Hari Bathini [Thu, 9 May 2024 11:57:54 +0000 (17:27 +0530)]
powerpc/fadump: setup additional parameters for dump capture kernel
For fadump case, passing additional parameters to dump capture kernel
helps in minimizing the memory footprint for it and also provides the
flexibility to disable components/modules, like hugepages, that are
hindering the boot process of the special dump capture environment.
Set up a dedicated parameter area to be passed to the capture kernel.
This area type is defined as RTAS_FADUMP_PARAM_AREA. Sysfs attribute
'/sys/kernel/fadump/bootargs_append' is exported to the userspace to
specify the additional parameters to be passed to the capture kernel
Hari Bathini [Thu, 9 May 2024 11:57:53 +0000 (17:27 +0530)]
powerpc/pseries/fadump: add support for multiple boot memory regions
Currently, fadump on pseries assumes a single boot memory region even
though f/w supports more than one boot memory region. Add support for
more boot memory regions to make the implementation flexible for any
enhancements that introduce other region types. For this, rtas memory
structure for fadump is updated to have multiple boot memory regions
instead of just one. Additionally, methods responsible for creating
the fadump memory structure during both the first and second kernel
boot have been modified to take these multiple boot memory regions
into account. Also, a new callback has been added to the fadump_ops
structure to get the maximum boot memory regions supported by the
platform.
Greg Kurz [Tue, 9 Mar 2021 18:11:10 +0000 (19:11 +0100)]
powerpc/xmon: Check cpu id in commands "c#", "dp#" and "dx#"
All these commands end up peeking into the PACA using the user
originated cpu id as an index. Check the cpu id is valid in order
to prevent xmon to crash. Instead of printing an error, this follows
the same behavior as the "lp s #" command : ignore the buggy cpu id
parameter and fall back to the #-less version of the command.
Benjamin Gray [Mon, 25 Mar 2024 05:28:15 +0000 (16:28 +1100)]
powerpc/code-patching: Use dedicated memory routines for patching
The patching page set up as a writable alias may be in quadrant 0
(userspace) if the temporary mm path is used. This causes sanitiser
failures if so. Sanitiser failures also occur on the non-mm path
because the plain memset family is instrumented, and KASAN treats the
patching window as poisoned.
Introduce locally defined patch_* variants of memset that perform an
uninstrumented lower level set, as well as detecting write errors like
the original single patch variant does.
copy_to_user() is not correct here, as the PTE makes it a proper kernel
page (the EAA is privileged access only, RW). It just happens to be in
quadrant 0 because that's the hardware's mechanism for using the current
PID vs PID 0 in translations. Importantly, it's incorrect to allow user
page accesses.
Now that the patching memsets are used, we also propagate a failure up
to the caller as the single patch variant does.
Benjamin Gray [Mon, 12 Feb 2024 04:50:20 +0000 (15:50 +1100)]
powerpc64/kasan: Pass virtual addresses to kasan_init_phys_region()
The kasan_init_phys_region() function maps shadow pages necessary for
the ranges of the linear map backed by physical pages. Currently
kasan_init_phys_region() is being passed physical addresses, but
kasan_mem_to_shadow() expects virtual addresses.
It works right now because the lower bits (12:64) of the
kasan_mem_to_shadow() calculation are the same for the real and virtual
addresses, so the actual PTE value is the same in the end. But virtual
addresses are the intended input, so fix it.
powerpc: rename SPRN_HID2 define to SPRN_HID2_750FX
This register number is hardware-specific, rename it for clarity.
FIXME comments are added in a few places where it seems like the wrong
register is used. As I can't test this, only the rename is done with no
functional change.
If both CONFIG_SENSORS_AMS_PMU and CONFIG_SENSORS_AMS_I2C are unset,
there is an unused variable warning in the ams driver:
drivers/macintosh/ams/ams-core.c: In function 'ams_init':
drivers/macintosh/ams/ams-core.c:181:29: warning: unused variable 'np'
181 | struct device_node *np;
The driver needs at least one of the configs enabled in order to
actually function. So fix the compiler warning by ensuring at least one
of the configs is enabled.
Naveen N Rao [Wed, 10 Jan 2024 14:12:37 +0000 (19:42 +0530)]
powerpc/Makefile: Remove bits related to the previous use of -mcmodel=large
All supported compilers today (gcc v5.1+ and clang v11+) have support for
-mcmodel=medium. As such, NO_MINIMAL_TOC is no longer being set. Remove
NO_MINIMAL_TOC as well as the fallback to -mminimal-toc.
GUO Zihua [Sat, 13 Jan 2024 08:05:09 +0000 (08:05 +0000)]
powerpc: Fix preserved memory size for int-vectors
The first 32k of memory is reserved for interrupt vectors, however for
powerpc64 this might not be enough. Fix this by reserving the maximum
size between 32k and the real size of interrupt vectors.
Xiaowei Bao [Fri, 19 Jan 2024 20:38:55 +0000 (15:38 -0500)]
powerpc: dts: p1010rdb: fix INTx interrupt issue on P1010RDB-PB
Due to the INTA is shared with the active-low PHY2 interrupt on
P1010RDB-PA board, so configure P1010RDB-PA's INTA with polarity as
active-low, the P1010RDB-PB board is used separately, so configure
P1010RDB-PB's INTA with polarity as active-high. The INTX in
P1010RDB-PB do not work because of the pcie@0 node fixup will be
overwrited by p1010si-post.dtsi file, so we move the pcie@0 node fixup
to p1010rdb-pb.dts and p1010rdb-pb_36b.dts.
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com> Signed-off-by: Li Yang <leoyang.li@nxp.com> Signed-off-by: Frank Li <Frank.Li@nxp.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240119203911.3143928-2-Frank.Li@nxp.com
Ran Wang [Fri, 19 Jan 2024 20:38:54 +0000 (15:38 -0500)]
powerpc: dts: add power management nodes to FSL chips
Enable Power Management feature on device tree, including MPC8536,
MPC8544, MPC8548, MPC8572, P1010, P1020, P1021, P1022, P2020, P2041,
P3041, T104X, T1024.
