Re: [RFC bpf-next 00/16] bpf: Speed up trampoline attach

From
Steven Rostedt <rostedt@goodmis.org>
Date
2020-10-22 16:21:57
Also in
bpf
Reply to
Steven Rostedt, 2020-10-22
On Thu, 22 Oct 2020 10:42:05 -0400
Steven Rostedt [off-list ref] wrote:
I'd like to see how batch functions will work. I guess I need to start
looking at the bpf trampoline, to see if we can modify the ftrace
trampoline to have a quick access to parameters. It would be much more
beneficial to update the existing generic function tracer to have access to
function parameters that all users could benefit from, than to tweak a
single use case into giving this feature to a single user.
Looking at the creation of the bpf trampoline, I think I can modify ftrace
to have a more flexible callback. Something that passes the callback the
following:

 the function being traced.
 a pointer to the parent caller (that could be modified)
 a pointer to the original stack frame (what the stack was when the
      function is entered)
 An array of the arguments of the function (which could also be modified)

This is a change I've been wanting to make for some time, because it would
allow function graph to be a user of function tracer, and would give
everything access to the arguments.

We would still need a helper function to store all regs to keep kprobes
working unmodified, but this would still only be done if asked.

The above change shouldn't hurt performance for either ftrace or bpf
because it appears they both do the same. If BPF wants to have a batch
processing of functions, then I think we should modify ftrace to do this
new approach instead of creating another set of function trampolines.

-- Steve
Thread (47 messages) 47 messages, 5 authors, last reply 2020-10-29
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help