[ Upstream commit 3840cbe24c ]
Brandon reports sporadic, non-sensical spikes in cumulative pressure
time (total=) when reading cpu.pressure at a high rate. This is due to
a race condition between reader aggregation and tasks changing states.
While it affects all states and all resources captured by PSI, in
practice it most likely triggers with CPU pressure, since scheduling
events are so frequent compared to other resource events.
The race context is the live snooping of ongoing stalls during a
pressure read. The read aggregates per-cpu records for stalls that
have concluded, but will also incorporate ad-hoc the duration of any
active state that hasn't been recorded yet. This is important to get
timely measurements of ongoing stalls. Those ad-hoc samples are
calculated on-the-fly up to the current time on that CPU; since the
stall hasn't concluded, it's expected that this is the minimum amount
of stall time that will enter the per-cpu records once it does.
The problem is that the path that concludes the state uses a CPU clock
read that is not synchronized against aggregators; the clock is read
outside of the seqlock protection. This allows aggregators to race and
snoop a stall with a longer duration than will actually be recorded.
With the recorded stall time being less than the last snapshot
remembered by the aggregator, a subsequent sample will underflow and
observe a bogus delta value, resulting in an erratic jump in pressure.
Fix this by moving the clock read of the state change into the seqlock
protection. This ensures no aggregation can snoop live stalls past the
time that's recorded when the state concludes.
Reported-by: Brandon Duffany <brandon@buildbuddy.io>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=219194
Link: https://lore.kernel.org/lkml/20240827121851.GB438928@cmpxchg.org/
Fixes: df77430639 ("psi: Reduce calls to sched_clock() in psi")
Cc: stable@vger.kernel.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Chengming Zhou <chengming.zhou@linux.dev>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 678379e1d4 upstream.
Cloning a descriptor table picks the size that would cover all currently
opened files. That's fine for clone() and unshare(), but for close_range()
there's an additional twist - we clone before we close, and it would be
a shame to have
close_range(3, ~0U, CLOSE_RANGE_UNSHARE)
leave us with a huge descriptor table when we are not going to keep
anything past stderr, just because some large file descriptor used to
be open before our call has taken it out.
Unfortunately, it had been dealt with in an inherently racy way -
sane_fdtable_size() gets a "don't copy anything past that" argument
(passed via unshare_fd() and dup_fd()), close_range() decides how much
should be trimmed and passes that to unshare_fd().
The problem is, a range that used to extend to the end of descriptor
table back when close_range() had looked at it might very well have stuff
grown after it by the time dup_fd() has allocated a new files_struct
and started to figure out the capacity of fdtable to be attached to that.
That leads to interesting pathological cases; at the very least it's a
QoI issue, since unshare(CLONE_FILES) is atomic in a sense that it takes
a snapshot of descriptor table one might have observed at some point.
Since CLOSE_RANGE_UNSHARE close_range() is supposed to be a combination
of unshare(CLONE_FILES) with plain close_range(), ending up with a
weird state that would never occur with unshare(2) is confusing, to put
it mildly.
It's not hard to get rid of - all it takes is passing both ends of the
range down to sane_fdtable_size(). There we are under ->files_lock,
so the race is trivially avoided.
So we do the following:
* switch close_files() from calling unshare_fd() to calling
dup_fd().
* undo the calling convention change done to unshare_fd() in
60997c3d45 "close_range: add CLOSE_RANGE_UNSHARE"
* introduce struct fd_range, pass a pointer to that to dup_fd()
and sane_fdtable_size() instead of "trim everything past that point"
they are currently getting. NULL means "we are not going to be punching
any holes"; NR_OPEN_MAX is gone.
* make sane_fdtable_size() use find_last_bit() instead of
open-coding it; it's easier to follow that way.
* while we are at it, have dup_fd() report errors by returning
ERR_PTR(), no need to use a separate int *errorp argument.
Fixes: 60997c3d45 "close_range: add CLOSE_RANGE_UNSHARE"
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 829e0c9f08 upstream.
There is another found exception that the "timerlat/1" thread was
scheduled on CPU0, and lead to timer corruption finally:
```
ODEBUG: init active (active state 0) object: ffff888237c2e108 object type: hrtimer hint: timerlat_irq+0x0/0x220
WARNING: CPU: 0 PID: 426 at lib/debugobjects.c:518 debug_print_object+0x7d/0xb0
Modules linked in:
CPU: 0 UID: 0 PID: 426 Comm: timerlat/1 Not tainted 6.11.0-rc7+ #45
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:debug_print_object+0x7d/0xb0
...
Call Trace:
<TASK>
? __warn+0x7c/0x110
? debug_print_object+0x7d/0xb0
? report_bug+0xf1/0x1d0
? prb_read_valid+0x17/0x20
? handle_bug+0x3f/0x70
? exc_invalid_op+0x13/0x60
? asm_exc_invalid_op+0x16/0x20
? debug_print_object+0x7d/0xb0
? debug_print_object+0x7d/0xb0
? __pfx_timerlat_irq+0x10/0x10
__debug_object_init+0x110/0x150
hrtimer_init+0x1d/0x60
timerlat_main+0xab/0x2d0
? __pfx_timerlat_main+0x10/0x10
kthread+0xb7/0xe0
? __pfx_kthread+0x10/0x10
ret_from_fork+0x2d/0x40
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
```
After tracing the scheduling event, it was discovered that the migration
of the "timerlat/1" thread was performed during thread creation. Further
analysis confirmed that it is because the CPU online processing for
osnoise is implemented through workers, which is asynchronous with the
offline processing. When the worker was scheduled to create a thread, the
CPU may has already been removed from the cpu_online_mask during the offline
process, resulting in the inability to select the right CPU:
T1 | T2
[CPUHP_ONLINE] | cpu_device_down()
osnoise_hotplug_workfn() |
| cpus_write_lock()
| takedown_cpu(1)
| cpus_write_unlock()
[CPUHP_OFFLINE] |
cpus_read_lock() |
start_kthread(1) |
cpus_read_unlock() |
To fix this, skip online processing if the CPU is already offline.
