Pull ring buffer fixes from Steven Rostedt:
- Revert: "ring-buffer: Do not have boot mapped buffers hook to CPU
hotplug"
A crash that happened on cpu hotplug was actually caused by the
incorrect ref counting that was fixed by commit 2cf9733891
("ring-buffer: Fix refcount setting of boot mapped buffers"). The
removal of calling cpu hotplug callbacks on memory mapped buffers was
not an issue even though the tests at the time pointed toward it. But
in fact, there's a check in that code that tests to see if the
buffers are already allocated or not, and will not allocate them
again if they are. Not calling the cpu hotplug callbacks ended up not
initializing the non boot CPU buffers.
Simply remove that change.
- Clear all CPU buffers when starting tracing in a boot mapped buffer
To properly process events from a previous boot, the address space
needs to be accounted for due to KASLR and the events in the buffer
are updated accordingly when read. This also requires that when the
buffer has tracing enabled again in the current boot that the buffers
are reset so that events from the previous boot do not interact with
the events of the current boot and cause confusing due to not having
the proper meta data.
It was found that if a CPU is taken offline, that its per CPU buffer
is not reset when tracing starts. This allows for events to be from
both the previous boot and the current boot to be in the buffer at
the same time. Clear all CPU buffers when tracing is started in a
boot mapped buffer.
* tag 'trace-ringbuffer-v6.12-rc7-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing/ring-buffer: Clear all memory mapped CPU ring buffers on first recording
Revert: "ring-buffer: Do not have boot mapped buffers hook to CPU hotplug"
The events of a memory mapped ring buffer from the previous boot should
not be mixed in with events from the current boot. There's meta data that
is used to handle KASLR so that function names can be shown properly.
Also, since the timestamps of the previous boot have no meaning to the
timestamps of the current boot, having them intermingled in a buffer can
also cause confusion because there could possibly be events in the future.
When a trace is activated the meta data is reset so that the pointers of
are now processed for the new address space. The trace buffers are reset
when tracing starts for the first time. The problem here is that the reset
only happens on online CPUs. If a CPU is offline, it does not get reset.
To demonstrate the issue, a previous boot had tracing enabled in the boot
mapped ring buffer on reboot. On the following boot, tracing has not been
started yet so the function trace from the previous boot is still visible.
# trace-cmd show -B boot_mapped -c 3 | tail
<idle>-0 [003] d.h2. 156.462395: __rcu_read_lock <-cpu_emergency_disable_virtualization
<idle>-0 [003] d.h2. 156.462396: vmx_emergency_disable_virtualization_cpu <-cpu_emergency_disable_virtualization
<idle>-0 [003] d.h2. 156.462396: __rcu_read_unlock <-__sysvec_reboot
<idle>-0 [003] d.h2. 156.462397: stop_this_cpu <-__sysvec_reboot
<idle>-0 [003] d.h2. 156.462397: set_cpu_online <-stop_this_cpu
<idle>-0 [003] d.h2. 156.462397: disable_local_APIC <-stop_this_cpu
<idle>-0 [003] d.h2. 156.462398: clear_local_APIC <-disable_local_APIC
<idle>-0 [003] d.h2. 156.462574: mcheck_cpu_clear <-stop_this_cpu
<idle>-0 [003] d.h2. 156.462575: mce_intel_feature_clear <-stop_this_cpu
<idle>-0 [003] d.h2. 156.462575: lmce_supported <-mce_intel_feature_clear
Now, if CPU 3 is taken offline, and tracing is started on the memory
mapped ring buffer, the events from the previous boot in the CPU 3 ring
buffer is not reset. Now those events are using the meta data from the
current boot and produces just hex values.
# echo 0 > /sys/devices/system/cpu/cpu3/online
# trace-cmd start -B boot_mapped -p function
# trace-cmd show -B boot_mapped -c 3 | tail
<idle>-0 [003] d.h2. 156.462395: 0xffffffff9a1e3194 <-0xffffffff9a0f655e
<idle>-0 [003] d.h2. 156.462396: 0xffffffff9a0a1d24 <-0xffffffff9a0f656f
<idle>-0 [003] d.h2. 156.462396: 0xffffffff9a1e6bc4 <-0xffffffff9a0f7323
<idle>-0 [003] d.h2. 156.462397: 0xffffffff9a0d12b4 <-0xffffffff9a0f732a
<idle>-0 [003] d.h2. 156.462397: 0xffffffff9a1458d4 <-0xffffffff9a0d12e2
<idle>-0 [003] d.h2. 156.462397: 0xffffffff9a0faed4 <-0xffffffff9a0d12e7
<idle>-0 [003] d.h2. 156.462398: 0xffffffff9a0faaf4 <-0xffffffff9a0faef2
<idle>-0 [003] d.h2. 156.462574: 0xffffffff9a0e3444 <-0xffffffff9a0d12ef
<idle>-0 [003] d.h2. 156.462575: 0xffffffff9a0e4964 <-0xffffffff9a0d12ef
<idle>-0 [003] d.h2. 156.462575: 0xffffffff9a0e3fb0 <-0xffffffff9a0e496f
Reset all CPUs when starting a boot mapped ring buffer for the first time,
and not just the online CPUs.
Fixes: 7a1d1e4b96 ("tracing/ring-buffer: Add last_boot_info file to boot instance")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
A ring buffer which has its buffered mapped at boot up to fixed memory
should not be freed. Other buffers can be. The ref counting setup was
wrong for both. It made the not mapped buffers ref count have zero, and the
boot mapped buffer a ref count of 1. But an normally allocated buffer
should be 1, where it can be removed.
