tracing: Fixes for 5.5
Subsystems (4)
Steven Rostedt
maintainercover tracing: Fixes for 5.5
Colin Ian King (1):
tracing: Fix indentation issue
Joel Fernandes (Google) (1):
tracing: Change offset type to s32 in preempt/irq tracepoints
Kaitao Cheng (1):
kernel/trace: Fix do not unregister tracepoints when register sched_migrate_task fail
Steven Rostedt (VMware) (3):
tracing: Initialize val to zero in parse_entry of inject code
tracing: Define MCOUNT_INSN_SIZE when not defined without direct calls
tracing: Have stack tracer compile when MCOUNT_INSN_SIZE is not defined
Wen Yang (1):
ftrace: Avoid potential division by zero in function profiler
----
include/trace/events/preemptirq.h | 8 ++++---- kernel/trace/fgraph.c | 14 ++++++++++++++ kernel/trace/ftrace.c | 6 +++--- kernel/trace/trace_events_inject.c | 2 +- kernel/trace/trace_sched_wakeup.c | 4 +++- kernel/trace/trace_seq.c | 2 +- kernel/trace/trace_stack.c | 5 +++++ 7 files changed, 31 insertions(+), 10 deletions(-)
Steven Rostedt
maintainerPATCH 1/7 tracing: Initialize val to zero in parse_entry of inject code
From: "Steven Rostedt (VMware)" <[email protected]>
gcc produces a variable may be uninitialized warning for "val" in parse_entry(). This is really a false positive, but the code is subtle enough to just initialize val to zero and it's not a fast path to worry about it.
Marked for stable to remove the warning in the stable trees as well.
--- kernel/trace/trace_events_inject.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/trace/trace_events_inject.c +1 -1
@@ -195,7 +195,7 @@ static int parse_entry(char *str, struct trace_event_call *call, void **pentry)unsigned long irq_flags;void *entry =NULL ;int entry_size; -u64 val; +u64 val =0 ;int len; entry =trace_alloc_entry (call, &entry_size);
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 2/7 tracing: Define MCOUNT_INSN_SIZE when not defined without direct calls
In reply to Steven RostedtFrom: "Steven Rostedt (VMware)" <[email protected]>
In order to handle direct calls along side of function graph tracer, a check is made to see if the address being traced by the function graph tracer is a direct call or not. To get the address used by direct callers, the return address is subtracted by MCOUNT_INSN_SIZE.
For some archs with certain configurations, MCOUNT_INSN_SIZE is undefined here. But these should not be using direct calls anyway. Just define MCOUNT_INSN_SIZE to zero in this case.
--- kernel/trace/fgraph.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
kernel/trace/fgraph.c +14 -0
@@ -96,6 +96,20 @@ ftrace_push_return_trace(unsigned long ret, unsigned long func,return 0 ; } +/* +* Not all archs define MCOUNT_INSN_SIZE which is used to look for direct +* functions. But those archs currently don't support direct functions +* anyway, and ftrace_find_rec_direct() is just a stub for them. +* Define MCOUNT_INSN_SIZE to keep those archs compiling. +*/ +#ifndef MCOUNT_INSN_SIZE +/* Make sure this only works without direct calls */ +# ifdef CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS +# error MCOUNT_INSN_SIZE not defined with direct calls enabled +# endif +# define MCOUNT_INSN_SIZE 0 +#endif +int function_graph_enter (unsigned long ret,unsigned long func,unsigned long frame_pointer,unsigned long *retp) {
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 3/7 tracing: Have stack tracer compile when MCOUNT_INSN_SIZE is not defined
In reply to Steven RostedtFrom: "Steven Rostedt (VMware)" <[email protected]>
On some archs with some configurations, MCOUNT_INSN_SIZE is not defined, and this makes the stack tracer fail to compile. Just define it to zero in this case.
--- kernel/trace/trace_stack.c | 5 +++++ 1 file changed, 5 insertions(+)
kernel/trace/trace_stack.c +5 -0
@@ -283,6 +283,11 @@ static void check_stack(unsigned long ip, unsigned long *stack)local_irq_restore (flags); } +/* Some archs may not define MCOUNT_INSN_SIZE */ +#ifndef MCOUNT_INSN_SIZE +# define MCOUNT_INSN_SIZE 0 +#endif +static void stack_trace_call (unsigned long ip,unsigned long parent_ip,struct ftrace_ops *op,struct pt_regs *pt_regs)
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 4/7 ftrace: Avoid potential division by zero in function profiler
In reply to Steven RostedtFrom: Wen Yang <[email protected]>
The ftrace_profile->counter is unsigned long and do_div truncates it to 32 bits, which means it can test non-zero and be truncated to zero for division. Fix this issue by using div64_ul() instead.