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20240924094515.3561410-4-liwei391@huawei.com
Fixes: c8895e271f ("trace/osnoise: Support hotplug operations")
Signed-off-by: Wei Li <liwei391@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b484a02c9c upstream.
stop_kthread() is the offline callback for "trace/osnoise:online", since
commit 5bfbcd1ee5 ("tracing/timerlat: Add interface_lock around clearing
of kthread in stop_kthread()"), the following ABBA deadlock scenario is
introduced:
T1 | T2 [BP] | T3 [AP]
osnoise_hotplug_workfn() | work_for_cpu_fn() | cpuhp_thread_fun()
| _cpu_down() | osnoise_cpu_die()
mutex_lock(&interface_lock) | | stop_kthread()
| cpus_write_lock() | mutex_lock(&interface_lock)
cpus_read_lock() | cpuhp_kick_ap() |
As the interface_lock here in just for protecting the "kthread" field of
the osn_var, use xchg() instead to fix this issue. Also use
for_each_online_cpu() back in stop_per_cpu_kthreads() as it can take
cpu_read_lock() again.
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20240924094515.3561410-3-liwei391@huawei.com
Fixes: 5bfbcd1ee5 ("tracing/timerlat: Add interface_lock around clearing of kthread in stop_kthread()")
Signed-off-by: Wei Li <liwei391@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b4afe4183e upstream.
On a system with CXL memory, the resource tree (/proc/iomem) related to
CXL memory may look like something as follows.
490000000-50fffffff : CXL Window 0
490000000-50fffffff : region0
490000000-50fffffff : dax0.0
490000000-50fffffff : System RAM (kmem)
Because drivers/dax/kmem.c calls add_memory_driver_managed() during
onlining CXL memory, which makes "System RAM (kmem)" a descendant of "CXL
Window X". This confuses region_intersects(), which expects all "System
RAM" resources to be at the top level of iomem_resource. This can lead to
bugs.
For example, when the following command line is executed to write some
memory in CXL memory range via /dev/mem,
$ dd if=data of=/dev/mem bs=$((1 << 10)) seek=$((0x490000000 >> 10)) count=1
dd: error writing '/dev/mem': Bad address
1+0 records in
0+0 records out
0 bytes copied, 0.0283507 s, 0.0 kB/s
the command fails as expected. However, the error code is wrong. It
should be "Operation not permitted" instead of "Bad address". More
seriously, the /dev/mem permission checking in devmem_is_allowed() passes
incorrectly. Although the accessing is prevented later because ioremap()
isn't allowed to map system RAM, it is a potential security issue. During
command executing, the following warning is reported in the kernel log for
calling ioremap() on system RAM.
ioremap on RAM at 0x0000000490000000 - 0x0000000490000fff
WARNING: CPU: 2 PID: 416 at arch/x86/mm/ioremap.c:216 __ioremap_caller.constprop.0+0x131/0x35d
Call Trace:
memremap+0xcb/0x184
xlate_dev_mem_ptr+0x25/0x2f
write_mem+0x94/0xfb
vfs_write+0x128/0x26d
ksys_write+0xac/0xfe
do_syscall_64+0x9a/0xfd
entry_SYSCALL_64_after_hwframe+0x4b/0x53
The details of command execution process are as follows. In the above
resource tree, "System RAM" is a descendant of "CXL Window 0" instead of a
top level resource. So, region_intersects() will report no System RAM
resources in the CXL memory region incorrectly, because it only checks the
top level resources. Consequently, devmem_is_allowed() will return 1
(allow access via /dev/mem) for CXL memory region incorrectly.
Fortunately, ioremap() doesn't allow to map System RAM and reject the
access.
So, region_intersects() needs to be fixed to work correctly with the
resource tree with "System RAM" not at top level as above. To fix it, if
we found a unmatched resource in the top level, we will continue to search
matched resources in its descendant resources. So, we will not miss any
matched resources in resource tree anymore.
In the new implementation, an example resource tree
|------------- "CXL Window 0" ------------|
|-- "System RAM" --|
will behave similar as the following fake resource tree for
region_intersects(, IORESOURCE_SYSTEM_RAM, ),
|-- "System RAM" --||-- "CXL Window 0a" --|
Where "CXL Window 0a" is part of the original "CXL Window 0" that
isn't covered by "System RAM".
Link: https://lkml.kernel.org/r/20240906030713.204292-2-ying.huang@intel.com
Fixes: c221c0b030 ("device-dax: "Hotplug" persistent memory for use like normal RAM")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Jonathan Cameron <jonathan.cameron@huawei.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Alison Schofield <alison.schofield@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 62c0b10615 upstream.
In perf_adjust_period, we will first calculate period, and then use
this period to calculate delta. However, when delta is less than 0,
there will be a deviation compared to when delta is greater than or
equal to 0. For example, when delta is in the range of [-14,-1], the
range of delta = delta + 7 is between [-7,6], so the final value of
delta/8 is 0. Therefore, the impact of -1 and -2 will be ignored.
This is unacceptable when the target period is very short, because
we will lose a lot of samples.
Here are some tests and analyzes:
before:
# perf record -e cs -F 1000 ./a.out
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.022 MB perf.data (518 samples) ]
# perf script
...
a.out 396 257.956048: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.957891: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.959730: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.961545: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.963355: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.965163: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.966973: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.968785: 23 cs: ffffffff81f4eeec schedul>
a.out 396 257.970593: 23 cs: ffffffff81f4eeec schedul>
...
after:
# perf record -e cs -F 1000 ./a.out
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.058 MB perf.data (1466 samples) ]
# perf script
...
a.out 395 59.338813: 11 cs: ffffffff81f4eeec schedul>
a.out 395 59.339707: 12 cs: ffffffff81f4eeec schedul>
a.out 395 59.340682: 13 cs: ffffffff81f4eeec schedul>
a.out 395 59.341751: 13 cs: ffffffff81f4eeec schedul>
a.out 395 59.342799: 12 cs: ffffffff81f4eeec schedul>
a.out 395 59.343765: 11 cs: ffffffff81f4eeec schedul>
a.out 395 59.344651: 11 cs: ffffffff81f4eeec schedul>
a.out 395 59.345539: 12 cs: ffffffff81f4eeec schedul>
a.out 395 59.346502: 13 cs: ffffffff81f4eeec schedul>
...
test.c
int main() {
for (int i = 0; i < 20000; i++)
usleep(10);
return 0;
}
# time ./a.out
real 0m1.583s
user 0m0.040s
sys 0m0.298s
The above results were tested on x86-64 qemu with KVM enabled using
test.c as test program. Ideally, we should have around 1500 samples,
but the previous algorithm had only about 500, whereas the modified
algorithm now has about 1400. Further more, the new version shows 1
sample per 0.001s, while the previous one is 1 sample per 0.002s.This
indicates that the new algorithm is more sensitive to small negative
values compared to old algorithm.