Signed-off-by: Zhao Chenhui <chenhui.zhao@freescale.com> Signed-off-by: Ran Wang <ran.wang_1@nxp.com> Signed-off-by: Frank Li <Frank.Li@nxp.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240119203911.3143928-1-Frank.Li@nxp.com
powerpc/64: Set _IO_BASE to POISON_POINTER_DELTA not 0 for CONFIG_PCI=n
There is code that builds with calls to IO accessors even when
CONFIG_PCI=n, but the actual calls are guarded by runtime checks.
If not those calls would be faulting, because the page at virtual
address zero is (usually) not mapped into the kernel. As Arnd pointed
out, it is possible a large port value could cause the address to be
above mmap_min_addr which would then access userspace, which would be
a bug.
To avoid any such issues, set _IO_BASE to POISON_POINTER_DELTA. That
is a value chosen to point into unmapped space between the kernel and
userspace, so any access will always fault.
Note that on 32-bit POISON_POINTER_DELTA is 0, so the patch only has an
effect on 64-bit.
With -Wextra clang warns about pointer arithmetic using a null pointer.
When building with CONFIG_PCI=n, that triggers a warning in the IO
accessors, eg:
In file included from linux/arch/powerpc/include/asm/io.h:672:
linux/arch/powerpc/include/asm/io-defs.h:23:1: warning: performing pointer arithmetic on a null pointer has undefined behavior [-Wnull-pointer-arithmetic]
23 | DEF_PCI_AC_RET(inb, u8, (unsigned long port), (port), pio, port)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...
linux/arch/powerpc/include/asm/io.h:591:53: note: expanded from macro '__do_inb'
591 | #define __do_inb(port) readb((PCI_IO_ADDR)_IO_BASE + port);
| ~~~~~~~~~~~~~~~~~~~~~ ^
That is because when CONFIG_PCI=n, _IO_BASE is defined as 0.
Although _IO_BASE is defined as plain 0, the cast (PCI_IO_ADDR) converts
it to void * before the addition with port happens.
Instead the addition can be done first, and then the cast. The resulting
value will be the same, but avoids the warning, and also avoids void
pointer arithmetic which is apparently non-standard.
Hari Bathini [Thu, 2 May 2024 17:32:05 +0000 (23:02 +0530)]
powerpc/bpf: enable kfunc call
Currently, bpf jit code on powerpc assumes all the bpf functions and
helpers to be part of core kernel text. This is false for kfunc case,
as function addresses may not be part of core kernel text area. So,
add support for addresses that are not within core kernel text area
too, to enable kfunc support. Emit instructions based on whether the
function address is within core kernel text address or not, to retain
optimized instruction sequence where possible.
In case of PCREL, as a bpf function that is not within core kernel
text area is likely to go out of range with relative addressing on
kernel base, use PC relative addressing. If that goes out of range,
load the full address with PPC_LI64().
With addresses that are not within core kernel text area supported,
override bpf_jit_supports_kfunc_call() to enable kfunc support. Also,
override bpf_jit_supports_far_kfunc_call() to enable 64-bit pointers,
as an address offset can be more than 32-bit long on PPC64.
Hari Bathini [Thu, 2 May 2024 17:32:04 +0000 (23:02 +0530)]
powerpc/64/bpf: fix tail calls for PCREL addressing
With PCREL addressing, there is no kernel TOC. So, it is not setup in
prologue when PCREL addressing is used. But the number of instructions
to skip on a tail call was not adjusted accordingly. That resulted in
not so obvious failures while using tailcalls. 'tailcalls' selftest
crashed the system with the below call trace:
Also, as bpf programs are always module addresses and a bpf helper in
general is a core kernel text address, using PC relative addressing
often fails with "out of range of pcrel address" error. Switch to
using kernel base for relative addressing to handle this better.
Fixes: 7e3a68be42e1 ("powerpc/64: vmlinux support building with PCREL addresing") Cc: stable@vger.kernel.org # v6.4+ Signed-off-by: Hari Bathini <hbathini@linux.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240502173205.142794-1-hbathini@linux.ibm.com
Benjamin Gray [Wed, 17 Apr 2024 11:23:20 +0000 (21:23 +1000)]
powerpc/dexcr: Add DEXCR prctl interface
Now that we track a DEXCR on a per-task basis, individual tasks are free
to configure it as they like.
The interface is a pair of getter/setter prctl's that work on a single
aspect at a time (multiple aspects at once is more difficult if there
are different rules applied for each aspect, now or in future). The
getter shows the current state of the process config, and the setter
allows setting/clearing the aspect.
Benjamin Gray [Wed, 17 Apr 2024 11:23:19 +0000 (21:23 +1000)]
powerpc/dexcr: Reset DEXCR value across exec
Inheriting the DEXCR across exec can have security and usability
concerns. If a program is compiled with hash instructions it generally
expects to run with NPHIE enabled. But if the parent process disables
NPHIE then if it's not careful it will be disabled for any children too
and the protection offered by hash checks is basically worthless.
This patch introduces a per-process reset value that new execs in a
particular process tree are initialized with. This enables fine grained
control over what DEXCR value child processes run with by default.
For example, containers running legacy binaries that expect hash
instructions to act as NOPs could configure the reset value of the
container root to control the default reset value for all members of
the container.
Benjamin Gray [Wed, 17 Apr 2024 11:23:18 +0000 (21:23 +1000)]
powerpc/dexcr: Track the DEXCR per-process
Add capability to make the DEXCR act as a per-process SPR.
We do not yet have an interface for changing the values per task. We
also expect the kernel to use a single DEXCR value across all tasks
while in privileged state, so there is no need to synchronize after
changing it (the userspace aspects will synchronize upon returning to
userspace).
Benjamin Gray [Wed, 17 Apr 2024 11:23:17 +0000 (21:23 +1000)]
selftests/powerpc/dexcr: Add -no-pie to hashchk tests
The hashchk tests want to verify that the hash key is changed over exec.
It does so by calculating hashes at the same address across an exec.
This is made simpler by disabling PIE functionality, so we can
re-execute ourselves and be using the same addresses in the child.
While -fno-pie is already added, -no-pie is also required.
powerpc/module: Remove arch specific module bug stuff
The last function to reference module_bug_list went in 2008's
commit b9754568ef17 ("powerpc: Remove dead module_find_bug code")
but I don't think that was called since 2006's
commit 73c9ceab40b1 ("[POWERPC] Generic BUG for powerpc")
Now that the list has gone, I think we can also clean up the bug
entries in mod_arch_specific.
