Hi Matt,
I'm still not happy with the init/exit alloc/free code:
On Thu, 2010-08-26 at 20:09 +0100, Matt Fleming wrote:
So here, if the perf_events allocation fails for a cpu, we free the
stuff we've already allocated [including counter_config] and return
-ENOMEM. Looking at drivers/oprofile/oprof.c:
static int __init oprofile_init(void)
{
int err;
err = oprofile_arch_init(&oprofile_ops);
if (err < 0 || timer) {
printk(KERN_INFO "oprofile: using timer interrupt.\n");
err = oprofile_timer_init(&oprofile_ops);
if (err)
goto out_arch;
}
err = oprofilefs_register();
if (err)
goto out_arch;
return 0;
out_arch:
oprofile_arch_exit();
return err;
}
So now, if timer_init fails or oprofilefs_register fails, we will
meaning that we will free everything again! This is what I
was trying to avoid in patch 1/4, by moving the counter_config
freeing into the *_exit function. Looking at it again, I think
all the freeing should be moved to the *_exit function and the init
function should just return error when allocation fails. What do you
typo :)
For what it's worth, I tested the series on my Cortex-A9 board and
everything seemed to work fine. I'll give the patches another spin when
we've sorted out these memory issues.
Cheers,
Will
--
