When switching from kthreads to vhost_tasks two bugs were added:
1. The vhost worker tasks's now show up as processes so scripts doing
ps or ps a would not incorrectly detect the vhost task as another
process. 2. kthreads disabled freeze by setting PF_NOFREEZE, but
vhost tasks's didn't disable or add support for them.
To fix both bugs, this switches the vhost task to be thread in the
process that does the VHOST_SET_OWNER ioctl, and has vhost_worker call
get_signal to support SIGKILL/SIGSTOP and freeze signals. Note that
SIGKILL/STOP support is required because CLONE_THREAD requires
CLONE_SIGHAND which requires those 2 signals to be supported.
This is a modified version of the patch written by Mike Christie
<michael.christie@oracle.com> which was a modified version of patch
originally written by Linus.
Much of what depended upon PF_IO_WORKER now depends on PF_USER_WORKER.
Including ignoring signals, setting up the register state, and having
get_signal return instead of calling do_group_exit.
Tidied up the vhost_task abstraction so that the definition of
vhost_task only needs to be visible inside of vhost_task.c. Making
it easier to review the code and tell what needs to be done where.
As part of this the main loop has been moved from vhost_worker into
vhost_task_fn. vhost_worker now returns true if work was done.
The main loop has been updated to call get_signal which handles
SIGSTOP, freezing, and collects the message that tells the thread to
exit as part of process exit. This collection clears
__fatal_signal_pending. This collection is not guaranteed to
clear signal_pending() so clear that explicitly so the schedule()
sleeps.
For now the vhost thread continues to exist and run work until the
last file descriptor is closed and the release function is called as
part of freeing struct file. To avoid hangs in the coredump
rendezvous and when killing threads in a multi-threaded exec. The
coredump code and de_thread have been modified to ignore vhost threads.
Remvoing the special case for exec appears to require teaching
vhost_dev_flush how to directly complete transactions in case
the vhost thread is no longer running.
Removing the special case for coredump rendezvous requires either the
above fix needed for exec or moving the coredump rendezvous into
get_signal.
Fixes: 6e890c5d50 ("vhost: use vhost_tasks for worker threads")
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Co-developed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull tracing updates from Steven Rostedt:
- User events are finally ready!
After lots of collaboration between various parties, we finally
locked down on a stable interface for user events that can also work
with user space only tracing.
This is implemented by telling the kernel (or user space library, but
that part is user space only and not part of this patch set), where
the variable is that the application uses to know if something is
listening to the trace.
There's also an interface to tell the kernel about these events,
which will show up in the /sys/kernel/tracing/events/user_events/
directory, where it can be enabled.
When it's enabled, the kernel will update the variable, to tell the
application to start writing to the kernel.
See https://lwn.net/Articles/927595/
- Cleaned up the direct trampolines code to simplify arm64 addition of
direct trampolines.
Direct trampolines use the ftrace interface but instead of jumping to
the ftrace trampoline, applications (mostly BPF) can register their
own trampoline for performance reasons.
- Some updates to the fprobe infrastructure. fprobes are more efficient
than kprobes, as it does not need to save all the registers that
kprobes on ftrace do. More work needs to be done before the fprobes
will be exposed as dynamic events.
- More updates to references to the obsolete path of
/sys/kernel/debug/tracing for the new /sys/kernel/tracing path.
- Add a seq_buf_do_printk() helper to seq_bufs, to print a large buffer
line by line instead of all at once.
There are users in production kernels that have a large data dump
that originally used printk() directly, but the data dump was larger
than what printk() allowed as a single print.
Using seq_buf() to do the printing fixes that.
- Add /sys/kernel/tracing/touched_functions that shows all functions
that was every traced by ftrace or a direct trampoline. This is used
for debugging issues where a traced function could have caused a
crash by a bpf program or live patching.
- Add a "fields" option that is similar to "raw" but outputs the fields
of the events. It's easier to read by humans.
- Some minor fixes and clean ups.
