Commit a21d1becaa ("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: a21d1becaa ("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
When we boot from open firmware (OF) using PPC_OF_BOOT_TRAMPOLINE, aka.
prom_init, we run parts of the kernel at an address other than the link
address. That happens because OF loads the kernel above zero (OF is at
zero) and we run prom_init before copying the kernel down to zero.
Currently that works even for non-relocatable kernels, because we do
various fixups to the prom_init code to make it run where it's loaded.
However those fixups are not sufficient if the kernel becomes large
enough. In that case prom_init()'s final call to __start() can end up
generating a plt branch:
bl c000000002000018 <00000078.plt_branch.__start>
That results in the kernel jumping to the linked address of __start,
0xc000000000000000, when really it needs to jump to the
0xc000000000000000 + the runtime address because the kernel is still
running at the load address.
We could do further shenanigans to handle that, see Jordan's patch for
example:
https://lore.kernel.org/linuxppc-dev/20210421021721.1539289-1-jniethe5@gmail.com
However it is much simpler to just require a kernel with prom_init() to
be built relocatable. The result works in all configurations without
further work, and requires less code.
This should have no effect on most people, as our defconfigs and
essentially all distro configs already have RELOCATABLE enabled.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210623130454.2542945-1-mpe@ellerman.id.au
It is sometimes desirable to run a command on all cpus in xmon. A
typical scenario is to obtain the backtrace from all cpus in xmon if
there is a soft lockup. Add rudimentary support for the same. The
command to be run on all cpus should be prefixed with 'c#'. As an
example, 'c#t' will run 't' command and produce a backtrace on all cpus
in xmon.
Since many xmon commands are not sensible for running in this manner, we
only allow a predefined list of commands -- 'r', 'S' and 't' for now.
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/20210601074801.617363-1-naveen.n.rao@linux.vnet.ibm.com
Persistent memory devices like NVDIMMs can loose cached writes in case
something prevents flush on power-fail. Such situations are termed as
dirty shutdown and are exposed to applications as
last-shutdown-state (LSS) flag and a dirty-shutdown-counter(DSC) as
described at [1]. The latter being useful in conditions where multiple
applications want to detect a dirty shutdown event without racing with
one another.
PAPR-NVDIMMs have so far only exposed LSS style flags to indicate a
dirty-shutdown-state. This patch further adds support for DSC via the
"ibm,persistence-failed-count" device tree property of an NVDIMM. This
property is a monotonic increasing 64-bit counter thats an indication
of number of times an NVDIMM has encountered a dirty-shutdown event
causing persistence loss.
Since this value is not expected to change after system-boot hence
papr_scm reads & caches its value during NVDIMM probe and exposes it
as a PAPR sysfs attributed named 'dirty_shutdown' to match the name of
similarly named NFIT sysfs attribute. Also this value is available to
libnvdimm via PAPR_PDSM_HEALTH payload. 'struct nd_papr_pdsm_health'
has been extended to add a new member called 'dimm_dsc' presence of
which is indicated by the newly introduced PDSM_DIMM_DSC_VALID flag.
References:
[1] https://pmem.io/documents/Dirty_Shutdown_Handling-V1.0.pdf
Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Reviewed-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/20210624080621.252038-1-vaibhav@linux.ibm.com
In case performance stats for an nvdimm are not available, reading the
'perf_stats' sysfs file returns an -ENOENT error. A better approach is
to make the 'perf_stats' file entirely invisible to indicate that
performance stats for an nvdimm are unavailable.
So this patch updates 'papr_nd_attribute_group' to add a 'is_visible'
callback implemented as newly introduced 'papr_nd_attribute_visible()'
that returns an appropriate mode in case performance stats aren't
supported in a given nvdimm.
Also the initialization of 'papr_scm_priv.stat_buffer_len' is moved
from papr_scm_nvdimm_init() to papr_scm_probe() so that it value is
available when 'papr_nd_attribute_visible()' is called during nvdimm
initialization.
Even though 'perf_stats' attribute is available since v5.9, there are
no known user-space tools/scripts that are dependent on presence of its
sysfs file. Hence I dont expect any user-space breakage with this
patch.