Keep the ref count of a normal boot buffer with its setup ref count (do
not decrement it), and increment the fixed memory boot mapped buffer's ref
count.
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20241011165224.33dd2624@gandalf.local.home
Fixes: e645535a95 ("tracing: Add option to use memmapped memory for trace boot instance")
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When the tp_printk kernel command line is used, the trace events go
directly to printk(). It is still checked via the trace_check_vprintf()
function to make sure the pointers of the trace event are legit.
The addition of reading buffers from previous boots required adding a
delta between the addresses of the previous boot and the current boot so
that the pointers in the old buffer can still be used. But this required
adding a trace_array pointer to acquire the delta offsets.
The tp_printk code does not provide a trace_array (tr) pointer, so when
the offsets were examined, a NULL pointer dereference happened and the
kernel crashed.
If the trace_array does not exist, just default the delta offsets to zero,
as that also means the trace event is not being read from a previous boot.
Link: https://lore.kernel.org/all/Zv3z5UsG_jsO9_Tb@aschofie-mobl2.lan/
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20241003104925.4e1b1fd9@gandalf.local.home
Fixes: 07714b4bb3 ("tracing: Handle old buffer mappings for event strings and functions")
Reported-by: Alison Schofield <alison.schofield@intel.com>
Tested-by: Alison Schofield <alison.schofield@intel.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
no_llseek had been defined to NULL two years ago, in commit 868941b144
("fs: remove no_llseek")
To quote that commit,
At -rc1 we'll need do a mechanical removal of no_llseek -
git grep -l -w no_llseek | grep -v porting.rst | while read i; do
sed -i '/\<no_llseek\>/d' $i
done
would do it.
Unfortunately, that hadn't been done. Linus, could you do that now, so
that we could finally put that thing to rest? All instances are of the
form
.llseek = no_llseek,
so it's obviously safe.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull ring-buffer updates from Steven Rostedt:
- tracing/ring-buffer: persistent buffer across reboots
This allows for the tracing instance ring buffer to stay persistent
across reboots. The way this is done is by adding to the kernel
command line:
trace_instance=boot_map@0x285400000:12M
This will reserve 12 megabytes at the address 0x285400000, and then
map the tracing instance "boot_map" ring buffer to that memory. This
will appear as a normal instance in the tracefs system:
/sys/kernel/tracing/instances/boot_map
A user could enable tracing in that instance, and on reboot or kernel
crash, if the memory is not wiped by the firmware, it will recreate
the trace in that instance. For example, if one was debugging a
shutdown of a kernel reboot:
# cd /sys/kernel/tracing
# echo function > instances/boot_map/current_tracer
# reboot
[..]
# cd /sys/kernel/tracing
# tail instances/boot_map/trace
swapper/0-1 [000] d..1. 164.549800: restore_boot_irq_mode <-native_machine_shutdown
swapper/0-1 [000] d..1. 164.549801: native_restore_boot_irq_mode <-native_machine_shutdown
swapper/0-1 [000] d..1. 164.549802: disconnect_bsp_APIC <-native_machine_shutdown
swapper/0-1 [000] d..1. 164.549811: hpet_disable <-native_machine_shutdown
swapper/0-1 [000] d..1. 164.549812: iommu_shutdown_noop <-native_machine_restart
swapper/0-1 [000] d..1. 164.549813: native_machine_emergency_restart <-__do_sys_reboot
swapper/0-1 [000] d..1. 164.549813: tboot_shutdown <-native_machine_emergency_restart
swapper/0-1 [000] d..1. 164.549820: acpi_reboot <-native_machine_emergency_restart
swapper/0-1 [000] d..1. 164.549821: acpi_reset <-acpi_reboot
swapper/0-1 [000] d..1. 164.549822: acpi_os_write_port <-acpi_reboot
On reboot, the buffer is examined to make sure it is valid. The
validation check even steps through every event to make sure the meta
data of the event is correct. If any test fails, it will simply reset
the buffer, and the buffer will be empty on boot.
- Allow the tracing persistent boot buffer to use the "reserve_mem"
option
Instead of having the admin find a physical address to store the
persistent buffer, which can be very tedious if they have to
administrate several different machines, allow them to use the
"reserve_mem" option that will find a location for them. It is not as
reliable because of KASLR, as the loading of the kernel in different
locations can cause the memory allocated to be inconsistent. Booting
with "nokaslr" can make reserve_mem more reliable.
- Have function graph tracer handle offsets from a previous boot.
The ring buffer output from a previous boot may have different
addresses due to kaslr. Have the function graph tracer handle these
by using the delta from the previous boot to the new boot address
space.
- Only reset the saved meta offset when the buffer is started or reset
In the persistent memory meta data, it holds the previous address
space information, so that it can calculate the delta to have
function tracing work. But this gets updated after being read to hold
the new address space. But if the buffer isn't used for that boot, on
reboot, the delta is now calculated from the previous boot and not
the boot that holds the data in the ring buffer. This causes the
functions not to be shown. Do not save the address space information
of the current kernel until it is being recorded.
- Add a magic variable to test the valid meta data
Add a magic variable in the meta data that can also be used for
validation. The validator of the previous buffer doesn't need this
magic data, but it can be used if the meta data is changed by a new
kernel, which may have the same format that passes the validator but
is used differently. This magic number can also be used as a
"versioning" of the meta data.
