Re: [PATCH] perf header: Fix lock/unlock imbalances
From: Gustavo A. R. Silva
Date: Mon Apr 08 2019 - 16:33:07 EST
On 4/8/19 3:00 PM, Arnaldo Carvalho de Melo wrote:
> Em Mon, Apr 08, 2019 at 02:52:52PM -0500, Gustavo A. R. Silva escreveu:
>>
>>
>> On 4/8/19 2:35 PM, Arnaldo Carvalho de Melo wrote:
>>> Em Mon, Apr 08, 2019 at 01:26:09PM -0500, Gustavo A. R. Silva escreveu:
>>>>
>>>>
>>>> On 4/8/19 1:22 PM, Song Liu wrote:
>>>>>
>>>>>
>>>>>> On Apr 8, 2019, at 10:33 AM, Gustavo A. R. Silva <gustavo@xxxxxxxxxxxxxx> wrote:
>>>>>>
>>>>>> Fix lock/unlock imbalances by refactoring the code a bit and adding
>>>>>> calls to up_write() before return.
>>>>>>
>>>>>> Addresses-Coverity-ID: 1444315 ("Missing unlock")
>>>>>> Addresses-Coverity-ID: 1444316 ("Missing unlock")
>>>>>> Fixes: a70a1123174a ("perf bpf: Save BTF information as headers to perf.data")
>>>>>> Fixes: 606f972b1361 ("perf bpf: Save bpf_prog_info information as headers to perf.data")
>>>>>> Signed-off-by: Gustavo A. R. Silva <gustavo@xxxxxxxxxxxxxx>
>>>>>
>>>>> Acked-by: Song Liu <songliubraving@xxxxxx>
>>>>>
>>>>> Thanks for the fix!
>>>>>
>>>>
>>>> Glad to help. :)
>>>
>>> Super cool, using the same idiom as the kernel and living in the kernel
>>> sources has its advantages 8-)
>>>
>>
>> :P
>>
>>> But see below,
>>>
>>>>>> +++ b/tools/perf/util/header.c
>>>>>> @@ -2606,6 +2606,7 @@ static int process_bpf_prog_info(struct feat_fd *ff, void *data __maybe_unused)
>>>>>> perf_env__insert_bpf_prog_info(env, info_node);
>>>>>> }
>>>>>>
>>>>>> + up_write(&env->bpf_progs.lock);
>>>>>> return 0;
>>>>>> out:
>>>>>> free(info_linear);
>>>>>> @@ -2623,7 +2624,9 @@ static int process_bpf_prog_info(struct feat_fd *ff __maybe_unused, void *data _
>>>>>> static int process_bpf_btf(struct feat_fd *ff, void *data __maybe_unused)
>>>>>> {
>>>>>> struct perf_env *env = &ff->ph->env;
>>>>>> + struct btf_node *node;
>>>>>> u32 count, i;
>>>>>> + int err = -1;
>>>
>>> Why are you using this 'err' variable? It is only set here and at the
>>> end, i.e. one write, one read. We could as well have that out: block
>>> return -1 straight away.
>>>
>>> Else we could do, see below
>>>
>>>>>>
>>>>>> if (ff->ph->needs_swap) {
>>>>>> pr_warning("interpreting btf from systems with endianity is not yet supported\n");
>>>>>> @@ -2636,31 +2639,33 @@ static int process_bpf_btf(struct feat_fd *ff, void *data __maybe_unused)
>>>>>> down_write(&env->bpf_progs.lock);
>>>>>>
>>>>>> for (i = 0; i < count; ++i) {
>>>>>> - struct btf_node *node;
>>>>>> u32 id, data_size;
>>>>>>
>>>>>> + node = NULL;
>>>>>> if (do_read_u32(ff, &id))
>>>>>> - return -1;
>>>>>> + goto out;
>>>>>> if (do_read_u32(ff, &data_size))
>>>>>> - return -1;
>>>>>> + goto out;
>>>>>>
>>>>>> node = malloc(sizeof(struct btf_node) + data_size);
>>>>>> if (!node)
>>>>>> - return -1;
>>>>>> + goto out;
>>>>>>
>>>>>> node->id = id;
>>>>>> node->data_size = data_size;
>>>>>>
>>>>>> - if (__do_read(ff, node->data, data_size)) {
>>>>>> - free(node);
>>>>>> - return -1;
>>>>>> - }
>>>>>> + if (__do_read(ff, node->data, data_size))
>>>>>> + goto out;
>>>>>>
>>>>>> perf_env__insert_btf(env, node);
>>>>>> }
>>>
>>> err = 0;
>>>
>>>>>>
>>>
>>> out:
>>>
>>>>>> up_write(&env->bpf_progs.lock);
>>>
>>> return err;
>>>
>>> And delete the rest.
