Move the libunwind register to perf register mapping functions in
arch/../util/unwind-libunwind.c into a new libunwind-arch
directory. Rename the functions to
__get_perf_regnum_for_unw_regnum_<arch>. Add untested ppc32 and s390
functions. Add a get_perf_regnum_for_unw_regnum function that takes an
ELF machine as well as a register number and chooses the appropriate
architecture implementation.
Split the x86 and powerpc 32 and 64-bit implementations apart so that
a single libunwind-<arch>.h header is included.
Move the e_machine into the unwind_info struct to make it easier to
pass.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alex@ghiti.fr>
Cc: Andrew Jones <andrew.jones@oss.qualcomm.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Dmitrii Dolgov <9erthalion6@gmail.com>
Cc: Florian Fainelli <florian.fainelli@broadcom.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Li Guan <guanli.oerv@isrc.iscas.ac.cn>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <pjw@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shimin Guo <shimin.guo@skydio.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Tomas Glozar <tglozar@redhat.com>
Cc: Will Deacon <will@kernel.org>
[ Map UNW_PPC32_NIP to PERF_REG_POWERPC_NIP like done for 64-bit, pointed out by a local sashiko ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Iterate LIBUNWIND_ARCHS when setting up CONFIG_ and HAVE_ definitions
rather than treating each architecture individually. This sets up the
libunwind build variables and C definitions beyond x86 and
arm/aarch64. The existing naming convention is followed for
compatibility.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alex@ghiti.fr>
Cc: Andrew Jones <andrew.jones@oss.qualcomm.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Dmitrii Dolgov <9erthalion6@gmail.com>
Cc: Florian Fainelli <florian.fainelli@broadcom.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Li Guan <guanli.oerv@isrc.iscas.ac.cn>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <pjw@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shimin Guo <shimin.guo@skydio.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Tomas Glozar <tglozar@redhat.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The separate test files only exist to pass a different #include,
instead have a single source file and pass -include to $(CC) to
include the relevant header file for the architecture being
tested. Generate the rules using a foreach loop. Include tests for all
current libunwind architectures.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alex@ghiti.fr>
Cc: Andrew Jones <andrew.jones@oss.qualcomm.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Dmitrii Dolgov <9erthalion6@gmail.com>
Cc: Florian Fainelli <florian.fainelli@broadcom.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Li Guan <guanli.oerv@isrc.iscas.ac.cn>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <pjw@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shimin Guo <shimin.guo@skydio.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Tomas Glozar <tglozar@redhat.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently, both libdw and libunwind define 'unwind__get_entries'. This
causes a duplicate symbol build failure when both are compiled into
perf.
This commit refactors the DWARF unwind post-processing to be
configurable at runtime via the .perfconfig file option
'unwind.style', or using the argument '--unwind-style' in the commands
'perf report', 'perf script' and 'perf inject', in a similar manner to
the addr2line or the disassembler style.
The file 'tools/perf/util/unwind.c' adds the top-level dispatch
function 'unwind__get_entries'. The backend implementations are
renamed to 'libdw__get_entries' and 'libunwind__get_entries'. Both are
attempted as fallbacks if not configured, or if the primary backend
fails.
Fixes: 2e9191573a ("perf build: Remove NO_LIBDW_DWARF_UNWIND option")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alex@ghiti.fr>
Cc: Andrew Jones <andrew.jones@oss.qualcomm.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Dmitrii Dolgov <9erthalion6@gmail.com>
Cc: Florian Fainelli <florian.fainelli@broadcom.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: libunwind-devel@nongnu.org
Cc: Li Guan <guanli.oerv@isrc.iscas.ac.cn>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <pjw@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shimin Guo <shimin.guo@skydio.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Tomas Glozar <tglozar@redhat.com>
Cc: Will Deacon <will@kernel.org>
[ Don't mix declarations and code, move 'entries' variable to the start of scope ]
[ Use pr_warning_once() instead of pr_err() in stubs for get_entries(), suggested by a local sashiko instance ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Only uncore PMUs can have an identifier, so add an optimized
perf_pmus__scan routine for that case to avoid all PMU types being
created.
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Leo Yan <leo.yan@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
l2_itlb_misses is a valid legacy cache event name, hence allowing it
in all_events in metric.py. l2_itlb_misses was also a json event for
AMD zen1, zen2 and zen3.
For zen4, zen5 and zen6 the checking that metric events are within the
json was skipping l2_itlb_misses as it is a valid legacy event, however,
the PMU driver lacks the event mapping causing it to be a bad event when
used in the metric.
Add bp_l1_tlb_miss_l2_tlb_miss.all as the l2 itlb miss event (bp =
branch predictor, the AMD way to say itlb), so that is used in
preference to l2_itlb_misses when the event exists.
Remove l2_itlb_misses from metric.py as the legacy event isn't used by
any metrics and having it is error prone for newer AMD zen models.
Fixes: e596f32966 ("perf jevents: Add itlb metric group for AMD")
Reviewed-by: Sandipan Das <sandipan.das@amd.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Code in tools/perf/arch causes portability issues/opaqueness and LTO
issues due to the use of weak symbols. Move the adding of LR to the
sample_user_regs into arm64-frame-pointer-unwind-support.c conditional
on EM_HOST == EM_AARCH64 (false on all non-ARM64 builds).
This also better encapsulates the use of the sampled registers by
get_leaf_frame_caller_aarch64 and the set up by the new
add_leaf_frame_caller_opts_aarch64, exposing opportunities for possibly
sampling PC and SP to help the unwinder.
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shimin Guo <shimin.guo@skydio.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
delegate_tool was missing the delegate overrides for schedstat_cpu
and schedstat_domain. As a result, when allocated with zalloc, these
callbacks defaulted to NULL, causing a segmentation fault crash if
any schedstat events were delivered during event processing.
Fix this by adding delegate_schedstat_cpu and delegate_schedstat_domain
via the CREATE_DELEGATE_OP2 macro, and ensuring delegate_tool__init
correctly registers them.
Additionally, delegate_tool__init completely omitted copying the
dont_split_sample_group property from the delegate. This would cause
wrapper tools to default the flag to false, which corrupts piped event
processing (e.g., in perf inject) by triggering duplicate event
deliveries on split sample values in deliver_sample_group().
Similarly, perf_tool__init() omitted the initialization of this
boolean field. On stack-allocated tools that rely on this initializer
(like intel-tpebs or __cmd_evlist), this could result in uninitialized
stack garbage evaluating to true—silently dropping non-leader event
members in deliver_sample_group().
Fix both issues by properly copying the field in delegate_tool__init
and initializing it to false in perf_tool__init.