* tag 'trace-v6.4' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (41 commits)
ring-buffer: Sync IRQ works before buffer destruction
tracing: Add missing spaces in trace_print_hex_seq()
ring-buffer: Ensure proper resetting of atomic variables in ring_buffer_reset_online_cpus
recordmcount: Fix memory leaks in the uwrite function
tracing/user_events: Limit max fault-in attempts
tracing/user_events: Prevent same address and bit per process
tracing/user_events: Ensure bit is cleared on unregister
tracing/user_events: Ensure write index cannot be negative
seq_buf: Add seq_buf_do_printk() helper
tracing: Fix print_fields() for __dyn_loc/__rel_loc
tracing/user_events: Set event filter_type from type
ring-buffer: Clearly check null ptr returned by rb_set_head_page()
tracing: Unbreak user events
tracing/user_events: Use print_format_fields() for trace output
tracing/user_events: Align structs with tabs for readability
tracing/user_events: Limit global user_event count
tracing/user_events: Charge event allocs to cgroups
tracing/user_events: Update documentation for ABI
tracing/user_events: Use write ABI in example
tracing/user_events: Add ABI self-test
...
During tracefs discussions it was decided instead of requiring a mapping
within a user-process to track the lifetime of memory descriptors we
should hook the appropriate calls. Do this by adding the minimal stubs
required for task fork, exec, and exit. Currently this is just a NOP.
Future patches will implement these calls fully.
Link: https://lkml.kernel.org/r/20230328235219.203-3-beaub@linux.microsoft.com
Suggested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Beau Belgrave <beaub@linux.microsoft.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull arm64 updates from Catalin Marinas:
- Support for arm64 SME 2 and 2.1. SME2 introduces a new 512-bit
architectural register (ZT0, for the look-up table feature) that
Linux needs to save/restore
- Include TPIDR2 in the signal context and add the corresponding
kselftests
- Perf updates: Arm SPEv1.2 support, HiSilicon uncore PMU updates, ACPI
support to the Marvell DDR and TAD PMU drivers, reset DTM_PMU_CONFIG
(ARM CMN) at probe time
- Support for DYNAMIC_FTRACE_WITH_CALL_OPS on arm64
- Permit EFI boot with MMU and caches on. Instead of cleaning the
entire loaded kernel image to the PoC and disabling the MMU and
caches before branching to the kernel bare metal entry point, leave
the MMU and caches enabled and rely on EFI's cacheable 1:1 mapping of
all of system RAM to populate the initial page tables
- Expose the AArch32 (compat) ELF_HWCAP features to user in an arm64
kernel (the arm32 kernel only defines the values)
- Harden the arm64 shadow call stack pointer handling: stash the shadow
stack pointer in the task struct on interrupt, load it directly from
this structure
- Signal handling cleanups to remove redundant validation of size
information and avoid reading the same data from userspace twice
- Refactor the hwcap macros to make use of the automatically generated
ID registers. It should make new hwcaps writing less error prone
- Further arm64 sysreg conversion and some fixes
- arm64 kselftest fixes and improvements
- Pointer authentication cleanups: don't sign leaf functions, unify
asm-arch manipulation
- Pseudo-NMI code generation optimisations
- Minor fixes for SME and TPIDR2 handling
- Miscellaneous updates: ARCH_FORCE_MAX_ORDER is now selectable,
replace strtobool() to kstrtobool() in the cpufeature.c code, apply
dynamic shadow call stack in two passes, intercept pfn changes in
set_pte_at() without the required break-before-make sequence, attempt
to dump all instructions on unhandled kernel faults
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (130 commits)
arm64: fix .idmap.text assertion for large kernels
kselftest/arm64: Don't require FA64 for streaming SVE+ZA tests
kselftest/arm64: Copy whole EXTRA context
arm64: kprobes: Drop ID map text from kprobes blacklist
perf: arm_spe: Print the version of SPE detected
perf: arm_spe: Add support for SPEv1.2 inverted event filtering
perf: Add perf_event_attr::config3
arm64/sme: Fix __finalise_el2 SMEver check
drivers/perf: fsl_imx8_ddr_perf: Remove set-but-not-used variable
arm64/signal: Only read new data when parsing the ZT context
arm64/signal: Only read new data when parsing the ZA context
arm64/signal: Only read new data when parsing the SVE context
arm64/signal: Avoid rereading context frame sizes
arm64/signal: Make interface for restore_fpsimd_context() consistent
arm64/signal: Remove redundant size validation from parse_user_sigframe()
arm64/signal: Don't redundantly verify FPSIMD magic
arm64/cpufeature: Use helper macros to specify hwcaps
arm64/cpufeature: Always use symbolic name for feature value in hwcaps
arm64/sysreg: Initial unsigned annotations for ID registers
arm64/sysreg: Initial annotation of signed ID registers
...
