve_size() is not inlined, there is a modpost
warning:
WARNING: modpost: vmlinux.o(.text+0x5196c): Section mismatch in
reference from the function fadump_calculate_reserve_size() to the
function .init.text:parse_crashkernel()
The function fadump_calculate_reserve_size() references
the function __init parse_crashkernel().
This is often because fadump_calculate_reserve_size lacks a __init
annotation or the annotation of parse_crashkernel is wrong.
fadump_calculate_reserve_size() calls parse_crashkernel(), which is
marked as __init and fadump_calculate_reserve_size() is called from
within fadump_reserve_mem(), which is also marked as __init.
Mark fadump_calculate_reserve_size() as __init to fix the section
mismatch. Additionally, remove the inline keyword as it is not necessary
to inline this function; the compiler is still free to do so if it feels
it is worthwhile since commit 889b3c1245 ("compiler: remove
CONFIG_OPTIMIZE_INLINING entirely").
Fixes: 11550dc0a0 ("powerpc/fadump: reuse crashkernel parameter for fadump memory reservation")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://github.com/ClangBuiltLinux/linux/issues/1300
Link: https://lore.kernel.org/r/20210302195013.2626335-1-nathan@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 5ae5bc12d0728db60a0aa9b62160ffc038875f1a upstream.
During the EEH MMIO error checking, the current implementation fails to map
the (virtual) MMIO address back to the pci device on radix with hugepage
mappings for I/O. This results into failure to dispatch EEH event with no
recovery even when EEH capability has been enabled on the device.
eeh_check_failure(token) # token = virtual MMIO address
addr = eeh_token_to_phys(token);
edev = eeh_addr_cache_get_dev(addr);
if (!edev)
return 0;
eeh_dev_check_failure(edev); <= Dispatch the EEH event
In case of hugepage mappings, eeh_token_to_phys() has a bug in virt -> phys
translation that results in wrong physical address, which is then passed to
eeh_addr_cache_get_dev() to match it against cached pci I/O address ranges
to get to a PCI device. Hence, it fails to find a match and the EEH event
never gets dispatched leaving the device in failed state.
The commit 3343962068 ("powerpc/eeh: Handle hugepages in ioremap space")
introduced following logic to translate virt to phys for hugepage mappings:
eeh_token_to_phys():
+ pa = pte_pfn(*ptep);
+
+ /* On radix we can do hugepage mappings for io, so handle that */
+ if (hugepage_shift) {
+ pa <<= hugepage_shift; <= This is wrong
+ pa |= token & ((1ul << hugepage_shift) - 1);
+ }
This patch fixes the virt -> phys translation in eeh_token_to_phys()
function.
$ cat /sys/kernel/debug/powerpc/eeh_address_cache
mem addr range [0x0000040080000000-0x00000400807fffff]: 0030:01:00.1
mem addr range [0x0000040080800000-0x0000040080ffffff]: 0030:01:00.1
mem addr range [0x0000040081000000-0x00000400817fffff]: 0030:01:00.0
mem addr range [0x0000040081800000-0x0000040081ffffff]: 0030:01:00.0
mem addr range [0x0000040082000000-0x000004008207ffff]: 0030:01:00.1
mem addr range [0x0000040082080000-0x00000400820fffff]: 0030:01:00.0
mem addr range [0x0000040082100000-0x000004008210ffff]: 0030:01:00.1
mem addr range [0x0000040082110000-0x000004008211ffff]: 0030:01:00.0
Above is the list of cached io address ranges of pci 0030:01:00.<fn>.