Michael Ellerman [Mon, 22 Apr 2024 13:34:53 +0000 (23:34 +1000)]
selftests/powerpc: Install tests in sub-directories
The sources for the powerpc selftests are arranged into sub-directories.
However when the tests are built and installed, the sub-directories are
squashed, losing the structure.
For example, with the current code the result of installing the selftests is:
All the powerpc tests are squashed into the single powerpc directory. In
particular, note that there is a single `settings` file, even though
there are multiple settings files in the powerpc selftest sources. One
of the settings files ends up installed, depending on install order,
even if they have different contents.
Similarly if there were two tests with the same name in different
sub-directories they would clobber each other.
Fix it by replicating the directory structure of the source tree into
the install directory. The result being for example:
selftests/powerpc: make sub-folders buildable on their own
Build breaks when executing make with run_tests for sub-folders
under powerpc. This is because, CFLAGS and GIT_VERSION macros are
defined in Makefile of toplevel powerpc folder.
make: Entering directory '/home/maddy/linux/tools/testing/selftests/powerpc/mm'
gcc hugetlb_vs_thp_test.c ../harness.c ../utils.c -o /home/maddy/selftest_output//hugetlb_vs_thp_test
hugetlb_vs_thp_test.c:6:10: fatal error: utils.h: No such file or directory
6 | #include "utils.h"
| ^~~~~~~~~
compilation terminated.
Fix this by adding the flags.mk in each sub-folder Makefile. Also remove
the CFLAGS and GIT_VERSION macros from powerpc/ folder Makefile since
the same is definied in flags.mk
selftests/powerpc: Add flags.mk to support pmu buildable
When running `make -C powerpc/pmu run_tests` from top level selftests
directory, currently this error is being reported:
make: Entering directory '/home/maddy/linux/tools/testing/selftests/powerpc/pmu'
Makefile:40: warning: overriding recipe for target 'emit_tests'
../../lib.mk:111: warning: ignoring old recipe for target 'emit_tests'
gcc -m64 count_instructions.c ../harness.c event.c lib.c ../utils.c loop.S -o /home/maddy/selftest_output//count_instructions
In file included from count_instructions.c:13:
event.h:12:10: fatal error: utils.h: No such file or directory
12 | #include "utils.h"
| ^~~~~~~~~
compilation terminated.
This is due to missing of include path in CFLAGS. That is, CFLAGS and
GIT_VERSION macros are defined in the powerpc/ folder Makefile which
in this case is not involved.
To address the failure in case of executing specific sub-folder test
directly, a new rule file has been addded by the patch called "flags.mk"
under selftest/powerpc/ folder and is linked to all the Makefile of
powerpc/pmu sub-folders.
Reported-by: Sachin Sant <sachinp@linux.ibm.com> Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com> Tested-by: Sachin Sant <sachinp@linux.ibm.com>
[mpe: Fixup ifeq, make GIT_VERSION simply expanded to avoid re-executing git describe] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240229093711.581230-2-maddy@linux.ibm.com
selftests/powerpc: Re-order *FLAGS to follow lib.mk
In some powerpc/ sub-folder Makefiles, CFLAGS are defined before lib.mk
include. Clean it up by re-ordering the flags to follow after the mk
include. This is needed to support sub-folders in powerpc/ buildable on
its own.
powerpc/pseries/vio: Don't return ENODEV if node or compatible missing
We noticed the following nuisance messages during boot process:
vio vio: uevent: failed to send synthetic uevent
vio 4000: uevent: failed to send synthetic uevent
vio 4001: uevent: failed to send synthetic uevent
vio 4002: uevent: failedto send synthetic uevent
vio 4004: uevent: failed to send synthetic uevent
It's caused by either vio_register_device_node() failing to set
dev->of_node or the node is missing a "compatible" property. To match
the definition of modalias in modalias_show(), remove the return of
ENODEV in such cases. The failure messages is also suppressed with this
change.
powerpc/pseries: Enforce hcall result buffer validity and size
plpar_hcall(), plpar_hcall9(), and related functions expect callers to
provide valid result buffers of certain minimum size. Currently this
is communicated only through comments in the code and the compiler has
no idea.
For example, if I write a bug like this:
long retbuf[PLPAR_HCALL_BUFSIZE]; // should be PLPAR_HCALL9_BUFSIZE
plpar_hcall9(H_ALLOCATE_VAS_WINDOW, retbuf, ...);
This compiles with no diagnostics emitted, but likely results in stack
corruption at runtime when plpar_hcall9() stores results past the end
of the array. (To be clear this is a contrived example and I have not
found a real instance yet.)
To make this class of error less likely, we can use explicitly-sized
array parameters instead of pointers in the declarations for the hcall
APIs. When compiled with -Warray-bounds[1], the code above now
provokes a diagnostic like this:
error: array argument is too small;
is of size 32, callee requires at least 72 [-Werror,-Warray-bounds]
60 | plpar_hcall9(H_ALLOCATE_VAS_WINDOW, retbuf,
| ^ ~~~~~~
[1] Enabled for LLVM builds but not GCC for now. See commit 0da6e5fd6c37 ("gcc: disable '-Warray-bounds' for gcc-13 too") and
related changes.
Which he bisected to a change in kmemleak, commit 23c2d497de21 ("mm: kmemleak: take a full lowmem check in kmemleak_*_phys()").
Because pmac_probe() is called before mem_topology_setup(), the min/
max PFN variables are still zero. That causes kmemleak_alloc_phys() to
ignore the allocation, because the checks against the PFN fail. Then
kmemleak_no_scan() can't find the allocation and prints warning.
Given that kmemleak_alloc_phys() is ignoring the allocation to begin
with, there's no need to call kmemleak_no_scan() at all, which avoids
the warning.
powerpc/eeh: Permanently disable the removed device
When a device is hot removed on powernv, the hotplug driver clears
the device's state. However, on pseries, if a device is removed by
phyp after reaching the error threshold, the kernel remains unaware,
leading to the device not being torn down. This prevents necessary
remediation actions like failover.
Permanently disable the device if the presence check fails.