Fixes: bd2b5b1284 ("perf_counter: More aggressive frequency adjustment")
Signed-off-by: Luo Gengkun <luogengkun@huaweicloud.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20240831074316.2106159-2-luogengkun@huaweicloud.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 7dd34d7b7d ]
Zac Ecob reported a problem where a bpf program may cause kernel crash due
to the following error:
Oops: divide error: 0000 [#1] PREEMPT SMP KASAN PTI
The failure is due to the below signed divide:
LLONG_MIN/-1 where LLONG_MIN equals to -9,223,372,036,854,775,808.
LLONG_MIN/-1 is supposed to give a positive number 9,223,372,036,854,775,808,
but it is impossible since for 64-bit system, the maximum positive
number is 9,223,372,036,854,775,807. On x86_64, LLONG_MIN/-1 will
cause a kernel exception. On arm64, the result for LLONG_MIN/-1 is
LLONG_MIN.
Further investigation found all the following sdiv/smod cases may trigger
an exception when bpf program is running on x86_64 platform:
- LLONG_MIN/-1 for 64bit operation
- INT_MIN/-1 for 32bit operation
- LLONG_MIN%-1 for 64bit operation
- INT_MIN%-1 for 32bit operation
where -1 can be an immediate or in a register.
On arm64, there are no exceptions:
- LLONG_MIN/-1 = LLONG_MIN
- INT_MIN/-1 = INT_MIN
- LLONG_MIN%-1 = 0
- INT_MIN%-1 = 0
where -1 can be an immediate or in a register.
Insn patching is needed to handle the above cases and the patched codes
produced results aligned with above arm64 result. The below are pseudo
codes to handle sdiv/smod exceptions including both divisor -1 and divisor 0
and the divisor is stored in a register.
sdiv:
tmp = rX
tmp += 1 /* [-1, 0] -> [0, 1]
if tmp >(unsigned) 1 goto L2
if tmp == 0 goto L1
rY = 0
L1:
rY = -rY;
goto L3
L2:
rY /= rX
L3:
smod:
tmp = rX
tmp += 1 /* [-1, 0] -> [0, 1]
if tmp >(unsigned) 1 goto L1
if tmp == 1 (is64 ? goto L2 : goto L3)
rY = 0;
goto L2
L1:
rY %= rX
L2:
goto L4 // only when !is64
L3:
wY = wY // only when !is64
L4:
[1] https://lore.kernel.org/bpf/tPJLTEh7S_DxFEqAI2Ji5MBSoZVg7_G-Py2iaZpAaWtM961fFTWtsnlzwvTbzBzaUzwQAoNATXKUlt0LZOFgnDcIyKCswAnAGdUF3LBrhGQ=@protonmail.com/
Reported-by: Zac Ecob <zacecob@protonmail.com>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20240913150326.1187788-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4cc8c50c9a ]
Currently we cannot pass the pointer returned by iter next method as
argument to KF_TRUSTED_ARGS or KF_RCU kfuncs, because the pointer
returned by iter next method is not "valid".
This patch sets the pointer returned by iter next method to be valid.
This is based on the fact that if the iterator is implemented correctly,
then the pointer returned from the iter next method should be valid.
This does not make NULL pointer valid. If the iter next method has
KF_RET_NULL flag, then the verifier will ask the ebpf program to
check NULL pointer.
KF_RCU_PROTECTED iterator is a special case, the pointer returned by
iter next method should only be valid within RCU critical section,
so it should be with MEM_RCU, not PTR_TRUSTED.
Another special case is bpf_iter_num_next, which returns a pointer with
base type PTR_TO_MEM. PTR_TO_MEM should not be combined with type flag
PTR_TRUSTED (PTR_TO_MEM already means the pointer is valid).
The pointer returned by iter next method of other types of iterators
is with PTR_TRUSTED.
In addition, this patch adds get_iter_from_state to help us get the
current iterator from the current state.
Signed-off-by: Juntong Deng <juntong.deng@outlook.com>
Link: https://lore.kernel.org/r/AM6PR03MB584869F8B448EA1C87B7CDA399962@AM6PR03MB5848.eurprd03.prod.outlook.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit cfa7f3d2c5 ]
When tracing user functions with uprobe functionality, it's common to
install the probe (e.g., a BPF program) at the first instruction of the
function. This is often going to be `push %rbp` instruction in function
preamble, which means that within that function frame pointer hasn't
been established yet. This leads to consistently missing an actual
caller of the traced function, because perf_callchain_user() only
records current IP (capturing traced function) and then following frame
pointer chain (which would be caller's frame, containing the address of
caller's caller).
So when we have target_1 -> target_2 -> target_3 call chain and we are
tracing an entry to target_3, captured stack trace will report
target_1 -> target_3 call chain, which is wrong and confusing.
This patch proposes a x86-64-specific heuristic to detect `push %rbp`
(`push %ebp` on 32-bit architecture) instruction being traced. Given
entire kernel implementation of user space stack trace capturing works
under assumption that user space code was compiled with frame pointer
register (%rbp/%ebp) preservation, it seems pretty reasonable to use
this instruction as a strong indicator that this is the entry to the
function. In that case, return address is still pointed to by %rsp/%esp,
so we fetch it and add to stack trace before proceeding to unwind the
rest using frame pointer-based logic.
We also check for `endbr64` (for 64-bit modes) as another common pattern
for function entry, as suggested by Josh Poimboeuf. Even if we get this
wrong sometimes for uprobes attached not at the function entry, it's OK
because stack trace will still be overall meaningful, just with one
extra bogus entry. If we don't detect this, we end up with guaranteed to
be missing caller function entry in the stack trace, which is worse
overall.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20240729175223.23914-1-andrii@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 11377947b5 ]
Currently, if the rcuscale module's async module parameter is specified
for RCU implementations that do not have async primitives such as RCU
Tasks Rude (which now lacks a call_rcu_tasks_rude() function), there
will be a series of splats due to calls to a NULL pointer. This commit
therefore warns of this situation, but switches to non-async testing.
Signed-off-by: "Paul E. McKenney" <paulmck@kernel.org>
Signed-off-by: Neeraj Upadhyay <neeraj.upadhyay@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fd70e9f1d8 ]
For kernels built with CONFIG_FORCE_NR_CPUS=y, the nr_cpu_ids is
defined as NR_CPUS instead of the number of possible cpus, this
will cause the following system panic:
smpboot: Allowing 4 CPUs, 0 hotplug CPUs
...
setup_percpu: NR_CPUS:512 nr_cpumask_bits:512 nr_cpu_ids:512 nr_node_ids:1
...
