- 11 Apr, 2018 5 commits
-
-
Steven Rostedt (VMware) authored
There's some inconsistency with what to set the output parameter filterp when passing to create_filter(..., struct event_filter **filterp). Whatever filterp points to, should be NULL when calling this function. The create_filter() calls create_filter_start() with a pointer to a local "filter" variable that is set to NULL. The create_filter_start() has a WARN_ON() if the passed in pointer isn't pointing to a value set to NULL. Ideally, create_filter() should pass the filterp variable it received to create_filter_start() and not hide it as with a local variable, this allowed create_filter() to fail, and not update the passed in filter, and the caller of create_filter() then tried to free filter, which was never initialized to anything, causing memory corruption. Link: http://lkml.kernel.org/r/00000000000032a0c30569916870@google.com Fixes: 80765597 ("tracing: Rewrite filter logic to be simpler and faster") Reported-by: syzbot+dadcc936587643d7f568@syzkaller.appspotmail.com Signed-off-by:
Steven Rostedt (VMware) <rostedt@goodmis.org>
-
Ravi Bangoria authored
Simplify probes_seq_show() function. No change in output before and after patch. Link: http://lkml.kernel.org/r/20180315082756.9050-2-ravi.bangoria@linux.vnet.ibm.comAcked-by:
Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by:
Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Signed-off-by:
-
Ravi Bangoria authored
tu->offset is unsigned long, not a pointer, thus %lx should be used to print it, not the %px. Link: http://lkml.kernel.org/r/20180315082756.9050-1-ravi.bangoria@linux.vnet.ibm.com Cc: stable@vger.kernel.org Acked-by:
Kees Cook <keescook@chromium.org> Signed-off-by:
-
Howard McLauchlan authored
uprobes cannot successfully attach to binaries located in a directory mounted with overlayfs. To verify, create directories for mounting overlayfs (upper,lower,work,merge), move some binary into merge/ and use readelf to obtain some known instruction of the binary. I used /bin/true and the entry instruction(0x13b0): $ mount -t overlay overlay -o lowerdir=lower,upperdir=upper,workdir=work merge $ cd /sys/kernel/debug/tracing $ echo 'p:true_entry PATH_TO_MERGE/merge/true:0x13b0' > uprobe_events $ echo 1 > events/uprobes/true_entry/enable This returns 'bash: echo: write error: Input/output error' and dmesg tells us 'event trace: Could not enable event true_entry' This change makes create_trace_uprobe() look for the real inode of a dentry. In the case of normal filesystems, this simplifies to just returning the inode. In the case of overlayfs(and similar fs) we will obtain the underlying dentry and corresponding inode, upon which uprobes can successfully register. Running the example above with the patch applied, we can see that the uprobe is enabled and will output to trace as expected. Link: http://lkml.kernel.org/r/20180410231030.2720-1-hmclauchlan@fb.comReviewed-by:
Josef Bacik <jbacik@fb.com> Reviewed-by:
Srikar Dronamraju <srikar@linux.vnet.ibm.com> Signed-off-by:
Howard McLauchlan <hmclauchlan@fb.com> Signed-off-by:
-
Jérémy Lefaure authored
It is useless to re-invent the ARRAY_SIZE macro so let's use it instead of DATA_CNT. Found with Coccinelle with the following semantic patch: @r depends on (org || report)@ type T; T[] E; position p; @@ ( (sizeof(E)@p /sizeof(*E)) | (sizeof(E)@p /sizeof(E[...])) | (sizeof(E)@p /sizeof(T)) ) Link: http://lkml.kernel.org/r/20171016012250.26453-1-jeremy.lefaure@lse.epita.frSigned-off-by:
Jérémy Lefaure <jeremy.lefaure@lse.epita.fr> [ Removed useless include of kernel.h ] Signed-off-by:
-
- 08 Apr, 2018 1 commit
-
-
Steven Rostedt (VMware) authored
Add macros around the initcall_debug tracepoint code to have the code to default back to the old method if CONFIG_TRACEPOINTS is not enabled. Signed-off-by:
-
- 06 Apr, 2018 19 commits
-
-
Steven Rostedt (VMware) authored