- Align user space mapped ring buffer sub buffers to improve TLB
entries
Linus mentioned that the mapped ring buffer sub buffers were
misaligned between the meta page and the sub-buffers, so that if the
sub-buffers were bigger than PAGE_SIZE, it wouldn't allow the TLB to
use bigger entries.
- Add new kernel command line "traceoff" to disable tracing on boot for
instances
If tracing is enabled for a boot instance, there needs a way to be
able to disable it on boot so that new events do not get entered into
the ring buffer and be mixed with events from a previous boot, as
that can be confusing.
- Allow trace_printk() to go to other instances
Currently, trace_printk() can only go to the top level instance. When
debugging with a persistent buffer, it is really useful to be able to
add trace_printk() to go to that buffer, so that you have access to
them after a crash.
- Do not use "bin_printk()" for traces to a boot instance
The bin_printk() saves only a pointer to the printk format in the
ring buffer, as the reader of the buffer can still have access to it.
But this is not the case if the buffer is from a previous boot. If
the trace_printk() is going to a "persistent" buffer, it will use the
slower version that writes the printk format into the buffer.
- Add command line option to allow trace_printk() to go to an instance
Allow the kernel command line to define which instance the
trace_printk() goes to, instead of forcing the admin to set it for
every boot via the tracefs options.
- Start a document that explains how to use tracefs to debug the kernel
- Add some more kernel selftests to test user mapped ring buffer
* tag 'trace-ring-buffer-v6.12' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (28 commits)
selftests/ring-buffer: Handle meta-page bigger than the system
selftests/ring-buffer: Verify the entire meta-page padding
tracing/Documentation: Start a document on how to debug with tracing
tracing: Add option to set an instance to be the trace_printk destination
tracing: Have trace_printk not use binary prints if boot buffer
tracing: Allow trace_printk() to go to other instance buffers
tracing: Add "traceoff" flag to boot time tracing instances
ring-buffer: Align meta-page to sub-buffers for improved TLB usage
ring-buffer: Add magic and struct size to boot up meta data
ring-buffer: Don't reset persistent ring-buffer meta saved addresses
tracing/fgraph: Have fgraph handle previous boot function addresses
tracing: Allow boot instances to use reserve_mem boot memory
tracing: Fix ifdef of snapshots to not prevent last_boot_info file
ring-buffer: Use vma_pages() helper function
tracing: Fix NULL vs IS_ERR() check in enable_instances()
tracing: Add last boot delta offset for stack traces
tracing: Update function tracing output for previous boot buffer
tracing: Handle old buffer mappings for event strings and functions
tracing/ring-buffer: Add last_boot_info file to boot instance
ring-buffer: Save text and data locations in mapped meta data
...
In __tracing_open(), when max latency tracers took place on the cpu,
the time start of its buffer would be updated, then event entries with
timestamps being earlier than start of the buffer would be skipped
(see tracing_iter_reset()).
Softlockup will occur if the kernel is non-preemptible and too many
entries were skipped in the loop that reset every cpu buffer, so add
cond_resched() to avoid it.
Cc: stable@vger.kernel.org
Fixes: 2f26ebd549 ("tracing: use timestamp to determine start of latency traces")
Link: https://lore.kernel.org/20240827124654.3817443-1-zhengyejian@huaweicloud.com
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Zheng Yejian <zhengyejian@huaweicloud.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Allow boot instances to use memory reserved by the reserve_mem boot
option.
reserve_mem=12M:4096:trace trace_instance=boot_mapped@trace
The above will allocate 12 megs with 4096 alignment and label it "trace".
The second parameter will create a "boot_mapped" instance and use the
memory reserved and labeled as "trace" as the memory for the ring buffer.
That will create an instance called "boot_mapped":
/sys/kernel/tracing/instances/boot_mapped
Note, because the ring buffer is using a defined memory ranged, it will
act just like a memory mapped ring buffer. It will not have a snapshot
buffer, as it can't swap out the buffer. The snapshot files as well as any
tracers that uses a snapshot will not be present in the boot_mapped
instance.
Also note that reserve_mem is not reliable in acquiring the same physical
memory at each soft reboot. It is possible that KALSR could map the kernel
at the previous boot memory location forcing the reserve_mem to return a
different memory location. In this case, the previous ring buffer will be
lost.
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ross Zwisler <zwisler@google.com>
Cc: Vincent Donnefort <vdonnefort@google.com>
Link: https://lore.kernel.org/20240815082811.669f7d8c@gandalf.local.home
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The mapping of the ring buffer to memory allocated at boot up will also
expose a "last_boot_info" to help tooling to read the raw data from the
last boot. As instances that have their ring buffer mapped to fixed
memory cannot perform snapshots, they can either have the "snapshot" file
or the "last_boot_info" file, but not both.
The code that added the "last_boot_info" file failed to notice that the
"snapshot" creation was inside a "#ifdef CONFIG_TRACER_SNAPSHOT" and
incorrectly placed the creation of the "last_boot_info" file within the
ifdef block. Not only does it cause a warning when CONFIG_TRACER_SNAPSHOT
is not enabled, it also incorrectly prevents the file from appearing.