Before this patch:
Tracing 'arg1' of function eeh_addr_cache_get_dev() during error injection
clearly shows that 'addr=' contains wrong physical address:
kworker/u16:0-7 [001] .... 108.883775: eeh_addr_cache_get_dev:
(eeh_addr_cache_get_dev+0xc/0xf0) addr=0x80103000a510
dmesg shows no EEH recovery messages:
[ 108.563768] bnx2x: [bnx2x_timer:5801(eth2)]MFW seems hanged: drv_pulse (0x9ae) != mcp_pulse (0x7fff)
[ 108.563788] bnx2x: [bnx2x_hw_stats_update:870(eth2)]NIG timer max (4294967295)
[ 108.883788] bnx2x: [bnx2x_acquire_hw_lock:2013(eth1)]lock_status 0xffffffff resource_bit 0x1
[ 108.884407] bnx2x 0030:01:00.0 eth1: MDC/MDIO access timeout
[ 108.884976] bnx2x 0030:01:00.0 eth1: MDC/MDIO access timeout
<..>
After this patch:
eeh_addr_cache_get_dev() trace shows correct physical address:
<idle>-0 [001] ..s. 1043.123828: eeh_addr_cache_get_dev:
(eeh_addr_cache_get_dev+0xc/0xf0) addr=0x40080bc7cd8
dmesg logs shows EEH recovery getting triggerred:
[ 964.323980] bnx2x: [bnx2x_timer:5801(eth2)]MFW seems hanged: drv_pulse (0x746f) != mcp_pulse (0x7fff)
[ 964.323991] EEH: Recovering PHB#30-PE#10000
[ 964.324002] EEH: PE location: N/A, PHB location: N/A
[ 964.324006] EEH: Frozen PHB#30-PE#10000 detected
<..>
Fixes: 3343962068 ("powerpc/eeh: Handle hugepages in ioremap space")
Cc: stable@vger.kernel.org # v5.3+
Reported-by: Dominic DeMarco <ddemarc@us.ibm.com>
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/161821396263.48361.2796709239866588652.stgit@jupiter
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 49c1d07fd04f54eb588c4a1dfcedc8d22c5ffd50 upstream.
Starting with ISA v3.1, LPCR[AIL] no longer controls the interrupt
mode for HV=1 interrupts. Instead, a new LPCR[HAIL] bit is defined
which behaves like AIL=3 for HV interrupts when set.
Set HAIL on bare metal to give us mmu-on interrupts and improve
performance.
This also fixes an scv bug: we don't implement scv real mode (AIL=0)
vectors because they are at an inconvenient location, so we just
disable scv support when AIL can not be set. However powernv assumes
that LPCR[AIL] will enable AIL mode so it enables scv support despite
HV interrupts being AIL=0, which causes scv interrupts to go off into
the weeds.
Fixes: 7fa95f9ada ("powerpc/64s: system call support for scv/rfscv instructions")
Cc: stable@vger.kernel.org # v5.9+
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210402024124.545826-1-npiggin@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit e3de1e291fa58a1ab0f471a4b458eff2514e4b5f ]
In commit bf13718bc57a ("powerpc: show registers when unwinding
interrupt frames") we changed our stack dumping logic to show the full
registers whenever we find an interrupt frame on the stack.
However we didn't notice that on 64-bit this doesn't show the final
frame, ie. the interrupt that brought us in from userspace, whereas on
32-bit it does.
That is due to confusion about the size of that last frame. The code
in show_stack() calls validate_sp(), passing it STACK_INT_FRAME_SIZE
to check the sp is at least that far below the top of the stack.
However on 64-bit that size is too large for the final frame, because
it includes the red zone, but we don't allocate a red zone for the
first frame.
So add a new define that encodes the correct size for 32-bit and
64-bit, and use it in show_stack().
