Re: [patch V2 28/29] stacktrace: Provide common infrastructure

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, 18 Apr 2019, Josh Poimboeuf wrote:

> On Thu, Apr 18, 2019 at 10:41:47AM +0200, Thomas Gleixner wrote:
> > All architectures which support stacktrace carry duplicated code and
> > do the stack storage and filtering at the architecture side.
> > 
> > Provide a consolidated interface with a callback function for consuming the
> > stack entries provided by the architecture specific stack walker. This
> > removes lots of duplicated code and allows to implement better filtering
> > than 'skip number of entries' in the future without touching any
> > architecture specific code.
> > 
> > Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > Cc: linux-arch@xxxxxxxxxxxxxxx
> 
> This is a step in the right direction, especially if it allows us to get
> rid of the 'skip' stuff.  But I'm not crazy about the callbacks.
> 
> Another idea I had (but never got a chance to work on) was to extend the
> x86 unwind interface to all arches.  So instead of the callbacks, each
> arch would implement something like this API:
> 
> 
> struct unwind_state state;
> 
> void unwind_start(struct unwind_state *state, struct task_struct *task,
> 		  struct pt_regs *regs, unsigned long *first_frame);
> 
> bool unwind_next_frame(struct unwind_state *state);
> 
> inline bool unwind_done(struct unwind_state *state);
> 
> 
> Not only would it avoid the callbacks (which is a nice benefit already),
> it would also allow the interfaces to be used outside of the
> stack_trace_*() interfaces.  That would come in handy in cases like the
> ftrace stack tracer code, which needs more than the stack_trace_*() API
> can give.

I surely thought about that, but after staring at all incarnations of
arch/*/stacktrace.c I just gave up.

Aside of that quite some archs already have callback based unwinders
because they use them for more than stacktracing and just have a single
implementation of that loop.

I'm fine either way. We can start with x86 and then let archs convert over
their stuff, but I wouldn't hold my breath that this will be completed in
the forseeable future.

> Of course, this may be more work than what you thought you signed up for
> ;-)

I did not sign up for anything. I tripped over that mess by accident and me
being me hated it strong enough to give it at least an initial steam blast.

Thanks,

	tglx
	



[Index of Archives]     [Linux Kernel]     [Kernel Newbies]     [x86 Platform Driver]     [Netdev]     [Linux Wireless]     [Netfilter]     [Bugtraq]     [Linux Filesystems]     [Yosemite Discussion]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]

  Powered by Linux