Link: https://lore.kernel.org/all/20240719102640.718554-1-arnd@kernel.org/
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reported-by: Arnd Bergmann <arnd@kernel.org>
Link: https://lore.kernel.org/20240719101312.3d4ac707@rorschach.local.home
Fixes: 7a1d1e4b96 ("tracing/ring-buffer: Add last_boot_info file to boot instance")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The "reserve_mem" kernel command line parameter has been pulled into
v6.11. Merge the latest -rc3 to allow the persistent ring buffer memory to
be able to be mapped at the address specified by the "reserve_mem" command
line parameter.
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When running the following:
# cd /sys/kernel/tracing/
# echo 1 > events/sched/sched_waking/enable
# echo 1 > events/sched/sched_switch/enable
# echo 0 > tracing_on
# dd if=per_cpu/cpu0/trace_pipe_raw of=/tmp/raw0.dat
The dd task would get stuck in an infinite loop in the kernel. What would
happen is the following:
When ring_buffer_read_page() returns -1 (no data) then a check is made to
see if the buffer is empty (as happens when the page is not full), it will
call wait_on_pipe() to wait until the ring buffer has data. When it is it
will try again to read data (unless O_NONBLOCK is set).
The issue happens when there's a reader and the file descriptor is closed.
The wait_on_pipe() will return when that is the case. But this loop will
continue to try again and wait_on_pipe() will again return immediately and
the loop will continue and never stop.
Simply check if the file was closed before looping and exit out if it is.
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20240808235730.78bf63e5@rorschach.local.home
Fixes: 2aa043a55b ("tracing/ring-buffer: Fix wait_on_pipe() race")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
const qualify the struct ctl_table argument in the proc_handler function
signatures. This is a prerequisite to moving the static ctl_table
structs into .rodata data which will ensure that proc_handler function
pointers cannot be modified.
This patch has been generated by the following coccinelle script:
```
virtual patch
@r1@
identifier ctl, write, buffer, lenp, ppos;
identifier func !~ "appldata_(timer|interval)_handler|sched_(rt|rr)_handler|rds_tcp_skbuf_handler|proc_sctp_do_(hmac_alg|rto_min|rto_max|udp_port|alpha_beta|auth|probe_interval)";
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int write, void *buffer, size_t *lenp, loff_t *ppos);
@r2@
identifier func, ctl, write, buffer, lenp, ppos;
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int write, void *buffer, size_t *lenp, loff_t *ppos)
{ ... }
@r3@
identifier func;
@@
int func(
- struct ctl_table *
+ const struct ctl_table *
,int , void *, size_t *, loff_t *);
@r4@
identifier func, ctl;
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int , void *, size_t *, loff_t *);
@r5@
identifier func, write, buffer, lenp, ppos;
@@
int func(
- struct ctl_table *
+ const struct ctl_table *
,int write, void *buffer, size_t *lenp, loff_t *ppos);
```
* Code formatting was adjusted in xfs_sysctl.c to comply with code
conventions. The xfs_stats_clear_proc_handler,
xfs_panic_mask_proc_handler and xfs_deprecated_dointvec_minmax where
adjusted.
* The ctl_table argument in proc_watchdog_common was const qualified.
This is called from a proc_handler itself and is calling back into
another proc_handler, making it necessary to change it as part of the
proc_handler migration.
Co-developed-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Co-developed-by: Joel Granados <j.granados@samsung.com>
Signed-off-by: Joel Granados <j.granados@samsung.com>
Pull tracing ring buffer updates from Steven Rostedt:
"Add ring_buffer memory mappings.
The tracing ring buffer was created based on being mostly used with
the splice system call. It is broken up into page ordered sub-buffers
and the reader swaps a new sub-buffer with an existing sub-buffer
that's part of the write buffer. It then has total access to the
swapped out sub-buffer and can do copyless movements of the memory
into other mediums (file system, network, etc).
The buffer is great for passing around the ring buffer contents in the
kernel, but is not so good for when the consumer is the user space
task itself.
A new interface is added that allows user space to memory map the ring
buffer. It will get all the write sub-buffers as well as reader
sub-buffer (that is not written to). It can send an ioctl to change
which sub-buffer is the new reader sub-buffer.
The ring buffer is read only to user space. It only needs to call the
ioctl when it is finished with a sub-buffer and needs a new sub-buffer
that the writer will not write over.
A self test program was also created for testing and can be used as an
example for the interface to user space. The libtracefs (external to
the kernel) also has code that interacts with this, although it is
disabled until the interface is in a official release. It can be
enabled by compiling the library with a special flag. This was used
for testing applications that perform better with the buffer being
mapped.
Memory mapped buffers have limitations. The main one is that it can
not be used with the snapshot logic. If the buffer is mapped,
snapshots will be disabled. If any logic is set to trigger snapshots
on a buffer, that buffer will not be allowed to be mapped"
* tag 'trace-ringbuffer-v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
ring-buffer: Add cast to unsigned long addr passed to virt_to_page()
ring-buffer: Have mmapped ring buffer keep track of missed events
ring-buffer/selftest: Add ring-buffer mapping test
Documentation: tracing: Add ring-buffer mapping
tracing: Allow user-space mapping of the ring-buffer
ring-buffer: Introducing ring-buffer mapping functions
ring-buffer: Allocate sub-buffers with __GFP_COMP
Currently, user-space extracts data from the ring-buffer via splice,
which is handy for storage or network sharing. However, due to splice
limitations, it is imposible to do real-time analysis without a copy.
A solution for that problem is to let the user-space map the ring-buffer
directly.
