linux-kernel - Re: [PATCH v7 8/8] perf record: Unmap overwrite evlist when event terminate

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <5767A3A9.5010708@huawei.com>
Date:	Mon, 20 Jun 2016 16:04:57 +0800
From:	"Wangnan (F)" <wangnan0@...wei.com>
To:	Arnaldo Carvalho de Melo <acme@...hat.com>
CC:	<acme@...nel.org>, <pi3orama@....com>,
	<linux-kernel@...r.kernel.org>, "He Kuang" <hekuang@...wei.com>,
	Jiri Olsa <jolsa@...nel.org>,
	Masami Hiramatsu <mhiramat@...nel.org>,
	Namhyung Kim <namhyung@...nel.org>,
	Zefan Li <lizefan@...wei.com>
Subject: Re: [PATCH v7 8/8] perf record: Unmap overwrite evlist when event
 terminate



On 2016/6/17 4:59, Arnaldo Carvalho de Melo wrote:
> Em Wed, Jun 15, 2016 at 02:23:35AM +0000, Wang Nan escreveu:
>> When see POLLERR or POLLHUP, unmap ring buffer from both the main
>> evlist and overwrite evlist.
> When you use an auxiliary evlist this makes evlist->parent be different
> than evlist, right? So can't we just hide this from tools and do it all
> in perf_evlist__filter_pollfd?

So we need to find all children from parent. Currently we can only track
parent from children. This is not enough. Although currently we can pass
rec->overwrite_evlist to the filter so perf_evlist__filter_pollfd knows
it should release mmaps from both evlists, if in future we have more than
one aux_evlists it doesn't work.

Thank you.


> This way we will not need to expose perf_evlist__mmap_put(), etc.
>
> I haven't tried to _actually_ do this, just asking from browsing the
> patchkit and with the goal of hiding the internals of this thing inside
> evlist.c as much as possible.
>
> - Arnaldo
>   
>> Signed-off-by: Wang Nan <wangnan0@...wei.com>
>> Cc: He Kuang <hekuang@...wei.com>
>> Cc: Arnaldo Carvalho de Melo <acme@...hat.com>
>> Cc: Jiri Olsa <jolsa@...nel.org>
>> Cc: Masami Hiramatsu <mhiramat@...nel.org>
>> Cc: Namhyung Kim <namhyung@...nel.org>
>> Cc: Zefan Li <lizefan@...wei.com>
>> Cc: pi3orama@....com
>> ---
>>   tools/perf/builtin-record.c | 30 +++++++++++++++++++++---------
>>   tools/perf/util/evlist.c    |  3 +--
>>   tools/perf/util/evlist.h    |  2 +-
>>   3 files changed, 23 insertions(+), 12 deletions(-)
>>
>> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
>> index b9094f0..ca6376c 100644
>> --- a/tools/perf/builtin-record.c
>> +++ b/tools/perf/builtin-record.c
>> @@ -914,6 +914,26 @@ out:
>>   	return err;
>>   }
>>   
>> +static void record__munmap_filtered(struct fdarray *fda, int fd, void *arg)
>> +{
>> +	struct record *rec = (struct record *)arg;
>> +
>> +	perf_evlist__mmap_put(rec->evlist, fda->priv[fd].idx);
>> +	if (rec->overwrite_evlist)
>> +		perf_evlist__mmap_put(rec->overwrite_evlist, fda->priv[fd].idx);
>> +}
>> +
>> +static int record__filter_pollfd(struct record *rec)
>> +{
>> +	/*
>> +	 * Although we may have auxiliray evlist, there is
>> +	 * only one pollfd, so we don't need to filter pollfd
>> +	 * for auxiliray evlist.
>> +	 */
>> +	return fdarray__filter(&rec->evlist->pollfd, POLLERR | POLLHUP,
>> +			       record__munmap_filtered, rec);
>> +}
>> +
>>   static int __cmd_record(struct record *rec, int argc, const char **argv)
>>   {
>>   	int err;
>> @@ -1150,15 +1170,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
>>   				err = 0;
>>   			waking++;
>>   
>> -			/*
>> -			 * Although we may have auxiliray evlist, there is
>> -			 * only one pollfd, so we don't need to filter pollfd
>> -			 * for auxiliray evlist.
>> -			 *
>> -			 * TODO: if an event is terminated (POLLERR | POLLHUP),
>> -			 * unmap mmaps for auxiliray evlist too.
>> -			 */
>> -			if (perf_evlist__filter_pollfd(rec->evlist, POLLERR | POLLHUP) == 0)
>> +			if (record__filter_pollfd(rec) == 0)
>>   				draining = true;
>>   		}
>>   
>> diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
>> index e8fcb22..d43ee81 100644
>> --- a/tools/perf/util/evlist.c
>> +++ b/tools/perf/util/evlist.c
>> @@ -27,7 +27,6 @@
>>   #include <linux/log2.h>
>>   #include <linux/err.h>
>>   
>> -static void perf_evlist__mmap_put(struct perf_evlist *evlist, int idx);
>>   static void __perf_evlist__munmap(struct perf_evlist *evlist, int idx);
>>   
>>   #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
>> @@ -863,7 +862,7 @@ static void perf_evlist__mmap_get(struct perf_evlist *evlist, int idx)
>>   	atomic_inc(&evlist->mmap[idx].refcnt);
>>   }
>>   
>> -static void perf_evlist__mmap_put(struct perf_evlist *evlist, int idx)
>> +void perf_evlist__mmap_put(struct perf_evlist *evlist, int idx)
>>   {
>>   	struct perf_mmap *md = &evlist->mmap[idx];
>>   
>> diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
>> index 41e65ac..ba5e006 100644
>> --- a/tools/perf/util/evlist.h
>> +++ b/tools/perf/util/evlist.h
>> @@ -141,7 +141,7 @@ union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist,
>>   void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx);
>>   
>>   void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx);
>> -
>> +void perf_evlist__mmap_put(struct perf_evlist *evlist, int idx);
>>   int perf_evlist__pause(struct perf_evlist *evlist);
>>   int perf_evlist__resume(struct perf_evlist *evlist);
>>   int perf_evlist__open(struct perf_evlist *evlist);
>> -- 
>> 1.8.3.4