Fixes: 6331b26693 ("perf tool: Add a delegate_tool that just delegates actions to another tool")
Fixes: 79bcd34e0f ("perf inject: Fix leader sampling inserting additional samples")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Gabriel Marin <gmx@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf_sched__timehist() registers event handlers for options using the
sched->tool struct. It registers handlers for MMAP, COMM, EXIT, FORK, etc.
but completely omits registering a handler for MMAP2 events.
Failing to register both MMAP and MMAP2 handlers causes modern systems
(which primarily output MMAP2 records) to silently drop VMA map mappings.
This results in uninitialized machine/thread mapping structures, making it
impossible to resolve shared library instruction pointers (IPs) to dynamic
symbols/DSOs during timehist callchain analysis.
Fix this by correctly registering perf_event__process_mmap2 in
sched->tool inside perf_sched__timehist().
Fixes: 49394a2a24 ("perf sched timehist: Introduce timehist command")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Gabriel Marin <gmx@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The patch addresses multiple segmentation fault vectors, out-of-bounds
reads, and memory leaks in perf sched stats by adding bounds checks,
NULL checks, proper error propagation, and robust memory cleanup.
1. In get_all_cpu_stats(), added assert(!list_empty(head)) to prevent
unsafe list_first_entry() calls on empty lists, added a missing NULL
check for summary_head->cpu_data allocation, and implemented a cleanup
ladder using a temporary list to prevent memory leaks on error paths.
2. In free_schedstat(), fixed memory leaks by ensuring internal domain_data
and cpu_data pointers are freed.
3. In show_schedstat_data(), fixed a stale pointer issue where ds2 retained
its value from a previous iteration when dptr2 became NULL, and added
proper propagation of errors from get_all_cpu_stats().
4. Propagated show_schedstat_data() errors up to perf_sched__schedstat_diff()
and perf_sched__schedstat_live() to prevent output corruption on failure.
5. In show_schedstat_data(), added NULL checks for cd_map1 and cd_map2
to gracefully handle invalid or empty data files.
6. Added parallel iteration termination checks using list_is_last() in
show_schedstat_data() for both domain and CPU lists to safely terminate
at the end of each list when files contain a different number of CPUs
or domains.
7. Added CPU bounds checks (cs1->cpu >= nr1 and cs2->cpu >= nr2) in
show_schedstat_data() to prevent out-of-bounds reads from cd_map1 and
cd_map2 when comparing files from machines with different CPU counts.
8. Added NULL checks for cd_info1 and cd_info2 to prevent crashes when
a CPU has data samples but no corresponding domain info in the header.
9. Added domain bounds checks (ds1->domain >= cd_info1->nr_domains and
ds2->domain >= cd_info2->nr_domains) to prevent out-of-bounds array
accesses in the domains array.
10. Added NULL checks for dinfo1 and dinfo2 in show_schedstat_data()
to prevent crashes when a domain has no corresponding domain info.
11. Zero-initialized the perf_data array in perf_sched__schedstat_diff()
to prevent stack garbage from causing perf_data_file__fd() to attempt
to use a NULL fptr when use_stdio happened to be non-zero.
Assisted-by: Gemini:gemini-3.1-pro-preview
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The trace summary test calls /bin/true and filters for open, read and
close events. These events are coming from shared library loads.
On a musl system, the loader and libc may point to the same file. true
needs only libc, no further shared libraries are loaded at startup. The
test fails since no open, read and close events are captured.
root@host:~# ldd /bin/true
/lib/ld-musl-riscv64.so.1 (0x3fb8882000)
libc.so => /lib/ld-musl-riscv64.so.1 (0x3fb8882000)
root@host:~# file /lib/ld-musl-riscv64.so.1
/lib/ld-musl-riscv64.so.1: symbolic link to /usr/lib/libc.so
root@host:~# strace -f /bin/true
execve("/bin/true", ["/bin/true", ...], ... /* 18 vars */) = 1
set_tid_address(0x3fa1f7bf70) = 330
mprotect(0x2ad6b8e000, 12288, PROT_READ) = 0
exit_group(0) = ?
+++ exited with 0 +++
Run "cat /dev/null" instead of "true". This creates the required events
regardless of the C library and it works for cat from busybox or from
coreutils.
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Acked-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The "syscalls:sys_enter_openat event fields" test calls
openat(AT_FDCWD, "/etc/passwd", O_RDONLY | O_DIRECTORY)
and verifies that the flags of the captured event are matching.
This fails for musl, where the openat syscall wrapper always adds
O_LARGEFILE.
Update the check to allow for additional flags, the access mode flags
must be unchanged.
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The perf ftrace test case runs:
perf ftrace profile --graph-opts depth=5 sleep 0.1
and checks that the output contains a *clock_nanosleep function with a
count of 1.
This fails on a risc-v system that uses musl as its C library. musl's
nanosleep syscall wrapper uses either the nanosleep or the
clock_nanosleep syscall.
Filter for sys_*nanosleep to allow both syscalls.
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
4751bddd3f ("perf tools: Make GTK2 support opt-in") changed GTK2
build to be opt-in.
So NO_GTK2 is meaningless and we need to pass GTK2=1 to enable it.
Let's update the build-test configuration for that.
Also make_no_ui is the same as make_no_slang since NO_GTK2 is no-op.
Let's get rid of it as well.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
On my system, `make GEN_VMLINUX_H=1` fails with a lot of error messages
like below:
./util/bpf_skel/vmlinux.h:134488:4: error: declaration does not declare anything [-Werror,-Wmissing-declarations]
134488 | struct freelist_counters;
| ^~~~~~~~~~~~~~~~~~~~~~~~
make[2]: *** [Makefile.perf:1249: linux/tools/perf/util/bpf_skel/.tmp/lock_contention.bpf.o] Error 1
I saw commit 835a507535 ("selftests/bpf: Add -fms-extensions to bpf
build flags") also added the same flags to bpf programs.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
It should say binutils-dev(el) instead of plain binutils package as it's
mostly installed already and can confuse people like me. :)
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Any IMPDEF events not printed out from a known core's IMPDEF list or for
a completely unknown core will still not be shown to the user. Fix this
by printing the remaining bits as comma separated raw numbers, e.g.
"IMPDEF:1,2,3,4".
Suggested-by: Al Grant <al.grant@arm.com>
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
>From the TRM [1], N1 has one IMPDEF event which isn't covered by the
common list. Add a framework so that more cores can be added in the
future and that the N1 IMPDEF event can be decoded. Also increase the
size of the buffer because we're adding more strings and if it gets
truncated it falls back to a hex dump only.