Contemporary versions of GCC (e.g. GCC 12.2.0) drop the alignment
specified by '-falign-functions=N' for functions marked with the
__cold__ attribute, and potentially for callees of __cold__ functions as
these may be implicitly marked as __cold__ by the compiler. LLVM appears
to respect '-falign-functions=N' in such cases.
This has been reported to GCC in bug 88345:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88345
... which also covers alignment being dropped when '-Os' is used, which
will be addressed in a separate patch.
Currently, use of '-falign-functions=N' is limited to
CONFIG_FUNCTION_ALIGNMENT, which is largely used for performance and/or
analysis reasons (e.g. with CONFIG_DEBUG_FORCE_FUNCTION_ALIGN_64B), but
isn't necessary for correct functionality. However, this dropped
alignment isn't great for the performance and/or analysis cases.
Subsequent patches will use CONFIG_FUNCTION_ALIGNMENT as part of arm64's
ftrace implementation, which will require all instrumented functions to
be aligned to at least 8-bytes.
This patch works around the dropped alignment by avoiding the use of the
__cold__ attribute when CONFIG_FUNCTION_ALIGNMENT is non-zero, and by
specifically aligning abort(), which GCC implicitly marks as __cold__.
As the __cold macro is now dependent upon config options (which is
against the policy described at the top of compiler_attributes.h), it is
moved into compiler_types.h.
I've tested this by building and booting a kernel configured with
defconfig + CONFIG_EXPERT=y + CONFIG_DEBUG_FORCE_FUNCTION_ALIGN_64B=y,
and looking for misaligned text symbols in /proc/kallsyms:
* arm64:
Before:
# uname -rm
6.2.0-rc3 aarch64
# grep ' [Tt] ' /proc/kallsyms | grep -iv '[048c]0 [Tt] ' | wc -l
5009
After:
# uname -rm
6.2.0-rc3-00001-g2a2bedf8bfa9 aarch64
# grep ' [Tt] ' /proc/kallsyms | grep -iv '[048c]0 [Tt] ' | wc -l
919
* x86_64:
Before:
# uname -rm
6.2.0-rc3 x86_64
# grep ' [Tt] ' /proc/kallsyms | grep -iv '[048c]0 [Tt] ' | wc -l
11537
After:
# uname -rm
6.2.0-rc3-00001-g2a2bedf8bfa9 x86_64
# grep ' [Tt] ' /proc/kallsyms | grep -iv '[048c]0 [Tt] ' | wc -l
2805
There's clearly a substantial reduction in the number of misaligned
symbols. From manual inspection, the remaining unaligned text labels are
a combination of ACPICA functions (due to the use of '-Os'), static call
trampolines, and non-function labels in assembly, which will be dealt
with in subsequent patches.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Florent Revest <revest@chromium.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Will Deacon <will@kernel.org>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Link: https://lore.kernel.org/r/20230123134603.1064407-3-mark.rutland@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
If a task oopses with irqs disabled, this can cause various cascading
problems in the oops path such as sleep-from-invalid warnings, and
potentially worse.
Since commit 0258b5fd7c ("coredump: Limit coredumps to a single
thread group"), the unconditional irq enable in coredump_task_exit()
will "fix" the irq state to be enabled early in do_exit(), so currently
this may not be triggerable, but that is coincidental and fragile.
Detect and fix the irqs_disabled() condition in the oops path before
calling do_exit(), similarly to the way in_atomic() is handled.
Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Link: https://lore.kernel.org/lkml/20221004094401.708299-1-npiggin@gmail.com/
Many Linux systems are configured to not panic on oops; but allowing an
attacker to oops the system **really** often can make even bugs that look
completely unexploitable exploitable (like NULL dereferences and such) if
each crash elevates a refcount by one or a lock is taken in read mode, and
this causes a counter to eventually overflow.
The most interesting counters for this are 32 bits wide (like open-coded
refcounts that don't use refcount_t). (The ldsem reader count on 32-bit
platforms is just 16 bits, but probably nobody cares about 32-bit platforms
that much nowadays.)