This results in the full trace being shown on 64-bit, eg:
sysrq: Trigger a crash
Kernel panic - not syncing: sysrq triggered crash
CPU: 0 PID: 83 Comm: sh Not tainted 5.11.0-rc2-gcc-8.2.0-00188-g571abcb96b10-dirty #649
Call Trace:
[c00000000a1c3ac0] [c000000000897b70] dump_stack+0xc4/0x114 (unreliable)
[c00000000a1c3b00] [c00000000014334c] panic+0x178/0x41c
[c00000000a1c3ba0] [c00000000094e600] sysrq_handle_crash+0x40/0x50
[c00000000a1c3c00] [c00000000094ef98] __handle_sysrq+0xd8/0x210
[c00000000a1c3ca0] [c00000000094f820] write_sysrq_trigger+0x100/0x188
[c00000000a1c3ce0] [c0000000005559dc] proc_reg_write+0x10c/0x1b0
[c00000000a1c3d10] [c000000000479950] vfs_write+0xf0/0x360
[c00000000a1c3d60] [c000000000479d9c] ksys_write+0x7c/0x140
[c00000000a1c3db0] [c00000000002bf5c] system_call_exception+0x19c/0x2c0
[c00000000a1c3e10] [c00000000000d35c] system_call_common+0xec/0x278
--- interrupt: c00 at 0x7fff9fbab428
NIP: 00007fff9fbab428 LR: 000000001000b724 CTR: 0000000000000000
REGS: c00000000a1c3e80 TRAP: 0c00 Not tainted (5.11.0-rc2-gcc-8.2.0-00188-g571abcb96b10-dirty)
MSR: 900000000280f033 <SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE> CR: 22002884 XER: 00000000
IRQMASK: 0
GPR00: 0000000000000004 00007fffc3cb8960 00007fff9fc59900 0000000000000001
GPR04: 000000002a4b32d0 0000000000000002 0000000000000063 0000000000000063
GPR08: 000000002a4b32d0 0000000000000000 0000000000000000 0000000000000000
GPR12: 0000000000000000 00007fff9fcca9a0 0000000000000000 0000000000000000
GPR16: 0000000000000000 0000000000000000 0000000000000000 00000000100b8fd0
GPR20: 000000002a4b3485 00000000100b8f90 0000000000000000 0000000000000000
GPR24: 000000002a4b0440 00000000100e77b8 0000000000000020 000000002a4b32d0
GPR28: 0000000000000001 0000000000000002 000000002a4b32d0 0000000000000001
NIP [00007fff9fbab428] 0x7fff9fbab428
LR [000000001000b724] 0x1000b724
--- interrupt: c00
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210209141627.2898485-1-mpe@ellerman.id.au
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 5537fcb319d016ce387f818dd774179bc03217f5 ]
On many powerpc platforms the discovery and initalisation of
pci_controllers (PHBs) happens inside of setup_arch(). This is very early
in boot (pre-initcalls) and means that we're initialising the PHB long
before many basic kernel services (slab allocator, debugfs, a real ioremap)
are available.
On PowerNV this causes an additional problem since we map the PHB registers
with ioremap(). As of commit d538aadc27 ("powerpc/ioremap: warn on early
use of ioremap()") a warning is printed because we're using the "incorrect"
API to setup and MMIO mapping in searly boot. The kernel does provide
early_ioremap(), but that is not intended to create long-lived MMIO
mappings and a seperate warning is printed by generic code if
early_ioremap() mappings are "leaked."
This is all fixable with dumb hacks like using early_ioremap() to setup
the initial mapping then replacing it with a real ioremap later on in
boot, but it does raise the question: Why the hell are we setting up the
PHB's this early in boot?
The old and wise claim it's due to "hysterical rasins." Aside from amused
grapes there doesn't appear to be any real reason to maintain the current
behaviour. Already most of the newer embedded platforms perform PHB
discovery in an arch_initcall and between the end of setup_arch() and the
start of initcalls none of the generic kernel code does anything PCI
related. On powerpc scanning PHBs occurs in a subsys_initcall so it should
be possible to move the PHB discovery to a core, postcore or arch initcall.
This patch adds the ppc_md.discover_phbs hook and a core_initcall stub that
calls it. The core_initcalls are the earliest to be called so this will
any possibly issues with dependency between initcalls. This isn't just an
academic issue either since on pseries and PowerNV EEH init occurs in an
arch_initcall and depends on the pci_controllers being available, similarly
the creation of pci_dns occurs at core_initcall_sync (i.e. between core and
postcore initcalls). These problems need to be addressed seperately.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
[mpe: Make discover_phbs() static]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20201103043523.916109-1-oohall@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit c119565a15a628efdfa51352f9f6c5186e506a1c upstream.
On book3s/32, page protection is defined by the PP bits in the PTE
which provide the following protection depending on the access
keys defined in the matching segment register:
- PP 00 means RW with key 0 and N/A with key 1.
- PP 01 means RW with key 0 and RO with key 1.
- PP 10 means RW with both key 0 and key 1.
- PP 11 means RO with both key 0 and key 1.