[1]: https://developer.arm.com/documentation/100616/0401/Statistical-Profiling-Extension/implementation-defined-features-of-SPE
Suggested-by: Al Grant <al.grant@arm.com>
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: https://developer.arm.com/documentation/100616/0401/Statistical-Profiling-Extension/implementation-defined-features-of-SPE
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This is so we can have a single function that prints events and can be
used with multiple mappings from different CPUs. Remove any bit that was
printed so that later we can print out the remaining unknown impdef
bits.
No functional changes intended.
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The MIDR will affect printing of arm_spe_pkts, so store a copy of it
there. Technically it's constant for each decoder, but there is no
decoder when doing a raw dump, so it has to be stored in every packet.
It will only be used in raw dump mode and not in normal decoding for
now, but to avoid any surprises, set MIDR properly on the decoder too.
Having both the MIDR and the arm_spe_pkt (which has a copy of it) in the
decoder seemed a bit weird, so remove arm_spe_pkt from the decoder. The
packet is only short lived anyway so probably shouldn't have been there
in the first place.
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Don't call strtol() with a null pointer to avoid undefined behavior.
I'm not sure of the exact scenario for missing CPU IDs but I don't think
it happens in practice. SPE decoding can continue without them with
reduced functionality, but print an error message anyway.
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
We'll need the MIDR to dump IMPDEF events in the next commits so extract
a function for it.
No functional changes intended.
Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf test 'perf inject to convert DWARF callchains to regular ones'
fails on s390. It was introduced with commit 92ea788d2a ("perf
inject: Add --convert-callchain option")
The failure comes the difference in output. Without the inject script to
convert DWARF the callchains is:
# perf record -F 999 --call-graph dwarf -- perf test -w noploop
# perf report -i perf.data --stdio --no-children -q \
--percent-limit=1 > /tmp/111
# cat /tmp/111
99.30% perf-noploop perf [.] noploop
|
---noploop
run_workload (inlined)
cmd_test
run_builtin (inlined)
handle_internal_command
run_argv (inlined)
main
__libc_start_call_main
__libc_start_main_impl (inlined)
_start
#
With the inject script step the output is:
# perf inject -i perf.data --convert-callchain -o /tmp/perf-inject-1.out
# perf report -i /tmp/perf-inject-1.out --stdio --no-children -q \
--percent-limit=1 > /tmp/222
# cat /tmp/222
99.40% perf-noploop perf [.] noploop
|
---noploop
run_workload (inlined)
cmd_test
run_builtin (inlined)
handle_internal_command
run_argv (inlined)
main
_start
# diff /tmp/111 /tmp/222
1c1
< 99.30% perf-noploop perf [.] noploop
---
> 99.40% perf-noploop perf [.] noploop
10,11d9
< __libc_start_call_main
< __libc_start_main_impl (inlined)
#
The difference are the symbols __libc_start_call_main and
__libc_start_main_impl.
On x86_64, kernel and user space share a single virtual address space,
with the kernel mapped to the upper end of memory. The instruction
pointer value alone is sufficient to distinguish between user space and
kernel space addresses.
This is not true for s390, which uses separate address spaces for user
and kernel.
The same virtual address can be valid in both address spaces, so the
instruction pointer value alone cannot determine whether an address
belongs to the kernel or user space.
Instead, perf must rely on the cpumode metadata derived from the
processor status word (PSW) at sample time.
In function perf_event__convert_sample_callchain() the first part
copies a kernel callchain and context entries, if any.
It then appends additional entries ignoring the address space
architecture. Taking that into account, the symbols at addresses
0x3ff970348cb __libc_start_call_main
0x3ff970349c5 __libc_start_main_impl
(located after the kernel address space on s390) are now included.
Output before:
# perf test 83
83: perf inject to convert DWARF callchains to regular ones : FAILED!
Output after:
# perf test 83
83: perf inject to convert DWARF callchains to regular ones : Ok
Question to Namhyung:
In function perf_event__convert_sample_callchain() just before the
for() loop this patch modifies, the kernel callchain is copied,
see this comment and the next 5 lines:
/* copy kernel callchain and context entries */
Then why is machine__kernel_ip() needed in the for() loop, when
the kernel entries have been copied just before the loop?
Note: This patch was tested on x86_64 virtual machine and succeeded.
Fixes: 92ea788d2a ("perf inject: Add --convert-callchain option")
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Jan Polensky <japo@linux.ibm.com>
Cc: linux-s390@vger.kernel.org
Cc: Sumanth Korikkar <sumanthk@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Check return value of `dwfl_report_end` during offline initialization.
Validate `dwfl_module_relocation_info` result before passing to `strcmp`
to avoid potential segmentation faults.
Additionally:
- Fix a file descriptor leak in `debuginfo__init_offline_dwarf()` when
`dwfl_report_offline()` or subsequent setup calls fail.
Fixes: 6f1b6291cf ("perf tools: Add util/debuginfo.[ch] files")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Check return values of `dwarf_aggregate_size` and `dwarf_formudata`.
Additionally:
- Avoid `vfprintf` undefined behavior with `NULL` strings by using
the `die_name()` helper for `dwarf_diename()` in `pr_*` calls.
- Use `die_get_data_member_location()` (updated to use
`dwarf_attr_integrate`) to correctly parse location expressions
for inherited member locations in the fallback path when
`dwarf_formudata()` fails.
Fixes: 2bc3cf575a ("perf annotate-data: Improve debug message with location info")
Fixes: 4a111cadac ("perf annotate-data: Add member field in the data type")
Fixes: 8b1042c425 ("perf annotate-data: Set bitfield member offset and size properly")
Fixes: fc044c53b9 ("perf annotate-data: Add dso->data_types tree")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Check return values of `dwarf_formsdata`, `dwarf_entrypc`,
`dwarf_highpc`, `dwarf_bytesize`, `dwarf_attr`, `dwarf_decl_line`,
`dwarf_getfuncs`, and `dwarf_formref_die`. Validate `dwarf_diename` and
`dwarf_diecu` results to prevent potential crashes. Fix C90 mixed
declarations.
Additionally:
- Avoid vfprintf undefined behavior with NULL strings by using the
`die_name()` helper for `dwarf_diename()` in `pr_*` calls,
including when warning about tail calls.
- Prevent NULL pointer dereference in `convert_variable_fields()`
when processing array elements for variables in registers.
- Fallback to offset 0 in `line_range_search_cb()` instead of
skipping functions without `DW_AT_decl_line`.
- Relax `dwarf_getfuncs` error checking in
`find_probe_point_by_func()` and `find_line_range_by_func()` to
prevent premature CU search aborts, ensuring robustness against
corrupted CUs.