The mapping is exposed via the per-CPU file trace_pipe_raw. The first
element of the mapping is the meta-page. It is followed by each
subbuffer constituting the ring-buffer, ordered by their unique page ID:
* Meta-page -- include/uapi/linux/trace_mmap.h for a description
* Subbuf ID 0
* Subbuf ID 1
...
It is therefore easy to translate a subbuf ID into an offset in the
mapping:
reader_id = meta->reader->id;
reader_offset = meta->meta_page_size + reader_id * meta->subbuf_size;
When new data is available, the mapper must call a newly introduced ioctl:
TRACE_MMAP_IOCTL_GET_READER. This will update the Meta-page reader ID to
point to the next reader containing unread data.
Mapping will prevent snapshot and buffer size modifications.
Link: https://lore.kernel.org/linux-trace-kernel/20240510140435.3550353-4-vdonnefort@google.com
CC: <linux-mm@kvack.org>
Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
As like '%pd' type, this patch supports print type '%pD' for print file's
name. For example "name=$arg1:%pD" casts the `$arg1` as (struct file*),
dereferences the "file.f_path.dentry.d_name.name" field and stores it to
"name" argument as a kernel string.
Here is an example:
[tracing]# echo 'p:testprobe vfs_read name=$arg1:%pD' > kprobe_event
[tracing]# echo 1 > events/kprobes/testprobe/enable
[tracing]# grep -q "1" events/kprobes/testprobe/enable
[tracing]# echo 0 > events/kprobes/testprobe/enable
[tracing]# grep "vfs_read" trace | grep "enable"
grep-15108 [003] ..... 5228.328609: testprobe: (vfs_read+0x4/0xbb0) name="enable"
Note that this expects the given argument (e.g. $arg1) is an address of struct
file. User must ensure it.
Link: https://lore.kernel.org/all/20240322064308.284457-3-yebin10@huawei.com/
[Masami: replaced "previous patch" with '%pd' type]
Signed-off-by: Ye Bin <yebin10@huawei.com>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
During fault locating, the file name needs to be printed based on the
dentry address. The offset needs to be calculated each time, which
is troublesome. Similar to printk, kprobe support print type '%pd' for
print dentry's name. For example "name=$arg1:%pd" casts the `$arg1`
as (struct dentry *), dereferences the "d_name.name" field and stores
it to "name" argument as a kernel string.
Here is an example:
[tracing]# echo 'p:testprobe dput name=$arg1:%pd' > kprobe_events
[tracing]# echo 1 > events/kprobes/testprobe/enable
[tracing]# grep -q "1" events/kprobes/testprobe/enable
[tracing]# echo 0 > events/kprobes/testprobe/enable
[tracing]# cat trace | grep "enable"
bash-14844 [002] ..... 16912.889543: testprobe: (dput+0x4/0x30) name="enable"
grep-15389 [003] ..... 16922.834182: testprobe: (dput+0x4/0x30) name="enable"
grep-15389 [003] ..... 16922.836103: testprobe: (dput+0x4/0x30) name="enable"
bash-14844 [001] ..... 16931.820909: testprobe: (dput+0x4/0x30) name="enable"
Note that this expects the given argument (e.g. $arg1) is an address of struct
dentry. User must ensure it.
Link: https://lore.kernel.org/all/20240322064308.284457-2-yebin10@huawei.com/
Signed-off-by: Ye Bin <yebin10@huawei.com>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Currently ftrace only dumps the global trace buffer on an OOPs. For
debugging a production usecase, instance trace will be helpful to
check specific problems since global trace buffer may be used for
other purposes.
This patch extend the ftrace_dump_on_oops parameter to dump a specific
or multiple trace instances:
- ftrace_dump_on_oops=0: as before -- don't dump
- ftrace_dump_on_oops[=1]: as before -- dump the global trace buffer
on all CPUs
- ftrace_dump_on_oops=2 or =orig_cpu: as before -- dump the global
trace buffer on CPU that triggered the oops
- ftrace_dump_on_oops=<instance_name>: new behavior -- dump the
tracing instance matching <instance_name>
- ftrace_dump_on_oops[=2/orig_cpu],<instance1_name>[=2/orig_cpu],
<instrance2_name>[=2/orig_cpu]: new behavior -- dump the global trace
buffer and multiple instance buffer on all CPUs, or only dump on CPU
that triggered the oops if =2 or =orig_cpu is given
Also, the sysctl node can handle the input accordingly.
Link: https://lore.kernel.org/linux-trace-kernel/20240223083126.1817731-1-quic_hyiwei@quicinc.com
Cc: Ross Zwisler <zwisler@google.com>
Cc: <mhiramat@kernel.org>
Cc: <mark.rutland@arm.com>
Cc: <mcgrof@kernel.org>
Cc: <keescook@chromium.org>
Cc: <j.granados@samsung.com>
Cc: <mathieu.desnoyers@efficios.com>
Cc: <corbet@lwn.net>
Signed-off-by: Huang Yiwei <quic_hyiwei@quicinc.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Running the ftrace selftests caused the ring buffer mapping test to fail.
Investigating, I found that the snapshot counter would be incremented
every time a tracer that uses the snapshot is enabled even if the snapshot
was used by the previous tracer.
That is:
# cd /sys/kernel/tracing
# echo wakeup_rt > current_tracer
# echo wakeup_dl > current_tracer
# echo nop > current_tracer
would leave the snapshot counter at 1 and not zero. That's because the
enabling of wakeup_dl would increment the counter again but the setting
the tracer to nop would only decrement it once.
Do not arm the snapshot for a tracer if the previous tracer already had it
armed.