So let's panic the system if the kernel is constantly oopsing.
The speed of oopsing 2^32 times probably depends on several factors, like
how long the stack trace is and which unwinder you're using; an empirically
important one is whether your console is showing a graphical environment or
a text console that oopses will be printed to.
In a quick single-threaded benchmark, it looks like oopsing in a vfork()
child with a very short stack trace only takes ~510 microseconds per run
when a graphical console is active; but switching to a text console that
oopses are printed to slows it down around 87x, to ~45 milliseconds per
run.
(Adding more threads makes this faster, but the actual oops printing
happens under &die_lock on x86, so you can maybe speed this up by a factor
of around 2 and then any further improvement gets eaten up by lock
contention.)
It looks like it would take around 8-12 days to overflow a 32-bit counter
with repeated oopsing on a multi-core X86 system running a graphical
environment; both me (in an X86 VM) and Seth (with a distro kernel on
normal hardware in a standard configuration) got numbers in that ballpark.
12 days aren't *that* short on a desktop system, and you'd likely need much
longer on a typical server system (assuming that people don't run graphical
desktop environments on their servers), and this is a *very* noisy and
violent approach to exploiting the kernel; and it also seems to take orders
of magnitude longer on some machines, probably because stuff like EFI
pstore will slow it down a ton if that's active.
Signed-off-by: Jann Horn <jannh@google.com>
Link: https://lore.kernel.org/r/20221107201317.324457-1-jannh@google.com
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221117234328.594699-2-keescook@chromium.org
Pull non-MM updates from Andrew Morton:
- hfs and hfsplus kmap API modernization (Fabio Francesco)
- make crash-kexec work properly when invoked from an NMI-time panic
(Valentin Schneider)
- ntfs bugfixes (Hawkins Jiawei)
- improve IPC msg scalability by replacing atomic_t's with percpu
counters (Jiebin Sun)
- nilfs2 cleanups (Minghao Chi)
- lots of other single patches all over the tree!
* tag 'mm-nonmm-stable-2022-10-11' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (71 commits)
include/linux/entry-common.h: remove has_signal comment of arch_do_signal_or_restart() prototype
proc: test how it holds up with mapping'less process
mailmap: update Frank Rowand email address
ia64: mca: use strscpy() is more robust and safer
init/Kconfig: fix unmet direct dependencies
ia64: update config files
nilfs2: replace WARN_ONs by nilfs_error for checkpoint acquisition failure
fork: remove duplicate included header files
init/main.c: remove unnecessary (void*) conversions
proc: mark more files as permanent
nilfs2: remove the unneeded result variable
nilfs2: delete unnecessary checks before brelse()
checkpatch: warn for non-standard fixes tag style
usr/gen_init_cpio.c: remove unnecessary -1 values from int file
ipc/msg: mitigate the lock contention with percpu counter
percpu: add percpu_counter_add_local and percpu_counter_sub_local
fs/ocfs2: fix repeated words in comments
relay: use kvcalloc to alloc page array in relay_alloc_page_array
proc: make config PROC_CHILDREN depend on PROC_FS
fs: uninline inode_maybe_inc_iversion()
...
Pull MM updates from Andrew Morton:
- Yu Zhao's Multi-Gen LRU patches are here. They've been under test in
linux-next for a couple of months without, to my knowledge, any
negative reports (or any positive ones, come to that).
- Also the Maple Tree from Liam Howlett. An overlapping range-based
tree for vmas. It it apparently slightly more efficient in its own
right, but is mainly targeted at enabling work to reduce mmap_lock
contention.
Liam has identified a number of other tree users in the kernel which
could be beneficially onverted to mapletrees.
Yu Zhao has identified a hard-to-hit but "easy to fix" lockdep splat
at [1]. This has yet to be addressed due to Liam's unfortunately
timed vacation. He is now back and we'll get this fixed up.
- Dmitry Vyukov introduces KMSAN: the Kernel Memory Sanitizer. It uses
clang-generated instrumentation to detect used-unintialized bugs down
to the single bit level.
KMSAN keeps finding bugs. New ones, as well as the legacy ones.
- Yang Shi adds a userspace mechanism (madvise) to induce a collapse of
memory into THPs.
- Zach O'Keefe has expanded Yang Shi's madvise(MADV_COLLAPSE) to
support file/shmem-backed pages.