>>>
>>> but I see, you used the same pattern in the first #ifdef HAVE_LIBBPF_SUPPORT
>>> block :-)
>>>
>>> Anyway, since we're fixing up that other case, we might as well
>>> streamline this, please check the patch below.
>>>
>>
>> Yeah. This is exactly how I would have coded this from the beginning. But, as you
>> correctly pointed out, I'm using the same pattern as in HAVE_LIBBPF_SUPPORT. :)
>>
>> Just a comment below...
>>
>>>>>> return 0;
>>>
>>>>>> +out:
>>>>>> + up_write(&env->bpf_progs.lock);
>>>>>> + free(node);
>>>>>> + return err;
>>>
>>> So, that is what I'm applying, please holler if I introduced some
>>> problem:
>>>
>>> diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
>>> index b9e693825873..2d2af2ac2b1e 100644
>>> --- a/tools/perf/util/header.c
>>> +++ b/tools/perf/util/header.c
>>> @@ -2606,6 +2606,7 @@ static int process_bpf_prog_info(struct feat_fd *ff, void *data __maybe_unused)
>>> perf_env__insert_bpf_prog_info(env, info_node);
>>> }
>>>
>>> + up_write(&env->bpf_progs.lock);
>>> return 0;
>>> out:
>>> free(info_linear);
>>> @@ -2623,7 +2624,9 @@ static int process_bpf_prog_info(struct feat_fd *ff __maybe_unused, void *data _
>>> static int process_bpf_btf(struct feat_fd *ff, void *data __maybe_unused)
>>> {
>>> struct perf_env *env = &ff->ph->env;
>>> + struct btf_node *node = NULL;
>>> u32 count, i;
>>> + int err = -1;
>>>
>>> if (ff->ph->needs_swap) {
>>> pr_warning("interpreting btf from systems with endianity is not yet supported\n");
>>> @@ -2636,31 +2639,32 @@ static int process_bpf_btf(struct feat_fd *ff, void *data __maybe_unused)
>>> down_write(&env->bpf_progs.lock);
>>>
>>> for (i = 0; i < count; ++i) {
>>> - struct btf_node *node;
>>> u32 id, data_size;
>>>
>>> if (do_read_u32(ff, &id))
>>> - return -1;
>>> + goto out;
>>> if (do_read_u32(ff, &data_size))
>>> - return -1;
>>> + goto out;
>>>
>>> node = malloc(sizeof(struct btf_node) + data_size);
>>> if (!node)
>>> - return -1;
>>> + goto out;
>>>
>>> node->id = id;
>>> node->data_size = data_size;
>>>
>>> - if (__do_read(ff, node->data, data_size)) {
>>> - free(node);
>>> - return -1;
>>> - }
>>> + if (__do_read(ff, node->data, data_size))
>>> + goto out;
>>>
>>> perf_env__insert_btf(env, node);
>>> + node = NULL;
>>
>> If we move this assignment to the beginning of the for loop, as in
>> the original patch, we avoid the same assignment while declaring
>> node at the beginning of the function.
>
> No, we don't, since the common exit path frees node, we better not free
> the last node in the success case, that is why I moved it to the end,
> i.e. after we're done with it, nullify it, so that the last btf_node
> isn't freed in the now uncoditionall free(node); call :-)
>
Yep. You're right.
So, everything is fine now. :)
Thanks
--
Gustavo