With trace events set before and after the initcall function calls, instead of having a separate routine for printing out the initcalls when initcall_debug is specified on the kernel command line, have the code register a callback to the tracepoints where the initcall trace events are. This removes the need for having a separate function to do the initcalls as the tracepoint callbacks can handle the printk. It also includes other initcalls that are not called by the do_one_initcall() which includes console and security initcalls. Signed-off-by: -
Abderrahmane Benbachir authored
Trace events have been added around the initcall functions defined in init/main.c. But console and security have their own initcalls. This adds the trace events associated for those initcall functions. Link: http://lkml.kernel.org/r/1521765208.19745.2.camel@polymtl.ca Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by:
Abderrahmane Benbachir <abderrahmane.benbachir@polymtl.ca> Signed-off-by:
-
Steven Rostedt (VMware) authored
Being able to trace the start and stop of initcalls is useful to see where the timings are an issue. There is already an "initcall_debug" parameter, but that can cause a large overhead itself, as the printing of the information may take longer than the initcall functions. Adding in a start and finish trace event around the initcall functions, as well as a trace event that records the level of the initcalls, one can get a much finer measurement of the times and interactions of the initcalls themselves, as trace events are much lighter than printk()s. Suggested-by:
-
Steven Rostedt (VMware) authored
A boot up test function update_pred_fn() dereferences filter->prog without the proper rcu annotation. To do this, we must also take the event_mutex first. Normally, this isn't needed because this test function can not race with other use cases that touch the event filters (it is disabled if any events are enabled). Reported-by:
kbuild test robot <fengguang.wu@intel.com> Fixes: 80765597 ("tracing: Rewrite filter logic to be simpler and faster") Signed-off-by:
-
Steven Rostedt (VMware) authored
ftrace_function_set_filter() referenences filter->prog without annotation and sparse complains about it. It needs a rcu_dereference_protected() wrapper. Reported-by:
-
Chris Wilson authored
In commit 932066a15335 ("tracing: Default to using trace_global_clock if sched_clock is unstable"), the logic for deciding to override the default clock if unstable was reversed from the earlier posting. I was trying to reduce the width of the message by using an early return rather than a if-block, but reverted back to using the if-block and accidentally left the predicate inverted. Link: http://lkml.kernel.org/r/20180404212450.26646-1-chris@chris-wilson.co.uk Fixes: 932066a15335 ("tracing: Default to using trace_global_clock if sched_clock is unstable") Signed-off-by:Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by:
-
Steven Rostedt (VMware) authored
Function tracing can trace in NMIs and such. If the TSC is determined to be unstable, the tracing clock will switch to the global clock on boot up, unless "trace_clock" is specified on the kernel command line. The global clock disables interrupts to access sched_clock_cpu(), and in doing so can be done within lockdep internals (because of function tracing and NMIs). This can trigger false lockdep splats. The trace_clock_global() is special, best not to trace the irq logic within it. Link: http://lkml.kernel.org/r/20180404145015.77bde42d@gandalf.local.homeSigned-off-by:
-
Steven Rostedt (VMware) authored
As si_mem_available() can say there is enough memory even though the memory available is not useable by the ring buffer, it is best to not kill innocent applications because the ring buffer is taking up all the memory while it is trying to allocate a great deal of memory. If the allocator is user space (because kernel threads can also increase the size of the kernel ring buffer on boot up), then after si_mem_available() says there is enough memory, set the OOM killer to kill the current task if an OOM triggers during the allocation. Link: http://lkml.kernel.org/r/20180404062340.GD6312@dhcp22.suse.czSuggested-by:
Michal Hocko <mhocko@kernel.org> Signed-off-by:
-
Steven Rostedt (VMware) authored