Fixes: 66f69b2197 ("perf probe: Support DW_AT_const_value constant value")
Fixes: 3d918a12a1 ("perf probe: Find fentry mcount fuzzed parameter location")
Fixes: bcfc082150 ("perf probe: Remove redundant dwarf functions")
Fixes: 221d061182 ("perf probe: Fix to search local variables in appropriate scope")
Fixes: b55a87ade3 ("perf probe: Remove die() from probe-finder code")
Fixes: 4c85935122 ("perf probe: Support glob wildcards for function name")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Check return values of `dwfl_report_end` and `dwfl_module_addrdie`
to prevent using uninitialized stack variables or reporting success on
failure.
Additionally:
- Ensure `*file` is freed and inline frames are cleared on error in
`libdw__addr2line()` to prevent memory leaks and duplicated
callchains when falling back to other unwinders.
- Use `die_name()` safe wrapper inside the inline function unwinding
callback (`libdw_a2l_cb`).
- Refactor `libdw_a2l_cb`'s repeated memory error handling/cleanup
paths using a cleaner goto control flow.
Fixes: b7a2b011e9 ("perf powerpc: Unify the skip-callchain-idx libdw with that for addr2line")
Fixes: 88c51002d0 ("perf addr2line: Add a libdw implementation")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Allow DWARF line 0 in `libdw_a2l_cb()`, as it is a valid
reference for compiler-generated code.
Filter `die_get_call_lineno` error codes (negative values), but
fallback to line 0 if `call_fname` is present to preserve the
caller's filename instead of discarding it entirely.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix the parent srcline lookup in `libdw_a2l_cb()` to target the
correct parent node depending on the callchain order
(ORDER_CALLER/ORDER_CALLEE).
This ensures inline callchains are not corrupted when nest depth > 2.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Introduce inline_node__clear_frames() to clean up partial allocations.
This is a prerequisite for error handling in libdw inline unwinding.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Check return values of `dwarf_decl_line` (where non-optional),
`dwarf_getfuncs`, and `dwarf_lineaddr` to prevent using uninitialized
stack variables or incorrectly reporting success on failure.
For the root DIE in `die_walk_lines()`, `dwarf_decl_line` and
`die_get_decl_file` are optional and their failures are handled
gracefully to avoid breaking line walking on valid functions.
Specifically, remove the strict `!decf` (declared file) check that
would prematurely abort line walking on generated or artificial
functions lacking this optional attribute.
Additionally:
- Add NULL pointer protection for `strcmp()` in `die_walk_lines()`
when `inf` or `decf` are NULL to prevent crashes on generated
code.
- Use `dwarf_attr_integrate` in `die_get_data_member_location` to
correctly resolve inherited member locations (e.g. via abstract
origin or specification).
Fixes: 57f95bf5f8 ("perf probe: Show correct statement line number by perf probe -l")
Fixes: 3f4460a28f ("perf probe: Filter out redundant inline-instances")
Fixes: 75186a9b09 ("perf probe: Fix to show lines of sys_ functions correctly")
Fixes: e0d153c690 ("perf-probe: Move dwarf library routines to dwarf-aux.{c, h}")
Fixes: 6243b9dc4c ("perf probe: Move dwarf specific functions to dwarf-aux.c")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
A segmentation fault was observed in `libdw` when running `perf kmem`
with `--page stat` on some workloads. The crash occurred deep inside
`libdw` (specifically in `dwarf_child` and `dwarf_diename`) when
processing DWARF information.
The root cause was improper error handling of `dwarf_getfuncs` in
`die_find_realfunc` and `die_find_tailfunc`.
`dwarf_getfuncs` returns:
- `0` on success (when all functions have been processed).
- A positive offset if the callback aborts early (e.g., via
`DWARF_CB_ABORT` when a match is found).
- `-1` on error.
The original code used `if (!dwarf_getfuncs(...)) return NULL;`. On
error (`-1`), `!-1` evaluates to `0` (false), bypassing the error
check. Execution then proceeded as if a match was found, returning
uninitialized stack memory (`die_mem`) to the caller
(`cu_walk_functions_at`). When `cu_walk_functions_at` passed this
uninitialized memory to `libdw` via `dwarf_diename`, it caused a
segmentation fault.
Fix this by correcting the error check to `if (dwarf_getfuncs(...) <= 0)`.
Fixes: e0d153c690 ("perf-probe: Move dwarf library routines to dwarf-aux.{c, h}")
Fixes: d4c537e6bf ("perf probe: Ignore tail calls to probed functions")
Assisted-by: Gemini-CLI:Google Gemini 3
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zecheng Li <zli94@ncsu.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
- Fix devm_alloc_workqueue() passing a va_list as a positional arg to the
variadic alloc_workqueue() macro, which garbled wq->name and skipped
lockdep init on the devm path. Fold both noprof entry points onto a
va_list helper. Also, annotate __printf(1, 0).
-----BEGIN PGP SIGNATURE-----
iIQEABYKACwWIQTfIjM1kS57o3GsC/uxYfJx3gVYGQUCafpheA4cdGpAa2VybmVs
Lm9yZwAKCRCxYfJx3gVYGeBeAQDKq3XmN45CaOB76k9wSmRYqVgTSoWWTL83O8km
nC2UhQEAiJBfqhP586coXyK6saaZX9QwFdPzTcB72LhZSAj13gQ=
=Zixt
-----END PGP SIGNATURE-----
Merge tag 'wq-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Pull workqueue fixes from Tejun Heo:
- Fix devm_alloc_workqueue() passing a va_list as a positional arg to
the variadic alloc_workqueue() macro, which garbled wq->name and
skipped lockdep init on the devm path. Fold both noprof entry points
onto a va_list helper.
Also, annotate it using __printf(1, 0)
* tag 'wq-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: Annotate alloc_workqueue_va() with __printf(1, 0)
workqueue: fix devm_alloc_workqueue() va_list misuse
- During v6.19, cgroup task unlink was moved from do_exit() to after the
final task switch to satisfy a controller invariant. That left the kernel
seeing tasks past exit_signals() longer than userspace expected, and
several v7.0 follow-ups tried to bridge the gap by making rmdir wait for
the kernel side. None held up. The latest is an A-A deadlock when rmdir
is invoked by the reaper of zombies whose pidns teardown the rmdir itself
is waiting on, which points at the synchronizing approach being
fundamentally wrong:
- Take a different approach: drop the wait, leave rmdir's user-visible
side returning as soon as cgroup.procs is empty, and defer the css
percpu_ref kill that drives ->css_offline() until the cgroup is fully
depopulated.
- Tagged for stable. Somewhat invasive but contained. The hope is that
fixing forward sticks. If not, the fallback is to revert the entire
chain and rework on the development branch.
- Doesn't plug a pre-existing analogous race in
cgroup_apply_control_disable() (controller disable via subtree_control).
Not a regression. The development branch will do the more invasive
restructuring needed for that.