Link: https://lore.kernel.org/linux-trace-kernel/20240223013344.570525723@goodmis.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vincent Donnefort <vdonnefort@google.com>
Fixes: 16f7e48ffc53a ("tracing: Add snapshot refcount")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When a ring-buffer is memory mapped by user-space, no trace or
ring-buffer swap is possible. This means the snapshot feature is
mutually exclusive with the memory mapping. Having a refcount on
snapshot users will help to know if a mapping is possible or not.
Instead of relying on the global trace_types_lock, a new spinlock is
introduced to serialize accesses to trace_array->snapshot. This intends
to allow access to that variable in a context where the mmap lock is
already held.
Link: https://lore.kernel.org/linux-trace-kernel/20240220202310.2489614-4-vdonnefort@google.com
Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The saved_cmdlines have three arrays for mapping PIDs to COMMs:
- map_pid_to_cmdline[]
- map_cmdline_to_pid[]
- saved_cmdlines
The map_pid_to_cmdline[] is PID_MAX_DEFAULT in size and holds the index
into the other arrays. The map_cmdline_to_pid[] is a mapping back to the
full pid as it can be larger than PID_MAX_DEFAULT. And the
saved_cmdlines[] just holds the COMMs associated to the pids.
Currently the map_pid_to_cmdline[] and saved_cmdlines[] are allocated
together (in reality the saved_cmdlines is just in the memory of the
rounding of the allocation of the structure as it is always allocated in
powers of two). The map_cmdline_to_pid[] array is allocated separately.
Since the rounding to a power of two is rather large (it allows for 8000
elements in saved_cmdlines), also include the map_cmdline_to_pid[] array.
(This drops it to 6000 by default, which is still plenty for most use
cases). This saves even more memory as the map_cmdline_to_pid[] array
doesn't need to be allocated.
Link: https://lore.kernel.org/linux-trace-kernel/20240212174011.068211d9@gandalf.local.home/
Link: https://lore.kernel.org/linux-trace-kernel/20240220140703.182330529@goodmis.org
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Vincent Donnefort <vdonnefort@google.com>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Mete Durlu <meted@linux.ibm.com>
Fixes: 44dc5c41b5 ("tracing: Fix wasted memory in saved_cmdlines logic")
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull tracing updates from Steven Rostedt:
- Do not update shortest_full in rb_watermark_hit() if the watermark is
hit. The shortest_full field was being updated regardless if the task
was going to wait or not. If the watermark is hit, then the task is
not going to wait, so do not update the shortest_full field (used by
the waker).
- Update shortest_full field before setting the full_waiters_pending
flag
In the poll logic, the full_waiters_pending flag was being set before
the shortest_full field was set. If the full_waiters_pending flag is
set, writers will check the shortest_full field which has the least
percentage of data that the ring buffer needs to be filled before
waking up. The writer will check shortest_full if
full_waiters_pending is set, and if the ring buffer percentage filled
is greater than shortest full, then it will call the irq_work to wake
up the waiters.
The problem was that the poll logic set the full_waiters_pending flag
before updating shortest_full, which when zero will always trigger
the writer to call the irq_work to wake up the waiters. The irq_work
will reset the shortest_full field back to zero as the woken waiters
is suppose to reset it.
- There's some optimized logic in the rb_watermark_hit() that is used
in ring_buffer_wait(). Use that helper function in the poll logic as
well.
- Restructure ring_buffer_wait() to use wait_event_interruptible()
The logic to wake up pending readers when the file descriptor is
closed is racy. Restructure ring_buffer_wait() to allow callers to
pass in conditions besides the ring buffer having enough data in it
by using wait_event_interruptible().
- Update the tracing_wait_on_pipe() to call ring_buffer_wait() with its
own conditions to exit the wait loop.
* tag 'trace-ring-buffer-v6.8-rc7-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing/ring-buffer: Fix wait_on_pipe() race
ring-buffer: Use wait_event_interruptible() in ring_buffer_wait()
ring-buffer: Reuse rb_watermark_hit() for the poll logic
ring-buffer: Fix full_waiters_pending in poll
ring-buffer: Do not set shortest_full when full target is hit
Pull probes updates from Masami Hiramatsu:
"x86 kprobes:
- Use boolean for some function return instead of 0 and 1
- Prohibit probing on INT/UD. This prevents user to put kprobe on
INTn/INT1/INT3/INTO and UD0/UD1/UD2 because these are used for a
special purpose in the kernel
- Boost Grp instructions. Because a few percent of kernel
instructions are Grp 2/3/4/5 and those are safe to be executed
without ip register fixup, allow those to be boosted (direct
execution on the trampoline buffer with a JMP)
tracing:
- Add function argument access from return events (kretprobe and
fprobe). This allows user to compare how a data structure field is
changed after executing a function. With BTF, return event also
accepts function argument access by name.
