lkml.org 
[lkml]   [2010]   [Aug]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/4] oprofile: Handle initialisation failure more gracefully
On 26.08.10 15:09:16, Matt Fleming wrote:
> From: Will Deacon <will.deacon@arm.com>
>
> The current implementation is not entirely safe in the case that
> oprofile_arch_init() fails. We need to make sure that we always call
> exit_driverfs() if we've called init_driverfs(). Also, avoid a potential
> double free when freeing 'counter_config', e.g. don't free
> 'counter_config' in both oprofile_arch_init() and oprofile_arch_exit().
>
> Signed-off-by: Will Deacon <will.deacon@arm.com>
> ---
> arch/arm/oprofile/common.c | 15 ++++++++-------
> 1 files changed, 8 insertions(+), 7 deletions(-)
>
> diff --git a/arch/arm/oprofile/common.c b/arch/arm/oprofile/common.c
> index 0691176..482779c 100644
> --- a/arch/arm/oprofile/common.c
> +++ b/arch/arm/oprofile/common.c
> @@ -275,10 +275,12 @@ out:
> return ret;
> }
>
> -static void exit_driverfs(void)
> +static void exit_driverfs(void)
> {
> - platform_device_unregister(oprofile_pdev);
> - platform_driver_unregister(&oprofile_driver);
> + if (!IS_ERR_OR_NULL(oprofile_pdev)) {
> + platform_device_unregister(oprofile_pdev);
> + platform_driver_unregister(&oprofile_driver);
> + }

The root cause that makes this check necessary is that
oprofile_arch_exit() is called though oprofile_arch_init() failed. We
should better fix this instead. I have to admit we will then have to
check all architectural implementations.

> }
> #else
> static int __init init_driverfs(void) { return 0; }
> @@ -363,10 +365,8 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
> }
>
> ret = init_driverfs();
> - if (ret) {
> - kfree(counter_config);

We should not return from oprofile_arch_init() with allocated
resources if the function fails. To fix duplicate kfrees, we should
free it here and then set counter_config to NULL. It should also be
freed if for_each_possible_cpu() or op_name_from_perf_id() fails.

Also, the pointer should be NULLed after freeing in
oprofile_arch_exit(). There, we also don't need the NULL pointer check
as it is save to call kfree(NULL).

-Robert

> + if (ret)
> return ret;
> - }
>
> for_each_possible_cpu(cpu) {
> perf_events[cpu] = kcalloc(perf_num_counters,
> @@ -401,8 +401,9 @@ void oprofile_arch_exit(void)
> int cpu, id;
> struct perf_event *event;
>
> + exit_driverfs();
> +
> if (*perf_events) {
> - exit_driverfs();
> for_each_possible_cpu(cpu) {
> for (id = 0; id < perf_num_counters; ++id) {
> event = perf_events[cpu][id];
> --
> 1.7.1
>
>

--
Advanced Micro Devices, Inc.
Operating System Research Center



\
 
 \ /
  Last update: 2010-08-27 14:59    [W:1.883 / U:0.580 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site