The ring buffer is made up of a link list of pages. When making the ring buffer bigger, it will allocate all the pages it needs before adding to the ring buffer, and if it fails, it frees them and returns an error. This makes increasing the ring buffer size an all or nothing action. When this was first created, the pages were allocated with "NORETRY". This was to not cause any Out-Of-Memory (OOM) actions from allocating the ring buffer. But NORETRY was too strict, as the ring buffer would fail to expand even when there's memory available, but was taken up in the page cache. Commit 84861885 ("tracing/ring_buffer: Try harder to allocate") changed the allocating from NORETRY to RETRY_MAYFAIL. The RETRY_MAYFAIL would allocate from the page cache, but if there was no memory available, it would simple fail the allocation and not trigger an OOM. This worked fine, but had one problem. As the ring buffer would allocate one page at a time, it could take up all memory in the system before it failed to allocate and free that memory. If the allocation is happening and the ring buffer allocates all memory and then tries to take more than available, its allocation will not trigger an OOM, but if there's any allocation that happens someplace else, that could trigger an OOM, even though once the ring buffer's allocation fails, it would free up all the previous memory it tried to allocate, and allow other memory allocations to succeed. Commit d02bd27b ("mm/page_alloc.c: calculate 'available' memory in a separate function") separated out si_mem_availble() as a separate function that could be used to see how much memory is available in the system. Using this function to make sure that the ring buffer could be allocated before it tries to allocate pages we can avoid allocating all memory in the system and making it vulnerable to OOMs if other allocations are taking place. Link: http://lkml.kernel.org/r/1522320104-6573-1-git-send-email-zhaoyang.huang@spreadtrum.com CC: stable@vger.kernel.org Cc: linux-mm@kvack.org Fixes: 84861885 ("tracing/ring_buffer: Try harder to allocate") Requires: d02bd27b ("mm/page_alloc.c: calculate 'available' memory in a separate function") Reported-by:
Zhaoyang Huang <huangzhaoyang@gmail.com> Tested-by:
Joel Fernandes <joelaf@google.com> Signed-off-by:
-
Steven Rostedt (VMware) authored
Running a test on a x86_32 kernel I triggered a bug that an interrupt disable/enable isn't being catched by lockdep. At least knowing where the last one was found would be helpful, but the warnings that are produced do not show this information. Even without debugging lockdep, having the WARN() display the last place hard and soft irqs were enabled or disabled is valuable. Signed-off-by: -
Steven Rostedt (VMware) authored
Commit 841a915d ("printf: Do not have bprintf dereference pointers") would preprocess various pointers that are dereferenced in the bprintf() because the recording and printing are done at two different times. Some pointers stayed dereferenced in the ring buffer because user space could handle them (namely "%pS" and friends). Pointers that are not dereferenced should not be processed immediately but instead just saved directly. Cc: stable@vger.kernel.org Fixes: 841a915d ("printf: Do not have bprintf dereference pointers") Signed-off-by:
-
Dan Carpenter authored
Smatch complains that idx can be used uninitialized when we check if (idx < 0). It has to be the first iteration through the loop and the HIST_FIELD_FL_STACKTRACE bit has to be clear and the HIST_FIELD_FL_VAR bit has to be set to reach the bug. Link: http://lkml.kernel.org/r/20180328114815.GC29050@mwanda Fixes: 30350d65 ("tracing: Add variable support to hist triggers") Acked-by:
Tom Zanussi <tom.zanussi@linux.intel.com> Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Tom Zanussi authored
The strncpy() currently being used for variable string fields can result in a lack of termination if the string length is equal to the field size. Use the safer strscpy() instead, which will guarantee termination. Link: http://lkml.kernel.org/r/fb97c1e518fb358c12a4057d7445ba2c46956cd7.1522256721.git.tom.zanussi@linux.intel.comSigned-off-by:
-
Tom Zanussi authored