BUG: unable to handle page fault for address: ffffffff9911c8c8
Oops: 0000 [#1] PREEMPT SMP PTI
CPU: 0 PID: 15 Comm: rcu_tasks_trace Tainted: G W
6.6.21 #1 5dc7acf91a5e8e9ac9dcfc35bee0245691283ea6
RIP: 0010:rcu_tasks_need_gpcb+0x25d/0x2c0
RSP: 0018:ffffa371c00a3e60 EFLAGS: 00010082
CR2: ffffffff9911c8c8 CR3: 000000040fa20005 CR4: 00000000001706f0
Call Trace:
<TASK>
? __die+0x23/0x80
? page_fault_oops+0xa4/0x180
? exc_page_fault+0x152/0x180
? asm_exc_page_fault+0x26/0x40
? rcu_tasks_need_gpcb+0x25d/0x2c0
? __pfx_rcu_tasks_kthread+0x40/0x40
rcu_tasks_one_gp+0x69/0x180
rcu_tasks_kthread+0x94/0xc0
kthread+0xe8/0x140
? __pfx_kthread+0x40/0x40
ret_from_fork+0x34/0x80
? __pfx_kthread+0x40/0x40
ret_from_fork_asm+0x1b/0x80
</TASK>
Considering that there may be holes in the CPU numbers, use the
maximum possible cpu number, instead of nr_cpu_ids, for configuring
enqueue and dequeue limits.
[ neeraj.upadhyay: Fix htmldocs build error reported by Stephen Rothwell ]
Closes: https://lore.kernel.org/linux-input/CALMA0xaTSMN+p4xUXkzrtR5r6k7hgoswcaXx7baR_z9r5jjskw@mail.gmail.com/T/#u
Reported-by: Zhixu Liu <zhixu.liu@gmail.com>
Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
Signed-off-by: Neeraj Upadhyay <neeraj.upadhyay@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 5e9629d0ae ]
Use perf_allow_kernel() for 'pa_enable' (physical addresses),
'pct_enable' (physical timestamps) and context IDs. This means that
perf_event_paranoid is now taken into account and LSM hooks can be used,
which is more consistent with other perf_event_open calls. For example
PERF_SAMPLE_PHYS_ADDR uses perf_allow_kernel() rather than just
perfmon_capable().
This also indirectly fixes the following error message which is
misleading because perf_event_paranoid is not taken into account by
perfmon_capable():
$ perf record -e arm_spe/pa_enable/
Error:
Access to performance monitoring and observability operations is
limited. Consider adjusting /proc/sys/kernel/perf_event_paranoid
setting ...
Suggested-by: Al Grant <al.grant@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Link: https://lore.kernel.org/r/20240827145113.1224604-1-james.clark@linaro.org
Link: https://lore.kernel.org/all/20240807120039.GD37996@noisy.programming.kicks-ass.net/
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1d7f856c2c ]
While commit 83ab38ef0a ("jump_label: Fix concurrency issues in
static_key_slow_dec()") fixed one problem, it created yet another,
notably the following is now possible:
slow_dec
if (try_dec) // dec_not_one-ish, false
// enabled == 1
slow_inc
if (inc_not_disabled) // inc_not_zero-ish
// enabled == 2
return
guard((mutex)(&jump_label_mutex);
if (atomic_cmpxchg(1,0)==1) // false, we're 2
slow_dec
if (try-dec) // dec_not_one, true
// enabled == 1
return
else
try_dec() // dec_not_one, false
WARN
Use dec_and_test instead of cmpxchg(), like it was prior to
83ab38ef0a. Add a few WARNs for the paranoid.
Fixes: 83ab38ef0a ("jump_label: Fix concurrency issues in static_key_slow_dec()")
Reported-by: "Darrick J. Wong" <djwong@kernel.org>
Tested-by: Klara Modin <klarasmodin@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fe513c2ef0 ]
static_call_module_notify() triggers a WARN_ON(), when memory allocation
fails in __static_call_add_module().
That's not really justified, because the failure case must be correctly
handled by the well known call chain and the error code is passed
through to the initiating userspace application.
A memory allocation fail is not a fatal problem, but the WARN_ON() takes
the machine out when panic_on_warn is set.
Replace it with a pr_warn().
Fixes: 9183c3f9ed ("static_call: Add inline static call infrastructure")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/8734mf7pmb.ffs@tglx
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4b30051c48 ]
Module insertion invokes static_call_add_module() to initialize the static
calls in a module. static_call_add_module() invokes __static_call_init(),
which allocates a struct static_call_mod to either encapsulate the built-in
static call sites of the associated key into it so further modules can be
added or to append the module to the module chain.
If that allocation fails the function returns with an error code and the
module core invokes static_call_del_module() to clean up eventually added
static_call_mod entries.
This works correctly, when all keys used by the module were converted over
to a module chain before the failure. If not then static_call_del_module()
causes a #GP as it blindly assumes that key::mods points to a valid struct
static_call_mod.
The problem is that key::mods is not a individual struct member of struct
static_call_key, it's part of a union to save space:
union {
/* bit 0: 0 = mods, 1 = sites */
unsigned long type;
struct static_call_mod *mods;
struct static_call_site *sites;
};
key::sites is a pointer to the list of built-in usage sites of the static
call. The type of the pointer is differentiated by bit 0. A mods pointer
has the bit clear, the sites pointer has the bit set.
As static_call_del_module() blidly assumes that the pointer is a valid
static_call_mod type, it fails to check for this failure case and
dereferences the pointer to the list of built-in call sites, which is
obviously bogus.
Cure it by checking whether the key has a sites or a mods pointer.
If it's a sites pointer then the key is not to be touched. As the sites are
walked in the same order as in __static_call_init() the site walk can be
terminated because all subsequent sites have not been touched by the init
code due to the error exit.
If it was converted before the allocation fail, then the inner loop which
searches for a module match will find nothing.
A fail in the second allocation in __static_call_init() is harmless and
does not require special treatment. The first allocation succeeded and
converted the key to a module chain. That first entry has mod::mod == NULL
and mod::next == NULL, so the inner loop of static_call_del_module() will
neither find a module match nor a module chain. The next site in the walk
was either already converted, but can't match the module, or it will exit
the outer loop because it has a static_call_site pointer and not a
static_call_mod pointer.