Also, in eeh_dev_check_failure in we may consider the error as false
positive if the device is hotpluged out as the get_state call returns
EEH_STATE_NOT_SUPPORT and we may end up not clearing the device state,
so log the event if the state is not moved to permanent failure state.
The patch titled ("powerpc: make fadump resilient with memory add/remove
events") has made significant changes to the implementation of fadump,
particularly on elfcorehdr creation and fadump crash info header
structure. Therefore, updating the fadump implementation documentation
to reflect those changes.
Following updates are done to firmware assisted dump documentation:
1. The elfcorehdr is no longer stored after fadump HDR in the reserved
dump area. Instead, the second kernel dynamically allocates memory
for the elfcorehdr within the address range from 0 to the boot memory
size. Therefore, update figures 1 and 2 of Memory Reservation during
the first and second kernels to reflect this change.
2. A version field has been added to the fadump header to manage the
future changes to fadump crash info header structure without changing
the fadump header magic number in the future. Therefore, remove the
corresponding TODO from the document.
The elfcorehdr describes the CPUs and memory of the crashed kernel to
the kernel that captures the dump, known as the second or fadump kernel.
The elfcorehdr needs to be updated if the system's memory changes due to
memory hotplug or online/offline events.
Currently, memory hotplug events are monitored in userspace by udev
rules, and fadump is re-registered, which recreates the elfcorehdr with
the latest available memory in the system.
However, the previous patch ("powerpc: make fadump resilient with memory
add/remove events") moved the creation of elfcorehdr to the second or
fadump kernel. This eliminates the need to regenerate the elfcorehdr
during memory hotplug or online/offline events.
Create a sysfs entry at /sys/kernel/fadump/hotplug_ready to let
userspace know that fadump re-registration is not required for memory
add/remove events.
powerpc: make fadump resilient with memory add/remove events
Due to changes in memory resources caused by either memory hotplug or
online/offline events, the elfcorehdr, which describes the CPUs and
memory of the crashed kernel to the kernel that collects the dump (known
as second/fadump kernel), becomes outdated. Consequently, attempting
dump collection with an outdated elfcorehdr can lead to failed or
inaccurate dump collection.
Memory hotplug or online/offline events is referred as memory add/remove
events in reset of the commit message.
The current solution to address the aforementioned issue is as follows:
Monitor memory add/remove events in userspace using udev rules, and
re-register fadump whenever there are changes in memory resources. This
leads to the creation of a new elfcorehdr with updated system memory
information.
There are several notable issues associated with re-registering fadump
for every memory add/remove events.
1. Bulk memory add/remove events with udev-based fadump re-registration
can lead to race conditions and, more importantly, it creates a wide
window during which fadump is inactive until all memory add/remove
events are settled.
2. Re-registering fadump for every memory add/remove event is
inefficient.
3. The memory for elfcorehdr is allocated based on the memblock regions
available during early boot and remains fixed thereafter. However, if
elfcorehdr is later recreated with additional memblock regions, its
size will increase, potentially leading to memory corruption.
Address the aforementioned challenges by shifting the creation of
elfcorehdr from the first kernel (also referred as the crashed kernel),
where it was created and frequently recreated for every memory
add/remove event, to the fadump kernel. As a result, the elfcorehdr only
needs to be created once, thus eliminating the necessity to re-register
fadump during memory add/remove events.
At present, the first kernel prepares fadump header and stores it in the
fadump reserved area. The fadump header includes the start address of
the elfcorehdr, crashing CPU details, and other relevant information. In
the event of a crash in the first kernel, the second/fadump boots and
accesses the fadump header prepared by the first kernel. It then
performs the following steps in a platform-specific function
[rtas|opal]_fadump_process:
1. Sanity check for fadump header
2. Update CPU notes in elfcorehdr
Along with the above, update the setup_fadump()/fadump.c to create
elfcorehdr and set its address to the global variable elfcorehdr_addr
for the vmcore module to process it in the second/fadump kernel.
Section below outlines the information required to create the elfcorehdr
and the changes made to make it available to the fadump kernel if it's
not already.
To create elfcorehdr, the following crashed kernel information is
required: CPU notes, vmcoreinfo, and memory ranges.
At present, the CPU notes are already prepared in the fadump kernel, so
no changes are needed in that regard. The fadump kernel has access to
all crashed kernel memory regions, including boot memory regions that
are relocated by firmware to fadump reserved areas, so no changes for
that either. However, it is necessary to add new members to the fadump
header, i.e., the 'fadump_crash_info_header' structure, in order to pass
the crashed kernel's vmcoreinfo address and its size to fadump kernel.
In addition to the vmcoreinfo address and size, there are a few other
attributes also added to the fadump_crash_info_header structure.
1. version:
It stores the fadump header version, which is currently set to 1.
This provides flexibility to update the fadump crash info header in
the future without changing the magic number. For each change in the
fadump header, the version will be increased. This will help the
updated kernel determine how to handle kernel dumps from older
kernels. The magic number remains relevant for checking fadump header
corruption.
2. pt_regs_sz/cpu_mask_sz:
Store size of pt_regs and cpu_mask structure of first kernel. These
attributes are used to prevent dump processing if the sizes of
pt_regs or cpu_mask structure differ between the first and fadump
kernels.
Note: if either first/crashed kernel or second/fadump kernel do not have
the changes introduced here then kernel fail to collect the dump and
prints relevant error message on the console.
When there are no options specified for lparstat, it is expected to
give reports since LPAR(Logical Partition) boot.
APP(Available Processor Pool) is an indicator of how many cores in the
shared pool are free to use in Shared Processor LPAR(SPLPAR). APP is
derived using pool_idle_time which is obtained using H_PIC call.
The interval based reports show correct APP value while since boot
report shows very high APP values. This happens because in that case APP
is obtained by dividing pool idle time by LPAR uptime. Since pool idle
time is reported by the PowerVM hypervisor since its boot, it need not
align with LPAR boot.
To fix that export boot pool idle time in lparcfg and powerpc-utils will
use this info to derive APP as below for since boot reports.