Actions also need to be considered when checking for matching triggers - triggers differing only by action should be allowed, but currently aren't because the matching check ignores the action and erroneously returns -EEXIST. Add and call an actions_match() function to address that. Here's an example using onmatch() actions. The first -EEXIST shouldn't occur because the onmatch() is different in the second wakeup_latency() param. The second -EEXIST shouldn't occur because it's a different action (in this case, it doesn't have an action, so shouldn't be seen as being the same and therefore rejected). In the after case, both are correctly accepted (and trying to add one of them again returns -EEXIST as it should). before: # echo 'wakeup_latency u64 lat; pid_t pid' >> /sys/kernel/debug/tracing/synthetic_events # echo 'hist:keys=pid:ts0=common_timestamp.usecs if comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_wakeup/trigger # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0 if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger # echo 'hist:keys=next_pid:onmatch(sched.sched_wakeup).wakeup_latency(sched.sched_switch.$wakeup_lat,next_pid) if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger # echo 'hist:keys=next_pid:onmatch(sched.sched_wakeup).wakeup_latency(sched.sched_switch.$wakeup_lat,prev_pid) if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger -su: echo: write error: File exists # echo 'hist:keys=next_pid if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger -su: echo: write error: File exists after: # echo 'wakeup_latency u64 lat; pid_t pid' >> /sys/kernel/debug/tracing/synthetic_events # echo 'hist:keys=pid:ts0=common_timestamp.usecs if comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_wakeup/trigger # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0 if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger # echo 'hist:keys=next_pid:onmatch(sched.sched_wakeup).wakeup_latency(sched.sched_switch.$wakeup_lat,next_pid) if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger # echo 'hist:keys=next_pid:onmatch(sched.sched_wakeup).wakeup_latency(sched.sched_switch.$wakeup_lat,prev_pid) if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger # echo 'hist:keys=next_pid if next_comm=="cyclictest"' >> /sys/kernel/debug/tracing/events/sched/sched_switch/trigger Link: http://lkml.kernel.org/r/a7fd668b87ec10736c8f016ac4279c8480d50c2b.1522256721.git.tom.zanussi@linux.intel.comTested-by:
-
Tom Zanussi authored
Variable references should never have flags appended when displayed - prevent that from happening. Before: # cat /sys/kernel/debug/tracing/events/sched/sched_switch/trigger hist:keys=next_pid:vals=hitcount:wakeup_lat=common_timestamp.usecs-$ts0.usecs:... After: hist:keys=next_pid:vals=hitcount:wakeup_lat=common_timestamp.usecs-$ts0:... Link: http://lkml.kernel.org/r/913318a5610ef6b24af2522575f671fa6ee19b6b.1522256721.git.tom.zanussi@linux.intel.comSigned-off-by:
-
Tom Zanussi authored
When displaying hist triggers, variable references that have the timestamp field flag set are erroneously displayed as common_timestamp rather than the variable reference. Additionally, timestamp expressions are displayed in the same way. Fix this by forcing the timestamp flag handling to follow variable reference and expression handling. Before: # cat /sys/kernel/debug/tracing/events/sched/sched_switch/trigger hist:keys=next_pid:vals=hitcount:wakeup_lat=common_timestamp.usecs:... After: # cat /sys/kernel/debug/tracing/events/sched/sched_switch/trigger hist:keys=next_pid:vals=hitcount:wakeup_lat=common_timestamp.usecs-$ts0.usecs:... Link: http://lkml.kernel.org/r/92746b06be67499c2a6217bd55395b350ad18fad.1522256721.git.tom.zanussi@linux.intel.comSigned-off-by:
-
Salvatore Mesoraca authored
Avoid a VLA by using a real constant expression instead of a variable. The compiler should be able to optimize the original code and avoid using an actual VLA. Anyway this change is useful because it will avoid a false positive with -Wvla, it might also help the compiler generating better code. Link: http://lkml.kernel.org/r/CA+55aFzCG-zNmZwX4A2FQpadafLfEzK6CC=qPXydAacU1RqZWA@mail.gmail.com Link: http://lkml.kernel.org/r/1522399988-8815-1-git-send-email-s.mesoraca16@gmail.comSigned-off-by:
Salvatore Mesoraca <s.mesoraca16@gmail.com> Signed-off-by:
-
Chris Wilson authored
Mention the alternative of adding trace_clock=global to the kernel command line when we detect that we've used an unstable clock across a suspend/resume cycle. Link: http://lkml.kernel.org/r/20180330150132.16903-2-chris@chris-wilson.co.ukSigned-off-by:
-
Chris Wilson authored
Across suspend, we may see a very large drift in timestamps if the sched clock is unstable, prompting the global trace's ringbuffer code to warn and suggest switching to the global clock. Preempt this request by detecting when the sched clock is unstable (determined during late_initcall) and automatically switching the default clock over to trace_global_clock. This should prevent requiring user interaction to resolve warnings such as: Delta way too big! 18446743856563626466 ts=18446744054496180323 write stamp = 197932553857 If you just came from a suspend/resume, please switch to the trace global clock: echo global > /sys/kernel/debug/tracing/trace_clock Link: http://lkml.kernel.org/r/20180330150132.16903-1-chris@chris-wilson.co.ukSigned-off-by:
-
- 23 Mar, 2018 2 commits
-
-
Steven Rostedt (VMware) authored
The early_initcall() functions get assigned to __initcall_start[]. These are called by do_pre_smp_initcalls(). The initcall_levels[] array starts with __initcall0_start[], and initcall_levels[] are to match the initcall_level_names[] array. The first name in that array is "early", but that is not correct. As pure_initcall() functions get assigned to __initcall0_start[] array. Change the first name in initcall_level_names[] array to "pure". Signed-off-by: -
Dan Carpenter authored
We forgot to set the error code on this path so we return ERR_PTR(0) which is NULL. It results in a NULL dereference in the caller. Link: http://lkml.kernel.org/r/20180323113735.GC28518@mwanda Fixes: 100719dc ("tracing: Add simple expression support to hist triggers") Acked-by:
-
- 14 Mar, 2018 3 commits
-
-
Steven Rostedt (VMware) authored
Al Viro reviewed the filter logic of ftrace trace events and found it to be very troubling. It creates a binary tree based on the logic operators and walks it during tracing. He sent myself and Tom Zanussi a long explanation (and formal proof) of how to do the string parsing better and end up with a program array that can be simply iterated to come up with the correct results. I took his ideas and his pseudo code and rewrote the filter logic based on them. In doing so, I was able to remove a lot of code, and have a much more condensed filter logic in the process. I wrote a very long comment describing the methadology that Al proposed in my own words. For more info on how this works, read the comment above predicate_parse(). Suggested-by:
Al Viro <viro@ZenIV.linux.org.uk> Signed-off-by:
-
Steven Rostedt (VMware) authored
The pred_funcs_##type arrays consist of five functions that are assigned based on the ops. The array must be in the same order of the ops each function represents. The PRED_FUNC_START macro denotes the op enum that starts the op that maps to the pred_funcs_##type arrays. This is all very subtle and prone to bugs if the code is changed. Add comments describing how PRED_FUNC_START and pred_funcs_##type array is used, and also a PRED_FUNC_MAX that is the maximum number of functions in the arrays. Clean up select_comparison_fn() that assigns the predicates to the pred_funcs_##type array function as well as add protection in case an op is passed in that does not map correctly to the array. Reviewed-by:
-
Steven Rostedt (VMware) authored
Instead of having a separate enum that is the index into another array, like a string array, make a single macro that combines them into a single list, and then the two can not get out of sync. This makes it easier to add and remove items. The macro trick is: #define DOGS \ C( JACK, "Jack Russell") \ C( ITALIAN, "Italian Greyhound") \ C( GERMAN, "German Shepherd") #undef C #define C(a, b) a enum { DOGS }; #undef C #define C(a, b) b static char dogs[] = { DOGS }; Reviewed-by:
-
- 10 Mar, 2018 10 commits
-
-
Steven Rostedt (VMware) authored
The replace_filter_string() frees the current string and then copies a given string. But in the two locations that it was used, the allocation happened right after the filter was allocated (nothing to replace). There's no need for this to be a helper function. Embedding the allocation in the two places where it was called will make changing the code in the future easier. Also make the variable consistent (always use "filter_string" as the name, as it was used in one instance as "filter_str") Signed-off-by: -
Steven Rostedt (VMware) authored
replace_system_preds() creates a filter list to free even when it doesn't really need to have it. Only save filters that require synchronize_sched() in the filter list to free. This will allow the code to be updated a bit easier in the future. Signed-off-by: -
Steven Rostedt (VMware) authored
The __alloc_filter() function does nothing more that allocate the filter. There's no reason to have it as a helper function. Signed-off-by: -
Steven Rostedt (VMware) authored
The filter code does open code string appending to produce an error message. Instead it can be simplified by using trace_seq function helpers. Signed-off-by: -
Steven Rostedt (VMware) authored
There's no reason to BUG if there's a bug in the filtering code. Simply do a warning and return. Signed-off-by: -
Rajvi Jingar authored
This adds inter-event hist triggers testcases which covers following: - create/remove synthetic event - disable histogram for synthetic event - extended error support - field variable support - histogram variables - histogram trigger onmatch action - histogram trigger onmax action - histogram trigger onmatch-onmax action - simple expression support - combined histogram Here is the test result. === Ftrace unit tests === [1] event trigger - test extended error support [PASS] [2] event trigger - test field variable support [PASS] [3] event trigger - test inter-event combined histogram trigger [PASS] [4] event trigger - test inter-event histogram trigger onmatch action [PASS] [5] event trigger - test inter-event histogram trigger onmatch-onmax action [PASS] [6] event trigger - test inter-event histogram trigger onmax action [PASS] [7] event trigger - test synthetic event create remove [PASS] Link: http://lkml.kernel.org/r/e07ef1e72f7bf0f84dc87c9b736d6dc91b4b0b49.1516069914.git.tom.zanussi@linux.intel.comSigned-off-by:Rajvi Jingar <rajvi.jingar@intel.com> Signed-off-by:
-
Tom Zanussi authored
So that users know that inter-event tracing is supported as part of the HIST_TRIGGERS option, include text to that effect in the help text. Link: http://lkml.kernel.org/r/a38e24231d8d980be636b56d35814570acfd167a.1516069914.git.tom.zanussi@linux.intel.comSigned-off-by:
-
Steven Rostedt (VMware) authored
Synthetic events can be done within the recording of other events. Notify the ring buffer via ring_buffer_nest_start() and ring_buffer_nest_end() that this is intended and not to block it due to its recursion protection. Signed-off-by: -
Steven Rostedt (VMware) authored
The ring-buffer code has recusion protection in case tracing ends up tracing itself, the ring-buffer will detect that it was called at the same context (normal, softirq, interrupt or NMI), and not continue to record the event. With the histogram synthetic events, they are called while tracing another event at the same context. The recusion protection triggers because it detects tracing at the same context and stops it. Add ring_buffer_nest_start() and ring_buffer_nest_end() that will notify the ring buffer that a trace is about to happen within another trace and that it is intended, and not to trigger the recursion blocking. Signed-off-by: -
Tom Zanussi authored
The default clock if timestamps are used in a histogram is "global". If timestamps aren't used, the clock is irrelevant. Use the "clock=" param only if you want to override the default "global" clock for a histogram with timestamps. Link: http://lkml.kernel.org/r/427bed1389c5d22aa40c3e0683e30cc3d151e260.1516069914.git.tom.zanussi@linux.intel.comSigned-off-by:
-