Since the implementation of kernel userspace access protection,
PP bits have been set as follows:
- PP00 for pages without _PAGE_USER
- PP01 for pages with _PAGE_USER and _PAGE_RW
- PP11 for pages with _PAGE_USER and without _PAGE_RW
For kernelspace segments, kernel accesses are performed with key 0
and user accesses are performed with key 1. As PP00 is used for
non _PAGE_USER pages, user can't access kernel pages not flagged
_PAGE_USER while kernel can.
For userspace segments, both kernel and user accesses are performed
with key 0, therefore pages not flagged _PAGE_USER are still
accessible to the user.
This shouldn't be an issue, because userspace is expected to be
accessible to the user. But unlike most other architectures, powerpc
implements PROT_NONE protection by removing _PAGE_USER flag instead of
flagging the page as not valid. This means that pages in userspace
that are not flagged _PAGE_USER shall remain inaccessible.
To get the expected behaviour, just mimic other architectures in the
TLB miss handler by checking _PAGE_USER permission on userspace
accesses as if it was the _PAGE_PRESENT bit.
Note that this problem only is only for 603 cores. The 604+ have
an hash table, and hash_page() function already implement the
verification of _PAGE_USER permission on userspace pages.
Fixes: f342adca3a ("powerpc/32s: Prepare Kernel Userspace Access Protection")
Cc: stable@vger.kernel.org # v5.2+
Reported-by: Christoph Plattner <christoph.plattner@thalesgroup.com>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/4a0c6e3bb8f0c162457bf54d9bc6fd8d7b55129f.1612160907.git.christophe.leroy@csgroup.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3642eb21256a317ac14e9ed560242c6d20cf06d9 upstream.
THREAD_ALIGN_SHIFT = THREAD_SHIFT + 1 = PAGE_SHIFT + 1
Maximum PAGE_SHIFT is 18 for 256k pages so
THREAD_ALIGN_SHIFT is 19 at the maximum.
No need to clobber cr1, it can be preserved when moving r1
into CR when we check stack overflow.
This reduces the number of instructions in Machine Check Exception
prolog and fixes a build failure reported by the kernel test robot
on v5.10 stable when building with RTAS + VMAP_STACK + KVM. That
build failure is due to too many instructions in the prolog hence
not fitting between 0x200 and 0x300. Allthough the problem doesn't
show up in mainline, it is still worth the change.
Fixes: 98bf2d3f4970 ("powerpc/32s: Fix RTAS machine check with VMAP stack")
Cc: stable@vger.kernel.org
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/5ae4d545e3ac58e133d2599e0deb88843cb494fc.1612768623.git.christophe.leroy@csgroup.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ed5b00a05c2ae95b59adc3442f45944ec632e794 upstream.
The "ibm,arch-vec-5-platform-support" property is a list of pairs of
bytes representing the options and values supported by the platform
firmware. At boot time, Linux scans this list and activates the
available features it recognizes : Radix and XIVE.
A recent change modified the number of entries to loop on and 8 bytes,
4 pairs of { options, values } entries are always scanned. This is
fine on KVM but not on PowerVM which can advertises less. As a
consequence on this platform, Linux reads extra entries pointing to
random data, interprets these as available features and tries to
activate them, leading to a firmware crash in
ibm,client-architecture-support.
Fix that by using the property length of "ibm,arch-vec-5-platform-support".
Fixes: ab91239942 ("powerpc/prom: Remove VLA in prom_check_platform_support()")
Cc: stable@vger.kernel.org # v4.20+
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Fabiano Rosas <farosas@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210122075029.797013-1-clg@kaod.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 60a707d0c99aff4eadb7fd334c5fd21df386723e ]
Since de78a9c42a ("powerpc: Add a framework for Kernel Userspace
Access Protection"), user access helpers call user_{read|write}_access_{begin|end}
when user space access is allowed.
Commit 890274c2dc ("powerpc/64s: Implement KUAP for Radix MMU") made
the mentioned helpers program a AMR special register to allow such
access for a short period of time, most of the time AMR is expected to
block user memory access by the kernel.