APP = (pool idle time - boot pool idle time) / (uptime * timebase)
powerpc/mm: Update the memory limit based on direct mapping restrictions
memory limit value specified by the user are further updated such that
the value is 16MB aligned. This is because hash translation mode use
16MB as direct mapping page size. Make sure we update the global
variable 'memory_limit' with the 16MB aligned value such that all kernel
components will see the new aligned value of the memory limit.
powerpc/fadump: Don't update the user-specified memory limit
If the user specifies the memory limit, the kernel should honor it such
that all allocation and reservations are made within the memory limit
specified. fadump was breaking that rule. Remove the code which updates
the memory limit such that fadump reservations are done within the
limit specified.
powerpc/mm: Align memory_limit value specified using mem= kernel parameter
The value specified for the memory limit is used to set a restriction on
memory usage. It is important to ensure that this restriction is within
the linear map kernel address space range. The hash page table
translation uses a 16MB page size to map the kernel linear map address
space. htab_bolt_mapping() function aligns down the size of the range
while mapping kernel linear address space. Since the memblock limit is
enforced very early during boot, before we can detect the type of memory
translation (radix vs hash), we align the memory limit value specified
as a kernel parameter to 16MB. This alignment value will work for both
hash and radix translations.
powerpc: Avoid nmi_enter/nmi_exit in real mode interrupt.
nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel
crash when invoked during real mode interrupt handling (e.g. early HMI/MCE
interrupt handler) if percpu allocation comes from vmalloc area.
Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI()
wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when
percpu allocation is from the embedded first chunk. However with
CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu
allocation can come from the vmalloc area.
With kernel command line "percpu_alloc=page" we can force percpu allocation
to come from vmalloc area and can see kernel crash in machine_check_early:
powerpc: Add static_key_feature_checks_initialized flag
JUMP_LABEL_FEATURE_CHECK_DEBUG used static_key_intialized to determine
whether {cpu,mmu}_has_feature() is used before static keys were
initialized. However, {cpu,mmu}_has_feature() should not be used before
setup_feature_keys() is called but static_key_initialized is set well
before this by the call to jump_label_init() in early_init_devtree().
This creates a window in which JUMP_LABEL_FEATURE_CHECK_DEBUG will not
detect misuse and report errors. Add a flag specifically to indicate
when {cpu,mmu}_has_feature() is safe to use.
The struct ps3_notification_device in the ps3_probe_thread routine
is too large to be on the stack, causing a warning for an
allmodconfig build with clang.
Change the struct ps3_notification_device from a variable on the stack
to a dynamically allocated variable.
Benjamin Gray [Tue, 26 Mar 2024 04:44:20 +0000 (15:44 +1100)]
powerpc: Error on assembly warnings
We currently enable -Werror on the arch/powerpc subtree. However this
only catches C warnings. Assembly warnings are logged, but the make
invocation will still succeed. This can allow incorrect syntax such as
ori r3, r4, r5
to be compiled without catching that the assembler is treating r5
as the immediate value 5.
To prevent this in assembly files and inline assembly, add the
-fatal-warnings option to assembler invocations.
Signed-off-by: Benjamin Gray <bgray@linux.ibm.com> Tested-by: Andrew Donnellan <ajd@linux.ibm.com> Reviewed-by: Andrew Donnellan <ajd@linux.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240326044420.577031-1-bgray@linux.ibm.com
vmpic_msi_feature is only used conditionally, which triggers a rare
-Werror=unused-const-variable= warning with gcc:
arch/powerpc/sysdev/fsl_msi.c:567:37: error: 'vmpic_msi_feature' defined but not used [-Werror=unused-const-variable=]
567 | static const struct fsl_msi_feature vmpic_msi_feature =
Hide this one in the same #ifdef as the reference so we can turn on
the warning by default.
Fixes: 305bcf26128e ("powerpc/fsl-soc: use CONFIG_EPAPR_PARAVIRT for hcalls") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://msgid.link/20240403080702.3509288-2-arnd@kernel.org
Linus Torvalds [Sun, 31 Mar 2024 18:23:51 +0000 (11:23 -0700)]
Merge tag 'kbuild-fixes-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild
Pull Kbuild fixes from Masahiro Yamada:
- Deduplicate Kconfig entries for CONFIG_CXL_PMU
- Fix unselectable choice entry in MIPS Kconfig, and forbid this
structure
- Remove unused include/asm-generic/export.h
- Fix a NULL pointer dereference bug in modpost
- Enable -Woverride-init warning consistently with W=1
- Drop KCSAN flags from *.mod.c files
* tag 'kbuild-fixes-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
kconfig: Fix typo HEIGTH to HEIGHT
Documentation/llvm: Note s390 LLVM=1 support with LLVM 18.1.0 and newer
kbuild: Disable KCSAN for autogenerated *.mod.c intermediaries
kbuild: make -Woverride-init warnings more consistent
modpost: do not make find_tosym() return NULL
export.h: remove include/asm-generic/export.h
kconfig: do not reparent the menu inside a choice block
MIPS: move unselectable FIT_IMAGE_FDT_EPM5 out of the "System type" choice
cxl: remove CONFIG_CXL_PMU entry in drivers/cxl/Kconfig
Linus Torvalds [Sun, 31 Mar 2024 18:15:32 +0000 (11:15 -0700)]
Merge tag 'edac_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras
Pull EDAC fixes from Borislav Petkov:
- Fix more issues in the AMD FMPM driver
* tag 'edac_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
RAS: Avoid build errors when CONFIG_DEBUG_FS=n
RAS/AMD/FMPM: Safely handle saved records of various sizes
RAS/AMD/FMPM: Avoid NULL ptr deref in get_saved_records()
Linus Torvalds [Sun, 31 Mar 2024 18:04:51 +0000 (11:04 -0700)]
Merge tag 'irq_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull irq fixes from Borislav Petkov:
- Fix an unused function warning on irqchip/irq-armada-370-xp
- Fix the IRQ sharing with pinctrl-amd and ACPI OSL
* tag 'irq_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqchip/armada-370-xp: Suppress unused-function warning
genirq: Introduce IRQF_COND_ONESHOT and use it in pinctrl-amd
Linus Torvalds [Sun, 31 Mar 2024 17:43:11 +0000 (10:43 -0700)]
Merge tag 'perf_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 perf fixes from Borislav Petkov:
- Define the correct set of default hw events on AMD Zen4
- Use the correct stalled cycles PMCs on AMD Zen2 and newer
- Fix detection of the LBR freeze feature on AMD
* tag 'perf_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/x86/amd/core: Define a proper ref-cycles event for Zen 4 and later
perf/x86/amd/core: Update and fix stalled-cycles-* events for Zen 2 and later
perf/x86/amd/lbr: Use freeze based on availability
x86/cpufeatures: Add new word for scattered features
Linus Torvalds [Sun, 31 Mar 2024 17:16:34 +0000 (10:16 -0700)]
Merge tag 'x86_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Borislav Petkov:
- Make sure single object builds in arch/x86/virt/ ala
make ... arch/x86/virt/vmx/tdx/seamcall.o
work again
- Do not do ROM range scans and memory validation when the kernel is
running as a SEV-SNP guest as those can get problematic and, before
that, are not really needed in such a guest
- Exclude the build-time generated vdso-image-x32.o object from objtool
validation and in particular the return sites in there due to a
warning which fires when an unpatched return thunk is being used
- Improve the NMI CPUs stall message to show additional information
about the state of each CPU wrt the NMI handler
- Enable gcc named address spaces support only on !KCSAN configs due to
compiler options incompatibility
- Revert a change which was trying to use GB pages for mapping regions
only when the regions would be large enough but that change lead to
kexec failing
- A documentation fixlet
* tag 'x86_urgent_for_v6.9_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/build: Use obj-y to descend into arch/x86/virt/
x86/sev: Skip ROM range scans and validation for SEV-SNP guests
x86/vdso: Fix rethunk patching for vdso-image-x32.o too
x86/nmi: Upgrade NMI backtrace stall checks & messages
x86/percpu: Disable named address spaces for KCSAN
Revert "x86/mm/ident_map: Use gbpages only where full GB page should be mapped."