- Fix a wrong comment (using "Kretprobe" in fprobe)
- Cleanup a big probe argument parser function into three parts, type
parser, post-processing function, and main parser
- Cleanup to set nr_args field when initializing trace_probe instead
of counting up it while parsing
- Cleanup a redundant #else block from tracefs/README source code
- Update selftests to check entry argument access from return probes
- Documentation update about entry argument access from return
probes"
* tag 'probes-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
Documentation: tracing: Add entry argument access at function exit
selftests/ftrace: Add test cases for entry args at function exit
tracing/probes: Support $argN in return probe (kprobe and fprobe)
tracing: Remove redundant #else block for BTF args from README
tracing/probes: cleanup: Set trace_probe::nr_args at trace_probe_init
tracing/probes: Cleanup probe argument parser
tracing/fprobe-event: cleanup: Fix a wrong comment in fprobe event
x86/kprobes: Boost more instructions from grp2/3/4/5
x86/kprobes: Prohibit kprobing on INT and UD
x86/kprobes: Refactor can_{probe,boost} return type to bool
When the trace_pipe_raw file is closed, there should be no new readers on
the file descriptor. This is mostly handled with the waking and wait_index
fields of the iterator. But there's still a slight race.
CPU 0 CPU 1
----- -----
wait_index++;
index = wait_index;
ring_buffer_wake_waiters();
wait_on_pipe()
ring_buffer_wait();
The ring_buffer_wait() will miss the wakeup from CPU 1. The problem is
that the ring_buffer_wait() needs the logic of:
prepare_to_wait();
if (!condition)
schedule();
Where the missing condition check is the iter->wait_index update.
Have the ring_buffer_wait() take a conditional callback function and a
data parameter that can be used within the wait_event_interruptible() of
the ring_buffer_wait() function.
In wait_on_pipe(), pass a condition function that will check if the
wait_index has been updated, if it has, it will return true to break out
of the wait_event_interruptible() loop.
Create a new field "closed" in the trace_iterator and set it in the
.flush() callback before calling ring_buffer_wake_waiters().
This will keep any new readers from waiting on a closed file descriptor.
Have the wait_on_pipe() condition callback also check the closed field.
Change the wait_index field of the trace_iterator to atomic_t. There's no
reason it needs to be 'long' and making it atomic and using
atomic_read_acquire() and atomic_fetch_inc_release() will provide the
necessary memory barriers.
Add a "woken" flag to tracing_buffers_splice_read() to exit the loop after
one more try to fetch data. That is, if it waited for data and something
woke it up, it should try to collect any new data and then exit back to
user space.
Link: https://lore.kernel.org/linux-trace-kernel/CAHk-=wgsNgewHFxZAJiAQznwPMqEtQmi1waeS2O1v6L4c_Um5A@mail.gmail.com/
Link: https://lore.kernel.org/linux-trace-kernel/20240312121703.557950713@goodmis.org
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linke li <lilinke99@qq.com>
Cc: Rabin Vincent <rabin@rab.in>
Fixes: f3ddb74ad0 ("tracing: Wake up ring buffer waiters on closing of the file")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The .release() function does not get called until all readers of a file
descriptor are finished.
If a thread is blocked on reading a file descriptor in ring_buffer_wait(),
and another thread closes the file descriptor, it will not wake up the
other thread as ring_buffer_wake_waiters() is called by .release(), and
that will not get called until the .read() is finished.
The issue originally showed up in trace-cmd, but the readers are actually
other processes with their own file descriptors. So calling close() would wake
up the other tasks because they are blocked on another descriptor then the
one that was closed(). But there's other wake ups that solve that issue.
When a thread is blocked on a read, it can still hang even when another
thread closed its descriptor.
This is what the .flush() callback is for. Have the .flush() wake up the
readers.
Link: https://lore.kernel.org/linux-trace-kernel/20240308202432.107909457@goodmis.org
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linke li <lilinke99@qq.com>
Cc: Rabin Vincent <rabin@rab.in>
Fixes: f3ddb74ad0 ("tracing: Wake up ring buffer waiters on closing of the file")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Limit the max print event of trace_marker to just 4K string size. This must
also be less than the amount that can be held by a trace_seq along with
the text that is before the output (like the task name, PID, CPU, state,
etc). As trace_seq is made to handle large events (some greater than 4K).
Make the max size of a trace_marker write event be 4K which is guaranteed
to fit in the trace_seq buffer.
Link: https://lore.kernel.org/linux-trace-kernel/20240304223433.4ba47dff@gandalf.local.home
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Support accessing $argN in the return probe events. This will help users to
record entry data in function return (exit) event for simplfing the function
entry/exit information in one event, and record the result values (e.g.
allocated object/initialized object) at function exit.
For example, if we have a function `int init_foo(struct foo *obj, int param)`
sometimes we want to check how `obj` is initialized. In such case, we can
define a new return event like below;
# echo 'r init_foo retval=$retval param=$arg2 field1=+0($arg1)' >> kprobe_events
Thus it records the function parameter `param` and its result `obj->field1`
(the dereference will be done in the function exit timing) value at once.
This also support fprobe, BTF args and'$arg*'. So if CONFIG_DEBUG_INFO_BTF
is enabled, we can trace both function parameters and the return value
by following command.
# echo 'f target_function%return $arg* $retval' >> dynamic_events
Link: https://lore.kernel.org/all/170952365552.229804.224112990211602895.stgit@devnote2/
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
While looking at improving the saved_cmdlines cache I found a huge amount
of wasted memory that should be used for the cmdlines.
The tracing data saves pids during the trace. At sched switch, if a trace
occurred, it will save the comm of the task that did the trace. This is
saved in a "cache" that maps pids to comms and exposed to user space via
the /sys/kernel/tracing/saved_cmdlines file. Currently it only caches by
default 128 comms.
The structure that uses this creates an array to store the pids using
PID_MAX_DEFAULT (which is usually set to 32768). This causes the structure
to be of the size of 131104 bytes on 64 bit machines.