Since the code accesses the user space memory, unsafe_get_user() calls
might_fault() which calls arch_local_irq_restore() if either
CONFIG_PROVE_LOCKING or CONFIG_DEBUG_ATOMIC_SLEEP is enabled.
arch_local_irq_restore() then attempts to replay pending soft
interrupts as KUAP regions have hardware interrupts enabled.
If a pending interrupt happens to do user access (performance
interrupts do that), it enables access for a short period of time so
after returning from the replay, the user access state remains blocked
and if a user page fault happens - "Bug: Read fault blocked by AMR!"
appears and SIGSEGV is sent.
An example trace:
Bug: Read fault blocked by AMR!
WARNING: CPU: 0 PID: 1603 at /home/aik/p/kernel/arch/powerpc/include/asm/book3s/64/kup-radix.h:145
CPU: 0 PID: 1603 Comm: amr Not tainted 5.10.0-rc6_v5.10-rc6_a+fstn1 #24
NIP: c00000000009ece8 LR: c00000000009ece4 CTR: 0000000000000000
REGS: c00000000dc63560 TRAP: 0700 Not tainted (5.10.0-rc6_v5.10-rc6_a+fstn1)
MSR: 8000000000021033 <SF,ME,IR,DR,RI,LE> CR: 28002888 XER: 20040000
CFAR: c0000000001fa928 IRQMASK: 1
GPR00: c00000000009ece4 c00000000dc637f0 c000000002397600 000000000000001f
GPR04: c0000000020eb318 0000000000000000 c00000000dc63494 0000000000000027
GPR08: c00000007fe4de68 c00000000dfe9180 0000000000000000 0000000000000001
GPR12: 0000000000002000 c0000000030a0000 0000000000000000 0000000000000000
GPR16: 0000000000000000 0000000000000000 0000000000000000 bfffffffffffffff
GPR20: 0000000000000000 c0000000134a4020 c0000000019c2218 0000000000000fe0
GPR24: 0000000000000000 0000000000000000 c00000000d106200 0000000040000000
GPR28: 0000000000000000 0000000000000300 c00000000dc63910 c000000001946730
NIP __do_page_fault+0xb38/0xde0
LR __do_page_fault+0xb34/0xde0
Call Trace:
__do_page_fault+0xb34/0xde0 (unreliable)
handle_page_fault+0x10/0x2c
--- interrupt: 300 at strncpy_from_user+0x290/0x440
LR = strncpy_from_user+0x284/0x440
strncpy_from_user+0x2f0/0x440 (unreliable)
getname_flags+0x88/0x2c0
do_sys_openat2+0x2d4/0x5f0
do_sys_open+0xcc/0x140
system_call_exception+0x160/0x240
system_call_common+0xf0/0x27c
To fix it save/restore the AMR when replaying interrupts, and also
add a check if AMR was not blocked prior to replaying interrupts.
Originally found by syzkaller.
Fixes: 890274c2dc ("powerpc/64s: Implement KUAP for Radix MMU")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Use normal commit citation format and add full oops log to
change log, move kuap_check_amr() into the restore routine to
avoid warnings about unreconciled IRQ state]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210202091541.36499-1-aik@ozlabs.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b709e32ef570b8b91dfbcb63cffac4324c87799f ]
When CONFIG_IRQ_TIME_ACCOUNTING and CONFIG_VIRT_CPU_ACCOUNTING_GEN, powerpc
does not enable "sched_clock_irqtime" and can not utilize irq time
accounting.
Like x86, powerpc does not use the sched_clock_register() interface. So it
needs an dedicated call to enable_sched_clock_irqtime() to enable irq time
accounting.
Fixes: 518470fe96 ("powerpc: Add HAVE_IRQ_TIME_ACCOUNTING")
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
[mpe: Add fixes tag]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1603349479-26185-1-git-send-email-kernelfans@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 24321ac668e452a4942598533d267805f291fdc9 upstream.
Commit 0138ba5783 ("powerpc/64/signal: Balance return predictor
stack in signal trampoline") changed __kernel_sigtramp_rt64() VDSO and
trampoline code, and introduced a regression in the way glibc's
backtrace()[1] detects the signal-handler stack frame. Apart from the
practical implications, __kernel_sigtramp_rt64() was a VDSO function
with the semantics that it is a function you can call from userspace
to end a signal handling. Now this semantics are no longer valid.