- userfaultfd updates from Axel Rasmussen
- zsmalloc cleanups from Alexey Romanov
- cleanups from Miaohe Lin: vmscan, hugetlb_cgroup, hugetlb and
memory-failure
- Huang Ying adds enhancements to NUMA balancing memory tiering mode's
page promotion, with a new way of detecting hot pages.
- memcg updates from Shakeel Butt: charging optimizations and reduced
memory consumption.
- memcg cleanups from Kairui Song.
- memcg fixes and cleanups from Johannes Weiner.
- Vishal Moola provides more folio conversions
- Zhang Yi removed ll_rw_block() :(
- migration enhancements from Peter Xu
- migration error-path bugfixes from Huang Ying
- Aneesh Kumar added ability for a device driver to alter the memory
tiering promotion paths. For optimizations by PMEM drivers, DRM
drivers, etc.
- vma merging improvements from Jakub Matěn.
- NUMA hinting cleanups from David Hildenbrand.
- xu xin added aditional userspace visibility into KSM merging
activity.
- THP & KSM code consolidation from Qi Zheng.
- more folio work from Matthew Wilcox.
- KASAN updates from Andrey Konovalov.
- DAMON cleanups from Kaixu Xia.
- DAMON work from SeongJae Park: fixes, cleanups.
- hugetlb sysfs cleanups from Muchun Song.
- Mike Kravetz fixes locking issues in hugetlbfs and in hugetlb core.
Link: https://lkml.kernel.org/r/CAOUHufZabH85CeUN-MEMgL8gJGzJEWUrkiM58JkTbBhh-jew0Q@mail.gmail.com [1]
* tag 'mm-stable-2022-10-08' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (555 commits)
hugetlb: allocate vma lock for all sharable vmas
hugetlb: take hugetlb vma_lock when clearing vma_lock->vma pointer
hugetlb: fix vma lock handling during split vma and range unmapping
mglru: mm/vmscan.c: fix imprecise comments
mm/mglru: don't sync disk for each aging cycle
mm: memcontrol: drop dead CONFIG_MEMCG_SWAP config symbol
mm: memcontrol: use do_memsw_account() in a few more places
mm: memcontrol: deprecate swapaccounting=0 mode
mm: memcontrol: don't allocate cgroup swap arrays when memcg is disabled
mm/secretmem: remove reduntant return value
mm/hugetlb: add available_huge_pages() func
mm: remove unused inline functions from include/linux/mm_inline.h
selftests/vm: add selftest for MADV_COLLAPSE of uffd-minor memory
selftests/vm: add file/shmem MADV_COLLAPSE selftest for cleared pmd
selftests/vm: add thp collapse shmem testing
selftests/vm: add thp collapse file and tmpfs testing
selftests/vm: modularize thp collapse memory operations
selftests/vm: dedup THP helpers
mm/khugepaged: add tracepoint to hpage_collapse_scan_file()
mm/madvise: add file and shmem support to MADV_COLLAPSE
...
Pull scheduler updates from Ingo Molnar:
"Debuggability:
- Change most occurances of BUG_ON() to WARN_ON_ONCE()
- Reorganize & fix TASK_ state comparisons, turn it into a bitmap
- Update/fix misc scheduler debugging facilities
Load-balancing & regular scheduling:
- Improve the behavior of the scheduler in presence of lot of
SCHED_IDLE tasks - in particular they should not impact other
scheduling classes.
- Optimize task load tracking, cleanups & fixes
- Clean up & simplify misc load-balancing code
Freezer:
- Rewrite the core freezer to behave better wrt thawing and be
simpler in general, by replacing PF_FROZEN with TASK_FROZEN &
fixing/adjusting all the fallout.