Fixes: 9183c3f9ed ("static_call: Add inline static call infrastructure")
Closes: https://lore.kernel.org/all/20230915082126.4187913-1-ruanjinjie@huawei.com
Reported-by: Jinjie Ruan <ruanjinjie@huawei.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Jinjie Ruan <ruanjinjie@huawei.com>
Link: https://lore.kernel.org/r/87zfon6b0s.ffs@tglx
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 9a22b28123 upstream.
When submitting more than 2^32 padata objects to padata_do_serial, the
current sorting implementation incorrectly sorts padata objects with
overflowed seq_nr, causing them to be placed before existing objects in
the reorder list. This leads to a deadlock in the serialization process
as padata_find_next cannot match padata->seq_nr and pd->processed
because the padata instance with overflowed seq_nr will be selected
next.
To fix this, we use an unsigned integer wrap around to correctly sort
padata objects in scenarios with integer overflow.
Fixes: bfde23ce20 ("padata: unbind parallel jobs from specific CPUs")
Cc: <stable@vger.kernel.org>
Co-developed-by: Christian Gafert <christian.gafert@rohde-schwarz.com>
Signed-off-by: Christian Gafert <christian.gafert@rohde-schwarz.com>
Co-developed-by: Max Ferger <max.ferger@rohde-schwarz.com>
Signed-off-by: Max Ferger <max.ferger@rohde-schwarz.com>
Signed-off-by: Van Giang Nguyen <vangiang.nguyen@rohde-schwarz.com>
Acked-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4b3786a6c5 ]
For all non-tracing helpers which formerly had ARG_PTR_TO_{LONG,INT} as input
arguments, zero the value for the case of an error as otherwise it could leak
memory. For tracing, it is not needed given CAP_PERFMON can already read all
kernel memory anyway hence bpf_get_func_arg() and bpf_get_func_ret() is skipped
in here.
Also, the MTU helpers mtu_len pointer value is being written but also read.
Technically, the MEM_UNINIT should not be there in order to always force init.
Removing MEM_UNINIT needs more verifier rework though: MEM_UNINIT right now
implies two things actually: i) write into memory, ii) memory does not have
to be initialized. If we lift MEM_UNINIT, it then becomes: i) read into memory,
ii) memory must be initialized. This means that for bpf_*_check_mtu() we're
readding the issue we're trying to fix, that is, it would then be able to
write back into things like .rodata BPF maps. Follow-up work will rework the
MEM_UNINIT semantics such that the intent can be better expressed. For now
just clear the *mtu_len on error path which can be lifted later again.
Fixes: 8a67f2de9b ("bpf: expose bpf_strtol and bpf_strtoul to all program types")
Fixes: d7a4cb9b67 ("bpf: Introduce bpf_strtol and bpf_strtoul helpers")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/e5edd241-59e7-5e39-0ee5-a51e31b6840a@iogearbox.net
Link: https://lore.kernel.org/r/20240913191754.13290-5-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 18752d73c1 ]
When checking malformed helper function signatures, also take other argument
types into account aside from just ARG_PTR_TO_UNINIT_MEM.
This concerns (formerly) ARG_PTR_TO_{INT,LONG} given uninitialized memory can
be passed there, too.
The func proto sanity check goes back to commit 435faee1aa ("bpf, verifier:
add ARG_PTR_TO_RAW_STACK type"), and its purpose was to detect wrong func protos
which had more than just one MEM_UNINIT-tagged type as arguments.
The reason more than one is currently not supported is as we mark stack slots with
STACK_MISC in check_helper_call() in case of raw mode based on meta.access_size to
allow uninitialized stack memory to be passed to helpers when they just write into
the buffer.
Probing for base type as well as MEM_UNINIT tagging ensures that other types do not
get missed (as it used to be the case for ARG_PTR_TO_{INT,LONG}).
Fixes: 57c3bb725a ("bpf: Introduce ARG_PTR_TO_{INT,LONG} arg types")
Reported-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Link: https://lore.kernel.org/r/20240913191754.13290-4-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 32556ce93b ]
Lonial found an issue that despite user- and BPF-side frozen BPF map
(like in case of .rodata), it was still possible to write into it from
a BPF program side through specific helpers having ARG_PTR_TO_{LONG,INT}
as arguments.
In check_func_arg() when the argument is as mentioned, the meta->raw_mode
is never set. Later, check_helper_mem_access(), under the case of
PTR_TO_MAP_VALUE as register base type, it assumes BPF_READ for the
subsequent call to check_map_access_type() and given the BPF map is
read-only it succeeds.
The helpers really need to be annotated as ARG_PTR_TO_{LONG,INT} | MEM_UNINIT
when results are written into them as opposed to read out of them. The
latter indicates that it's okay to pass a pointer to uninitialized memory
as the memory is written to anyway.
However, ARG_PTR_TO_{LONG,INT} is a special case of ARG_PTR_TO_FIXED_SIZE_MEM
just with additional alignment requirement. So it is better to just get
rid of the ARG_PTR_TO_{LONG,INT} special cases altogether and reuse the
fixed size memory types. For this, add MEM_ALIGNED to additionally ensure
alignment given these helpers write directly into the args via *<ptr> = val.
The .arg*_size has been initialized reflecting the actual sizeof(*<ptr>).
MEM_ALIGNED can only be used in combination with MEM_FIXED_SIZE annotated
argument types, since in !MEM_FIXED_SIZE cases the verifier does not know
the buffer size a priori and therefore cannot blindly write *<ptr> = val.
Fixes: 57c3bb725a ("bpf: Introduce ARG_PTR_TO_{INT,LONG} arg types")
Reported-by: Lonial Con <kongln9170@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Link: https://lore.kernel.org/r/20240913191754.13290-3-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit cfe69c50b0 ]
The bpf_strtol() and bpf_strtoul() helpers are currently broken on 32bit:
The argument type ARG_PTR_TO_LONG is BPF-side "long", not kernel-side "long"
and therefore always considered fixed 64bit no matter if 64 or 32bit underlying
architecture.
This contract breaks in case of the two mentioned helpers since their BPF_CALL
definition for the helpers was added with {unsigned,}long *res. Meaning, the
transition from BPF-side "long" (BPF program) to kernel-side "long" (BPF helper)
breaks here.
Both helpers call __bpf_strtoll() with "long long" correctly, but later assigning
the result into 32-bit "*(long *)" on 32bit architectures. From a BPF program
point of view, this means upper bits will be seen as uninitialised.