Fixes: 2d02bf835e ("powerpc/papr_scm: Fetch nvdimm performance stats from PHYP")
Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210513092349.285021-1-vaibhav@linux.ibm.com
Trying to use a kprobe on ppc32 results in the below splat:
BUG: Unable to handle kernel data access on read at 0x7c0802a6
Faulting instruction address: 0xc002e9f0
Oops: Kernel access of bad area, sig: 11 [#1]
BE PAGE_SIZE=4K PowerPC 44x Platform
Modules linked in:
CPU: 0 PID: 89 Comm: sh Not tainted 5.13.0-rc1-01824-g3a81c0495fdb #7
NIP: c002e9f0 LR: c0011858 CTR: 00008a47
REGS: c292fd50 TRAP: 0300 Not tainted (5.13.0-rc1-01824-g3a81c0495fdb)
MSR: 00009000 <EE,ME> CR: 24002002 XER: 20000000
DEAR: 7c0802a6 ESR: 00000000
<snip>
NIP [c002e9f0] emulate_step+0x28/0x324
LR [c0011858] optinsn_slot+0x128/0x10000
Call Trace:
opt_pre_handler+0x7c/0xb4 (unreliable)
optinsn_slot+0x128/0x10000
ret_from_syscall+0x0/0x28
The offending instruction is:
81 24 00 00 lwz r9,0(r4)
Here, we are trying to load the second argument to emulate_step():
struct ppc_inst, which is the instruction to be emulated. On ppc64,
structures are passed in registers when passed by value. However, per
the ppc32 ABI, structures are always passed to functions as pointers.
This isn't being adhered to when setting up the call to emulate_step()
in the optprobe trampoline. Fix the same.
Fixes: eacf4c0202 ("powerpc: Enable OPTPROBES on PPC32")
Cc: stable@vger.kernel.org
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.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/5bdc8cbc9a95d0779e27c9ddbf42b40f51f883c0.1624425798.git.christophe.leroy@csgroup.eu
To remove code duplication, use the binary stats descriptors in the
implementation of the debugfs interface for statistics. This unifies
the definition of statistics for the binary and debugfs interfaces.
Signed-off-by: Jing Zhang <jingzhangos@google.com>
Message-Id: <20210618222709.1858088-8-jingzhangos@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This commit defines the API for userspace and prepare the common
functionalities to support per VM/VCPU binary stats data readings.
The KVM stats now is only accessible by debugfs, which has some
shortcomings this change series are supposed to fix:
1. The current debugfs stats solution in KVM could be disabled
when kernel Lockdown mode is enabled, which is a potential
rick for production.
2. The current debugfs stats solution in KVM is organized as "one
stats per file", it is good for debugging, but not efficient
for production.
3. The stats read/clear in current debugfs solution in KVM are
protected by the global kvm_lock.
Besides that, there are some other benefits with this change:
1. All KVM VM/VCPU stats can be read out in a bulk by one copy
to userspace.
2. A schema is used to describe KVM statistics. From userspace's
perspective, the KVM statistics are self-describing.
3. With the fd-based solution, a separate telemetry would be able
to read KVM stats in a less privileged environment.
4. After the initial setup by reading in stats descriptors, a
telemetry only needs to read the stats data itself, no more
parsing or setup is needed.
Reviewed-by: David Matlack <dmatlack@google.com>
Reviewed-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Fuad Tabba <tabba@google.com>
Tested-by: Fuad Tabba <tabba@google.com> #arm64
Signed-off-by: Jing Zhang <jingzhangos@google.com>
Message-Id: <20210618222709.1858088-3-jingzhangos@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Generic KVM stats are those collected in architecture independent code
or those supported by all architectures; put all generic statistics in
a separate structure. This ensures that they are defined the same way
in the statistics API which is being added, removing duplication among
different architectures in the declaration of the descriptors.
No functional change intended.
Reviewed-by: David Matlack <dmatlack@google.com>
Reviewed-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Jing Zhang <jingzhangos@google.com>
Message-Id: <20210618222709.1858088-2-jingzhangos@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
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
Commit aaa2295292 ("powerpc/mm: Add physical address to Linux page
table dump") changed range coalescing to only combine ranges that are
both virtually and physically contiguous, in order to avoid erroneous
combination of unrelated mappings in IOREMAP space.
But in the VMALLOC space, mappings almost never have contiguous
physical pages, so the commit mentionned above leads to dumping one
line per page for vmalloc mappings.
Taking into account the vmalloc always leave a gap between two areas,
we never have two mappings dumped as a single combination even if they
have the exact same flags. The only space that may have encountered
such an issue was the early IOREMAP which is not using vmalloc engine.