Deadline scheduler:
- Fix the DL capacity-aware code
- Factor out dl_task_is_earliest_deadline() &
replenish_dl_new_period()
- Relax/optimize locking in task_non_contending()
Cleanups:
- Factor out the update_current_exec_runtime() helper
- Various cleanups, simplifications"
* tag 'sched-core-2022-10-07' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (41 commits)
sched: Fix more TASK_state comparisons
sched: Fix TASK_state comparisons
sched/fair: Move call to list_last_entry() in detach_tasks
sched/fair: Cleanup loop_max and loop_break
sched/fair: Make sure to try to detach at least one movable task
sched: Show PF_flag holes
freezer,sched: Rewrite core freezer logic
sched: Widen TAKS_state literals
sched/wait: Add wait_event_state()
sched/completion: Add wait_for_completion_state()
sched: Add TASK_ANY for wait_task_inactive()
sched: Change wait_task_inactive()s match_state
freezer,umh: Clean up freezer/initrd interaction
freezer: Have {,un}lock_system_sleep() save/restore flags
sched: Rename task_running() to task_on_cpu()
sched/fair: Cleanup for SIS_PROP
sched/fair: Default to false in test_idle_cores()
sched/fair: Remove useless check in select_idle_core()
sched/fair: Avoid double search on same cpu
sched/fair: Remove redundant check in select_idle_smt()
...
Pull ptrace update from Eric Biederman:
"ptrace: Stop supporting SIGKILL for PTRACE_EVENT_EXIT
Recently I had a conversation where it was pointed out to me that
SIGKILL sent to a tracee stropped in PTRACE_EVENT_EXIT is quite
difficult for a tracer to handle.
Keeping SIGKILL working after the process has been killed is pain from
an implementation point of view.
So since the debuggers don't want this behavior let's see if we can
remove this wart for the userspace API
If a regression is detected it should only need to be the last change
that is the reverted. The other two are just general cleanups that
make the last patch simpler"
* tag 'signal-for-v5.20' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
signal: Drop signals received after a fatal signal has been processed
signal: Guarantee that SIGNAL_GROUP_EXIT is set on process exit
signal: Ensure SIGNAL_GROUP_EXIT gets set in do_group_exit
To further exploit spatial locality, the aging prefers to walk page tables
to search for young PTEs and promote hot pages. A kill switch will be
added in the next patch to disable this behavior. When disabled, the
aging relies on the rmap only.
NB: this behavior has nothing similar with the page table scanning in the
2.4 kernel [1], which searches page tables for old PTEs, adds cold pages
to swapcache and unmaps them.
To avoid confusion, the term "iteration" specifically means the traversal
of an entire mm_struct list; the term "walk" will be applied to page
tables and the rmap, as usual.
An mm_struct list is maintained for each memcg, and an mm_struct follows
its owner task to the new memcg when this task is migrated. Given an
lruvec, the aging iterates lruvec_memcg()->mm_list and calls
walk_page_range() with each mm_struct on this list to promote hot pages
before it increments max_seq.
When multiple page table walkers iterate the same list, each of them gets
a unique mm_struct; therefore they can run concurrently. Page table
walkers ignore any misplaced pages, e.g., if an mm_struct was migrated,
pages it left in the previous memcg will not be promoted when its current
memcg is under reclaim. Similarly, page table walkers will not promote
pages from nodes other than the one under reclaim.
This patch uses the following optimizations when walking page tables:
1. It tracks the usage of mm_struct's between context switches so that
page table walkers can skip processes that have been sleeping since
the last iteration.
2. It uses generational Bloom filters to record populated branches so
that page table walkers can reduce their search space based on the
query results, e.g., to skip page tables containing mostly holes or
misplaced pages.
3. It takes advantage of the accessed bit in non-leaf PMD entries when
CONFIG_ARCH_HAS_NONLEAF_PMD_YOUNG=y.
4. It does not zigzag between a PGD table and the same PMD table
spanning multiple VMAs. IOW, it finishes all the VMAs within the
range of the same PMD table before it returns to a PGD table. This
improves the cache performance for workloads that have large
numbers of tiny VMAs [2], especially when CONFIG_PGTABLE_LEVELS=5.
