Re: [PATCH RESEND v3] perf/core: Fix installing cgroup event into cpu

[kbuild test robot <lkp@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 3831 bytes --]

Hi leilei.lin,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on tip/perf/core]
[also build test ERROR on v4.15 next-20180202]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/linxiulei-gmail-com/perf-core-Fix-installing-cgroup-event-into-cpu/20180203-133110
config: i386-randconfig-s0-201804 (attached as .config)
compiler: gcc-6 (Debian 6.4.0-9) 6.4.0 20171026
reproduce:
        # save the attached .config to linux build tree
        make ARCH=i386 

All errors (new ones prefixed by >>):

   kernel/events/core.c: In function '__perf_install_in_context':
>> kernel/events/core.c:2332:10: error: implicit declaration of function 'perf_cgroup_from_task' [-Werror=implicit-function-declaration]
      cgrp = perf_cgroup_from_task(current, ctx);
             ^~~~~~~~~~~~~~~~~~~~~
   kernel/events/core.c:2332:8: warning: assignment makes pointer from integer without a cast [-Wint-conversion]
      cgrp = perf_cgroup_from_task(current, ctx);
           ^
   kernel/events/core.c:2333:40: error: dereferencing pointer to incomplete type 'struct perf_cgroup'
      reprogram = cgroup_is_descendant(cgrp->css.cgroup,
                                           ^~
   kernel/events/core.c:2334:11: error: 'struct perf_event' has no member named 'cgrp'
         event->cgrp->css.cgroup);
              ^~
   cc1: some warnings being treated as errors

vim +/perf_cgroup_from_task +2332 kernel/events/core.c

  2284	
  2285	/*
  2286	 * Cross CPU call to install and enable a performance event
  2287	 *
  2288	 * Very similar to remote_function() + event_function() but cannot assume that
  2289	 * things like ctx->is_active and cpuctx->task_ctx are set.
  2290	 */
  2291	static int  __perf_install_in_context(void *info)
  2292	{
  2293		struct perf_event *event = info;
  2294		struct perf_event_context *ctx = event->ctx;
  2295		struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
  2296		struct perf_event_context *task_ctx = cpuctx->task_ctx;
  2297		struct perf_cgroup *cgrp;
  2298		bool reprogram = true;
  2299		int ret = 0;
  2300	
  2301		raw_spin_lock(&cpuctx->ctx.lock);
  2302		if (ctx->task) {
  2303			raw_spin_lock(&ctx->lock);
  2304			task_ctx = ctx;
  2305	
  2306			reprogram = (ctx->task == current);
  2307	
  2308			/*
  2309			 * If the task is running, it must be running on this CPU,
  2310			 * otherwise we cannot reprogram things.
  2311			 *
  2312			 * If its not running, we don't care, ctx->lock will
  2313			 * serialize against it becoming runnable.
  2314			 */
  2315			if (task_curr(ctx->task) && !reprogram) {
  2316				ret = -ESRCH;
  2317				goto unlock;
  2318			}
  2319	
  2320			WARN_ON_ONCE(reprogram && cpuctx->task_ctx && cpuctx->task_ctx != ctx);
  2321		} else if (task_ctx) {
  2322			raw_spin_lock(&task_ctx->lock);
  2323		}
  2324	
  2325		if (is_cgroup_event(event)) {
  2326			/*
  2327			 * Only care about cgroup events.
  2328			 *
  2329			 * If only the task belongs to cgroup of this event,
  2330			 * we will continue the installment
  2331			 */
> 2332			cgrp = perf_cgroup_from_task(current, ctx);
  2333			reprogram = cgroup_is_descendant(cgrp->css.cgroup,
  2334						event->cgrp->css.cgroup);
  2335		}
  2336	
  2337		if (reprogram) {
  2338			ctx_sched_out(ctx, cpuctx, EVENT_TIME);
  2339			add_event_to_ctx(event, ctx);
  2340			ctx_resched(cpuctx, task_ctx, get_event_type(event));
  2341		} else {
  2342			add_event_to_ctx(event, ctx);
  2343		}
  2344	
  2345	unlock:
  2346		perf_ctx_unlock(cpuctx, task_ctx);
  2347	
  2348		return ret;
  2349	}
  2350	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 30878 bytes --]