Documentation/x86: Fix title underline length
Documentation/llvm: Note s390 LLVM=1 support with LLVM 18.1.0 and newer
As of the first s390 pull request during the 6.9 merge window,
commit 691632f0e869 ("Merge tag 's390-6.9-1' of
git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux"), s390 can be
built with LLVM=1 when using LLVM 18.1.0, which is the first version
that has SystemZ support implemented in ld.lld and llvm-objcopy.
Update the supported architectures table in the Kbuild LLVM
documentation to note this explicitly to make it more discoverable by
users and other developers. Additionally, this brings s390 in line with
the rest of the architectures in the table, which all support LLVM=1.
kbuild: Disable KCSAN for autogenerated *.mod.c intermediaries
When KCSAN and CONSTRUCTORS are enabled, one can trigger the
"Unpatched return thunk in use. This should not happen!"
catch-all warning.
Usually, when objtool runs on the .o objects, it does generate a section
.return_sites which contains all offsets in the objects to the return
thunks of the functions present there. Those return thunks then get
patched at runtime by the alternatives.
KCSAN and CONSTRUCTORS add this to the object file's .text.startup
section:
-------------------
Disassembly of section .text.startup:
which, if it is built as a module goes through the intermediary stage of
creating a <module>.mod.c file which, when translated, receives a second
constructor:
-------------------
Disassembly of section .text.startup:
Objtool has run already so that second constructor's return thunk cannot
be added to the .return_sites section and thus the return thunk remains
unpatched and the warning rightfully fires.
Drop KCSAN flags from the mod.c generation stage as those constructors
do not contain data races one would be interested about.
Debugged together with David Kaplan <David.Kaplan@amd.com> and Nikolay
Borisov <nik.borisov@suse.com>.
Reported-by: Paul Menzel <pmenzel@molgen.mpg.de> Closes: https://lore.kernel.org/r/0851a207-7143-417e-be31-8bf2b3afb57d@molgen.mpg.de Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Tested-by: Paul Menzel <pmenzel@molgen.mpg.de> # Dell XPS 13 Reviewed-by: Nikolay Borisov <nik.borisov@suse.com> Reviewed-by: Marco Elver <elver@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Arnd Bergmann [Tue, 26 Mar 2024 14:47:16 +0000 (15:47 +0100)]
kbuild: make -Woverride-init warnings more consistent
The -Woverride-init warn about code that may be intentional or not,
but the inintentional ones tend to be real bugs, so there is a bit of
disagreement on whether this warning option should be enabled by default
and we have multiple settings in scripts/Makefile.extrawarn as well as
individual subsystems.
Older versions of clang only supported -Wno-initializer-overrides with
the same meaning as gcc's -Woverride-init, though all supported versions
now work with both. Because of this difference, an earlier cleanup of
mine accidentally turned the clang warning off for W=1 builds and only
left it on for W=2, while it's still enabled for gcc with W=1.
There is also one driver that only turns the warning off for newer
versions of gcc but not other compilers, and some but not all the
Makefiles still use a cc-disable-warning conditional that is no
longer needed with supported compilers here.
Address all of the above by removing the special cases for clang
and always turning the warning off unconditionally where it got
in the way, using the syntax that is supported by both compilers.
Fixes: 2cd3271b7a31 ("kbuild: avoid duplicate warning options") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Acked-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Andrew Jeffery <andrew@codeconstruct.com.au> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Reviewed-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Mikulas Patocka [Sat, 30 Mar 2024 19:23:08 +0000 (20:23 +0100)]
objtool: Fix compile failure when using the x32 compiler
When compiling the v6.9-rc1 kernel with the x32 compiler, the following
errors are reported. The reason is that we take an "unsigned long"
variable and print it using "PRIx64" format string.
In file included from check.c:16:
check.c: In function ‘add_dead_ends’:
/usr/src/git/linux-2.6/tools/objtool/include/objtool/warn.h:46:17: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 5 has type ‘long unsigned int’ [-Werror=format=]
46 | "%s: warning: objtool: " format "\n", \
| ^~~~~~~~~~~~~~~~~~~~~~~~
check.c:613:33: note: in expansion of macro ‘WARN’
613 | WARN("can't find unreachable insn at %s+0x%" PRIx64,
| ^~~~
...