Server benchmark results:
Single workload:
fio (buffered I/O): no change
Single workload:
memcached (anon): +[8, 10]%
Ops/sec KB/sec
patch1-7: 1147696.57 44640.29
patch1-8: 1245274.91 48435.66
Configurations:
no change
Client benchmark results:
kswapd profiles:
patch1-7
48.16% lzo1x_1_do_compress (real work)
8.20% page_vma_mapped_walk (overhead)
7.06% _raw_spin_unlock_irq
2.92% ptep_clear_flush
2.53% __zram_bvec_write
2.11% do_raw_spin_lock
2.02% memmove
1.93% lru_gen_look_around
1.56% free_unref_page_list
1.40% memset
patch1-8
49.44% lzo1x_1_do_compress (real work)
6.19% page_vma_mapped_walk (overhead)
5.97% _raw_spin_unlock_irq
3.13% get_pfn_folio
2.85% ptep_clear_flush
2.42% __zram_bvec_write
2.08% do_raw_spin_lock
1.92% memmove
1.44% alloc_zspage
1.36% memset
Configurations:
no change
Thanks to the following developers for their efforts [3].
kernel test robot <lkp@intel.com>
[1] https://lwn.net/Articles/23732/
[2] https://llvm.org/docs/ScudoHardenedAllocator.html
[3] https://lore.kernel.org/r/202204160827.ekEARWQo-lkp@intel.com/
Link: https://lkml.kernel.org/r/20220918080010.2920238-9-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Rewrite the core freezer to behave better wrt thawing and be simpler
in general.
By replacing PF_FROZEN with TASK_FROZEN, a special block state, it is
ensured frozen tasks stay frozen until thawed and don't randomly wake
up early, as is currently possible.
As such, it does away with PF_FROZEN and PF_FREEZER_SKIP, freeing up
two PF_flags (yay!).
Specifically; the current scheme works a little like:
freezer_do_not_count();
schedule();
freezer_count();
And either the task is blocked, or it lands in try_to_freezer()
through freezer_count(). Now, when it is blocked, the freezer
considers it frozen and continues.
However, on thawing, once pm_freezing is cleared, freezer_count()
stops working, and any random/spurious wakeup will let a task run
before its time.
That is, thawing tries to thaw things in explicit order; kernel
threads and workqueues before doing bringing SMP back before userspace
etc.. However due to the above mentioned races it is entirely possible
for userspace tasks to thaw (by accident) before SMP is back.
This can be a fatal problem in asymmetric ISA architectures (eg ARMv9)
where the userspace task requires a special CPU to run.
As said; replace this with a special task state TASK_FROZEN and add
the following state transitions:
TASK_FREEZABLE -> TASK_FROZEN
__TASK_STOPPED -> TASK_FROZEN
__TASK_TRACED -> TASK_FROZEN
The new TASK_FREEZABLE can be set on any state part of TASK_NORMAL
(IOW. TASK_INTERRUPTIBLE and TASK_UNINTERRUPTIBLE) -- any such state
is already required to deal with spurious wakeups and the freezer
causes one such when thawing the task (since the original state is
lost).
The special __TASK_{STOPPED,TRACED} states *can* be restored since
their canonical state is in ->jobctl.
With this, frozen tasks need an explicit TASK_FROZEN wakeup and are
free of undue (early / spurious) wakeups.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/20220822114649.055452969@infradead.org
Track how many threads have not started exiting and when the last
thread starts exiting set SIGNAL_GROUP_EXIT.
This guarantees that SIGNAL_GROUP_EXIT will get set when a process
exits. In practice this achieves nothing as glibc's implementation of
_exit calls sys_group_exit then sys_exit. While glibc's implemenation
of pthread_exit calls exit (which cleansup and calls _exit) if it is
the last thread and sys_exit if it is the last thread.
This means the only way the kernel might observe a process that does
not set call exit_group is if the language runtime does not use glibc.
With more cleanups I hope to move the decrement of quick_threads
earlier.
Link: https://lkml.kernel.org/r/87bkukd4tc.fsf_-_@email.froward.int.ebiederm.org
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
The function do_group_exit has an optimization that avoids taking
siglock and doing the work to find other threads in the signal group
and shutting them down.
It is very desirable for SIGNAL_GROUP_EXIT to always been set whenever
it is decided for the process to exit. That ensures only a single
place needs to be tested, and a single bit of state needs to be looked
at. This makes the optimization in do_group_exit counter productive.
Make the code and maintenance simpler by removing this unnecessary
option.
Link: https://lkml.kernel.org/r/87letod4v3.fsf_-_@email.froward.int.ebiederm.org
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
As Chris explains, the comment above exit_itimers() is not correct,
we can race with proc_timers_seq_ops. Change exit_itimers() to clear
signal->posix_timers with ->siglock held.
Cc: <stable@vger.kernel.org>
Reported-by: chris@accessvector.net
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