Therefore, fix both BPF_CALL signatures to {s,u}64 types to fix this situation.
Now, changing also uapi/bpf.h helper documentation which generates bpf_helper_defs.h
for BPF programs is tricky: Changing signatures there to __{s,u}64 would trigger
compiler warnings (incompatible pointer types passing 'long *' to parameter of type
'__s64 *' (aka 'long long *')) for existing BPF programs.
Leaving the signatures as-is would be fine as from BPF program point of view it is
still BPF-side "long" and thus equivalent to __{s,u}64 on 64 or 32bit underlying
architectures.
Note that bpf_strtol() and bpf_strtoul() are the only helpers with this issue.
Fixes: d7a4cb9b67 ("bpf: Introduce bpf_strtol and bpf_strtoul helpers")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/481fcec8-c12c-9abb-8ecb-76c71c009959@iogearbox.net
Link: https://lore.kernel.org/r/20240913191754.13290-1-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 84d265281d ]
commit 97450eb909 ("sched/pelt: Remove shift of thermal clock")
removed the decay_shift for hw_pressure. This commit uses the
sched_clock_task() in sched_tick() while it replaces the
sched_clock_task() with rq_clock_pelt() in __update_blocked_others().
This could bring inconsistence. One possible scenario I can think of
is in ___update_load_sum():
u64 delta = now - sa->last_update_time
'now' could be calculated by rq_clock_pelt() from
__update_blocked_others(), and last_update_time was calculated by
rq_clock_task() previously from sched_tick(). Usually the former
chases after the latter, it cause a very large 'delta' and brings
unexpected behavior.
Fixes: 97450eb909 ("sched/pelt: Remove shift of thermal clock")
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Hongyan Xia <hongyan.xia2@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20240827112607.181206-1-yu.c.chen@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f22cde4371 ]
Problem statement:
Since commit fc137c0dda ("sched/numa: enhance vma scanning logic"), the
Numa vma scan overhead has been reduced a lot. Meanwhile, the reducing of
the vma scan might create less Numa page fault information. The
insufficient information makes it harder for the Numa balancer to make
decision. Later, commit b7a5b537c5 ("sched/numa: Complete scanning of
partial VMAs regardless of PID activity") and commit 84db47ca71
("sched/numa: Fix mm numa_scan_seq based unconditional scan") are found to
bring back part of the performance.
Recently when running SPECcpu omnetpp_r on a 320 CPUs/2 Sockets system, a
long duration of remote Numa node read was observed by PMU events: A few
cores having ~500MB/s remote memory access for ~20 seconds. It causes
high core-to-core variance and performance penalty. After the
investigation, it is found that many vmas are skipped due to the active
PID check. According to the trace events, in most cases,
vma_is_accessed() returns false because the history access info stored in
pids_active array has been cleared.
Proposal:
The main idea is to adjust vma_is_accessed() to let it return true easier.
Thus compare the diff between mm->numa_scan_seq and
vma->numab_state->prev_scan_seq. If the diff has exceeded the threshold,
scan the vma.
This patch especially helps the cases where there are small number of
threads, like the process-based SPECcpu. Without this patch, if the
SPECcpu process access the vma at the beginning, then sleeps for a long
time, the pid_active array will be cleared. A a result, if this process
is woken up again, it never has a chance to set prot_none anymore.
Because only the first 2 times of access is granted for vma scan:
(current->mm->numa_scan_seq) - vma->numab_state->start_scan_seq) < 2 to be
worse, no other threads within the task can help set the prot_none. This
causes information lost.
Raghavendra helped test current patch and got the positive result
on the AMD platform:
autonumabench NUMA01
base patched
Amean syst-NUMA01 194.05 ( 0.00%) 165.11 * 14.92%*
Amean elsp-NUMA01 324.86 ( 0.00%) 315.58 * 2.86%*
Duration User 380345.36 368252.04
Duration System 1358.89 1156.23
Duration Elapsed 2277.45 2213.25
autonumabench NUMA02
Amean syst-NUMA02 1.12 ( 0.00%) 1.09 * 2.93%*
Amean elsp-NUMA02 3.50 ( 0.00%) 3.56 * -1.84%*
Duration User 1513.23 1575.48
Duration System 8.33 8.13
Duration Elapsed 28.59 29.71
kernbench
Amean user-256 22935.42 ( 0.00%) 22535.19 * 1.75%*
Amean syst-256 7284.16 ( 0.00%) 7608.72 * -4.46%*
Amean elsp-256 159.01 ( 0.00%) 158.17 * 0.53%*
Duration User 68816.41 67615.74
Duration System 21873.94 22848.08
Duration Elapsed 506.66 504.55
Intel 256 CPUs/2 Sockets:
autonuma benchmark also shows improvements:
v6.10-rc5 v6.10-rc5
+patch
Amean syst-NUMA01 245.85 ( 0.00%) 230.84 * 6.11%*
Amean syst-NUMA01_THREADLOCAL 205.27 ( 0.00%) 191.86 * 6.53%*
Amean syst-NUMA02 18.57 ( 0.00%) 18.09 * 2.58%*
Amean syst-NUMA02_SMT 2.63 ( 0.00%) 2.54 * 3.47%*
Amean elsp-NUMA01 517.17 ( 0.00%) 526.34 * -1.77%*
Amean elsp-NUMA01_THREADLOCAL 99.92 ( 0.00%) 100.59 * -0.67%*
Amean elsp-NUMA02 15.81 ( 0.00%) 15.72 * 0.59%*
Amean elsp-NUMA02_SMT 13.23 ( 0.00%) 12.89 * 2.53%*
v6.10-rc5 v6.10-rc5
+patch
Duration User 1064010.16 1075416.23
Duration System 3307.64 3104.66
Duration Elapsed 4537.54 4604.73
The SPECcpu remote node access issue disappears with the patch applied.
Link: https://lkml.kernel.org/r/20240827112958.181388-1-yu.c.chen@intel.com
Fixes: fc137c0dda ("sched/numa: enhance vma scanning logic")
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Co-developed-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Yujie Liu <yujie.liu@intel.com>
Reported-by: Xiaoping Zhou <xiaoping.zhou@intel.com>
Reviewed-and-tested-by: Raghavendra K T <raghavendra.kt@amd.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: "Chen, Tim C" <tim.c.chen@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Raghavendra K T <raghavendra.kt@amd.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e16c7b0778 ]
When analyzing a kernel waring message, Peter pointed out that there is a
race condition when the kworker is being frozen and falls into
try_to_freeze() with TASK_INTERRUPTIBLE, which could trigger a
might_sleep() warning in try_to_freeze(). Although the root cause is not
related to freeze()[1], it is still worthy to fix this issue ahead.
