> syscall nr and pid at minimum then.

oprofile already supports logging the pid I believe. Otherwise
the pid filter in opreport could hardly work.

quoted text
> Still doesn't work for modules either.

oprofile works fine for modules.

quoted text
> 
> what it ends up doing is using an entirely different interface for 
> basically the
> same code / operation inside the kernel.

Well rather it uses an existing framework for something that fits
it well.

Also the way I proposed is very cheap and would be possible
to use in production kernels without special configs.

quoted text
> The current interface code is maybe 80 lines of /proc code... and very 
> simple to
> use (unlike the oprofile interface)

The oprofile interface is per CPU (so you wouldn't need to reinvent
that to fix your locking) and if you add the syscall
logging feature to it it would apply to all profile events
e.g. you could then do things like "matching cache misses to syscalls" 

-andi

--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/