I believe the aforementioned change affects all releases since 5.9.
This patch tries to fix both the semantics and practical aspect of
__kernel_sigtramp_rt64() returning it to the previous code, whilst
keeping the intended behaviour of 0138ba5783 by adding a new symbol
to serve as the jump target from the kernel to the trampoline. Now the
trampoline has two parts, a new entry point and the old return point.
[1] https://lists.ozlabs.org/pipermail/linuxppc-dev/2021-January/223194.html
Fixes: 0138ba5783 ("powerpc/64/signal: Balance return predictor stack in signal trampoline")
Cc: stable@vger.kernel.org # v5.9+
Signed-off-by: Raoni Fassina Firmino <raoni@linux.ibm.com>
Acked-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Minor tweaks to change log formatting, add stable tag]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210201200505.iz46ubcizipnkcxe@work-tp
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4025c784c573cab7e3f84746cc82b8033923ec62 ]
When an asynchronous interrupt calls irq_exit, it checks for softirqs
that may have been created, and runs them. Running softirqs enables
local irqs, which can replay pending interrupts causing recursion in
replay_soft_interrupts. This abridged trace shows how this can occur:
! NIP replay_soft_interrupts
LR interrupt_exit_kernel_prepare
Call Trace:
interrupt_exit_kernel_prepare (unreliable)
interrupt_return
--- interrupt: ea0 at __rb_reserve_next
NIP __rb_reserve_next
LR __rb_reserve_next
Call Trace:
ring_buffer_lock_reserve
trace_function
function_trace_call
ftrace_call
__do_softirq
irq_exit
timer_interrupt
! replay_soft_interrupts
interrupt_exit_kernel_prepare
interrupt_return
--- interrupt: ea0 at arch_local_irq_restore
This can not be prevented easily, because softirqs must not block hard
irqs, so it has to be dealt with.
The recursion is bounded by design in the softirq code because softirq
replay disables softirqs and loops around again to check for new
softirqs created while it ran, so that's not a problem.
However it does mess up interrupt replay state, causing superfluous
interrupts when the second replay_soft_interrupts clears a pending
interrupt, leaving it still set in the first call in the 'happened'
local variable.
Fix this by not caching a copy of irqs_happened across interrupt
handler calls.
Fixes: 3282a3da25 ("powerpc/64: Implement soft interrupt replay in C")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210123061244.2076145-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 08685be7761d69914f08c3d6211c543a385a5b9c upstream.
The L1D flush fallback functions are not recoverable vs interrupts,
yet the scv entry flush runs with MSR[EE]=1. This can result in a
timer (soft-NMI) or MCE or SRESET interrupt hitting here and overwriting
the EXRFI save area, which ends up corrupting userspace registers for
scv return.
Fix this by disabling RI and EE for the scv entry fallback flush.
Fixes: f79643787e ("powerpc/64s: flush L1D on kernel entry")
Cc: stable@vger.kernel.org # 5.9+ which also have flush L1D patch backport
Reported-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210111062408.287092-1-npiggin@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 2225a8dda263edc35a0e8b858fe2945cf6240fde ]
This is a bug that causes early crashes in builds with an .exit.text
section smaller than a page and an .init.text section that ends in the
beginning of a physical page (this is kinda random, which might
explain why this wasn't really encountered before).
The init sections are ordered like this:
.init.text
.exit.text
.init.data
Currently, these sections aren't page aligned.
Because the init code might become read-only at runtime and because
the .init.text section can potentially reside on the same physical
page as .init.data, the beginning of .init.data might be mapped
read-only along with .init.text.
Then when the kernel tries to modify a variable in .init.data (like
kthreadd_done, used in kernel_init()) the kernel panics.
To avoid this, make _einittext page aligned and also align .exit.text
to make sure .init.data is always seperated from the text segments.
Fixes: 060ef9d89d ("powerpc32: PAGE_EXEC required for inittext")
Signed-off-by: Ariel Marcovitch <ariel.marcovitch@gmail.com>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210102201156.10805-1-ariel.marcovitch@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