One possible race scenario:
CPU 0 CPU 1
----- -----
// kthread_worker_fn
set_current_state(TASK_INTERRUPTIBLE);
suspend_freeze_processes()
freeze_processes
static_branch_inc(&freezer_active);
freeze_kernel_threads
pm_nosig_freezing = true;
if (work) { //false
__set_current_state(TASK_RUNNING);
} else if (!freezing(current)) //false, been frozen
freezing():
if (static_branch_unlikely(&freezer_active))
if (pm_nosig_freezing)
return true;
schedule()
}
// state is still TASK_INTERRUPTIBLE
try_to_freeze()
might_sleep() <--- warning
Fix this by explicitly set the TASK_RUNNING before entering
try_to_freeze().
Link: https://lore.kernel.org/lkml/Zs2ZoAcUsZMX2B%2FI@chenyu5-mobl2/ [1]
Link: https://lkml.kernel.org/r/20240827112308.181081-1-yu.c.chen@intel.com
Fixes: b56c0d8937 ("kthread: implement kthread_worker")
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: David Gow <davidgow@google.com>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mickaël Salaün <mic@digikod.net>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 3d2786d65a ]
In case of malformed relocation record of kind BPF_CORE_TYPE_ID_LOCAL
referencing a non-existing BTF type, function bpf_core_calc_relo_insn
would cause a null pointer deference.
Fix this by adding a proper check upper in call stack, as malformed
relocation records could be passed from user space.
Simplest reproducer is a program:
r0 = 0
exit
With a single relocation record:
.insn_off = 0, /* patch first instruction */
.type_id = 100500, /* this type id does not exist */
.access_str_off = 6, /* offset of string "0" */
.kind = BPF_CORE_TYPE_ID_LOCAL,
See the link for original reproducer or next commit for a test case.
Fixes: 74753e1462 ("libbpf: Replace btf__type_by_id() with btf_type_by_id().")
Reported-by: Liu RuiTong <cnitlrt@gmail.com>
Closes: https://lore.kernel.org/bpf/CAK55_s6do7C+DVwbwY_7nKfUz0YLDoiA1v6X3Y9+p0sWzipFSA@mail.gmail.com/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20240822080124.2995724-2-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 92de36080c ]
syzbot reported a kernel crash due to
commit 1f1e864b65 ("bpf: Handle sign-extenstin ctx member accesses").
The reason is due to sign-extension of 32-bit load for
packet data/data_end/data_meta uapi field.
The original code looks like:
r2 = *(s32 *)(r1 + 76) /* load __sk_buff->data */
r3 = *(u32 *)(r1 + 80) /* load __sk_buff->data_end */
r0 = r2
r0 += 8
if r3 > r0 goto +1
...
Note that __sk_buff->data load has 32-bit sign extension.
After verification and convert_ctx_accesses(), the final asm code looks like:
r2 = *(u64 *)(r1 +208)
r2 = (s32)r2
r3 = *(u64 *)(r1 +80)
r0 = r2
r0 += 8
if r3 > r0 goto pc+1
...
Note that 'r2 = (s32)r2' may make the kernel __sk_buff->data address invalid
which may cause runtime failure.
Currently, in C code, typically we have
void *data = (void *)(long)skb->data;
void *data_end = (void *)(long)skb->data_end;
...
and it will generate
r2 = *(u64 *)(r1 +208)
r3 = *(u64 *)(r1 +80)
r0 = r2
r0 += 8
if r3 > r0 goto pc+1
If we allow sign-extension,
void *data = (void *)(long)(int)skb->data;
void *data_end = (void *)(long)skb->data_end;
...
the generated code looks like
r2 = *(u64 *)(r1 +208)
r2 <<= 32
r2 s>>= 32
r3 = *(u64 *)(r1 +80)
r0 = r2
r0 += 8
if r3 > r0 goto pc+1
and this will cause verification failure since "r2 <<= 32" is not allowed
as "r2" is a packet pointer.
To fix this issue for case
r2 = *(s32 *)(r1 + 76) /* load __sk_buff->data */
this patch added additional checking in is_valid_access() callback
function for packet data/data_end/data_meta access. If those accesses
are with sign-extenstion, the verification will fail.
[1] https://lore.kernel.org/bpf/000000000000c90eee061d236d37@google.com/
Reported-by: syzbot+ad9ec60c8eaf69e6f99c@syzkaller.appspotmail.com
Fixes: 1f1e864b65 ("bpf: Handle sign-extenstin ctx member accesses")
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20240723153439.2429035-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 763aa759d3 ]
After checking lsm hook return range in verifier, the test case
"test_progs -t test_lsm" failed, and the failure log says:
libbpf: prog 'test_int_hook': BPF program load failed: Invalid argument
libbpf: prog 'test_int_hook': -- BEGIN PROG LOAD LOG --
0: R1=ctx() R10=fp0
; int BPF_PROG(test_int_hook, struct vm_area_struct *vma, @ lsm.c:89
0: (79) r0 = *(u64 *)(r1 +24) ; R0_w=scalar(smin=smin32=-4095,smax=smax32=0) R1=ctx()
[...]
24: (b4) w0 = -1 ; R0_w=0xffffffff
; int BPF_PROG(test_int_hook, struct vm_area_struct *vma, @ lsm.c:89
25: (95) exit
At program exit the register R0 has smin=4294967295 smax=4294967295 should have been in [-4095, 0]
It can be seen that instruction "w0 = -1" zero extended -1 to 64-bit
register r0, setting both smin and smax values of r0 to 4294967295.
This resulted in a false reject when r0 was checked with range [-4095, 0].
Given bpf lsm does not return 64-bit values, this patch fixes it by changing
the compare between r0 and return range from 64-bit operation to 32-bit
operation for bpf lsm.
Fixes: 8fa4ecd49b ("bpf: enforce exact retval range on subprog/callback exit")
Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
Acked-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Link: https://lore.kernel.org/r/20240719110059.797546-5-xukuohai@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 5d99e198be ]
A bpf prog returning a positive number attached to file_alloc_security
hook makes kernel panic.
This happens because file system can not filter out the positive number
returned by the LSM prog using IS_ERR, and misinterprets this positive
number as a file pointer.