- Documentation update for cgroup-v1 charge-commit section that still
referenced functions removed when the memcg hugetlb try-commit-cancel
protocol was retired.
-----BEGIN PGP SIGNATURE-----
iIQEABYKACwWIQTfIjM1kS57o3GsC/uxYfJx3gVYGQUCafphbw4cdGpAa2VybmVs
Lm9yZwAKCRCxYfJx3gVYGbydAQDxlEIeJPdJlwbU6X4PBW/7DYeDHABG7OdrFf5K
Fq4ECAD/ZHsFyCNEOcZym6t2/FCZR0xbaPGQibLt3er6AkLRFwM=
=3Jra
-----END PGP SIGNATURE-----
Merge tag 'cgroup-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup fixes from Tejun Heo:
- During v6.19, cgroup task unlink was moved from do_exit() to after the
final task switch to satisfy a controller invariant. That left the kernel
seeing tasks past exit_signals() longer than userspace expected, and
several v7.0 follow-ups tried to bridge the gap by making rmdir wait for
the kernel side. None held up.
The latest is an A-A deadlock when rmdir is invoked by the reaper of
zombies whose pidns teardown the rmdir itself is waiting on, which
points at the synchronizing approach being fundamentally wrong.
Take a different approach: drop the wait, leave rmdir's user-visible
side returning as soon as cgroup.procs is empty, and defer the css
percpu_ref kill that drives ->css_offline() until the cgroup is fully
depopulated.
Tagged for stable. Somewhat invasive but contained. The hope is that
fixing forward sticks. If not, the fallback is to revert the entire
chain and rework on the development branch.
Note that this doesn't plug a pre-existing analogous race in
cgroup_apply_control_disable() (controller disable via
subtree_control). Not a regression. The development branch will do
the more invasive restructuring needed for that.
- Documentation update for cgroup-v1 charge-commit section that still
referenced functions removed when the memcg hugetlb try-commit-cancel
protocol was retired.
* tag 'cgroup-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
docs: cgroup-v1: Update charge-commit section
cgroup: Defer css percpu_ref kill on rmdir until cgroup is depopulated
- Fix idle CPU selection returning prev_cpu outside the task's cpus_ptr
when the BPF caller's allowed mask was wider. Stable backport.
- Two opposite-direction gaps in scx_task_iter's cgroup-scoped mode
versus the global mode:
- Tasks past exit_signals() are filtered by the cgroup walk but kept by
global. Sub-scheduler enable abort leaked __scx_init_task() state.
Add a CSS_TASK_ITER_WITH_DEAD flag to cgroup's task iterator
(scx_task_iter is its only user) and use it.
- Tasks past sched_ext_dead() are still returned, tripping
WARN_ON_ONCE() in callers or making them touch torn-down state. Mark
and skip under the per-task rq lock.
-----BEGIN PGP SIGNATURE-----
iIQEABYKACwWIQTfIjM1kS57o3GsC/uxYfJx3gVYGQUCafphXA4cdGpAa2VybmVs
Lm9yZwAKCRCxYfJx3gVYGbI/AP4nRHDusUuYDSFBLyHODvLZXfMM3Nb0yzS7euQJ
qvx6OQEA1p5AyRWA2apFvKjjQrl1dOb5vUlro1Fj8VF51X7Spwc=
=olGB
-----END PGP SIGNATURE-----
Merge tag 'sched_ext-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext
Pull sched_ext fixes from Tejun Heo:
- Fix idle CPU selection returning prev_cpu outside the task's cpus_ptr
when the BPF caller's allowed mask was wider. Stable backport.
- Two opposite-direction gaps in scx_task_iter's cgroup-scoped mode
versus the global mode:
- Tasks past exit_signals() are filtered by the cgroup walk but kept
by global. Sub-scheduler enable abort leaked __scx_init_task()
state. Add a CSS_TASK_ITER_WITH_DEAD flag to cgroup's task
iterator (scx_task_iter is its only user) and use it.
- Tasks past sched_ext_dead() are still returned, tripping
WARN_ON_ONCE() in callers or making them touch torn-down state.
Mark and skip under the per-task rq lock.
* tag 'sched_ext-for-7.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext:
sched_ext: idle: Recheck prev_cpu after narrowing allowed mask
sched_ext: Skip past-sched_ext_dead() tasks in scx_task_iter_next_locked()
cgroup, sched_ext: Include exiting tasks in cgroup iter
All in drivers. The largest change is the ufs one which has to
introduce a new function to check the power state before doing the
update and the most widely encountered one is the obvious change to sg
to not use GFP_ATOMIC.
Signed-off-by: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
-----BEGIN PGP SIGNATURE-----
iLgEABMIAGAWIQTnYEDbdso9F2cI+arnQslM7pishQUCafpc1xsUgAAAAAAEAA5t
YW51MiwyLjUrMS4xMiwyLDImHGphbWVzLmJvdHRvbWxleUBoYW5zZW5wYXJ0bmVy
c2hpcC5jb20ACgkQ50LJTO6YrIUGSgEA7UBQbSsaUO3ikG6BjXw9ter8VMEcqMg5
qRFgGYFCnTYA/iiiYJsm2/WBIpyUgOaIioSdm4rCuq/vNhaJo87q5MCM
=tBNu
-----END PGP SIGNATURE-----
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"All in drivers.