Linus Torvalds [Sat, 30 Mar 2024 20:51:58 +0000 (13:51 -0700)]
Merge tag 'xfs-6.9-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs fixes from Chandan Babu:
- Allow stripe unit/width value passed via mount option to be written
over existing values in the super block
- Do not set current->journal_info to avoid its value from being miused
by another filesystem context
* tag 'xfs-6.9-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
xfs: don't use current->journal_info
xfs: allow sunit mount option to repair bad primary sb stripe values
Linus Torvalds [Sat, 30 Mar 2024 20:44:52 +0000 (13:44 -0700)]
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes and updates from James Bottomley:
"Fully half this pull is updates to lpfc and qla2xxx which got
committed just as the merge window opened. A sizeable fraction of the
driver updates are simple bug fixes (and lock reworks for bug fixes in
the case of lpfc), so rather than splitting the few actual
enhancements out, we're just adding the drivers to the -rc1 pull.
The enhancements for lpfc are log message removals, copyright updates
and three patches redefining types. For qla2xxx it's just removing a
debug message on module removal and the manufacturer detail update.
The two major fixes are the sg teardown race and a core error leg
problem with the procfs directory not being removed if we destroy a
created host that never got to the running state. The rest are minor
fixes and constifications"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (41 commits)
scsi: bnx2fc: Remove spin_lock_bh while releasing resources after upload
scsi: core: Fix unremoved procfs host directory regression
scsi: mpi3mr: Avoid memcpy field-spanning write WARNING
scsi: sd: Fix TCG OPAL unlock on system resume
scsi: sg: Avoid sg device teardown race
scsi: lpfc: Copyright updates for 14.4.0.1 patches
scsi: lpfc: Update lpfc version to 14.4.0.1
scsi: lpfc: Define types in a union for generic void *context3 ptr
scsi: lpfc: Define lpfc_dmabuf type for ctx_buf ptr
scsi: lpfc: Define lpfc_nodelist type for ctx_ndlp ptr
scsi: lpfc: Use a dedicated lock for ras_fwlog state
scsi: lpfc: Release hbalock before calling lpfc_worker_wake_up()
scsi: lpfc: Replace hbalock with ndlp lock in lpfc_nvme_unregister_port()
scsi: lpfc: Update lpfc_ramp_down_queue_handler() logic
scsi: lpfc: Remove IRQF_ONESHOT flag from threaded IRQ handling
scsi: lpfc: Move NPIV's transport unregistration to after resource clean up
scsi: lpfc: Remove unnecessary log message in queuecommand path
scsi: qla2xxx: Update version to 10.02.09.200-k
scsi: qla2xxx: Delay I/O Abort on PCI error
scsi: qla2xxx: Change debug message during driver unload
...
Linus Torvalds [Sat, 30 Mar 2024 19:59:00 +0000 (12:59 -0700)]
Merge tag 'staging-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
Pull staging driver fixes from Greg KH:
"Here are two small staging driver fixes for the vc04_services driver
that resolve reported problems:
- strncpy fix for information leak
- another information leak discovered by the previous strncpy fix
Both of these have been in linux-next all this past week with no
reported issues"
* tag 'staging-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
staging: vc04_services: fix information leak in create_component()
staging: vc04_services: changen strncpy() to strscpy_pad()
Wolfram Sang [Sat, 30 Mar 2024 14:37:54 +0000 (15:37 +0100)]
Merge tag 'i2c-host-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/andi.shyti/linux into i2c/for-current
One fix in the i801 driver where a bug caused touchpad
malfunctions on some Lenovo P1 models by incorrectly overwriting
a status variable during successful SMBUS transactions.
Masahiro Yamada [Sat, 30 Mar 2024 06:05:54 +0000 (15:05 +0900)]
x86/build: Use obj-y to descend into arch/x86/virt/
Commit c33621b4c5ad ("x86/virt/tdx: Wire up basic SEAMCALL functions")
introduced a new instance of core-y instead of the standardized obj-y
syntax.
X86 Makefiles descend into subdirectories of arch/x86/virt inconsistently;
into arch/x86/virt/ via core-y defined in arch/x86/Makefile, but into
arch/x86/virt/svm/ via obj-y defined in arch/x86/Kbuild.
This is problematic when you build a single object in parallel because
multiple threads attempt to build the same file.
$ make -j$(nproc) arch/x86/virt/vmx/tdx/seamcall.o
[ snip ]
AS arch/x86/virt/vmx/tdx/seamcall.o
AS arch/x86/virt/vmx/tdx/seamcall.o
fixdep: error opening file: arch/x86/virt/vmx/tdx/.seamcall.o.d: No such file or directory
make[4]: *** [scripts/Makefile.build:362: arch/x86/virt/vmx/tdx/seamcall.o] Error 2
Linus Torvalds [Fri, 29 Mar 2024 22:51:15 +0000 (15:51 -0700)]
Merge tag 'drm-fixes-2024-03-30' of https://gitlab.freedesktop.org/drm/kernel
Pull drm fixes from Dave Airlie:
"Regular fixes for rc2, quite a few i915/amdgpu as usual, some xe, and
then mostly scattered around. rc3 might be quieter with the holidays
but we shall see.