Given that hook file_alloc_security never returned positive number
before the introduction of BPF LSM, and other BPF LSM hooks may
encounter similar issues, this patch adds LSM return value check
in verifier, to ensure no unexpected value is returned.
Fixes: 520b7aa00d ("bpf: lsm: Initialize the BPF LSM hooks")
Reported-by: Xin Liu <liuxin350@huawei.com>
Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20240719110059.797546-3-xukuohai@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit faa42d2941 ]
Consider the following cgroup:
root
|
------------------------
| |
normal_cgroup idle_cgroup
| |
SCHED_IDLE task_A SCHED_NORMAL task_B
According to the cgroup hierarchy, A should preempt B. But current
check_preempt_wakeup_fair() treats cgroup se and task separately, so B
will preempt A unexpectedly.
Unify the wakeup logic by {c,p}se_is_idle only. This makes SCHED_IDLE of
a task a relative policy that is effective only within its own cgroup,
similar to the behavior of NICE.
Also fix se_is_idle() definition when !CONFIG_FAIR_GROUP_SCHED.
Fixes: 304000390f ("sched: Cgroup SCHED_IDLE support")
Signed-off-by: Tianchen Ding <dtcccc@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Josh Don <joshdon@google.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20240626023505.1332596-1-dtcccc@linux.alibaba.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 9139f93209 ]
After a CPU is marked offline and until it reaches its final trip to
idle, rcuo has several opportunities to be woken up, either because
a callback has been queued in the meantime or because
rcutree_report_cpu_dead() has issued the final deferred NOCB wake up.
If RCU-boosting is enabled, RCU kthreads are set to SCHED_FIFO policy.
And if RT-bandwidth is enabled, the related hrtimer might be armed.
However this then happens after hrtimers have been migrated at the
CPUHP_AP_HRTIMERS_DYING stage, which is broken as reported by the
following warning:
Call trace:
enqueue_hrtimer+0x7c/0xf8
hrtimer_start_range_ns+0x2b8/0x300
enqueue_task_rt+0x298/0x3f0
enqueue_task+0x94/0x188
ttwu_do_activate+0xb4/0x27c
try_to_wake_up+0x2d8/0x79c
wake_up_process+0x18/0x28
__wake_nocb_gp+0x80/0x1a0
do_nocb_deferred_wakeup_common+0x3c/0xcc
rcu_report_dead+0x68/0x1ac
cpuhp_report_idle_dead+0x48/0x9c
do_idle+0x288/0x294
cpu_startup_entry+0x34/0x3c
secondary_start_kernel+0x138/0x158
Fix this with waking up rcuo using an IPI if necessary. Since the
existing API to deal with this situation only handles swait queue, rcuo
is only woken up from offline CPUs if it's not already waiting on a
grace period. In the worst case some callbacks will just wait for a
grace period to complete before being assigned to a subsequent one.
Reported-by: "Cheng-Jui Wang (王正睿)" <Cheng-Jui.Wang@mediatek.com>
Fixes: 5c0930ccaa ("hrtimers: Push pending hrtimers away from outgoing CPU earlier")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Neeraj Upadhyay <neeraj.upadhyay@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 24cc57d8fa ]
In the case where we are forcing the ps.chunk_size to be at least 1,
we are ignoring the caller's alignment.
Move the forcing of ps.chunk_size to be at least 1 before rounding it
up to caller's alignment, so that caller's alignment is honored.
While at it, use max() to force the ps.chunk_size to be at least 1 to
improve readability.
Fixes: 6d45e1c948 ("padata: Fix possible divide-by-0 panic in padata_mt_helper()")
Signed-off-by: Kamlesh Gurudasani <kamlesh@ti.com>
Acked-by: Waiman Long <longman@redhat.com>
Acked-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit ff0ce721ec ]
It was found that some hotplug operations may cause multiple
rebuild_sched_domains_locked() calls. Some of those intermediate calls
may use cpuset states not in the final correct form leading to incorrect
sched domain setting.
Fix this problem by using the existing force_rebuild flag to inhibit
immediate rebuild_sched_domains_locked() calls if set and only doing
one final call at the end. Also renaming the force_rebuild flag to
force_sd_rebuild to make its meaning for clear.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 98f887f820 ]
On a ~2000 CPU powerpc system, hard lockups have been observed in the
workqueue code when stop_machine runs (in this case due to CPU hotplug).
This is due to lots of CPUs spinning in multi_cpu_stop, calling
touch_nmi_watchdog() which ends up calling wq_watchdog_touch().
wq_watchdog_touch() writes to the global variable wq_watchdog_touched,
and that can find itself in the same cacheline as other important
workqueue data, which slows down operations to the point of lockups.
In the case of the following abridged trace, worker_pool_idr was in
the hot line, causing the lockups to always appear at idr_find.
watchdog: CPU 1125 self-detected hard LOCKUP @ idr_find
Call Trace:
get_work_pool
__queue_work
call_timer_fn
run_timer_softirq
__do_softirq
do_softirq_own_stack
irq_exit
timer_interrupt
decrementer_common_virt
* interrupt: 900 (timer) at multi_cpu_stop
multi_cpu_stop
cpu_stopper_thread
smpboot_thread_fn
kthread
Fix this by having wq_watchdog_touch() only write to the line if the
last time a touch was recorded exceeds 1/4 of the watchdog threshold.
Reported-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 18e24deb1c ]
Warn in the case it is called with cpu == -1. This does not appear
to happen anywhere.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Stable-dep-of: 98f887f820 ("workqueue: Improve scalability of workqueue watchdog touch")
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 2ab9d83026 upstream.
Ole reported that event->mmap_mutex is strictly insufficient to
serialize the AUX buffer, add a per RB mutex to fully serialize it.
Note that in the lock order comment the perf_event::mmap_mutex order
was already wrong, that is, it nesting under mmap_lock is not new with
this patch.
Fixes: 45bfb2e504 ("perf: Add AUX area to ring buffer for raw data streams")
Reported-by: Ole <ole@binarygecko.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 77aeb1b685 ]
For CONFIG_DEBUG_OBJECTS_WORK=y kernels sscs.work defined by
INIT_WORK_ONSTACK() is initialized by debug_object_init_on_stack() for
the debug check in __init_work() to work correctly.
But this lacks the counterpart to remove the tracked object from debug
objects again, which will cause a debug object warning once the stack is
freed.
Add the missing destroy_work_on_stack() invocation to cure that.
[ tglx: Massaged changelog ]
Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/20240704065213.13559-1-qiang.zhang1211@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