The largest change is the ufs one which has to introduce a new
function to check the power state before doing the update and the most
widely encountered one is the obvious change to sg to not use
GFP_ATOMIC"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: target: iscsi: reject invalid size Extended CDB AHS
scsi: ufs: core: Fix bRefClkFreq write failure in HS-LSS mode
scsi: hisi_sas: Fix sparse warnings in prep_ata_v3_hw()
scsi: pmcraid: Fix typo in comments
scsi: scsi_dh_alua: Increase default ALUA timeout to maximum spec value
scsi: smartpqi: Silence a recursive lock warning
scsi: mpt3sas: Limit NVMe request size to 2 MiB
scsi: sg: Don't use GFP_ATOMIC in sg_start_req()
scsi: target: configfs: Bound snprintf() return in tg_pt_gp_members_show()
- Fix rotating fonts by 180 degrees [Thomas Zimmermann]
- Drop duplicate include of linux/module.h in fb_defio [Chen Ni]
- Add vm_ops in udlfb to prevent use-after-free [Rajat Gupta]
- ipu-v3: clean up kernel-doc warnings [Randy Dunlap]
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQS86RI+GtKfB8BJu973ErUQojoPXwUCafpViwAKCRD3ErUQojoP
X503AQCoi8Lq/xUAYzVq7xsoGrqBzkIjJ7x+OkVUCdmsuWRQIAEA8tjtzN6ZUdo7
cJMadOEen9zyLN3qpksqID4sVA//FwM=
=T+Gn
-----END PGP SIGNATURE-----
Merge tag 'fbdev-for-7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev
Pull fbdev fixes from Helge Deller:
"Four small patches for fbdev, of which two are important: One fixes
the bitmap font generation and the other prevents a possible
use-after-free in udlfb:
- Fix rotating fonts by 180 degrees (Thomas Zimmermann)
- Drop duplicate include of linux/module.h in fb_defio (Chen Ni)
- Add vm_ops in udlfb to prevent use-after-free (Rajat Gupta)
- ipu-v3: clean up kernel-doc warnings (Randy Dunlap)"
* tag 'fbdev-for-7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev:
fbdev: udlfb: add vm_ops to dlfb_ops_mmap to prevent use-after-free
lib/fonts: Fix bit position when rotating by 180 degrees
fbdev: defio: Remove duplicate include of linux/module.h
fbdev: ipu-v3: clean up kernel-doc warnings
A collection of bugs, many found with inspection tools:
- Several error unwind misses on system calls in mlx5, mana,, ocrdma,
vmw_pvrdma, mlx4, and hns
- More rxe bugs processing network packets
- User triggerable races in mlx5 when destroying and creating the same
same object when the FW returns the same object ID
- Incorrect passing of an IPv6 address through netlink
RDMA_NL_LS_OP_IP_RESOLVE
- Add memory ordering for mlx5's lock avoidance pattenr
- Protect mana from kernel memory overflow
- Use safe patterns for xarray/radix_tree look up in mlx5 and hns
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQRRRCHOFoQz/8F5bUaFwuHvBreFYQUCafoC6AAKCRCFwuHvBreF
YSHCAP9AU6NnDa+ltlY65HrKXNP3R7kgKAKM3x0e7keZrj4ahQEA9zCBNEJ/PY/Q
jM0jgMor2W1aWfuuqn47g82rPN5e4QY=
=lqDa
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Pull rdma fixes from Jason Gunthorpe:
- Several error unwind misses on system calls in mlx5, mana, ocrdma,
vmw_pvrdma, mlx4, and hns
- More rxe bugs processing network packets
- User triggerable races in mlx5 when destroying and creating the same
same object when the FW returns the same object ID
- Incorrect passing of an IPv6 address through netlink
RDMA_NL_LS_OP_IP_RESOLVE
- Add memory ordering for mlx5's lock avoidance pattenr
- Protect mana from kernel memory overflow
- Use safe patterns for xarray/radix_tree look up in mlx5 and hns
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (24 commits)
RDMA/hns: Fix unlocked call to hns_roce_qp_remove()
RDMA/hns: Fix xarray race in hns_roce_create_qp_common()
RDMA/hns: Fix xarray race in hns_roce_create_srq()
RDMA/mlx4: Fix mis-use of RCU in mlx4_srq_event()
RDMA/mlx4: Fix resource leak on error in mlx4_ib_create_srq()
RDMA/vmw_pvrdma: Fix double free on pvrdma_alloc_ucontext() error path
RDMA/ocrdma: Don't NULL deref uctx on errors in ocrdma_copy_pd_uresp()
RDMA/ocrdma: Clarify the mm_head searching
RDMA/mana: Fix error unwind in mana_ib_create_qp_rss()
RDMA/mana: Fix mana_destroy_wq_obj() cleanup in mana_ib_create_qp_rss()
RDMA/mana: Remove user triggerable WARN_ON() in mana_ib_create_qp_rss()
RDMA/mana: Validate rx_hash_key_len
RDMA/mlx5: Add missing store/release for lock elision pattern
RDMA/mlx5: Restore zero-init to mlx5_ib_modify_qp() ucmd
RDMA/ionic: Fix typo in format string
RDMA/mlx5: Fix null-ptr-deref in Raw Packet QP creation
RDMA/core: Fix rereg_mr use-after-free race
IB/core: Fix IPv6 netlink message size in ib_nl_ip_send_msg()
RDMA/mlx5: Fix UAF in DCT destroy due to race with create
RDMA/mlx5: Fix UAF in SRQ destroy due to race with create
...
Fixes extra test number increment in ksft_exit_skip() that results in
incorrect KTAP result.
Fixes regression introduced by addition of explicit constructor orders
for fixture tests. This addition broke the ordering of those relative to
non-fixture tests and the reverse-constructor-order detection.
-----BEGIN PGP SIGNATURE-----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=u8Qv
-----END PGP SIGNATURE-----
Merge tag 'linux_kselftest-fixes-7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest
Pull kselftest fixes from Shuah Khan:
- Fix extra test number increment in ksft_exit_skip() that results in
incorrect KTAP result
- Fix regression introduced by addition of explicit constructor orders
for fixture tests. This addition broke the ordering of those relative
to non-fixture tests and the reverse-constructor-order detection
* tag 'linux_kselftest-fixes-7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
selftests: harness: Restore order of test functions
selftests: kselftest: fix wrong test number in ksft_exit_skip
Commit 1d8f136a42 ("memcg/hugetlb: remove memcg hugetlb
try-commit-cancel protocol") removed mem_cgroup_commit_charge() and
mem_cgroup_cancel_charge(), but the docs still refer to those functions.
There is no longer any charge cancellation.
Update the docs to match the code.
Signed-off-by: T.J. Mercier <tjmercier@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
scx_select_cpu_dfl() narrows @allowed to @cpus_allowed & @p->cpus_ptr
when the BPF caller supplies a @cpus_allowed that differs from
@p->cpus_ptr and @p doesn't have full affinity. However,
@is_prev_allowed was computed against the original (wider)
@cpus_allowed, so the prev_cpu fast paths could pick a @prev_cpu that
is in @cpus_allowed but not in @p->cpus_ptr, violating the intended
invariant that the returned CPU is always usable by @p. The kernel
masks this via the SCX_EV_SELECT_CPU_FALLBACK fallback, but the
behavior contradicts the documented contract.
Move the @is_prev_allowed evaluation past the narrowing block so it
tests against the final @allowed mask.
Fixes: ee9a4e9279 ("sched_ext: idle: Properly handle invalid prev_cpu during idle selection")
Cc: stable@vger.kernel.org # v6.16+
Assisted-by: Claude <noreply@anthropic.com>
Signed-off-by: David Carlier <devnexen@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The first two fixes are workarounds for buggy IPMI hardware. The
hardware says it has data for the IPMI driver to read constantly, so the
driver reads the data constantly, causing any new requests to be
blocked.
The first fix was to check for invalid data right when the data was read
from the device and stop the operation there (there was a later check
for invalid data, but it could not stop the operation at that point).
It turned out the device was providing good data, so that didn't fix the
issue, but it's still a good check.