bridge:
- select DRM_KMS_HELPER
dma-buf:
- fix NULL-pointer deref
dp:
- fix div-by-zero in DP MST unplug code
fbdev:
- select FB_IOMEM_FOPS for SBus
sched:
- fix NULL-pointer deref
xe:
- Fix build on mips
- Fix wrong bound checks
- Fix use of msec rather than jiffies
- Remove dead code
i915:
- Fix for BUG_ON/BUILD_BUG_ON IN I915_memcpy.c
- Update a MTL workaround
- Fix locking inversion in hwmon's sysfs
- Remove a bogus error message around PXP
- Fix UAF on VMA
- Reset queue_priority_hint on parking
- Display Fixes:
- Remove duplicated audio enable/disable on SDVO and DP
- Disable AuxCCS for Xe driver
- Revert init order of MIPI DSI
- DRRS debugfs fix with an extra refactor patch
- VRR related fixes
- Fix a JSL eDP corruption
- Fix the cursor physical dma address
- BIOS VBT related fix
rockchip:
- vop2: remove support for AR30 and AB30 formats
vmwgfx:
- debugfs: create ttm_resource_manager entry only if needed"
* tag 'drm-fixes-2024-03-30' of https://gitlab.freedesktop.org/drm/kernel: (55 commits)
drm/i915/bios: Tolerate devdata==NULL in intel_bios_encoder_supports_dp_dual_mode()
drm/i915: Pre-populate the cursor physical dma address
drm/i915/gt: Reset queue_priority_hint on parking
drm/i915/vma: Fix UAF on destroy against retire race
drm/i915: Do not print 'pxp init failed with 0' when it succeed
drm/i915: Do not match JSL in ehl_combo_pll_div_frac_wa_needed()
drm/i915/hwmon: Fix locking inversion in sysfs getter
drm/i915/dsb: Fix DSB vblank waits when using VRR
drm/i915/vrr: Generate VRR "safe window" for DSB
drm/i915/display/debugfs: Fix duplicate checks in i915_drrs_status
drm/i915/drrs: Refactor CPU transcoder DRRS check
drm/i915/mtl: Update workaround 14018575942
drm/i915/dsi: Go back to the previous INIT_OTP/DISPLAY_ON order, mostly
drm/i915/display: Disable AuxCCS framebuffers if built for Xe
drm/i915: Stop doing double audio enable/disable on SDVO and g4x+ DP
drm/i915: Add includes for BUG_ON/BUILD_BUG_ON in i915_memcpy.c
drm/qxl: remove unused variable from `qxl_process_single_command()`
drm/qxl: remove unused `count` variable from `qxl_surface_id_alloc()`
drm/i915: add bug.h include to i915_memcpy.c
drm/vmwgfx: Create debugfs ttm_resource_manager entry only if needed
...
Linus Torvalds [Fri, 29 Mar 2024 22:38:29 +0000 (15:38 -0700)]
Merge tag 'linux_kselftest-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest
Pull kselftest fixes from Shuah Khan:
"Fixes to seccomp and ftrace tests and a change to add config file for
dmabuf-heap test to increase coverage"
* tag 'linux_kselftest-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
selftests: dmabuf-heap: add config file for the test
selftests/seccomp: Try to fit runtime of benchmark into timeout
selftests/ftrace: Fix event filter target_func selection
Linus Torvalds [Fri, 29 Mar 2024 22:35:12 +0000 (15:35 -0700)]
Merge tag 'linux_kselftest-kunit-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest
Pull KUnit fixes from Shuah Khan:
"One urgent fix for --alltests build failure related to renaming of
CONFIG_DAMON_DBGFS to DAMON_DBGFS_DEPRECATED to the missing config
option"
* tag 'linux_kselftest-kunit-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
kunit: configs: Enable CONFIG_DAMON_DBGFS_DEPRECATED for --alltests
Mark Brown [Mon, 25 Mar 2024 16:57:59 +0000 (16:57 +0000)]
selftests/seccomp: Try to fit runtime of benchmark into timeout
The seccomp benchmark runs five scenarios, one calibration run with no
seccomp filters enabled then four further runs each adding a filter. The
calibration run times itself for 15s and then each additional run executes
for the same number of times.
Currently the seccomp tests, including the benchmark, run with an extended
120s timeout but this is not sufficient to robustly run the tests on a lot
of platforms. Sample timings from some recent runs:
Platform Run 1 Run 2 Run 3 Run 4
--------- ----- ----- ----- -----
PowerEdge R200 16.6s 16.6s 31.6s 37.4s
BBB (arm) 20.4s 20.4s 54.5s
Synquacer (arm64) 20.7s 23.7s 40.3s
The x86 runs from the PowerEdge are quite marginal and routinely fail, for
the successful run reported here the timed portions of the run are at
117.2s leaving less than 3s of margin which is frequently breached. The
added overhead of adding filters on the other platforms is such that there
is no prospect of their runs fitting into the 120s timeout, especially
on 32 bit arm where there is no BPF JIT.
While we could lower the time we calibrate for I'm also already seeing the
currently completing runs reporting issues with the per filter overheads
not matching expectations:
Let's instead raise the timeout to 180s which is only a 50% increase on the
current timeout which is itself not *too* large given that there's only two
tests in this suite.
Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The event filter function test has been failing in our internal test
farm:
| # not ok 33 event filter function - test event filtering on functions
Running the test in verbose mode indicates that this is because the test
erroneously determines that kmem_cache_free() is the most common caller
of kmem_cache_free():
... and as kmem_cache_free() doesn't call itself, setting this as the
filter function for kmem_cache_free() results in no hits, and
consequently the test fails:
This seems to be because the system in question has tasks with ':' in
their name (which a number of kernel worker threads have). These show up
in the trace, e.g.
... and so when we try to extact the call_site with:
cut -d: -f3 trace | sed 's/call_site=\([^+]*\)+0x.*/\1/'
... the 'cut' command will extrace the column containing
'kmem_cache_free' rather than the column containing 'call_site=...', and
the 'sed' command will leave this unchanged. Consequently, the test will
decide to use 'kmem_cache_free' as the filter function, resulting in the
failure seen above.
Fix this by matching the 'call_site=<func>' part specifically to extract
the function name.
The original version of the mitigation would patch in the calls to the
untraining routines directly. That is, the alternative() in UNTRAIN_RET
will patch in the CALL to srso_alias_untrain_ret() directly.
However, even if commit e7c25c441e9e ("x86/cpu: Cleanup the untrain
mess") meant well in trying to clean up the situation, due to micro-
architectural reasons, the untraining routine srso_alias_untrain_ret()
must be the target of a CALL instruction and not of a JMP instruction as
it is done now.
Reshuffle the alternative macros to accomplish that.
Linus Torvalds [Fri, 29 Mar 2024 18:50:38 +0000 (11:50 -0700)]
Merge tag 'thermal-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull thermal control fixes from Rafael Wysocki:
"These revert a problematic optimization commit and address a devfreq
cooling device issue.
Specifics:
- Revert thermal core optimization that introduced a functional issue
causing a critical trip point to be crossed in some cases (Daniel
Lezcano)
- Add missing conversion between different state ranges to the
devfreq cooling device driver (Ye Zhang)"
* tag 'thermal-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
thermal: devfreq_cooling: Fix perf state when calculate dfc res_util
Revert "thermal: core: Don't update trip points inside the hysteresis range"
projects / thirdparty / kernel / linux.git / log
