ftrace not showing stack trace data during boot processes

ftrace not showing stack trace data during boot processes

[Subodh Nijsure]

I am trying to debug some boot sequences during kernel bootup and
trying to use ftrace to discover function flow.

I am booting kernel with following options.

stacktrace trace_buf_size=30M ftrace=function
ftrace_filter=do_sync_write,usbfs*,ext2_*,vfs_write,blkdev_*,vfs_write,sys_write
ftrace_notrace=ext2_permission

I have compiled kernel with following options

CONFIG_NOP_TRACER=y
CONFIG_HAVE_FUNCTION_TRACER=y
CONFIG_HAVE_DYNAMIC_FTRACE=y
CONFIG_HAVE_FTRACE_MCOUNT_RECORD=y
CONFIG_RING_BUFFER=y
CONFIG_TRACING=y
CONFIG_FUNCTION_TRACER=y
CONFIG_SYSPROF_TRACER=y
CONFIG_CONTEXT_SWITCH_TRACER=y
CONFIG_BOOT_TRACER=y
CONFIG_STACK_TRACER=y
CONFIG_DYNAMIC_FTRACE=y
CONFIG_FTRACE_MCOUNT_RECORD=y

However when I look at  /sys/kernel/debug/tracing/trace

I see output similar to this and  I don't see the "stack trace"
            init-1     [000]    65.317499: sys_write <-sysenter_do_call
            init-1     [000]    65.317504: vfs_write <-sys_write
            init-1     [000]    65.317617: sys_write <-sysenter_do_call
            init-1     [000]    65.317619: vfs_write <-sys_write

Is there option I am missing while compiling kernel or some part of
kernel parameters is missing? Is there way to explicitly specify
iter_ctrl on kernel boot cmdline?

/Subodh

Re: ftrace not showing stack trace data during boot processes

[Steven Rostedt]

Note, please Cc maintainers directly if you want a response. This mailing
list has a traffic of 600 emails a day, and things sent here can easily
be lost.

On Sun, Nov 29, 2009 at 06:20:35PM -0800, Subodh Nijsure wrote:
> I am trying to debug some boot sequences during kernel bootup and
> trying to use ftrace to discover function flow.
> 
> I am booting kernel with following options.
> 
> stacktrace trace_buf_size=30M ftrace=function

Note, the stacktrace is the stack tracer, not the tracing option of "stacktrace"
or the function trace option "func_stack_trace".

The output of the stack tracer just shows the max stack trace and is listed in

debugfs/tracing/stack_trace

> ftrace_filter=do_sync_write,usbfs*,ext2_*,vfs_write,blkdev_*,vfs_write,sys_write
> ftrace_notrace=ext2_permission
> 
> I have compiled kernel with following options
> 
> CONFIG_NOP_TRACER=y
> CONFIG_HAVE_FUNCTION_TRACER=y
> CONFIG_HAVE_DYNAMIC_FTRACE=y
> CONFIG_HAVE_FTRACE_MCOUNT_RECORD=y
> CONFIG_RING_BUFFER=y
> CONFIG_TRACING=y
> CONFIG_FUNCTION_TRACER=y
> CONFIG_SYSPROF_TRACER=y
> CONFIG_CONTEXT_SWITCH_TRACER=y
> CONFIG_BOOT_TRACER=y
> CONFIG_STACK_TRACER=y
> CONFIG_DYNAMIC_FTRACE=y
> CONFIG_FTRACE_MCOUNT_RECORD=y
> 
> However when I look at  /sys/kernel/debug/tracing/trace
> 
> I see output similar to this and  I don't see the "stack trace"
>             init-1     [000]    65.317499: sys_write <-sysenter_do_call
>             init-1     [000]    65.317504: vfs_write <-sys_write
>             init-1     [000]    65.317617: sys_write <-sysenter_do_call
>             init-1     [000]    65.317619: vfs_write <-sys_write
> 
> Is there option I am missing while compiling kernel or some part of
> kernel parameters is missing? Is there way to explicitly specify
> iter_ctrl on kernel boot cmdline?

Currently there is no way to enable an option from the kernel command line. I could
write something up though.

-- Steve