The second fix stops fetching this data after a few fetches and allows
other operations to occur. The driver won't work very well, but at
least it won't wedge. This seems to fix the issue.
The third issue is a problem I spotted while working on the previous
issue where if a certain memory allocation failed the driver would stop
working.
The fourth issue is a problem was a missing set to NULL on a PTR_ERR()
return, introduced in the previous series for 7.1.
-----BEGIN PGP SIGNATURE-----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=jA6/
-----END PGP SIGNATURE-----
Merge tag 'for-linus-7.1-2' of https://github.com/cminyard/linux-ipmi
Pull IPMI fixes from Corey Minyard:
"Fix a number of issues that came up recently
The first two fixes are workarounds for buggy IPMI hardware. The
hardware says it has data for the IPMI driver to read constantly, so
the driver reads the data constantly, causing any new requests to be
blocked.
The first fix was to check for invalid data right when the data was
read from the device and stop the operation there (there was a later
check for invalid data, but it could not stop the operation at that
point). It turned out the device was providing good data, so that
didn't fix the issue, but it's still a good check.
The second fix stops fetching this data after a few fetches and allows
other operations to occur. The driver won't work very well, but at
least it won't wedge. This seems to fix the issue.
The third issue is a problem I spotted while working on the previous
issue where if a certain memory allocation failed the driver would
stop working.
The fourth issue is a problem was a missing set to NULL on a PTR_ERR()
return, introduced in the previous series for 7.1"
* tag 'for-linus-7.1-2' of https://github.com/cminyard/linux-ipmi:
ipmi:ssif: NULL thread on error
ipmi:si: Return state to normal if message allocation fails
ipmi: Add limits to event and receive message requests
ipmi: Check event message buffer response for bad data
scx_task_iter's cgroup-scoped mode can return tasks whose
sched_ext_dead() has already completed: cgroup_task_dead() removes
from cset->tasks after sched_ext_dead() in finish_task_switch() and is
irq-work deferred on PREEMPT_RT. The global mode is fine -
sched_ext_dead() removes from scx_tasks via list_del_init() first.
Callers (sub-sched enable prep/abort/apply, scx_sub_disable(),
scx_fail_parent()) assume returned tasks are still on @sch and trip
WARN_ON_ONCE() or operate on torn-down state otherwise.
Set %SCX_TASK_OFF_TASKS in sched_ext_dead() under @p's rq lock and
have scx_task_iter_next_locked() skip flagged tasks under the same
lock. Setter and reader serialize on the per-task rq lock - no race.
Signed-off-by: Tejun Heo <tj@kernel.org>
a72f73c4dd ("cgroup: Don't expose dead tasks in cgroup") made
css_task_iter_advance() skip exiting tasks so cgroup.procs stays consistent
with waitpid() visibility. Unfortunately, this broke scx_task_iter.
scx_task_iter walks either scx_tasks (global) or a cgroup subtree via
css_task_iter() and the two modes are expected to cover the same set of
tasks. After the above change the cgroup-scoped mode silently skips tasks
past exit_signals() that are still on scx_tasks.
scx_sub_enable_workfn()'s abort path is one of the symptoms: an exiting
SCX_TASK_SUB_INIT task can race past the cgroup iter leaking
__scx_init_task() state. Other iterations share the same gap.
Add CSS_TASK_ITER_WITH_DEAD to opt out of the skip and use it from
scx_task_iter().
Fixes: b0e4c2f8a0 ("sched_ext: Implement cgroup subtree iteration for scx_task_iter")
Reported-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
A chain of commits going back to v7.0 reworked rmdir to satisfy the
controller invariant that a subsystem's ->css_offline() must not run while
tasks are still doing kernel-side work in the cgroup.
[1] d245698d72 ("cgroup: Defer task cgroup unlink until after the task is done switching out")
[2] a72f73c4dd ("cgroup: Don't expose dead tasks in cgroup")
[3] 1b164b876c ("cgroup: Wait for dying tasks to leave on rmdir")
[4] 4c56a8ac68 ("cgroup: Fix cgroup_drain_dying() testing the wrong condition")
[5] 13e786b64b ("cgroup: Increment nr_dying_subsys_* from rmdir context")
[1] moved task cset unlink from do_exit() to finish_task_switch() so a
task's cset link drops only after the task has fully stopped scheduling.
That made tasks past exit_signals() linger on cset->tasks until their final
context switch, which led to a series of problems as what userspace expected
to see after rmdir diverged from what the kernel needs to wait for. [2]-[5]
tried to bridge that divergence: [2] filtered the exiting tasks from
cgroup.procs; [3] had rmdir(2) sleep in TASK_UNINTERRUPTIBLE for them; [4]
fixed the wait's condition; [5] made nr_dying_subsys_* visible
synchronously.
The cgroup_drain_dying() wait in [3] turned out to be a dead end. When the
rmdir caller is also the reaper of a zombie that pins a pidns teardown (e.g.
host PID 1 systemd reaping orphan pids that were re-parented to it during
the same teardown), rmdir blocks in TASK_UNINTERRUPTIBLE waiting for those
pids to free, the pids can't free because PID 1 is the reaper and it's stuck
in rmdir, and the system A-A deadlocks. No internal lock ordering breaks
this; the wait itself is the bug.
The css killing side that drove the original reorder, however, can be made
cleanly asynchronous: ->css_offline() is already async, run from
css_killed_work_fn() driven by percpu_ref_kill_and_confirm(). The fix is to
make that chain start only after all tasks have left the cgroup. rmdir's
user-visible side then returns as soon as cgroup.procs and friends are
empty, while ->css_offline() still runs only after the cgroup is fully
drained.
Verified by the original reproducer (pidns teardown + zombie reaper, runs
under vng) which hangs vanilla and succeeds here, and by per-commit
deterministic repros for [2], [3], [4], [5] with a boot parameter that
widens the post-exit_signals() window so each state is reliably reachable.
Some stress tests on top of that.
cgroup_apply_control_disable() has the same shape of pre-existing race:
when a controller is disabled via subtree_control, kill_css() ran
synchronously while tasks past exit_signals() could still be linked to
the cgroup's csets, and ->css_offline() could fire before they drained.
This patch preserves the existing synchronous behavior at that call site
(kill_css_sync() + kill_css_finish() back-to-back) and a follow-up patch
will defer kill_css_finish() there using a per-css trigger.
This seems like the right approach and I don't see problems with it. The
changes are somewhat invasive but not excessively so, so backporting to
-stable should be okay. If something does turn out to be wrong, the fallback
is to revert the entire chain ([1]-[5]) and rework in the development branch
instead.
v2: Pin cgrp across the deferred destroy work with explicit
cgroup_get()/cgroup_put() around queue_work() and the work_fn. v1
wasn't actually broken (ordered cgroup_offline_wq + queue_work order
in cgroup_task_dead() saved it) but the explicit ref removes the
dependency on those non-obvious invariants. Also note the
pre-existing cgroup_apply_control_disable() race in the description;
a follow-up will defer kill_css_finish() there.