But previous commits added gaps between early IO mappings, so it is
not an issue anymore.
That commit created some difficulties with KASAN mappings, see
commit cabe8138b2 ("powerpc: dump as a single line areas mapping a
single physical page.") and with huge page, see
commit b00ff6d8c1 ("powerpc/ptdump: Properly handle non standard
page size").
So, almost revert commit aaa2295292 to properly coalesce pages
mapped with the same flags as before, only keep the display of the
first physical address of the range, as it can be usefull especially
for IO mappings.
It brings back powerpc at the same level as other architectures and
simplifies the conversion to GENERIC PTDUMP.
With the patch:
---[ kasan shadow mem start ]---
0xf8000000-0xf8ffffff 0x07000000 16M huge rw present dirty accessed
0xf9000000-0xf91fffff 0x01434000 2M r present accessed
0xf9200000-0xf95affff 0x02104000 3776K rw present dirty accessed
0xfef5c000-0xfeffffff 0x01434000 656K r present accessed
---[ kasan shadow mem end ]---
Before:
---[ kasan shadow mem start ]---
0xf8000000-0xf8ffffff 0x07000000 16M huge rw present dirty accessed
0xf9000000-0xf91fffff 0x01434000 16K r present accessed
0xf9200000-0xf9203fff 0x02104000 16K rw present dirty accessed
0xf9204000-0xf9207fff 0x0213c000 16K rw present dirty accessed
0xf9208000-0xf920bfff 0x02174000 16K rw present dirty accessed
0xf920c000-0xf920ffff 0x02188000 16K rw present dirty accessed
0xf9210000-0xf9213fff 0x021dc000 16K rw present dirty accessed
0xf9214000-0xf9217fff 0x02220000 16K rw present dirty accessed
0xf9218000-0xf921bfff 0x023c0000 16K rw present dirty accessed
0xf921c000-0xf921ffff 0x023d4000 16K rw present dirty accessed
0xf9220000-0xf9227fff 0x023ec000 32K rw present dirty accessed
...
0xf93b8000-0xf93e3fff 0x02614000 176K rw present dirty accessed
0xf93e4000-0xf94c3fff 0x027c0000 896K rw present dirty accessed
0xf94c4000-0xf94c7fff 0x0236c000 16K rw present dirty accessed
0xf94c8000-0xf94cbfff 0x041f0000 16K rw present dirty accessed
0xf94cc000-0xf94cffff 0x029c0000 16K rw present dirty accessed
0xf94d0000-0xf94d3fff 0x041ec000 16K rw present dirty accessed
0xf94d4000-0xf94d7fff 0x0407c000 16K rw present dirty accessed
0xf94d8000-0xf94f7fff 0x041c0000 128K rw present dirty accessed
...
0xf95ac000-0xf95affff 0x042b0000 16K rw present dirty accessed
0xfef5c000-0xfeffffff 0x01434000 16K r present accessed
---[ kasan shadow mem end ]---
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c56ce1f5c3c75adc9811b1a5f9c410fa74183a8d.1618828806.git.christophe.leroy@csgroup.eu
The validation done at the start of dlpar_memory_add_by_ic() is an all
of nothing scenario - if any LMBs in the range is marked as RESERVED we
can fail right away.
We then can remove the 'lmbs_available' var and its check with
'lmbs_to_add' since the whole LMB range was already validated in the
previous step.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210622133923.295373-4-danielhb413@gmail.com
After a successful dlpar_add_lmb() call the LMB is marked as reserved.
Later on, depending whether we added enough LMBs or not, we rely on
the marked LMBs to see which ones might need to be removed, and we
remove the reservation of all of them.
These are done in for_each_drmem_lmb() loops without any break
condition. This means that we're going to check all LMBs of the partition
even after going through all the reserved ones.
This patch adds break conditions in both loops to avoid this. The
'lmbs_added' variable was renamed to 'lmbs_reserved', and it's now
being decremented each time a lmb reservation is removed, indicating
if there are still marked LMBs to be processed.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210622133923.295373-3-danielhb413@gmail.com
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
PowerVM will not arbitrarily oversubscribe or stop guests, page out the
guest kernel text to a NFS volume connected by carrier pigeon to abacus
based storage, etc., as a KVM host might. So PowerVM guests are not
likely to be killed by the hard lockup watchdog in normal operation,
even with shared processor LPARs which still get a minimum allotment of
CPU time.