--- kernel/trace/ftrace.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/trace/ftrace.c +3 -3
@@ -526,8 +526,7 @@ static int function_stat_show(struct seq_file *m, void *v) }#ifdef CONFIG_FUNCTION_GRAPH_TRACER - avg = rec->time; -do_div (avg, rec->counter); + avg =div64_ul (rec->time, rec->counter);if (tracing_thresh && (avg < tracing_thresh)) goto out;#endif
@@ -553,7 +552,8 @@ static int function_stat_show(struct seq_file *m, void *v)* Divide only 1000 for ns^2 -> us^2 conversion. * trace_print_graph_duration will divide 1000 again. */ -do_div (stddev, rec->counter * (rec->counter -1 ) *1000 ); + stddev =div64_ul (stddev, + rec->counter * (rec->counter -1 ) *1000 ); }trace_seq_init (&s);
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 5/7 tracing: Change offset type to s32 in preempt/irq tracepoints
In reply to Steven RostedtFrom: "Joel Fernandes (Google)" <[email protected]>
Discussion in the below link reported that symbols in modules can appear to be before _stext on ARM architecture, causing wrapping with the offsets of this tracepoint. Change the offset type to s32 to fix this.
--- include/trace/events/preemptirq.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
include/trace/events/preemptirq.h +4 -4
@@ -18,13 +18,13 @@ DECLARE_EVENT_CLASS(preemptirq_template,TP_ARGS (ip, parent_ip),TP_STRUCT__entry ( - __field(u32, caller_offs) -__field (u32, parent_offs) + __field(s32, caller_offs) +__field (s32, parent_offs) ),TP_fast_assign ( - __entry->caller_offs = (u32)(ip - (unsigned long)_stext); - __entry->parent_offs = (u32)(parent_ip - (unsigned long )_stext); + __entry->caller_offs = (s32)(ip - (unsigned long)_stext); + __entry->parent_offs = (s32)(parent_ip - (unsigned long )_stext); ), TP_printk("caller=%pS parent=%pS",
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 6/7 kernel/trace: Fix do not unregister tracepoints when register sched_migrate_task fail
In reply to Steven RostedtFrom: Kaitao Cheng <[email protected]>
In the function, if register_trace_sched_migrate_task() returns error, sched_switch/sched_wakeup_new/sched_wakeup won't unregister. That is why fail_deprobe_sched_switch was added.
--- kernel/trace/trace_sched_wakeup.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/trace/trace_sched_wakeup.c +3 -1
@@ -630,7 +630,7 @@ static void start_wakeup_tracer(struct trace_array *tr)if (ret) {pr_info ("wakeup trace: Couldn't activate tracepoint" " probe to kernel_sched_migrate_task\n" ); -return ; + goto fail_deprobe_sched_switch; }wakeup_reset (tr);
@@ -648,6 +648,8 @@ static void start_wakeup_tracer(struct trace_array *tr)printk (KERN_ERR "failed to start wakeup tracer\n" );return ; +fail_deprobe_sched_switch: +unregister_trace_sched_switch (probe_wakeup_sched_switch,NULL ); fail_deprobe_wake_new:unregister_trace_sched_wakeup_new (probe_wakeup,NULL ); fail_deprobe:
Signature · 1 line
2.24.0
Steven Rostedt
maintainerPATCH 7/7 tracing: Fix indentation issue
In reply to Steven RostedtFrom: Colin Ian King <[email protected]>
There is a declaration that is indented one level too deeply, remove the extraneous tab.
--- kernel/trace/trace_seq.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/trace/trace_seq.c +1 -1
@@ -381,7 +381,7 @@ int trace_seq_hex_dump(struct trace_seq *s, const char *prefix_str, int prefix_type, int rowsize, int groupsize, const void *buf, size_t len, bool ascii) { - unsigned int save_len = s->seq.len; + unsigned int save_len = s->seq.len;if (s->full)return 0 ;
Signature · 1 line
2.24.0