In hex: 131104 = 0x20020, and since the kernel allocates generic memory in
powers of two, the kernel would allocate 0x40000 or 262144 bytes to store
this structure. That leaves 131040 bytes of wasted space.
Worse, the structure points to an allocated array to store the comm names,
which is 16 bytes times the amount of names to save (currently 128), which
is 2048 bytes. Instead of allocating a separate array, make the structure
end with a variable length string and use the extra space for that.
This is similar to a recommendation that Linus had made about eventfs_inode names:
https://lore.kernel.org/all/20240130190355.11486-5-torvalds@linux-foundation.org/
Instead of allocating a separate string array to hold the saved comms,
have the structure end with: char saved_cmdlines[]; and round up to the
next power of two over sizeof(struct saved_cmdline_buffers) + num_cmdlines * TASK_COMM_LEN
It will use this extra space for the saved_cmdline portion.
Now, instead of saving only 128 comms by default, by using this wasted
space at the end of the structure it can save over 8000 comms and even
saves space by removing the need for allocating the other array.
Link: https://lore.kernel.org/linux-trace-kernel/20240209063622.1f7b6d5f@rorschach.local.home
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Vincent Donnefort <vdonnefort@google.com>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Mete Durlu <meted@linux.ibm.com>
Fixes: 939c7a4f04 ("tracing: Introduce saved_cmdlines_size file")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull tracing updates from Steven Rostedt:
- Allow kernel trace instance creation to specify what events are
created
Inside the kernel, a subsystem may create a tracing instance that it
can use to send events to user space. This sub-system may not care
about the thousands of events that exist in eventfs. Allow the
sub-system to specify what sub-systems of events it cares about, and
only those events are exposed to this instance.
- Allow the ring buffer to be broken up into bigger sub-buffers than
just the architecture page size.
A new tracefs file called "buffer_subbuf_size_kb" is created. The
user can now specify a minimum size the sub-buffer may be in
kilobytes. Note, that the implementation currently make the
sub-buffer size a power of 2 pages (1, 2, 4, 8, 16, ...) but the user
only writes in kilobyte size, and the sub-buffer will be updated to
the next size that it will can accommodate it. If the user writes in
10, it will change the size to be 4 pages on x86 (16K), as that is
the next available size that can hold 10K pages.
- Update the debug output when a corrupt time is detected in the ring
buffer. If the ring buffer detects inconsistent timestamps, there's a
debug config options that will dump the contents of the meta data of
the sub-buffer that is used for debugging. Add some more information
to this dump that helps with debugging.
- Add more timestamp debugging checks (only triggers when the config is
enabled)
- Increase the trace_seq iterator to 2 page sizes.
- Allow strings written into tracefs_marker to be larger. Up to just
under 2 page sizes (based on what trace_seq can hold).
- Increase the trace_maker_raw write to be as big as a sub-buffer can
hold.
- Remove 32 bit time stamp logic, now that the rb_time_cmpxchg() has
been removed.
- More selftests were added.
- Some code clean ups as well.
* tag 'trace-v6.8' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (29 commits)
ring-buffer: Remove stale comment from ring_buffer_size()
tracing histograms: Simplify parse_actions() function
tracing/selftests: Remove exec permissions from trace_marker.tc test
ring-buffer: Use subbuf_order for buffer page masking
tracing: Update subbuffer with kilobytes not page order
ringbuffer/selftest: Add basic selftest to test changing subbuf order
ring-buffer: Add documentation on the buffer_subbuf_order file
ring-buffer: Just update the subbuffers when changing their allocation order
ring-buffer: Keep the same size when updating the order
tracing: Stop the tracing while changing the ring buffer subbuf size
tracing: Update snapshot order along with main buffer order
ring-buffer: Make sure the spare sub buffer used for reads has same size
ring-buffer: Do no swap cpu buffers if order is different
ring-buffer: Clear pages on error in ring_buffer_subbuf_order_set() failure
ring-buffer: Read and write to ring buffers with custom sub buffer size
ring-buffer: Set new size of the ring buffer sub page
ring-buffer: Add interface for configuring trace sub buffer size
ring-buffer: Page size per ring buffer
ring-buffer: Have ring_buffer_print_page_header() be able to access ring_buffer_iter
ring-buffer: Check if absolute timestamp goes backwards
...
If an application blocks on the snapshot or snapshot_raw files, expecting
to be woken up when a snapshot occurs, it will not happen. Or it may
happen with an unexpected result.
That result is that the application will be reading the main buffer
instead of the snapshot buffer. That is because when the snapshot occurs,
the main and snapshot buffers are swapped. But the reader has a descriptor
still pointing to the buffer that it originally connected to.
This is fine for the main buffer readers, as they may be blocked waiting
for a watermark to be hit, and when a snapshot occurs, the data that the
main readers want is now on the snapshot buffer.
But for waiters of the snapshot buffer, they are waiting for an event to
occur that will trigger the snapshot and they can then consume it quickly
to save the snapshot before the next snapshot occurs. But to do this, they
need to read the new snapshot buffer, not the old one that is now
receiving new data.
Also, it does not make sense to have a watermark "buffer_percent" on the
snapshot buffer, as the snapshot buffer is static and does not receive new
data except all at once.
Link: https://lore.kernel.org/linux-trace-kernel/20231228095149.77f5b45d@gandalf.local.home
Cc: stable@vger.kernel.org
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Fixes: debdd57f51 ("tracing: Make a snapshot feature available from userspace")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