Enable the hard lockup detector by default on !KVM guests, which we will
assume is PowerVM. It has been useful in finding problems on bare metal
kernels.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210623021528.702241-1-npiggin@gmail.com
The PPC_RFI_SRR_DEBUG check added by patch "powerpc/64s: avoid reloading
(H)SRR registers if they are still valid" has a few deficiencies. It
does not fix the actual problem, it's not enabled by default, and it
causes a program check interrupt which can cause more difficulties.
However there are a lot of paths which may clobber SRRs or change return
regs, and difficult to have a high confidence that all paths are covered
without wider testing.
Add a relatively low overhead always-enabled check that catches most
such cases, reports once, and fixes it so the kernel can continue.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Rebase, use switch & INT names, squash in race fix from Nick]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Use the restart table facility to return from interrupt or system calls
without disabling MSR[EE] or MSR[RI].
Interrupt return asm is put into the low soft-masked region, to prevent
interrupts being processed here, although they are still taken as masked
interrupts which causes SRRs to be clobbered, and a pending soft-masked
interrupt to require replaying.
The return code uses restart table regions to redirct to a fixup handler
rather than continue with the exit, if such an interrupt happens. In
this case the interrupt return is redirected to a fixup handler which
reloads r1 for the interrupt stack and reloads registers and sets state
up to replay the soft-masked interrupt and try the exit again.
Some types of security exit fallback flushes and barriers are currently
unable to cope with reentrant interrupts, e.g., because they store some
state in the scratch SPR which would be clobbered even by masked
interrupts. For now the interrupts-enabled exits are disabled when these
flushes are used.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Guard unused exit_must_hard_disable() as reported by lkp]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210617155116.2167984-13-npiggin@gmail.com
Prevent interrupt restore from allowing racing hard interrupts going
ahead of previous soft-pending ones, by using the soft-masked restart
handler to allow a store to clear the soft-mask while knowing nothing
is soft-pending.
This probably doesn't matter much in practice, but it's a simple
demonstrator / test case to exercise the restart table logic.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210617155116.2167984-11-npiggin@gmail.com
The exception table fixup adjusts a failed page fault's interrupt return
location if it was taken at an address specified in the exception table,
to a corresponding fixup handler address.
Introduce a variation of that idea which adds a fixup table for NMIs and
soft-masked asynchronous interrupts. This will be used to protect
certain critical sections that are sensitive to being clobbered by
interrupts coming in (due to using the same SPRs and/or irq soft-mask
state).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210617155116.2167984-10-npiggin@gmail.com
When an interrupt is taken, the SRR registers are set to return to where
it left off. Unless they are modified in the meantime, or the return
address or MSR are modified, there is no need to reload these registers
when returning from interrupt.
Introduce per-CPU flags that track the validity of SRR and HSRR
registers. These are cleared when returning from interrupt, when
using the registers for something else (e.g., OPAL calls), when
adjusting the return address or MSR of a context, and when context
switching (which changes the return address and MSR).
This improves the performance of interrupt returns.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Fold in fixup patch from Nick]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210617155116.2167984-5-npiggin@gmail.com
Pass the value of linux_banner to firmware via option vector 7.
Option vector 7 is described in "LoPAR" Linux on Power Architecture
Reference v2.9, in table B.7 on page 824:
An ASCII character formatted null terminated string that describes
the client operating system. The string shall be human readable and
may be displayed on the console.
The string can be up to 256 bytes total, including the nul terminator.
linux_banner contains lots of information, and should make it possible
to identify the exact kernel version that is running:
const char linux_banner[] =
"Linux version " UTS_RELEASE " (" LINUX_COMPILE_BY "@"
LINUX_COMPILE_HOST ") (" LINUX_COMPILER ") " UTS_VERSION "\n";
For example:
Linux version 4.15.0-144-generic (buildd@bos02-ppc64el-018) (gcc
version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)) #148-Ubuntu SMP Sat May 8
02:32:13 UTC 2021 (Ubuntu 4.15.0-144.148-generic 4.15.18)
It's also printed at boot to the console/dmesg, which should make it
possible to correlate what firmware receives with the console/dmesg on
the machine.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210621064938.2021419-2-mpe@ellerman.id.au