Fixes: 1b164b876c ("cgroup: Wait for dying tasks to leave on rmdir")
Cc: stable@vger.kernel.org # v7.0+
Reported-and-tested-by: Martin Pitt <martin@piware.de>
Link: https://lore.kernel.org/all/afHNg2VX2jy9bW7y@piware.de/
Link: https://lore.kernel.org/all/35e0670adb4abeab13da2c321582af9f@kernel.org/
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
dlfb_ops_mmap() uses remap_pfn_range() to map vmalloc framebuffer pages
to userspace but sets no vm_ops on the VMA. This means the kernel cannot
track active mmaps. When dlfb_realloc_framebuffer() replaces the backing
buffer via FBIOPUT_VSCREENINFO, existing mmap PTEs are not invalidated.
On USB disconnect, dlfb_ops_destroy() calls vfree() on the old pages
while userspace PTEs still reference them, resulting in a use-after-free:
the process retains read/write access to freed kernel pages.
Add vm_operations_struct with open/close callbacks that maintain an
atomic mmap_count on struct dlfb_data. In dlfb_realloc_framebuffer(),
check mmap_count and return -EBUSY if the buffer is currently mapped,
preventing buffer replacement while userspace holds stale PTEs.
Tested with PoC using dummy_hcd + raw_gadget USB device emulation.
Signed-off-by: Rajat Gupta <rajgupt@qti.qualcomm.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org
Signed-off-by: Helge Deller <deller@gmx.de>
* Check that nEPT/nNPT is enabled in slow flush hypercalls. If it is
not, the hypercalls can be processed as usual even while running a
nested guest.
* Fix shadow paging use-after-free due to page tables changing outside
execution of the guest. A bug that is 16 years old and stems from an
imprecision in the very first KVM series.
* Scan IRR whenever PID.ON is true, even if PIR is empty, which avoids
a somewhat rare WARN.
-----BEGIN PGP SIGNATURE-----
iQFIBAABCgAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmn3sQoUHHBib256aW5p
QHJlZGhhdC5jb20ACgkQv/vSX3jHroNMdQf/WuwNbJCb2bVWCoNeQUWvtxrlmZ5p
IqQrnEACYOkn20eWDrXrAiEqPoIj0fPTYSewU9mYoC5DhyGFXzB5TEwMMHUn9kNn
QVBQS/IlfuhO5BGrov4dbdmzUmXaTuUe29tjoOnr1IGNapf6naA9+m03u5/TaZcd
FtS28EPy5Z2ZRm3KiKrZahsMMVr9ZXI3xfiEFVht9lDwiYpZquxhH2bju3QnBQ71
0x02ZxMKbqwi5uJLCGmF62AacCSqag1+eNs05WFPThL5RNNkwpac1lceOetYFm9/
y+iqOfev4SYeL3rACyQ/nStpgdkhpL2cbkPISRvjQ0WKVVawk+IAzNzvsQ==
=sGxH
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull kvm fixes from Paolo Bonzini:
"Three bug fixes for x86:
- Check that nEPT/nNPT is enabled in slow flush hypercalls. If it is
not, the hypercalls can be processed as usual even while running a
nested guest
- Fix shadow paging use-after-free due to page tables changing
outside execution of the guest. A bug that is 16 years old and
stems from an imprecision in the very first KVM series
- Scan IRR whenever PID.ON is true, even if PIR is empty, which
avoids a somewhat rare WARN"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: x86: Fix shadow paging use-after-free due to unexpected GFN
KVM: x86: Fix misleading variable names and add more comments for PIR=>IRR flow
KVM: x86: Do IRR scan in __kvm_apic_update_irr even if PIR is empty
KVM: x86: check for nEPT/nNPT in slow flush hypercalls
The shadow MMU computes GFNs for direct shadow pages using sp->gfn plus
the SPTE index. This assumption breaks for shadow paging if the guest
page tables are modified between VM entries (similar to commit
aad885e774, "KVM: x86/mmu: Drop/zap existing present SPTE even
when creating an MMIO SPTE", 2026-03-27). The flow is as follows:
- a PDE is installed for a 2MB mapping, and a page in that area is
accessed. KVM creates a kvm_mmu_page consisting of 512 4KB pages;
the kvm_mmu_page is marked by FNAME(fetch) as direct-mapped because
the guest's mapping is a huge page (and thus contiguous).
- the PDE mapping is changed from outside the guest.
- the guest accesses another page in the same 2MB area. KVM installs
a new leaf SPTE and rmap entry; the SPTE uses the "correct" GFN
(i.e. based on the new mapping, as changed in the previous step) but
that GFN is outside of the [sp->gfn, sp->gfn + 511] range; therefore
the rmap entry cannot be found and removed when the kvm_mmu_page
is zapped.
- the memslot that covers the first 2MB mapping is deleted, and the
kvm_mmu_page for the now-invalid GPA is zapped. However, rmap_remove()
only looks at the [sp->gfn, sp->gfn + 511] range established in step 1,
and fails to find the rmap entry that was recorded by step 3.
- any operation that causes an rmap walk for the same page accessed
by step 3 then walks a stale rmap and dereferences a freed kvm_mmu_page.
This includes dirty logging or MMU notifier invalidations (e.g., from
MADV_DONTNEED).
The underlying issue is that KVM's walking of shadow PTEs assumes that
if a SPTE is present when KVM wants to install a non-leaf SPTE, then the
existing kvm_mmu_page must be for the correct gfn. Because the only way
for the gfn to be wrong is if KVM messed up and failed to zap a SPTE...
which shouldn't happen, but *actually* only happens in response to a
guest write.
That bug dates back literally forever, as even the first version of KVM
assumes that the GFN matches and walks into the "wrong" shadow page.
However, that was only an imprecision until 2032a93d66 ("KVM: MMU:
Don't allocate gfns page for direct mmu pages") came along.
Fix it by checking for a target gfn mismatch and zapping the existing
SPTE. That way the old SP and rmap entries are gone, KVM installs
the rmap in the right location, and everyone is happy.
Fixes: 2032a93d66 ("KVM: MMU: Don't allocate gfns page for direct mmu pages")
Fixes: 6aa8b732ca ("kvm: userspace interface")
Reported-by: Alexander Bulekov <bkov@amazon.com>
Reported-by: Fred Griffoul <fgriffo@amazon.co.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Link: https://patch.msgid.link/20260503201029.106481-1-pbonzini@redhat.com/
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
